COWARDS AND HEROES GROUP LOYALTY IN THEAMERICAN CIVIL WAR

DORA L COSTA AND MATTHEW E KAHN

What motivated men to risk death in the most horric war in U S historywhen pay was low and irregular and military punishment strategies were weakIn such a situation creating group loyalty by promoting social capital is of para-mount importance and in the Civil War was the cement of both armies We ndthat individual and company socioeconomic and demographic characteristicsideology and morale were important predictors of group loyalty in the UnionArmy Company characteristics were more important than ideology or moraleSoldiers in companies that were more homogeneous in ethnicity occupation andage were less likely to shirk

I INTRODUCTION

Decisive battle in which two opposing forces meet face to faceuntil annihilation or surrender dominates western warfare strat-egy [Hanson 2001] Winning this type of battle requires amassingsufcient numbers of soldiers who will stand their ground Butthroughout history soldiers have frequently deserted and theirleaders have had to devote a great deal of attention to preventingdesertions At Agincourt a large number of the French cavalrysought refuge from the rain of arrows in a nearby wood AtWaterloo the Dutch-Belgian and minor German regiments delib-erately stayed out of the battle which was lost when Napoleanrsquosfamed Guard collapsed and ed from the steady musket reDuring World War I the main participants all lost their will toghtmdash over half of the French divisions on the Western Frontrebelled in May 1917 the Russian Army refused to ght in July1917 the Italian Second Army collapsed in November 1917 theBritish Fifth Army fell apart in March 1918 and decisively theGerman army in the west refused to continue the ght in October1918

What motivates soldiers to stand their ground Mercenaryarmies have been motivated by pay professional armies by pro-

We have beneted from the comments of Daron Acemoglu Eli BermanStanley Engerman Edward Glaeser Daniel Hamermesh Dean Karlan LawrenceKatz Joanna Lahey John Quigley Jesse Shapiro Peter Temin three anonymousreferees and seminar participants at the 2001 NBERDAE Summer InstituteMassachusetts Institute of Technology Boston University Harvard Universitythe University of Chicago and the University of California Berkeley Dora Costagratefully acknowledges the support of NIH grants AG12658 and AG10120

copy 2003 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics May 2003

519

motions and volunteers and draftees by punishments Battlepolice or even menrsquos commanding ofcers have stood behind themto prevent their running away During World War II not only didStalinrsquos armies have special detachments who formed a secondline to shoot at any soldiers in the rst line who ed but thefamilies of all deserters were also arrested [Beevor 1998] Democ-racies cannot inict such punishments and when ghting majorwars have never been very generous with pay Based in partupon questionnaires administered to World War II U S soldiersmany sociologists psychologists and military historians haveargued that soldiersrsquo primary motivation for ghting is intenseloyalty to the point of self-sacrice to a small band of comrades[McPherson 1997 p 86 Stouffer et al 1949 p 109] Becausesoldiers live with the same men for so long endangering thegroup leads to personal guilt and ostracism within the groupOliver Wendell Holmes who served as an ofcer in the Civil Warwept at not being able to be with his comrades at the battle ofFredericksburg where his regiment lost more men than in anyother engagement of the war [Menand 2001 p 43] Ideologicalfervor bolsters this loyalty Hanson [1999] argues that the moralvision commanders such as Sherman imparted to their troops ledto their victories Questionnaires administered to American vol-unteers in the Spanish Civil War found that ideology was thesingle most important factor helping men to overcome fear inbattle [Dollard 1943 p 555] Morale also matters The BritishFrench Italian and Russian armies of World War I cracked whenthe total number of deaths equaled the number of ghting infan-try in the divisions The Germans cracked later but only aftertheir armies were no longer victorious [Keegan 1976 p 276]Individual characteristics matter because they determine a sol-dierrsquos productivity Studies of American soldiers in World War IIfound combat performance to correlate positively with social classand education age and being married [Stouffer et al 1949 pp36ndash37]

This paper investigates the determinants of group loyaltyamong Union Army soldiers in the American Civil War studyingthe relative importance of individual and community character-istics of ideology and of morale to group loyalty among UnionArmy soldiers The Civil War was the most horric war in UnitedStates history The total number of deaths in the Civil Warequaled the total number killed in almost all other wars com-bined and more than one out of every ve white men participat-

520 QUARTERLY JOURNAL OF ECONOMICS

ing died over half of them from disease [Vinovskis 1990] Thecombatants faced death the hardships and monotony of camplife and distance from loved ones all for low and irregular pay Ifa Union Army soldier had deserted he would have faced only a 40percent chance of being caught and a negligible risk of death ifarrested [Linderman 1987 pp 174 176] A self-interested soldierwould have deserted But over 90 percent of all Union Armysoldiers did not [Linderman 1987] and among Union Army sol-diers whose three-year enlistment terms were up half of themreenlisted [McPherson 1997 pp 81ndash82]1 What motivated thesemen to remain loyal to the Union

This paper provides the rst large-scale quantitative assess-ment of the correlates of cowardice and heroism based uponsoldiersrsquo deeds rather than their words Loyalty is expressedthrough such actions as desertion arrests and absences withoutleave An unusually rich data set provides us with detailed de-mographic and economic characteristics of individuals of compa-nies and of the geographical areas from which individuals cameBecause companies contained only 100 men who were in constantclose contact we have a better measure of community than thoseoften used in the social capital literature Another advantage ofstudying group loyalty in this setting is that the stakes are highIt is costly for a military company if an individual shirks It is alsocostly for soldiers to do their duty thus allowing researchers toobtain a better measure of commitment than those commonlyused in the social capital and organizational behavior literature

Our analysis contributes to ongoing research on group loy-alty social capital and organizational design A growing litera-ture has examined loyalty to organizations as diverse as gangsHasidic Jews and corporations [Levitt and Venkatesh 2000 Ber-man 2000 Pfeffer 1997] A distinguishing characteristic betweenthe military and the modern rm is the militaryrsquos inability (ex-cept for a mercenary army) to fully compensate individuals forrisk and to link pay to performance In an organization whereworkers have discretion and unobserved effort matters altruismfor others and the need for othersrsquo respect will mitigate the

1 In contrast in the rst half of the eighteenth century around 20 percent ofthe French Army deserted and though no estimates are available the leaders ofother nations voiced laments about extremely high desertion rates [Sikora 1998]

521COWARDS AND HEROES

agency problem Social capital is therefore an important inputinto having a productive organization2

II EMPIRICAL FRAMEWORK

The Union Army like all organizations faced agency prob-lems The usual solutions for mitigating these problems such asbackloading pay using promotions as an incentive and payingbonuses to individuals [Lazear 1979 Gibbons 1998] were unlikelyto have been effective in the Civil War Army Soldiers who sur-vived expected to be discharged from the wartime military whentheir enlistment term was up were lucky if their pay arrived ontime and faced a higher risk of death on the battleeld if pro-moted because ofcers led the charges3 In addition militaryoutcomes are produced in a team setting in which one or moreregiments win or lose a battle In such a case where only teamoutput is observed and individual effort is not a for prot can usepay for performance incentives to induce the efcient level ofindividual effort [Holmstrom 1982] Unlike such an organizationthe military substitutes loyalty for high-powered incentives (seeKandel and Lazear [1992] for a theoretical analysis) This loyaltyneeds to be built within each company and cannot be purchasedin the market place

The four hypotheses that we will examine are that loyalty tothe Union was built through 1) soldiersrsquo ghting ability (as prox-ied by the individual characteristics of soldiers) 2) loyalty to asmall group (the community) 3) loyalty to a cause (ideology) and4) morale The empirical framework that we outline below willenable us to investigate the relative importance of each of thesehypotheses

Our empirical framework can be thought of in terms of thefollowing equations

(1) individual loyalty

5 f~social capital individual characteristics ideology morale

2 Social capital is dened as aspects of the social structure such as trustnetworks and conventions that encourage collaboration and coordination be-tween friends and strangers [Coleman 1990] OrsquoReilly Caldwell and Barnett[1989] nd that in work units where social integration is high turnover is low

3 While there may have been career benets to some men from beingperceived as war heroes this is unlikely to be true for farmers and they were inthe majority

522 QUARTERLY JOURNAL OF ECONOMICS

(2) social capital 5 g~community characteristics

where equation (1) represents an individualrsquos choice to be loyaland equation (2) models the determinants of social capital withina community Several recent studies emphasize that participa-tion is lower in more heterogeneous communities [Alesina and LaFerrara 2000 Costa and Kahn 2003] Since we do not explicitlymeasure the social capital embodied in the community we sub-stitute equation (2) into equation (1) and model loyalty as afunction of individual characteristics community characteristicsideology and morale Table I lists the sets of variables determin-ing group loyalty We will examine how these variables affect theconditional probability of desertion arrest or AWOL

Socioeconomic and demographic characteristics of soldierssuch as age or literacy may proxy for soldiersrsquo productivity (egolder soldiers may be more disciplined) whereas other character-istics such as social status or birthplace may affect group loyaltybecause they inuence ideas of patriotism honor and duty andshape soldiersrsquo ideology (see Table I for a list of individual char-acteristics that determine ghting ability) Married men may beeither more or less motivated to ght by the thought of loved onesIn the case of Civil War soldiers the sense of duty and honor andthe potential for public shame was greater among the more so-cially prominent Germans who ed the revolutions of 1848 weremore likely than Irish or British immigrants who migrated for

TABLE IDETERMINANTS OF GROUP LOYALTY

Individual Community Ideology Morale

Social statusOccupationFamily wealthLiteracy

NativityNative-bornGermanIrishEnglishOther

AgeMarital status

Birthplacefragmentation

Occupationalfragmentation

Age diversitySize of city of

enlistmentBrother in

companyPercent of own

nativityPercent of own

occupation

Year mustered inVolunteer statusFrom pro-Lincoln

county

Percent in companydying

Fraction Unionvictories

523COWARDS AND HEROES

economic reasons to view the United States as the best hope forthe survival of a form of republican government Protestant Ger-mans were more likely to be Republican than the Irish because alarge proportion of Republican voters were anti-Catholic Know-Nothings [Fogel 1989 p 384] Financial hardship at home ledsome married men to desert but this was truer of Confederatesoldiers whose families faced food shortages [McPherson 1997p 138]

Community characteristics inuence group participationWithin heterogeneous units team production may be harder be-cause there is less social integration and informal communica-tion If social capital is low team production may also be harderbecause social sanctions are less effective Our primary measureof a soldierrsquos community is which of the 303 companies in oursample he was in We examine the effect of such company char-acteristics as birthplace fragmentation economic fragmentation(proxied by occupational fragmentation) age diversity and thepercent of the company of own ethnicity and occupation on grouployalty4 Companies could increase social integration among like-minded individuals because soldiers formed their own groupswithin companies ranging from debate societies to Christianassociations We also investigate the impact of other denitions ofcommunity including whether the soldier had a brother fatheror son in the same company and population size of city of enlist-ment Among Civil War soldiers feelings of loyalty were com-pounded by community pressure since fellow soldiers from thesame hometown could and did report on othersrsquo behavior[McPherson 1997 pp 77ndash89] The size of the soldierrsquos town ofenlistment provides some indication whether the soldier facedthis kind of community pressure

The formation of communities (companies) during the warcan be thought of as an assignment problem An unusual featureof the Civil War military is that the federal government did notexplicitly control this assignmentmdashall company formation wasdone at the local level Because as we discuss later men hadsome control over what company to join this may raise concernsabout the exogeneity of community attributes If identical excel-lent ghters could Tiebout sort to form an exclusive company in

4 We cannot tell a priori whether such measures of community heterogene-ity as fragmentation indexes are better predictors of group loyalty than thepercent of the company of own ethnicity or occupation

524 QUARTERLY JOURNAL OF ECONOMICS

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

motions and volunteers and draftees by punishments Battlepolice or even menrsquos commanding ofcers have stood behind themto prevent their running away During World War II not only didStalinrsquos armies have special detachments who formed a secondline to shoot at any soldiers in the rst line who ed but thefamilies of all deserters were also arrested [Beevor 1998] Democ-racies cannot inict such punishments and when ghting majorwars have never been very generous with pay Based in partupon questionnaires administered to World War II U S soldiersmany sociologists psychologists and military historians haveargued that soldiersrsquo primary motivation for ghting is intenseloyalty to the point of self-sacrice to a small band of comrades[McPherson 1997 p 86 Stouffer et al 1949 p 109] Becausesoldiers live with the same men for so long endangering thegroup leads to personal guilt and ostracism within the groupOliver Wendell Holmes who served as an ofcer in the Civil Warwept at not being able to be with his comrades at the battle ofFredericksburg where his regiment lost more men than in anyother engagement of the war [Menand 2001 p 43] Ideologicalfervor bolsters this loyalty Hanson [1999] argues that the moralvision commanders such as Sherman imparted to their troops ledto their victories Questionnaires administered to American vol-unteers in the Spanish Civil War found that ideology was thesingle most important factor helping men to overcome fear inbattle [Dollard 1943 p 555] Morale also matters The BritishFrench Italian and Russian armies of World War I cracked whenthe total number of deaths equaled the number of ghting infan-try in the divisions The Germans cracked later but only aftertheir armies were no longer victorious [Keegan 1976 p 276]Individual characteristics matter because they determine a sol-dierrsquos productivity Studies of American soldiers in World War IIfound combat performance to correlate positively with social classand education age and being married [Stouffer et al 1949 pp36ndash37]

This paper investigates the determinants of group loyaltyamong Union Army soldiers in the American Civil War studyingthe relative importance of individual and community character-istics of ideology and of morale to group loyalty among UnionArmy soldiers The Civil War was the most horric war in UnitedStates history The total number of deaths in the Civil Warequaled the total number killed in almost all other wars com-bined and more than one out of every ve white men participat-

520 QUARTERLY JOURNAL OF ECONOMICS

ing died over half of them from disease [Vinovskis 1990] Thecombatants faced death the hardships and monotony of camplife and distance from loved ones all for low and irregular pay Ifa Union Army soldier had deserted he would have faced only a 40percent chance of being caught and a negligible risk of death ifarrested [Linderman 1987 pp 174 176] A self-interested soldierwould have deserted But over 90 percent of all Union Armysoldiers did not [Linderman 1987] and among Union Army sol-diers whose three-year enlistment terms were up half of themreenlisted [McPherson 1997 pp 81ndash82]1 What motivated thesemen to remain loyal to the Union

This paper provides the rst large-scale quantitative assess-ment of the correlates of cowardice and heroism based uponsoldiersrsquo deeds rather than their words Loyalty is expressedthrough such actions as desertion arrests and absences withoutleave An unusually rich data set provides us with detailed de-mographic and economic characteristics of individuals of compa-nies and of the geographical areas from which individuals cameBecause companies contained only 100 men who were in constantclose contact we have a better measure of community than thoseoften used in the social capital literature Another advantage ofstudying group loyalty in this setting is that the stakes are highIt is costly for a military company if an individual shirks It is alsocostly for soldiers to do their duty thus allowing researchers toobtain a better measure of commitment than those commonlyused in the social capital and organizational behavior literature

Our analysis contributes to ongoing research on group loy-alty social capital and organizational design A growing litera-ture has examined loyalty to organizations as diverse as gangsHasidic Jews and corporations [Levitt and Venkatesh 2000 Ber-man 2000 Pfeffer 1997] A distinguishing characteristic betweenthe military and the modern rm is the militaryrsquos inability (ex-cept for a mercenary army) to fully compensate individuals forrisk and to link pay to performance In an organization whereworkers have discretion and unobserved effort matters altruismfor others and the need for othersrsquo respect will mitigate the

1 In contrast in the rst half of the eighteenth century around 20 percent ofthe French Army deserted and though no estimates are available the leaders ofother nations voiced laments about extremely high desertion rates [Sikora 1998]

521COWARDS AND HEROES

agency problem Social capital is therefore an important inputinto having a productive organization2

II EMPIRICAL FRAMEWORK

The Union Army like all organizations faced agency prob-lems The usual solutions for mitigating these problems such asbackloading pay using promotions as an incentive and payingbonuses to individuals [Lazear 1979 Gibbons 1998] were unlikelyto have been effective in the Civil War Army Soldiers who sur-vived expected to be discharged from the wartime military whentheir enlistment term was up were lucky if their pay arrived ontime and faced a higher risk of death on the battleeld if pro-moted because ofcers led the charges3 In addition militaryoutcomes are produced in a team setting in which one or moreregiments win or lose a battle In such a case where only teamoutput is observed and individual effort is not a for prot can usepay for performance incentives to induce the efcient level ofindividual effort [Holmstrom 1982] Unlike such an organizationthe military substitutes loyalty for high-powered incentives (seeKandel and Lazear [1992] for a theoretical analysis) This loyaltyneeds to be built within each company and cannot be purchasedin the market place

The four hypotheses that we will examine are that loyalty tothe Union was built through 1) soldiersrsquo ghting ability (as prox-ied by the individual characteristics of soldiers) 2) loyalty to asmall group (the community) 3) loyalty to a cause (ideology) and4) morale The empirical framework that we outline below willenable us to investigate the relative importance of each of thesehypotheses

Our empirical framework can be thought of in terms of thefollowing equations

(1) individual loyalty

5 f~social capital individual characteristics ideology morale

2 Social capital is dened as aspects of the social structure such as trustnetworks and conventions that encourage collaboration and coordination be-tween friends and strangers [Coleman 1990] OrsquoReilly Caldwell and Barnett[1989] nd that in work units where social integration is high turnover is low

3 While there may have been career benets to some men from beingperceived as war heroes this is unlikely to be true for farmers and they were inthe majority

522 QUARTERLY JOURNAL OF ECONOMICS

(2) social capital 5 g~community characteristics

where equation (1) represents an individualrsquos choice to be loyaland equation (2) models the determinants of social capital withina community Several recent studies emphasize that participa-tion is lower in more heterogeneous communities [Alesina and LaFerrara 2000 Costa and Kahn 2003] Since we do not explicitlymeasure the social capital embodied in the community we sub-stitute equation (2) into equation (1) and model loyalty as afunction of individual characteristics community characteristicsideology and morale Table I lists the sets of variables determin-ing group loyalty We will examine how these variables affect theconditional probability of desertion arrest or AWOL

Socioeconomic and demographic characteristics of soldierssuch as age or literacy may proxy for soldiersrsquo productivity (egolder soldiers may be more disciplined) whereas other character-istics such as social status or birthplace may affect group loyaltybecause they inuence ideas of patriotism honor and duty andshape soldiersrsquo ideology (see Table I for a list of individual char-acteristics that determine ghting ability) Married men may beeither more or less motivated to ght by the thought of loved onesIn the case of Civil War soldiers the sense of duty and honor andthe potential for public shame was greater among the more so-cially prominent Germans who ed the revolutions of 1848 weremore likely than Irish or British immigrants who migrated for

TABLE IDETERMINANTS OF GROUP LOYALTY

Individual Community Ideology Morale

Social statusOccupationFamily wealthLiteracy

NativityNative-bornGermanIrishEnglishOther

AgeMarital status

Birthplacefragmentation

Occupationalfragmentation

Age diversitySize of city of

enlistmentBrother in

companyPercent of own

nativityPercent of own

occupation

Year mustered inVolunteer statusFrom pro-Lincoln

county

Percent in companydying

Fraction Unionvictories

523COWARDS AND HEROES

economic reasons to view the United States as the best hope forthe survival of a form of republican government Protestant Ger-mans were more likely to be Republican than the Irish because alarge proportion of Republican voters were anti-Catholic Know-Nothings [Fogel 1989 p 384] Financial hardship at home ledsome married men to desert but this was truer of Confederatesoldiers whose families faced food shortages [McPherson 1997p 138]

Community characteristics inuence group participationWithin heterogeneous units team production may be harder be-cause there is less social integration and informal communica-tion If social capital is low team production may also be harderbecause social sanctions are less effective Our primary measureof a soldierrsquos community is which of the 303 companies in oursample he was in We examine the effect of such company char-acteristics as birthplace fragmentation economic fragmentation(proxied by occupational fragmentation) age diversity and thepercent of the company of own ethnicity and occupation on grouployalty4 Companies could increase social integration among like-minded individuals because soldiers formed their own groupswithin companies ranging from debate societies to Christianassociations We also investigate the impact of other denitions ofcommunity including whether the soldier had a brother fatheror son in the same company and population size of city of enlist-ment Among Civil War soldiers feelings of loyalty were com-pounded by community pressure since fellow soldiers from thesame hometown could and did report on othersrsquo behavior[McPherson 1997 pp 77ndash89] The size of the soldierrsquos town ofenlistment provides some indication whether the soldier facedthis kind of community pressure

The formation of communities (companies) during the warcan be thought of as an assignment problem An unusual featureof the Civil War military is that the federal government did notexplicitly control this assignmentmdashall company formation wasdone at the local level Because as we discuss later men hadsome control over what company to join this may raise concernsabout the exogeneity of community attributes If identical excel-lent ghters could Tiebout sort to form an exclusive company in

4 We cannot tell a priori whether such measures of community heterogene-ity as fragmentation indexes are better predictors of group loyalty than thepercent of the company of own ethnicity or occupation

524 QUARTERLY JOURNAL OF ECONOMICS

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

ing died over half of them from disease [Vinovskis 1990] Thecombatants faced death the hardships and monotony of camplife and distance from loved ones all for low and irregular pay Ifa Union Army soldier had deserted he would have faced only a 40percent chance of being caught and a negligible risk of death ifarrested [Linderman 1987 pp 174 176] A self-interested soldierwould have deserted But over 90 percent of all Union Armysoldiers did not [Linderman 1987] and among Union Army sol-diers whose three-year enlistment terms were up half of themreenlisted [McPherson 1997 pp 81ndash82]1 What motivated thesemen to remain loyal to the Union

This paper provides the rst large-scale quantitative assess-ment of the correlates of cowardice and heroism based uponsoldiersrsquo deeds rather than their words Loyalty is expressedthrough such actions as desertion arrests and absences withoutleave An unusually rich data set provides us with detailed de-mographic and economic characteristics of individuals of compa-nies and of the geographical areas from which individuals cameBecause companies contained only 100 men who were in constantclose contact we have a better measure of community than thoseoften used in the social capital literature Another advantage ofstudying group loyalty in this setting is that the stakes are highIt is costly for a military company if an individual shirks It is alsocostly for soldiers to do their duty thus allowing researchers toobtain a better measure of commitment than those commonlyused in the social capital and organizational behavior literature

Our analysis contributes to ongoing research on group loy-alty social capital and organizational design A growing litera-ture has examined loyalty to organizations as diverse as gangsHasidic Jews and corporations [Levitt and Venkatesh 2000 Ber-man 2000 Pfeffer 1997] A distinguishing characteristic betweenthe military and the modern rm is the militaryrsquos inability (ex-cept for a mercenary army) to fully compensate individuals forrisk and to link pay to performance In an organization whereworkers have discretion and unobserved effort matters altruismfor others and the need for othersrsquo respect will mitigate the

1 In contrast in the rst half of the eighteenth century around 20 percent ofthe French Army deserted and though no estimates are available the leaders ofother nations voiced laments about extremely high desertion rates [Sikora 1998]

521COWARDS AND HEROES

agency problem Social capital is therefore an important inputinto having a productive organization2

II EMPIRICAL FRAMEWORK

The Union Army like all organizations faced agency prob-lems The usual solutions for mitigating these problems such asbackloading pay using promotions as an incentive and payingbonuses to individuals [Lazear 1979 Gibbons 1998] were unlikelyto have been effective in the Civil War Army Soldiers who sur-vived expected to be discharged from the wartime military whentheir enlistment term was up were lucky if their pay arrived ontime and faced a higher risk of death on the battleeld if pro-moted because ofcers led the charges3 In addition militaryoutcomes are produced in a team setting in which one or moreregiments win or lose a battle In such a case where only teamoutput is observed and individual effort is not a for prot can usepay for performance incentives to induce the efcient level ofindividual effort [Holmstrom 1982] Unlike such an organizationthe military substitutes loyalty for high-powered incentives (seeKandel and Lazear [1992] for a theoretical analysis) This loyaltyneeds to be built within each company and cannot be purchasedin the market place

The four hypotheses that we will examine are that loyalty tothe Union was built through 1) soldiersrsquo ghting ability (as prox-ied by the individual characteristics of soldiers) 2) loyalty to asmall group (the community) 3) loyalty to a cause (ideology) and4) morale The empirical framework that we outline below willenable us to investigate the relative importance of each of thesehypotheses

Our empirical framework can be thought of in terms of thefollowing equations

(1) individual loyalty

5 f~social capital individual characteristics ideology morale

2 Social capital is dened as aspects of the social structure such as trustnetworks and conventions that encourage collaboration and coordination be-tween friends and strangers [Coleman 1990] OrsquoReilly Caldwell and Barnett[1989] nd that in work units where social integration is high turnover is low

3 While there may have been career benets to some men from beingperceived as war heroes this is unlikely to be true for farmers and they were inthe majority

522 QUARTERLY JOURNAL OF ECONOMICS

(2) social capital 5 g~community characteristics

where equation (1) represents an individualrsquos choice to be loyaland equation (2) models the determinants of social capital withina community Several recent studies emphasize that participa-tion is lower in more heterogeneous communities [Alesina and LaFerrara 2000 Costa and Kahn 2003] Since we do not explicitlymeasure the social capital embodied in the community we sub-stitute equation (2) into equation (1) and model loyalty as afunction of individual characteristics community characteristicsideology and morale Table I lists the sets of variables determin-ing group loyalty We will examine how these variables affect theconditional probability of desertion arrest or AWOL

Socioeconomic and demographic characteristics of soldierssuch as age or literacy may proxy for soldiersrsquo productivity (egolder soldiers may be more disciplined) whereas other character-istics such as social status or birthplace may affect group loyaltybecause they inuence ideas of patriotism honor and duty andshape soldiersrsquo ideology (see Table I for a list of individual char-acteristics that determine ghting ability) Married men may beeither more or less motivated to ght by the thought of loved onesIn the case of Civil War soldiers the sense of duty and honor andthe potential for public shame was greater among the more so-cially prominent Germans who ed the revolutions of 1848 weremore likely than Irish or British immigrants who migrated for

TABLE IDETERMINANTS OF GROUP LOYALTY

Individual Community Ideology Morale

Social statusOccupationFamily wealthLiteracy

NativityNative-bornGermanIrishEnglishOther

AgeMarital status

Birthplacefragmentation

Occupationalfragmentation

Age diversitySize of city of

enlistmentBrother in

companyPercent of own

nativityPercent of own

occupation

Year mustered inVolunteer statusFrom pro-Lincoln

county

Percent in companydying

Fraction Unionvictories

523COWARDS AND HEROES

economic reasons to view the United States as the best hope forthe survival of a form of republican government Protestant Ger-mans were more likely to be Republican than the Irish because alarge proportion of Republican voters were anti-Catholic Know-Nothings [Fogel 1989 p 384] Financial hardship at home ledsome married men to desert but this was truer of Confederatesoldiers whose families faced food shortages [McPherson 1997p 138]

Community characteristics inuence group participationWithin heterogeneous units team production may be harder be-cause there is less social integration and informal communica-tion If social capital is low team production may also be harderbecause social sanctions are less effective Our primary measureof a soldierrsquos community is which of the 303 companies in oursample he was in We examine the effect of such company char-acteristics as birthplace fragmentation economic fragmentation(proxied by occupational fragmentation) age diversity and thepercent of the company of own ethnicity and occupation on grouployalty4 Companies could increase social integration among like-minded individuals because soldiers formed their own groupswithin companies ranging from debate societies to Christianassociations We also investigate the impact of other denitions ofcommunity including whether the soldier had a brother fatheror son in the same company and population size of city of enlist-ment Among Civil War soldiers feelings of loyalty were com-pounded by community pressure since fellow soldiers from thesame hometown could and did report on othersrsquo behavior[McPherson 1997 pp 77ndash89] The size of the soldierrsquos town ofenlistment provides some indication whether the soldier facedthis kind of community pressure

The formation of communities (companies) during the warcan be thought of as an assignment problem An unusual featureof the Civil War military is that the federal government did notexplicitly control this assignmentmdashall company formation wasdone at the local level Because as we discuss later men hadsome control over what company to join this may raise concernsabout the exogeneity of community attributes If identical excel-lent ghters could Tiebout sort to form an exclusive company in

4 We cannot tell a priori whether such measures of community heterogene-ity as fragmentation indexes are better predictors of group loyalty than thepercent of the company of own ethnicity or occupation

524 QUARTERLY JOURNAL OF ECONOMICS

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

agency problem Social capital is therefore an important inputinto having a productive organization2

II EMPIRICAL FRAMEWORK

The Union Army like all organizations faced agency prob-lems The usual solutions for mitigating these problems such asbackloading pay using promotions as an incentive and payingbonuses to individuals [Lazear 1979 Gibbons 1998] were unlikelyto have been effective in the Civil War Army Soldiers who sur-vived expected to be discharged from the wartime military whentheir enlistment term was up were lucky if their pay arrived ontime and faced a higher risk of death on the battleeld if pro-moted because ofcers led the charges3 In addition militaryoutcomes are produced in a team setting in which one or moreregiments win or lose a battle In such a case where only teamoutput is observed and individual effort is not a for prot can usepay for performance incentives to induce the efcient level ofindividual effort [Holmstrom 1982] Unlike such an organizationthe military substitutes loyalty for high-powered incentives (seeKandel and Lazear [1992] for a theoretical analysis) This loyaltyneeds to be built within each company and cannot be purchasedin the market place

The four hypotheses that we will examine are that loyalty tothe Union was built through 1) soldiersrsquo ghting ability (as prox-ied by the individual characteristics of soldiers) 2) loyalty to asmall group (the community) 3) loyalty to a cause (ideology) and4) morale The empirical framework that we outline below willenable us to investigate the relative importance of each of thesehypotheses

Our empirical framework can be thought of in terms of thefollowing equations

(1) individual loyalty

5 f~social capital individual characteristics ideology morale

2 Social capital is dened as aspects of the social structure such as trustnetworks and conventions that encourage collaboration and coordination be-tween friends and strangers [Coleman 1990] OrsquoReilly Caldwell and Barnett[1989] nd that in work units where social integration is high turnover is low

3 While there may have been career benets to some men from beingperceived as war heroes this is unlikely to be true for farmers and they were inthe majority

522 QUARTERLY JOURNAL OF ECONOMICS

(2) social capital 5 g~community characteristics

where equation (1) represents an individualrsquos choice to be loyaland equation (2) models the determinants of social capital withina community Several recent studies emphasize that participa-tion is lower in more heterogeneous communities [Alesina and LaFerrara 2000 Costa and Kahn 2003] Since we do not explicitlymeasure the social capital embodied in the community we sub-stitute equation (2) into equation (1) and model loyalty as afunction of individual characteristics community characteristicsideology and morale Table I lists the sets of variables determin-ing group loyalty We will examine how these variables affect theconditional probability of desertion arrest or AWOL

Socioeconomic and demographic characteristics of soldierssuch as age or literacy may proxy for soldiersrsquo productivity (egolder soldiers may be more disciplined) whereas other character-istics such as social status or birthplace may affect group loyaltybecause they inuence ideas of patriotism honor and duty andshape soldiersrsquo ideology (see Table I for a list of individual char-acteristics that determine ghting ability) Married men may beeither more or less motivated to ght by the thought of loved onesIn the case of Civil War soldiers the sense of duty and honor andthe potential for public shame was greater among the more so-cially prominent Germans who ed the revolutions of 1848 weremore likely than Irish or British immigrants who migrated for

TABLE IDETERMINANTS OF GROUP LOYALTY

Individual Community Ideology Morale

Social statusOccupationFamily wealthLiteracy

NativityNative-bornGermanIrishEnglishOther

AgeMarital status

Birthplacefragmentation

Occupationalfragmentation

Age diversitySize of city of

enlistmentBrother in

companyPercent of own

nativityPercent of own

occupation

Year mustered inVolunteer statusFrom pro-Lincoln

county

Percent in companydying

Fraction Unionvictories

523COWARDS AND HEROES

economic reasons to view the United States as the best hope forthe survival of a form of republican government Protestant Ger-mans were more likely to be Republican than the Irish because alarge proportion of Republican voters were anti-Catholic Know-Nothings [Fogel 1989 p 384] Financial hardship at home ledsome married men to desert but this was truer of Confederatesoldiers whose families faced food shortages [McPherson 1997p 138]

Community characteristics inuence group participationWithin heterogeneous units team production may be harder be-cause there is less social integration and informal communica-tion If social capital is low team production may also be harderbecause social sanctions are less effective Our primary measureof a soldierrsquos community is which of the 303 companies in oursample he was in We examine the effect of such company char-acteristics as birthplace fragmentation economic fragmentation(proxied by occupational fragmentation) age diversity and thepercent of the company of own ethnicity and occupation on grouployalty4 Companies could increase social integration among like-minded individuals because soldiers formed their own groupswithin companies ranging from debate societies to Christianassociations We also investigate the impact of other denitions ofcommunity including whether the soldier had a brother fatheror son in the same company and population size of city of enlist-ment Among Civil War soldiers feelings of loyalty were com-pounded by community pressure since fellow soldiers from thesame hometown could and did report on othersrsquo behavior[McPherson 1997 pp 77ndash89] The size of the soldierrsquos town ofenlistment provides some indication whether the soldier facedthis kind of community pressure

The formation of communities (companies) during the warcan be thought of as an assignment problem An unusual featureof the Civil War military is that the federal government did notexplicitly control this assignmentmdashall company formation wasdone at the local level Because as we discuss later men hadsome control over what company to join this may raise concernsabout the exogeneity of community attributes If identical excel-lent ghters could Tiebout sort to form an exclusive company in

4 We cannot tell a priori whether such measures of community heterogene-ity as fragmentation indexes are better predictors of group loyalty than thepercent of the company of own ethnicity or occupation

524 QUARTERLY JOURNAL OF ECONOMICS

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

(2) social capital 5 g~community characteristics

where equation (1) represents an individualrsquos choice to be loyaland equation (2) models the determinants of social capital withina community Several recent studies emphasize that participa-tion is lower in more heterogeneous communities [Alesina and LaFerrara 2000 Costa and Kahn 2003] Since we do not explicitlymeasure the social capital embodied in the community we sub-stitute equation (2) into equation (1) and model loyalty as afunction of individual characteristics community characteristicsideology and morale Table I lists the sets of variables determin-ing group loyalty We will examine how these variables affect theconditional probability of desertion arrest or AWOL

Socioeconomic and demographic characteristics of soldierssuch as age or literacy may proxy for soldiersrsquo productivity (egolder soldiers may be more disciplined) whereas other character-istics such as social status or birthplace may affect group loyaltybecause they inuence ideas of patriotism honor and duty andshape soldiersrsquo ideology (see Table I for a list of individual char-acteristics that determine ghting ability) Married men may beeither more or less motivated to ght by the thought of loved onesIn the case of Civil War soldiers the sense of duty and honor andthe potential for public shame was greater among the more so-cially prominent Germans who ed the revolutions of 1848 weremore likely than Irish or British immigrants who migrated for

TABLE IDETERMINANTS OF GROUP LOYALTY

Individual Community Ideology Morale

Social statusOccupationFamily wealthLiteracy

NativityNative-bornGermanIrishEnglishOther

AgeMarital status

Birthplacefragmentation

Occupationalfragmentation

Age diversitySize of city of

enlistmentBrother in

companyPercent of own

nativityPercent of own

occupation

Year mustered inVolunteer statusFrom pro-Lincoln

county

Percent in companydying

Fraction Unionvictories

523COWARDS AND HEROES

economic reasons to view the United States as the best hope forthe survival of a form of republican government Protestant Ger-mans were more likely to be Republican than the Irish because alarge proportion of Republican voters were anti-Catholic Know-Nothings [Fogel 1989 p 384] Financial hardship at home ledsome married men to desert but this was truer of Confederatesoldiers whose families faced food shortages [McPherson 1997p 138]

Community characteristics inuence group participationWithin heterogeneous units team production may be harder be-cause there is less social integration and informal communica-tion If social capital is low team production may also be harderbecause social sanctions are less effective Our primary measureof a soldierrsquos community is which of the 303 companies in oursample he was in We examine the effect of such company char-acteristics as birthplace fragmentation economic fragmentation(proxied by occupational fragmentation) age diversity and thepercent of the company of own ethnicity and occupation on grouployalty4 Companies could increase social integration among like-minded individuals because soldiers formed their own groupswithin companies ranging from debate societies to Christianassociations We also investigate the impact of other denitions ofcommunity including whether the soldier had a brother fatheror son in the same company and population size of city of enlist-ment Among Civil War soldiers feelings of loyalty were com-pounded by community pressure since fellow soldiers from thesame hometown could and did report on othersrsquo behavior[McPherson 1997 pp 77ndash89] The size of the soldierrsquos town ofenlistment provides some indication whether the soldier facedthis kind of community pressure

The formation of communities (companies) during the warcan be thought of as an assignment problem An unusual featureof the Civil War military is that the federal government did notexplicitly control this assignmentmdashall company formation wasdone at the local level Because as we discuss later men hadsome control over what company to join this may raise concernsabout the exogeneity of community attributes If identical excel-lent ghters could Tiebout sort to form an exclusive company in

4 We cannot tell a priori whether such measures of community heterogene-ity as fragmentation indexes are better predictors of group loyalty than thepercent of the company of own ethnicity or occupation

524 QUARTERLY JOURNAL OF ECONOMICS

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

economic reasons to view the United States as the best hope forthe survival of a form of republican government Protestant Ger-mans were more likely to be Republican than the Irish because alarge proportion of Republican voters were anti-Catholic Know-Nothings [Fogel 1989 p 384] Financial hardship at home ledsome married men to desert but this was truer of Confederatesoldiers whose families faced food shortages [McPherson 1997p 138]

Community characteristics inuence group participationWithin heterogeneous units team production may be harder be-cause there is less social integration and informal communica-tion If social capital is low team production may also be harderbecause social sanctions are less effective Our primary measureof a soldierrsquos community is which of the 303 companies in oursample he was in We examine the effect of such company char-acteristics as birthplace fragmentation economic fragmentation(proxied by occupational fragmentation) age diversity and thepercent of the company of own ethnicity and occupation on grouployalty4 Companies could increase social integration among like-minded individuals because soldiers formed their own groupswithin companies ranging from debate societies to Christianassociations We also investigate the impact of other denitions ofcommunity including whether the soldier had a brother fatheror son in the same company and population size of city of enlist-ment Among Civil War soldiers feelings of loyalty were com-pounded by community pressure since fellow soldiers from thesame hometown could and did report on othersrsquo behavior[McPherson 1997 pp 77ndash89] The size of the soldierrsquos town ofenlistment provides some indication whether the soldier facedthis kind of community pressure

The formation of communities (companies) during the warcan be thought of as an assignment problem An unusual featureof the Civil War military is that the federal government did notexplicitly control this assignmentmdashall company formation wasdone at the local level Because as we discuss later men hadsome control over what company to join this may raise concernsabout the exogeneity of community attributes If identical excel-lent ghters could Tiebout sort to form an exclusive company in

4 We cannot tell a priori whether such measures of community heterogene-ity as fragmentation indexes are better predictors of group loyalty than thepercent of the company of own ethnicity or occupation

524 QUARTERLY JOURNAL OF ECONOMICS

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

order to maximize their survival probabilities then both commu-nity heterogeneity and desertion would be low and we wouldmislabel this sorting on unobservables as social capital In thiscase a personrsquos desertion probability and the community hetero-geneity measure would be simultaneously determined ratherthan community social capital having a causal impact on laterwar effort5 While we recognize this possibility our empiricaldesign minimizes its relevance The Civil War Army was com-posed of civilians Enlistees could not know whether their friendshad any combat skills6 Although there was some sorting alongethnic lines nding a company that was a good match waspartially a matter of luck7 As the war progressed some individ-uals would enlist away from home to receive another townrsquoshigher enlistment bounty thus providing additional variation Inour regression models we will control for individual attributesand will perform robustness tests where we control for meancompany observable skill proxies such as percent farmer or per-cent foreign-born

Ideology mitigates the agency problem because it raises loy-alty During the American Civil War not just own ideology butalso ideology of the soldiersrsquo hometown was an important factorSoldiersrsquo morale depended not just upon good news from the frontbut also upon their familiesrsquo and communitiesrsquo support We mea-sure ideology using year of enlistment volunteer status andpercent of the county voting for Lincoln Men who enlisted after1862 were commonly described as being without patriotismhonor or interest in the cause [McPherson 1997 p 9] We recog-nize that this variable might be measuring factors other thanideology such as an inux of inferior recruits or an inux ofrecruits who did not enlist together However we nd that ourresults remain unchanged when we analyze late or early recruits

5 Ichino and Maggi [2000] used the records of an Italian rm to examine howindividuals who entered an organization performed In the Civil War men rarelytransferred and men who died were not replaced We only have information onthe men within a given company

6 We have not been able to nd any references in any of the regimentalhistories to men sorting into companies on the basis of combat skills

7 One soldier wrote home ldquoWe have a remarkable civil and Religious com-pany I think it is a providencial circumstance that I enlisted in this companyfor I hear that there is desperate wickedness in very regiments I came so nearenlisting inrdquo (Letter of David Close November 4 1862 126th Ohio VolunteerInfantry Company D httpwwwiwaynetedulsci

525COWARDS AND HEROES

only8 The constituencies voting for Lincoln were diverse consist-ing of anti-Catholics farmers and land reformers among othersopposed to slavery on both economic and moral grounds [Fogel1989 pp 369ndash387]9 Soldiersrsquo commitment to the cause may havegrown the longer they served in the army When Lincoln ran forreelection he received 78 percent of the soldier vote comparedwith 53 percent of the civilian vote despite some 40 to 45 percentof soldiers having come from Democratic families in 1860[McPherson 1997 p 176] We can test whether soldiersrsquo commit-ment increased by examining whether cowardice hazards de-crease with time

Another important determinant of group loyalty is the mo-rale of the troops Morale will depend upon support from thehome front leadership and also upon the unitrsquos recent fatalitiesand the entire Armyrsquos success on the battleeld Morale is adynamic variable World War I soldiers rebelled when casualtygures became too high [Keegan 1976 p 276] Past deaths proxyfor the perceived costs of ghting on We capture the dynamicaspects of morale by using the company mortality rate and thefraction of major Union victories within each half year that therecruit was in service Of course these variables may also reectthe competence of the ofcers and the troops In 1865 desertionreached epidemic levels in the Confederate Army when it wasclear that the Confederacy could not win In the Union Armydesertion reached a high point after the removal of McClellan inNovember 1862 (despite his procrastination he was respected asa professional soldier) the defeats at Fredericksburg and atChickasaw Bluffs in December 1862 the rise of the peace Dem-ocrats at home and the controversy over emancipation Moralerevived with victories at Gettysburg and at Vicksburg in July1863 though continued gyrations were in store for the troops[McPherson 1997 pp 155ndash162]

8 Margo and Steckel [1983] nd that while some skewing in the heightdistribution (and therefore arguably the health or productivity distribution) ofsoldiers appeared as the war progressed this effect was not statisticallysignicant

9 Controlling for other county characteristics does not affect our coefcienton the percent of the county voting for Lincoln suggesting that we cannot distin-guish between an antislavery vote on moral versus on economic grounds Wecannot distinguish between a pro-Union and antislavery vote The effect of thepercent voting for Lincoln was statistically indistinguishable from the effect of thepercent voting for Bell on desertion rates

526 QUARTERLY JOURNAL OF ECONOMICS

III THE UNION ARMY

On the eve of the Civil War the regular army consisted ofonly 15000 enlisted men10 By the end of the war over two millionmen had served in the Union Army with four out of ve men bornin the prime birth cohorts of 1837ndash1845 serving From April 1861to July 1862 the army depended solely upon volunteers enlistingfor low pay In July 1862 the Militia Act assigned quotas to eachstate to ll and states in turn assigned quotas to towns Whenpatriotic appeals failed states and towns began offering menbounties to induce them to enlist so that they could ll theirquotas11 In March 1863 the Enrollment Act created a conscrip-tion system administered by the federal government Quotaswere assigned to each congressional district and then brokendown into subdistricts within each district When towns failed tomeet their quotas every able-bodied male citizen between theages of 20 and 45 became eligible for the draft though marriedmen were less likely to be called Draftees could hire a substituteto take their place or they could pay a commutation fee of $300(equal to the yearly wage of an average worker) to be exempt fromthat particular draft though not from another Draftees andsubstitutes were relatively rare constituting no more than 10percent of all soldiers Paying a commutation fee was also rareOnly 87000 men became exempt in this way

This paper investigates the motivations of the men whofought in the Civil War The sample that we use is representativeof the Union Army However because a large fraction of themilitary age population served it is also representative of thenorthern population of military age Sixty-ve to 98 percent of thecohorts born between 1838 and 1845 were examined for militaryservice and 48 to 81 percent of these cohorts served the remain-der rejected for poor health The men who served are represen-tative of the northern population of military age in terms of realestate and personal property wealth in 1860 [Fogel 2001] Theyare also representative in terms of literacy rates (98 percent in

10 See Hattaway [1997] Gould [1869] and U S Provost Marshall General[1866] for a detailed discussion of the organization of the Civil War Armies andLinderman [1987] Kemp [1990] Mitchell [1990] and McPherson [1997] for dis-cussions of soldiers and their communities

11 Although higher bounties were paid to men in counties where birthplaceheterogeneity was greater (controlling for state xed effects) the effect was notstatistically signicant

527COWARDS AND HEROES

the Union Army sample compared with 95 percent for the north-ern population of military age)

States and individuals played a large role in the formation ofregiments of volunteers the basic units of the armies The vol-unteer infantry regiments consisted of 10 companies each con-taining roughly 100 men commanded by a captain and two lieu-tenants often volunteer ofcers drawn from state militias men ofpolitical signicance or assorted prominent men in the commu-nity Regiments were typically formed from men who came fromthe same area Each company would generally contain bands ofmen who had known each other in civilian life Because of thestrong loyalties men felt toward their companies a company wasnot replenished with new men when disease military casualtiesand expirations of enlistment terms whittled down a companyrsquosnumbers If a companyrsquos numbers were sufciently reduced thecompany disappeared and the men who continued to ght wouldtransfer to another company

The Union Army was not held together by discipline Whenofcers were men soldiers had known all their lives the men hadtrouble thinking of ofcers as their superiors and were slow to orrefused to follow orders Ofcers who commanded contempt be-cause of their cowardice or disregard for the welfare of their menresigned their commissions driven out by their menrsquos ill will

The Armyrsquos coercive powers were limited As the war pro-gressed the Army designated units of provost guards to drivestragglers (men who milled at the rear) into line However be-cause they were reluctant to shoot soldiers wearing the sameuniform they were not always effective Similarly executions forsuch serious penalties as desertion were relatively rare Out ofroughly 200000 deserters 80000 were caught and returned tothe army and 147 were executed for desertion [Linderman 1987pp 174 176]12 The penalties for desertion and also AWOLgenerally ranged from nes and loss of pay to imprisonment(including with hard labor) to performance of the more onerousduties in the company to the social sanctions of menrsquos homecommunities

12 In contrast of the roughly 35000 German soldiers tried for desertion bythe Third Reich about 22750 were executed [Knippschild 1998] Hanson [1999 p320] puts the total number of executions for either desertion or cowardice at50000

528 QUARTERLY JOURNAL OF ECONOMICS

IV DATA

Our data consist of 31854 white enlisted men in 303 UnionArmy infantry companies13 The sample represents roughly 13percent of all whites mustered into the Union Army and 8 percentof all regiments that comprised the Union Army The data arebased upon a 100 percent sample of all enlisted men in 331companies picked at random thus allowing us to create commu-nity variables for each company14 Ninety-one percent of thesample consists of volunteers with the remainder evenly dividedbetween draftees and substitutes The primary data source con-sists of menrsquos military service records These records provide suchbasic information as year of muster age birthplace and height ininches and also information on what happened to the soldierduring his military service Desertions arrests and AWOLs werehandled by military courts convened in the eld Men were linkedto the manuscript schedules of the 1860 census which providesinformation on the value of personal property for all individualsin the household and on illiteracy and allows us to infer maritalstatus (Linkage details are provided in the Appendix) Wemerged data on population in city of enlistment and voting in the1860 presidential election (see the Appendix for sources)

Table II illustrates the wide variation in shirking and mor-tality rates by state Shirking was high in the border states ofKentucky and Maryland and also in New York and New Jersey(two of the more urban states) suggesting that ideology andcommunity characteristics matter

We constructed variables describing recruitsrsquo individualcharacteristics the characteristics of their communities theirideological fervor and their morale (see the Appendix for details)In addition to the variables listed in Table I our regressionscontrol for height in inches (a measure of productivity) regionxed effects for New England Middle Atlantic East North Cen-tral Border and West and dummies for missing information(occupation at enlistment not linked to the 1860 census andtherefore missing information on marital status and on wealth

13 The data were collected by Robert Fogel and are available from httpwwwcpeuchicagoedu

14 Our sample is limited to 303 companies because complete data have notyet been collected on all 331 companies Among the original 331 companies NewEngland is underrepresented and the Midwest overrepresented relative to thearmy as a whole The companies that have not yet been collected are from Indianaand Wisconsin states that were very committed to the Union cause

529COWARDS AND HEROES

III THE UNION ARMY

On the eve of the Civil War the regular army consisted ofonly 15000 enlisted men10 By the end of the war over two millionmen had served in the Union Army with four out of ve men bornin the prime birth cohorts of 1837ndash1845 serving From April 1861to July 1862 the army depended solely upon volunteers enlistingfor low pay In July 1862 the Militia Act assigned quotas to eachstate to ll and states in turn assigned quotas to towns Whenpatriotic appeals failed states and towns began offering menbounties to induce them to enlist so that they could ll theirquotas11 In March 1863 the Enrollment Act created a conscrip-tion system administered by the federal government Quotaswere assigned to each congressional district and then brokendown into subdistricts within each district When towns failed tomeet their quotas every able-bodied male citizen between theages of 20 and 45 became eligible for the draft though marriedmen were less likely to be called Draftees could hire a substituteto take their place or they could pay a commutation fee of $300(equal to the yearly wage of an average worker) to be exempt fromthat particular draft though not from another Draftees andsubstitutes were relatively rare constituting no more than 10percent of all soldiers Paying a commutation fee was also rareOnly 87000 men became exempt in this way

This paper investigates the motivations of the men whofought in the Civil War The sample that we use is representativeof the Union Army However because a large fraction of themilitary age population served it is also representative of thenorthern population of military age Sixty-ve to 98 percent of thecohorts born between 1838 and 1845 were examined for militaryservice and 48 to 81 percent of these cohorts served the remain-der rejected for poor health The men who served are represen-tative of the northern population of military age in terms of realestate and personal property wealth in 1860 [Fogel 2001] Theyare also representative in terms of literacy rates (98 percent in

10 See Hattaway [1997] Gould [1869] and U S Provost Marshall General[1866] for a detailed discussion of the organization of the Civil War Armies andLinderman [1987] Kemp [1990] Mitchell [1990] and McPherson [1997] for dis-cussions of soldiers and their communities

11 Although higher bounties were paid to men in counties where birthplaceheterogeneity was greater (controlling for state xed effects) the effect was notstatistically signicant

527COWARDS AND HEROES

the Union Army sample compared with 95 percent for the north-ern population of military age)

States and individuals played a large role in the formation ofregiments of volunteers the basic units of the armies The vol-unteer infantry regiments consisted of 10 companies each con-taining roughly 100 men commanded by a captain and two lieu-tenants often volunteer ofcers drawn from state militias men ofpolitical signicance or assorted prominent men in the commu-nity Regiments were typically formed from men who came fromthe same area Each company would generally contain bands ofmen who had known each other in civilian life Because of thestrong loyalties men felt toward their companies a company wasnot replenished with new men when disease military casualtiesand expirations of enlistment terms whittled down a companyrsquosnumbers If a companyrsquos numbers were sufciently reduced thecompany disappeared and the men who continued to ght wouldtransfer to another company

The Union Army was not held together by discipline Whenofcers were men soldiers had known all their lives the men hadtrouble thinking of ofcers as their superiors and were slow to orrefused to follow orders Ofcers who commanded contempt be-cause of their cowardice or disregard for the welfare of their menresigned their commissions driven out by their menrsquos ill will

The Armyrsquos coercive powers were limited As the war pro-gressed the Army designated units of provost guards to drivestragglers (men who milled at the rear) into line However be-cause they were reluctant to shoot soldiers wearing the sameuniform they were not always effective Similarly executions forsuch serious penalties as desertion were relatively rare Out ofroughly 200000 deserters 80000 were caught and returned tothe army and 147 were executed for desertion [Linderman 1987pp 174 176]12 The penalties for desertion and also AWOLgenerally ranged from nes and loss of pay to imprisonment(including with hard labor) to performance of the more onerousduties in the company to the social sanctions of menrsquos homecommunities

12 In contrast of the roughly 35000 German soldiers tried for desertion bythe Third Reich about 22750 were executed [Knippschild 1998] Hanson [1999 p320] puts the total number of executions for either desertion or cowardice at50000

528 QUARTERLY JOURNAL OF ECONOMICS

IV DATA

Our data consist of 31854 white enlisted men in 303 UnionArmy infantry companies13 The sample represents roughly 13percent of all whites mustered into the Union Army and 8 percentof all regiments that comprised the Union Army The data arebased upon a 100 percent sample of all enlisted men in 331companies picked at random thus allowing us to create commu-nity variables for each company14 Ninety-one percent of thesample consists of volunteers with the remainder evenly dividedbetween draftees and substitutes The primary data source con-sists of menrsquos military service records These records provide suchbasic information as year of muster age birthplace and height ininches and also information on what happened to the soldierduring his military service Desertions arrests and AWOLs werehandled by military courts convened in the eld Men were linkedto the manuscript schedules of the 1860 census which providesinformation on the value of personal property for all individualsin the household and on illiteracy and allows us to infer maritalstatus (Linkage details are provided in the Appendix) Wemerged data on population in city of enlistment and voting in the1860 presidential election (see the Appendix for sources)

Table II illustrates the wide variation in shirking and mor-tality rates by state Shirking was high in the border states ofKentucky and Maryland and also in New York and New Jersey(two of the more urban states) suggesting that ideology andcommunity characteristics matter

We constructed variables describing recruitsrsquo individualcharacteristics the characteristics of their communities theirideological fervor and their morale (see the Appendix for details)In addition to the variables listed in Table I our regressionscontrol for height in inches (a measure of productivity) regionxed effects for New England Middle Atlantic East North Cen-tral Border and West and dummies for missing information(occupation at enlistment not linked to the 1860 census andtherefore missing information on marital status and on wealth

13 The data were collected by Robert Fogel and are available from httpwwwcpeuchicagoedu

14 Our sample is limited to 303 companies because complete data have notyet been collected on all 331 companies Among the original 331 companies NewEngland is underrepresented and the Midwest overrepresented relative to thearmy as a whole The companies that have not yet been collected are from Indianaand Wisconsin states that were very committed to the Union cause

529COWARDS AND HEROES

the Union Army sample compared with 95 percent for the north-ern population of military age)

States and individuals played a large role in the formation ofregiments of volunteers the basic units of the armies The vol-unteer infantry regiments consisted of 10 companies each con-taining roughly 100 men commanded by a captain and two lieu-tenants often volunteer ofcers drawn from state militias men ofpolitical signicance or assorted prominent men in the commu-nity Regiments were typically formed from men who came fromthe same area Each company would generally contain bands ofmen who had known each other in civilian life Because of thestrong loyalties men felt toward their companies a company wasnot replenished with new men when disease military casualtiesand expirations of enlistment terms whittled down a companyrsquosnumbers If a companyrsquos numbers were sufciently reduced thecompany disappeared and the men who continued to ght wouldtransfer to another company

The Union Army was not held together by discipline Whenofcers were men soldiers had known all their lives the men hadtrouble thinking of ofcers as their superiors and were slow to orrefused to follow orders Ofcers who commanded contempt be-cause of their cowardice or disregard for the welfare of their menresigned their commissions driven out by their menrsquos ill will

The Armyrsquos coercive powers were limited As the war pro-gressed the Army designated units of provost guards to drivestragglers (men who milled at the rear) into line However be-cause they were reluctant to shoot soldiers wearing the sameuniform they were not always effective Similarly executions forsuch serious penalties as desertion were relatively rare Out ofroughly 200000 deserters 80000 were caught and returned tothe army and 147 were executed for desertion [Linderman 1987pp 174 176]12 The penalties for desertion and also AWOLgenerally ranged from nes and loss of pay to imprisonment(including with hard labor) to performance of the more onerousduties in the company to the social sanctions of menrsquos homecommunities

12 In contrast of the roughly 35000 German soldiers tried for desertion bythe Third Reich about 22750 were executed [Knippschild 1998] Hanson [1999 p320] puts the total number of executions for either desertion or cowardice at50000

528 QUARTERLY JOURNAL OF ECONOMICS

IV DATA

Our data consist of 31854 white enlisted men in 303 UnionArmy infantry companies13 The sample represents roughly 13percent of all whites mustered into the Union Army and 8 percentof all regiments that comprised the Union Army The data arebased upon a 100 percent sample of all enlisted men in 331companies picked at random thus allowing us to create commu-nity variables for each company14 Ninety-one percent of thesample consists of volunteers with the remainder evenly dividedbetween draftees and substitutes The primary data source con-sists of menrsquos military service records These records provide suchbasic information as year of muster age birthplace and height ininches and also information on what happened to the soldierduring his military service Desertions arrests and AWOLs werehandled by military courts convened in the eld Men were linkedto the manuscript schedules of the 1860 census which providesinformation on the value of personal property for all individualsin the household and on illiteracy and allows us to infer maritalstatus (Linkage details are provided in the Appendix) Wemerged data on population in city of enlistment and voting in the1860 presidential election (see the Appendix for sources)

Table II illustrates the wide variation in shirking and mor-tality rates by state Shirking was high in the border states ofKentucky and Maryland and also in New York and New Jersey(two of the more urban states) suggesting that ideology andcommunity characteristics matter

We constructed variables describing recruitsrsquo individualcharacteristics the characteristics of their communities theirideological fervor and their morale (see the Appendix for details)In addition to the variables listed in Table I our regressionscontrol for height in inches (a measure of productivity) regionxed effects for New England Middle Atlantic East North Cen-tral Border and West and dummies for missing information(occupation at enlistment not linked to the 1860 census andtherefore missing information on marital status and on wealth

13 The data were collected by Robert Fogel and are available from httpwwwcpeuchicagoedu

14 Our sample is limited to 303 companies because complete data have notyet been collected on all 331 companies Among the original 331 companies NewEngland is underrepresented and the Midwest overrepresented relative to thearmy as a whole The companies that have not yet been collected are from Indianaand Wisconsin states that were very committed to the Union cause

529COWARDS AND HEROES

IV DATA

Our data consist of 31854 white enlisted men in 303 UnionArmy infantry companies13 The sample represents roughly 13percent of all whites mustered into the Union Army and 8 percentof all regiments that comprised the Union Army The data arebased upon a 100 percent sample of all enlisted men in 331companies picked at random thus allowing us to create commu-nity variables for each company14 Ninety-one percent of thesample consists of volunteers with the remainder evenly dividedbetween draftees and substitutes The primary data source con-sists of menrsquos military service records These records provide suchbasic information as year of muster age birthplace and height ininches and also information on what happened to the soldierduring his military service Desertions arrests and AWOLs werehandled by military courts convened in the eld Men were linkedto the manuscript schedules of the 1860 census which providesinformation on the value of personal property for all individualsin the household and on illiteracy and allows us to infer maritalstatus (Linkage details are provided in the Appendix) Wemerged data on population in city of enlistment and voting in the1860 presidential election (see the Appendix for sources)

Table II illustrates the wide variation in shirking and mor-tality rates by state Shirking was high in the border states ofKentucky and Maryland and also in New York and New Jersey(two of the more urban states) suggesting that ideology andcommunity characteristics matter

We constructed variables describing recruitsrsquo individualcharacteristics the characteristics of their communities theirideological fervor and their morale (see the Appendix for details)In addition to the variables listed in Table I our regressionscontrol for height in inches (a measure of productivity) regionxed effects for New England Middle Atlantic East North Cen-tral Border and West and dummies for missing information(occupation at enlistment not linked to the 1860 census andtherefore missing information on marital status and on wealth

13 The data were collected by Robert Fogel and are available from httpwwwcpeuchicagoedu

14 Our sample is limited to 303 companies because complete data have notyet been collected on all 331 companies Among the original 331 companies NewEngland is underrepresented and the Midwest overrepresented relative to thearmy as a whole The companies that have not yet been collected are from Indianaand Wisconsin states that were very committed to the Union cause

529COWARDS AND HEROES

literacy and county voting) Note that both of our morale mea-sures (the fraction of the company who died and the fraction ofUnion victories to all major battles) are time-varying covariatesThe fraction of Union victories does not vary across companiesand only varies across individuals who were mustered in atdifferent dates We do not treat the other company variables astime-varying covariates because there was very little change incompany characteristics from the start to the end of their serviceWe cannot include company leader characteristics as a variablebecause we know leader characteristics only for internalpromotions

TABLE IIPERCENT SERVING BY STATE AND PERCENT DESERTED ARRESTED

AND AWOL AND DIED IN WAR BY STATE

Number ofobservations

Samplein state

Desertedarrested or

AWOL

Deserted

Arrested

AWOL

Died

Connecticut 525 165 2476 1962 229 610 1638Maine 415 131 962 529 361 216 2019Massachusetts 526 165 989 589 209 209 1692New Hampshire 588 185 2262 1718 238 646 2364Vermont 307 096 000 000 000 000 619Delaware 444 139 1554 1261 270 135 900New Jersey 881 277 2860 2486 375 159 806New York 6309 1981 1986 1360 319 499 1522Pennsylvania 2999 941 1230 1040 083 170 1214Illinois 3879 1218 1196 918 106 219 1624Indiana 1344 422 1198 759 104 402 1540Michigan 1433 450 1096 809 154 223 1654Ohio 5567 1748 1128 814 092 300 1538Wisconsin 1389 436 569 317 130 130 1037Iowa 1377 432 574 240 196 232 2193Kansas 260 082 500 308 154 077 346Minnesota 295 093 576 339 102 169 407Missouri 1020 320 1206 892 196 245 1804Kentucky 905 284 2773 1801 110 1050 1657Maryland 294 092 2959 2109 306 748 1327Washington DC 117 037 2650 1368 940 1111 085West Virginia 334 105 719 269 120 389 689New Mexico 95 030 3789 2421 1368 105 105California 550 173 3182 1836 1382 164 527Total 31854 10000 1451 1033 203 329 1468

The column labeled deserted arrested or AWOL uses only the rst instance or either desertion arrestor AWOL Individual arrests and AWOLs are those preceding desertion only The columns individuallylabeled deserted arrested and AWOL therefore do not sum to the single column labeled deserted arrestedor AWOL We do not have an explanation for why shirking rates are 0 for Vermont However our resultsremain the same when we exclude Vermont from our regressions

530 QUARTERLY JOURNAL OF ECONOMICS

Table III lists all variables used in the regression tables andshows that the sample means for those who deserted were ar-rested and were AWOL differ substantially from those for theentire sample (To simplify the tables we do not include ascovariates the fraction of the company that is of the soldierrsquos ownethnicity or occupation or whether the soldier had a brother in thecompany instead we describe the results in the text)

V ECONOMETRIC FRAMEWORK

Our measures of cowardice and heroism are desertion ar-rest and AWOL We combine these three as one summary mea-sure of loyalty and also examine each of these measures individ-ually Desertion is the best measure of shirking Arrests for minorinfractions depend upon ofcer decisions Desertion is a moreserious offense than AWOL and because 10 percent of the sampledeserted it also is the measure with the largest number of eventsAbsences without leave were generally failing to return fromfurlough on time and straggling from the company The determi-nation of whether a case was AWOL or desertion was made by amilitary court convened in the eld If a soldier was determined tohave deserted the time that he deserted was noted as when hewas rst missing Arrests that were not for desertions (and be-cause we censor on desertion we do not examine these) or AWOLwere for drunkenness assault robbery insubordination andsleeping while on picket duty

Our empirical strategy uses four time-varying independentcompeting risk hazard models to estimate days from entry intothe company (muster-in) until 1) the rst case of desertion ar-rest or AWOL 2) desertion 3) arrests preceding desertion and 4)AWOLs preceding desertion We use a competing risk frameworkbecause morale varies over time because men can become morecommitted soldiers and because of censoringmdashsome men mayhave died been discharged changed company become prisonersof war or be missing in action before they could desert We treatthese men as censored in our estimation strategy When weexamine time until rst arrest or AWOL we also treat men whodeserted as censored (see Figure I) Note that we are assumingthat the risk of desertion arrest or AWOL is independent of theoutcomes such as death that we censor on Hazard models providea framework to estimate the micro and macro determinants of

531COWARDS AND HEROES

TABLE IIIVARIABLE MEANS FOR ALL MEN FOR DESERTED ARRESTED AND AWOL COMBINED

AND FOR DESERTED ARRESTED AND AWOL SEPARATELY

CombinedStddev

Alloutcomes Deserted Arrested AWOL

Days from muster until 237181 190644 385175 356181Dummy 5 1 if occupation

Farmer 0511 0369Dagger 0326Dagger 0387Dagger 0493Artisan 0200 0266Dagger 0243Dagger 0195 0191Professionalproprietor 0075 0084Dagger 0086Dagger 0085 0076Laborer 0207 0315Dagger 0338Dagger 0330Dagger 0236daggerUnknown 0007 0006 0007 0003 0004

Dummy 5 1 if born inU S 0755 0663Dagger 0599Dagger 0591Dagger 0708DaggerGermany 0071 0077 0082Dagger 0065 0066Ireland 0084 0160Dagger 0165Dagger 0203Dagger 0125DaggerGreat Britain 0038 0066Dagger 0071Dagger 0074Dagger 0051daggerOther 0052 0075Dagger 0082Dagger 0067 0050

Age at enlistment 25774 7622 25645 25530dagger 25735 25844Dummy 5 1 if married 0133 0094Dagger 0084Dagger 0080Dagger 0135Log (total household personal

property) 18601639 2699 0914Dagger 0743Dagger 1014Dagger 1364Dagger

Dummy 5 1 if illiterate 0017 0021dagger 0018 0015 0031DaggerCompany characteristics

Birthplace fragmentation 0564 0204 0612Dagger 0614Dagger 0648Dagger 0591DaggerOccupational fragmentation 0549 0549 0606Dagger 0620Dagger 0613Dagger 0566DaggerCoefcient of variation for

age 3 10028373 0284 0283 28394 27492Dagger 28596dagger

Log (population) cityenlistment

8599 1874 9224Dagger 9377Dagger 9067Dagger 8870Dagger

Dummy 5 1 if mustered in1861 0210 0234Dagger 0190Dagger 0302Dagger 0354Dagger1862 0351 0346 0347 0339 03311863 0064 0107Dagger 0126Dagger 0096Dagger 00571864 0254 0200Dagger 0197Dagger 0215dagger 0204Dagger1865 0120 0113Dagger 0140Dagger 0048Dagger 0054Dagger

Dummy 5 1 if volunteer 0907 0861Dagger 0842Dagger 0895 0893Percent in county of

enlistment voting forLincoln 35525 26607 33829Dagger 35465 31033Dagger 30281DaggerVote for other 34777 26187 39089Dagger 40455Dagger 35686 37212DaggerUnknown 29698 29698 27082Dagger 24080Dagger 33282dagger 32507dagger

Percent in company dying 13712 8667 12976Dagger 12880Dagger 11595Dagger 13982Fraction Union victories in 6

months of event0450 0388 0308Dagger 0273Dagger 0410Dagger 0393Dagger

Number of observations 31854 4623 3289 646 1049

The symbols dagger and Dagger indicate that the mean is signicantly different from the mean for those not inthe category at the 10 5 and 1 percent level respectively Combined outcomes refer to the rst case ofdesertion arrest or AWOL Arrests and AWOLs are those preceding desertion only The logarithm ofpersonal property wealth is set equal to zero for those for whom this information is missing The standarddeviations of log (total household personal property) birthplace fragmentation occupational fragmentationthe coefcient of variation for age the percent in the company dying and log (population) are 2699 02040181 3193 8667 and 1874 respectively

532 QUARTERLY JOURNAL OF ECONOMICS

cowardice and heroism Our estimated hazard li(t) for one ofour four models (i) is

(3) l i~t 5 exp~x9IbI 1 x9CbC 1 x9DbD 1 x9MbMli0~t

where I indexes the individual variables C indexes the commu-nity variables D indexes the ideology variables M indexes themorale variables and liO (t) is the baseline hazard which weassume to be Weibull The survival function thus takes the formexp((2l i jtj)

p ) for subject j where p is the duration dependenceparameter and can be interpreted as representing whether menwho were in the war longer became more or less committedsoldiers15 We present results both with and without the moralevariable The hazard ratios that we report indicate whether aone-unit change in an independent variable gives an increasedecrease in the odds of an event Thus a hazard ratio of 13 on ourIrish-born dummy variable indicates that the Irish were 13 timesas likely as the native-born to desert We account for unobservedcompany-level correlation by using variance correction models[Lee Wei and Amato 1992 Cai Wei and Wilcox 2000] Cluster-ing on companies provides us with an upper bound on the stan-dard error of company characteristics

15 Because some men may be so loyal that they would never desert we alsoestimated models that account for individual heterogeneity These yielded virtu-ally identical results We also tested whether censoring men who served beyondthree years affected the results We found that the magnitude of the coefcientsand of the duration dependence parameter was similar but that the standarderror of some of our coefcients (eg occupational fragmentation percent of thecounty voting for Lincoln) rose while on others it fell (eg age diversity) Thecoefcients on our morale variables remained strongly signicant

FIGURE ISchematic of Events Studied

533COWARDS AND HEROES

VI RESULTS

Our results show that individual characteristics communitycharacteristics ideology and morale were all important predic-tors of cowardice and heroism Table IV presents results for oursummary measure (time until rst desertion arrest or AWOL)The relative importance of our variable categories depends uponwhether we examine desertions arrests or AWOLs (see Table V)However the results for our summary measure are very similarto those for desertion because desertion is by far the most com-mon rst outcome

Consider rst individual characteristics that proxy for ght-ing ability In the case of desertion men who were farmers whowere older who came from a household with high property wealthin 1860 and who were literate were less likely to desert16 Mar-ried men were signicantly more likely to desert but the inter-action term on married and personal property wealth was insig-nicant suggesting that nancial hardship at home did notnecessarily lead to disproportionate desertions among marriedmen Married men were more likely to be AWOL (but not signi-cantly so) probably because furloughs were generally grantedonly to married men thus providing them with an opportunity togo AWOL Whether a soldier was owed a bounty (as was true formany volunteers after 1862) decreased desertion rates (notshown) but the effect was not statistically signicant17 Relativeto the native-born the Irish and British were more likely todesert They were also twice as likely to be arrested as thenative-born18

Community characteristics were also important predictors ofcowardice and of heroism Men who came from companies inwhich birthplace occupation and age heterogeneity was high andmen who came from larger cities were all more likely desertAlthough birthplace fragmentation was not a statistically signi-cant predictor of desertion it became a statistically signicant

16 If all men in the sample had come from the wealthiest household (one inwhich the logarithm of personal property wealth was 108) the average predictedprobability of desertion would have been 0056 instead of 0094

17 God was not necessarily a better motivator than mammon The higherthe ratio of church seats to county of enlistment population the higher thedesertion rate However this ratio is probably a proxy for urbanization We couldnd no clear pattern by type of religion

18 We do not have a good explanation for high disloyalty rates among theIrish and British These results persist even when the Irish and British were inthe majority in a company

534 QUARTERLY JOURNAL OF ECONOMICS

TABLE IVCOMBINED COMPETING RISK HAZARD MODEL FOR DESERTION ARREST OR AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1240Dagger 0065 1236Dagger 0065Professionalproprietor 1267Dagger 0082 1261Dagger 0082Laborer 1374Dagger 0085 1372Dagger 0085

Dummy 5 1 if born inU SGermany 0896 0123 0894 0123Ireland 1376Dagger 0094 1374Dagger 0094Great Britain 1399Dagger 0120 1400Dagger 0119Other 1205dagger 0093 1204dagger 0093

Age at enlistment 0989Dagger 0003 0989Dagger 0003Dummy 5 1 if married 1287Dagger 0092 1286Dagger 0092Log (total household personal

property) 18600964Dagger 0013 0964Dagger 0013

Dummy 5 1 if illiterate 1582Dagger 0192 1586Dagger 0192Company-level measures

Birthplace fragmentation 1612 0462 1619 0464Occupational fragmentation 2239dagger 0844 2245dagger 0844Coefcient of variation for

age 3 1001027dagger 0014 1028dagger 0014

Log (population) cityenlistment

1048dagger 0023 1048dagger 0023

Dummy 5 1 if mustered in18611862 1287Dagger 0131 1309Dagger 01331863 1649Dagger 0264 1702Dagger 02741864 1291dagger 0136 1330Dagger 01431865 2089Dagger 0308 2060Dagger 0305

Dummy 5 1 if volunteer 0751Dagger 0085 0752Dagger 0085Percent in county of enlistment

voting for Lincoln0993Dagger 0002 0993Dagger 0002

Percent in company dying 3100 (time-varying)

1036Dagger 0009 1035Dagger 0009

Fraction Union victories (time-varying)

0737Dagger 0070

Duration dependenceparameter

0752 0028 0782 0030

x2(32)x2(33) forSignicance of all coefcients 75249 75259

Days until rst desertion arrest or AWOL are measured from rst mustering in The rst instance ofeither is an event Standard errors are clustered on the company The symbols dagger and Dagger indicate that thecoefcient is signicantly different from 1 at the 10 5 and 1 percent level respectively Signicance of allcoefcients is for equality of all coefcients to 1 Men who died became POWs were discharged were missingin action or changed companies before the rst desertion arrest or AWOL are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

535COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

Table III lists all variables used in the regression tables andshows that the sample means for those who deserted were ar-rested and were AWOL differ substantially from those for theentire sample (To simplify the tables we do not include ascovariates the fraction of the company that is of the soldierrsquos ownethnicity or occupation or whether the soldier had a brother in thecompany instead we describe the results in the text)

V ECONOMETRIC FRAMEWORK

Our measures of cowardice and heroism are desertion ar-rest and AWOL We combine these three as one summary mea-sure of loyalty and also examine each of these measures individ-ually Desertion is the best measure of shirking Arrests for minorinfractions depend upon ofcer decisions Desertion is a moreserious offense than AWOL and because 10 percent of the sampledeserted it also is the measure with the largest number of eventsAbsences without leave were generally failing to return fromfurlough on time and straggling from the company The determi-nation of whether a case was AWOL or desertion was made by amilitary court convened in the eld If a soldier was determined tohave deserted the time that he deserted was noted as when hewas rst missing Arrests that were not for desertions (and be-cause we censor on desertion we do not examine these) or AWOLwere for drunkenness assault robbery insubordination andsleeping while on picket duty

Our empirical strategy uses four time-varying independentcompeting risk hazard models to estimate days from entry intothe company (muster-in) until 1) the rst case of desertion ar-rest or AWOL 2) desertion 3) arrests preceding desertion and 4)AWOLs preceding desertion We use a competing risk frameworkbecause morale varies over time because men can become morecommitted soldiers and because of censoringmdashsome men mayhave died been discharged changed company become prisonersof war or be missing in action before they could desert We treatthese men as censored in our estimation strategy When weexamine time until rst arrest or AWOL we also treat men whodeserted as censored (see Figure I) Note that we are assumingthat the risk of desertion arrest or AWOL is independent of theoutcomes such as death that we censor on Hazard models providea framework to estimate the micro and macro determinants of

531COWARDS AND HEROES

TABLE IIIVARIABLE MEANS FOR ALL MEN FOR DESERTED ARRESTED AND AWOL COMBINED

AND FOR DESERTED ARRESTED AND AWOL SEPARATELY

CombinedStddev

Alloutcomes Deserted Arrested AWOL

Days from muster until 237181 190644 385175 356181Dummy 5 1 if occupation

Farmer 0511 0369Dagger 0326Dagger 0387Dagger 0493Artisan 0200 0266Dagger 0243Dagger 0195 0191Professionalproprietor 0075 0084Dagger 0086Dagger 0085 0076Laborer 0207 0315Dagger 0338Dagger 0330Dagger 0236daggerUnknown 0007 0006 0007 0003 0004

Dummy 5 1 if born inU S 0755 0663Dagger 0599Dagger 0591Dagger 0708DaggerGermany 0071 0077 0082Dagger 0065 0066Ireland 0084 0160Dagger 0165Dagger 0203Dagger 0125DaggerGreat Britain 0038 0066Dagger 0071Dagger 0074Dagger 0051daggerOther 0052 0075Dagger 0082Dagger 0067 0050

Age at enlistment 25774 7622 25645 25530dagger 25735 25844Dummy 5 1 if married 0133 0094Dagger 0084Dagger 0080Dagger 0135Log (total household personal

property) 18601639 2699 0914Dagger 0743Dagger 1014Dagger 1364Dagger

Dummy 5 1 if illiterate 0017 0021dagger 0018 0015 0031DaggerCompany characteristics

Birthplace fragmentation 0564 0204 0612Dagger 0614Dagger 0648Dagger 0591DaggerOccupational fragmentation 0549 0549 0606Dagger 0620Dagger 0613Dagger 0566DaggerCoefcient of variation for

age 3 10028373 0284 0283 28394 27492Dagger 28596dagger

Log (population) cityenlistment

8599 1874 9224Dagger 9377Dagger 9067Dagger 8870Dagger

Dummy 5 1 if mustered in1861 0210 0234Dagger 0190Dagger 0302Dagger 0354Dagger1862 0351 0346 0347 0339 03311863 0064 0107Dagger 0126Dagger 0096Dagger 00571864 0254 0200Dagger 0197Dagger 0215dagger 0204Dagger1865 0120 0113Dagger 0140Dagger 0048Dagger 0054Dagger

Dummy 5 1 if volunteer 0907 0861Dagger 0842Dagger 0895 0893Percent in county of

enlistment voting forLincoln 35525 26607 33829Dagger 35465 31033Dagger 30281DaggerVote for other 34777 26187 39089Dagger 40455Dagger 35686 37212DaggerUnknown 29698 29698 27082Dagger 24080Dagger 33282dagger 32507dagger

Percent in company dying 13712 8667 12976Dagger 12880Dagger 11595Dagger 13982Fraction Union victories in 6

months of event0450 0388 0308Dagger 0273Dagger 0410Dagger 0393Dagger

Number of observations 31854 4623 3289 646 1049

The symbols dagger and Dagger indicate that the mean is signicantly different from the mean for those not inthe category at the 10 5 and 1 percent level respectively Combined outcomes refer to the rst case ofdesertion arrest or AWOL Arrests and AWOLs are those preceding desertion only The logarithm ofpersonal property wealth is set equal to zero for those for whom this information is missing The standarddeviations of log (total household personal property) birthplace fragmentation occupational fragmentationthe coefcient of variation for age the percent in the company dying and log (population) are 2699 02040181 3193 8667 and 1874 respectively

532 QUARTERLY JOURNAL OF ECONOMICS

cowardice and heroism Our estimated hazard li(t) for one ofour four models (i) is

(3) l i~t 5 exp~x9IbI 1 x9CbC 1 x9DbD 1 x9MbMli0~t

where I indexes the individual variables C indexes the commu-nity variables D indexes the ideology variables M indexes themorale variables and liO (t) is the baseline hazard which weassume to be Weibull The survival function thus takes the formexp((2l i jtj)

p ) for subject j where p is the duration dependenceparameter and can be interpreted as representing whether menwho were in the war longer became more or less committedsoldiers15 We present results both with and without the moralevariable The hazard ratios that we report indicate whether aone-unit change in an independent variable gives an increasedecrease in the odds of an event Thus a hazard ratio of 13 on ourIrish-born dummy variable indicates that the Irish were 13 timesas likely as the native-born to desert We account for unobservedcompany-level correlation by using variance correction models[Lee Wei and Amato 1992 Cai Wei and Wilcox 2000] Cluster-ing on companies provides us with an upper bound on the stan-dard error of company characteristics

15 Because some men may be so loyal that they would never desert we alsoestimated models that account for individual heterogeneity These yielded virtu-ally identical results We also tested whether censoring men who served beyondthree years affected the results We found that the magnitude of the coefcientsand of the duration dependence parameter was similar but that the standarderror of some of our coefcients (eg occupational fragmentation percent of thecounty voting for Lincoln) rose while on others it fell (eg age diversity) Thecoefcients on our morale variables remained strongly signicant

FIGURE ISchematic of Events Studied

533COWARDS AND HEROES

VI RESULTS

Our results show that individual characteristics communitycharacteristics ideology and morale were all important predic-tors of cowardice and heroism Table IV presents results for oursummary measure (time until rst desertion arrest or AWOL)The relative importance of our variable categories depends uponwhether we examine desertions arrests or AWOLs (see Table V)However the results for our summary measure are very similarto those for desertion because desertion is by far the most com-mon rst outcome

Consider rst individual characteristics that proxy for ght-ing ability In the case of desertion men who were farmers whowere older who came from a household with high property wealthin 1860 and who were literate were less likely to desert16 Mar-ried men were signicantly more likely to desert but the inter-action term on married and personal property wealth was insig-nicant suggesting that nancial hardship at home did notnecessarily lead to disproportionate desertions among marriedmen Married men were more likely to be AWOL (but not signi-cantly so) probably because furloughs were generally grantedonly to married men thus providing them with an opportunity togo AWOL Whether a soldier was owed a bounty (as was true formany volunteers after 1862) decreased desertion rates (notshown) but the effect was not statistically signicant17 Relativeto the native-born the Irish and British were more likely todesert They were also twice as likely to be arrested as thenative-born18

Community characteristics were also important predictors ofcowardice and of heroism Men who came from companies inwhich birthplace occupation and age heterogeneity was high andmen who came from larger cities were all more likely desertAlthough birthplace fragmentation was not a statistically signi-cant predictor of desertion it became a statistically signicant

16 If all men in the sample had come from the wealthiest household (one inwhich the logarithm of personal property wealth was 108) the average predictedprobability of desertion would have been 0056 instead of 0094

17 God was not necessarily a better motivator than mammon The higherthe ratio of church seats to county of enlistment population the higher thedesertion rate However this ratio is probably a proxy for urbanization We couldnd no clear pattern by type of religion

18 We do not have a good explanation for high disloyalty rates among theIrish and British These results persist even when the Irish and British were inthe majority in a company

534 QUARTERLY JOURNAL OF ECONOMICS

TABLE IVCOMBINED COMPETING RISK HAZARD MODEL FOR DESERTION ARREST OR AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1240Dagger 0065 1236Dagger 0065Professionalproprietor 1267Dagger 0082 1261Dagger 0082Laborer 1374Dagger 0085 1372Dagger 0085

Dummy 5 1 if born inU SGermany 0896 0123 0894 0123Ireland 1376Dagger 0094 1374Dagger 0094Great Britain 1399Dagger 0120 1400Dagger 0119Other 1205dagger 0093 1204dagger 0093

Age at enlistment 0989Dagger 0003 0989Dagger 0003Dummy 5 1 if married 1287Dagger 0092 1286Dagger 0092Log (total household personal

property) 18600964Dagger 0013 0964Dagger 0013

Dummy 5 1 if illiterate 1582Dagger 0192 1586Dagger 0192Company-level measures

Birthplace fragmentation 1612 0462 1619 0464Occupational fragmentation 2239dagger 0844 2245dagger 0844Coefcient of variation for

age 3 1001027dagger 0014 1028dagger 0014

Log (population) cityenlistment

1048dagger 0023 1048dagger 0023

Dummy 5 1 if mustered in18611862 1287Dagger 0131 1309Dagger 01331863 1649Dagger 0264 1702Dagger 02741864 1291dagger 0136 1330Dagger 01431865 2089Dagger 0308 2060Dagger 0305

Dummy 5 1 if volunteer 0751Dagger 0085 0752Dagger 0085Percent in county of enlistment

voting for Lincoln0993Dagger 0002 0993Dagger 0002

Percent in company dying 3100 (time-varying)

1036Dagger 0009 1035Dagger 0009

Fraction Union victories (time-varying)

0737Dagger 0070

Duration dependenceparameter

0752 0028 0782 0030

x2(32)x2(33) forSignicance of all coefcients 75249 75259

Days until rst desertion arrest or AWOL are measured from rst mustering in The rst instance ofeither is an event Standard errors are clustered on the company The symbols dagger and Dagger indicate that thecoefcient is signicantly different from 1 at the 10 5 and 1 percent level respectively Signicance of allcoefcients is for equality of all coefcients to 1 Men who died became POWs were discharged were missingin action or changed companies before the rst desertion arrest or AWOL are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

535COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

TABLE IIIVARIABLE MEANS FOR ALL MEN FOR DESERTED ARRESTED AND AWOL COMBINED

AND FOR DESERTED ARRESTED AND AWOL SEPARATELY

CombinedStddev

Alloutcomes Deserted Arrested AWOL

Days from muster until 237181 190644 385175 356181Dummy 5 1 if occupation

Farmer 0511 0369Dagger 0326Dagger 0387Dagger 0493Artisan 0200 0266Dagger 0243Dagger 0195 0191Professionalproprietor 0075 0084Dagger 0086Dagger 0085 0076Laborer 0207 0315Dagger 0338Dagger 0330Dagger 0236daggerUnknown 0007 0006 0007 0003 0004

Dummy 5 1 if born inU S 0755 0663Dagger 0599Dagger 0591Dagger 0708DaggerGermany 0071 0077 0082Dagger 0065 0066Ireland 0084 0160Dagger 0165Dagger 0203Dagger 0125DaggerGreat Britain 0038 0066Dagger 0071Dagger 0074Dagger 0051daggerOther 0052 0075Dagger 0082Dagger 0067 0050

Age at enlistment 25774 7622 25645 25530dagger 25735 25844Dummy 5 1 if married 0133 0094Dagger 0084Dagger 0080Dagger 0135Log (total household personal

property) 18601639 2699 0914Dagger 0743Dagger 1014Dagger 1364Dagger

Dummy 5 1 if illiterate 0017 0021dagger 0018 0015 0031DaggerCompany characteristics

Birthplace fragmentation 0564 0204 0612Dagger 0614Dagger 0648Dagger 0591DaggerOccupational fragmentation 0549 0549 0606Dagger 0620Dagger 0613Dagger 0566DaggerCoefcient of variation for

age 3 10028373 0284 0283 28394 27492Dagger 28596dagger

Log (population) cityenlistment

8599 1874 9224Dagger 9377Dagger 9067Dagger 8870Dagger

Dummy 5 1 if mustered in1861 0210 0234Dagger 0190Dagger 0302Dagger 0354Dagger1862 0351 0346 0347 0339 03311863 0064 0107Dagger 0126Dagger 0096Dagger 00571864 0254 0200Dagger 0197Dagger 0215dagger 0204Dagger1865 0120 0113Dagger 0140Dagger 0048Dagger 0054Dagger

Dummy 5 1 if volunteer 0907 0861Dagger 0842Dagger 0895 0893Percent in county of

enlistment voting forLincoln 35525 26607 33829Dagger 35465 31033Dagger 30281DaggerVote for other 34777 26187 39089Dagger 40455Dagger 35686 37212DaggerUnknown 29698 29698 27082Dagger 24080Dagger 33282dagger 32507dagger

Percent in company dying 13712 8667 12976Dagger 12880Dagger 11595Dagger 13982Fraction Union victories in 6

months of event0450 0388 0308Dagger 0273Dagger 0410Dagger 0393Dagger

Number of observations 31854 4623 3289 646 1049

The symbols dagger and Dagger indicate that the mean is signicantly different from the mean for those not inthe category at the 10 5 and 1 percent level respectively Combined outcomes refer to the rst case ofdesertion arrest or AWOL Arrests and AWOLs are those preceding desertion only The logarithm ofpersonal property wealth is set equal to zero for those for whom this information is missing The standarddeviations of log (total household personal property) birthplace fragmentation occupational fragmentationthe coefcient of variation for age the percent in the company dying and log (population) are 2699 02040181 3193 8667 and 1874 respectively

532 QUARTERLY JOURNAL OF ECONOMICS

cowardice and heroism Our estimated hazard li(t) for one ofour four models (i) is

(3) l i~t 5 exp~x9IbI 1 x9CbC 1 x9DbD 1 x9MbMli0~t

where I indexes the individual variables C indexes the commu-nity variables D indexes the ideology variables M indexes themorale variables and liO (t) is the baseline hazard which weassume to be Weibull The survival function thus takes the formexp((2l i jtj)

p ) for subject j where p is the duration dependenceparameter and can be interpreted as representing whether menwho were in the war longer became more or less committedsoldiers15 We present results both with and without the moralevariable The hazard ratios that we report indicate whether aone-unit change in an independent variable gives an increasedecrease in the odds of an event Thus a hazard ratio of 13 on ourIrish-born dummy variable indicates that the Irish were 13 timesas likely as the native-born to desert We account for unobservedcompany-level correlation by using variance correction models[Lee Wei and Amato 1992 Cai Wei and Wilcox 2000] Cluster-ing on companies provides us with an upper bound on the stan-dard error of company characteristics

15 Because some men may be so loyal that they would never desert we alsoestimated models that account for individual heterogeneity These yielded virtu-ally identical results We also tested whether censoring men who served beyondthree years affected the results We found that the magnitude of the coefcientsand of the duration dependence parameter was similar but that the standarderror of some of our coefcients (eg occupational fragmentation percent of thecounty voting for Lincoln) rose while on others it fell (eg age diversity) Thecoefcients on our morale variables remained strongly signicant

FIGURE ISchematic of Events Studied

533COWARDS AND HEROES

VI RESULTS

Our results show that individual characteristics communitycharacteristics ideology and morale were all important predic-tors of cowardice and heroism Table IV presents results for oursummary measure (time until rst desertion arrest or AWOL)The relative importance of our variable categories depends uponwhether we examine desertions arrests or AWOLs (see Table V)However the results for our summary measure are very similarto those for desertion because desertion is by far the most com-mon rst outcome

Consider rst individual characteristics that proxy for ght-ing ability In the case of desertion men who were farmers whowere older who came from a household with high property wealthin 1860 and who were literate were less likely to desert16 Mar-ried men were signicantly more likely to desert but the inter-action term on married and personal property wealth was insig-nicant suggesting that nancial hardship at home did notnecessarily lead to disproportionate desertions among marriedmen Married men were more likely to be AWOL (but not signi-cantly so) probably because furloughs were generally grantedonly to married men thus providing them with an opportunity togo AWOL Whether a soldier was owed a bounty (as was true formany volunteers after 1862) decreased desertion rates (notshown) but the effect was not statistically signicant17 Relativeto the native-born the Irish and British were more likely todesert They were also twice as likely to be arrested as thenative-born18

Community characteristics were also important predictors ofcowardice and of heroism Men who came from companies inwhich birthplace occupation and age heterogeneity was high andmen who came from larger cities were all more likely desertAlthough birthplace fragmentation was not a statistically signi-cant predictor of desertion it became a statistically signicant

16 If all men in the sample had come from the wealthiest household (one inwhich the logarithm of personal property wealth was 108) the average predictedprobability of desertion would have been 0056 instead of 0094

17 God was not necessarily a better motivator than mammon The higherthe ratio of church seats to county of enlistment population the higher thedesertion rate However this ratio is probably a proxy for urbanization We couldnd no clear pattern by type of religion

18 We do not have a good explanation for high disloyalty rates among theIrish and British These results persist even when the Irish and British were inthe majority in a company

534 QUARTERLY JOURNAL OF ECONOMICS

TABLE IVCOMBINED COMPETING RISK HAZARD MODEL FOR DESERTION ARREST OR AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1240Dagger 0065 1236Dagger 0065Professionalproprietor 1267Dagger 0082 1261Dagger 0082Laborer 1374Dagger 0085 1372Dagger 0085

Dummy 5 1 if born inU SGermany 0896 0123 0894 0123Ireland 1376Dagger 0094 1374Dagger 0094Great Britain 1399Dagger 0120 1400Dagger 0119Other 1205dagger 0093 1204dagger 0093

Age at enlistment 0989Dagger 0003 0989Dagger 0003Dummy 5 1 if married 1287Dagger 0092 1286Dagger 0092Log (total household personal

property) 18600964Dagger 0013 0964Dagger 0013

Dummy 5 1 if illiterate 1582Dagger 0192 1586Dagger 0192Company-level measures

Birthplace fragmentation 1612 0462 1619 0464Occupational fragmentation 2239dagger 0844 2245dagger 0844Coefcient of variation for

age 3 1001027dagger 0014 1028dagger 0014

Log (population) cityenlistment

1048dagger 0023 1048dagger 0023

Dummy 5 1 if mustered in18611862 1287Dagger 0131 1309Dagger 01331863 1649Dagger 0264 1702Dagger 02741864 1291dagger 0136 1330Dagger 01431865 2089Dagger 0308 2060Dagger 0305

Dummy 5 1 if volunteer 0751Dagger 0085 0752Dagger 0085Percent in county of enlistment

voting for Lincoln0993Dagger 0002 0993Dagger 0002

Percent in company dying 3100 (time-varying)

1036Dagger 0009 1035Dagger 0009

Fraction Union victories (time-varying)

0737Dagger 0070

Duration dependenceparameter

0752 0028 0782 0030

x2(32)x2(33) forSignicance of all coefcients 75249 75259

Days until rst desertion arrest or AWOL are measured from rst mustering in The rst instance ofeither is an event Standard errors are clustered on the company The symbols dagger and Dagger indicate that thecoefcient is signicantly different from 1 at the 10 5 and 1 percent level respectively Signicance of allcoefcients is for equality of all coefcients to 1 Men who died became POWs were discharged were missingin action or changed companies before the rst desertion arrest or AWOL are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

535COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

cowardice and heroism Our estimated hazard li(t) for one ofour four models (i) is

(3) l i~t 5 exp~x9IbI 1 x9CbC 1 x9DbD 1 x9MbMli0~t

where I indexes the individual variables C indexes the commu-nity variables D indexes the ideology variables M indexes themorale variables and liO (t) is the baseline hazard which weassume to be Weibull The survival function thus takes the formexp((2l i jtj)

p ) for subject j where p is the duration dependenceparameter and can be interpreted as representing whether menwho were in the war longer became more or less committedsoldiers15 We present results both with and without the moralevariable The hazard ratios that we report indicate whether aone-unit change in an independent variable gives an increasedecrease in the odds of an event Thus a hazard ratio of 13 on ourIrish-born dummy variable indicates that the Irish were 13 timesas likely as the native-born to desert We account for unobservedcompany-level correlation by using variance correction models[Lee Wei and Amato 1992 Cai Wei and Wilcox 2000] Cluster-ing on companies provides us with an upper bound on the stan-dard error of company characteristics

15 Because some men may be so loyal that they would never desert we alsoestimated models that account for individual heterogeneity These yielded virtu-ally identical results We also tested whether censoring men who served beyondthree years affected the results We found that the magnitude of the coefcientsand of the duration dependence parameter was similar but that the standarderror of some of our coefcients (eg occupational fragmentation percent of thecounty voting for Lincoln) rose while on others it fell (eg age diversity) Thecoefcients on our morale variables remained strongly signicant

FIGURE ISchematic of Events Studied

533COWARDS AND HEROES

VI RESULTS

Our results show that individual characteristics communitycharacteristics ideology and morale were all important predic-tors of cowardice and heroism Table IV presents results for oursummary measure (time until rst desertion arrest or AWOL)The relative importance of our variable categories depends uponwhether we examine desertions arrests or AWOLs (see Table V)However the results for our summary measure are very similarto those for desertion because desertion is by far the most com-mon rst outcome

Consider rst individual characteristics that proxy for ght-ing ability In the case of desertion men who were farmers whowere older who came from a household with high property wealthin 1860 and who were literate were less likely to desert16 Mar-ried men were signicantly more likely to desert but the inter-action term on married and personal property wealth was insig-nicant suggesting that nancial hardship at home did notnecessarily lead to disproportionate desertions among marriedmen Married men were more likely to be AWOL (but not signi-cantly so) probably because furloughs were generally grantedonly to married men thus providing them with an opportunity togo AWOL Whether a soldier was owed a bounty (as was true formany volunteers after 1862) decreased desertion rates (notshown) but the effect was not statistically signicant17 Relativeto the native-born the Irish and British were more likely todesert They were also twice as likely to be arrested as thenative-born18

Community characteristics were also important predictors ofcowardice and of heroism Men who came from companies inwhich birthplace occupation and age heterogeneity was high andmen who came from larger cities were all more likely desertAlthough birthplace fragmentation was not a statistically signi-cant predictor of desertion it became a statistically signicant

16 If all men in the sample had come from the wealthiest household (one inwhich the logarithm of personal property wealth was 108) the average predictedprobability of desertion would have been 0056 instead of 0094

17 God was not necessarily a better motivator than mammon The higherthe ratio of church seats to county of enlistment population the higher thedesertion rate However this ratio is probably a proxy for urbanization We couldnd no clear pattern by type of religion

18 We do not have a good explanation for high disloyalty rates among theIrish and British These results persist even when the Irish and British were inthe majority in a company

534 QUARTERLY JOURNAL OF ECONOMICS

TABLE IVCOMBINED COMPETING RISK HAZARD MODEL FOR DESERTION ARREST OR AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1240Dagger 0065 1236Dagger 0065Professionalproprietor 1267Dagger 0082 1261Dagger 0082Laborer 1374Dagger 0085 1372Dagger 0085

Dummy 5 1 if born inU SGermany 0896 0123 0894 0123Ireland 1376Dagger 0094 1374Dagger 0094Great Britain 1399Dagger 0120 1400Dagger 0119Other 1205dagger 0093 1204dagger 0093

Age at enlistment 0989Dagger 0003 0989Dagger 0003Dummy 5 1 if married 1287Dagger 0092 1286Dagger 0092Log (total household personal

property) 18600964Dagger 0013 0964Dagger 0013

Dummy 5 1 if illiterate 1582Dagger 0192 1586Dagger 0192Company-level measures

Birthplace fragmentation 1612 0462 1619 0464Occupational fragmentation 2239dagger 0844 2245dagger 0844Coefcient of variation for

age 3 1001027dagger 0014 1028dagger 0014

Log (population) cityenlistment

1048dagger 0023 1048dagger 0023

Dummy 5 1 if mustered in18611862 1287Dagger 0131 1309Dagger 01331863 1649Dagger 0264 1702Dagger 02741864 1291dagger 0136 1330Dagger 01431865 2089Dagger 0308 2060Dagger 0305

Dummy 5 1 if volunteer 0751Dagger 0085 0752Dagger 0085Percent in county of enlistment

voting for Lincoln0993Dagger 0002 0993Dagger 0002

Percent in company dying 3100 (time-varying)

1036Dagger 0009 1035Dagger 0009

Fraction Union victories (time-varying)

0737Dagger 0070

Duration dependenceparameter

0752 0028 0782 0030

x2(32)x2(33) forSignicance of all coefcients 75249 75259

Days until rst desertion arrest or AWOL are measured from rst mustering in The rst instance ofeither is an event Standard errors are clustered on the company The symbols dagger and Dagger indicate that thecoefcient is signicantly different from 1 at the 10 5 and 1 percent level respectively Signicance of allcoefcients is for equality of all coefcients to 1 Men who died became POWs were discharged were missingin action or changed companies before the rst desertion arrest or AWOL are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

535COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

VI RESULTS

Our results show that individual characteristics communitycharacteristics ideology and morale were all important predic-tors of cowardice and heroism Table IV presents results for oursummary measure (time until rst desertion arrest or AWOL)The relative importance of our variable categories depends uponwhether we examine desertions arrests or AWOLs (see Table V)However the results for our summary measure are very similarto those for desertion because desertion is by far the most com-mon rst outcome

Consider rst individual characteristics that proxy for ght-ing ability In the case of desertion men who were farmers whowere older who came from a household with high property wealthin 1860 and who were literate were less likely to desert16 Mar-ried men were signicantly more likely to desert but the inter-action term on married and personal property wealth was insig-nicant suggesting that nancial hardship at home did notnecessarily lead to disproportionate desertions among marriedmen Married men were more likely to be AWOL (but not signi-cantly so) probably because furloughs were generally grantedonly to married men thus providing them with an opportunity togo AWOL Whether a soldier was owed a bounty (as was true formany volunteers after 1862) decreased desertion rates (notshown) but the effect was not statistically signicant17 Relativeto the native-born the Irish and British were more likely todesert They were also twice as likely to be arrested as thenative-born18

Community characteristics were also important predictors ofcowardice and of heroism Men who came from companies inwhich birthplace occupation and age heterogeneity was high andmen who came from larger cities were all more likely desertAlthough birthplace fragmentation was not a statistically signi-cant predictor of desertion it became a statistically signicant

16 If all men in the sample had come from the wealthiest household (one inwhich the logarithm of personal property wealth was 108) the average predictedprobability of desertion would have been 0056 instead of 0094

17 God was not necessarily a better motivator than mammon The higherthe ratio of church seats to county of enlistment population the higher thedesertion rate However this ratio is probably a proxy for urbanization We couldnd no clear pattern by type of religion

18 We do not have a good explanation for high disloyalty rates among theIrish and British These results persist even when the Irish and British were inthe majority in a company

534 QUARTERLY JOURNAL OF ECONOMICS

TABLE IVCOMBINED COMPETING RISK HAZARD MODEL FOR DESERTION ARREST OR AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1240Dagger 0065 1236Dagger 0065Professionalproprietor 1267Dagger 0082 1261Dagger 0082Laborer 1374Dagger 0085 1372Dagger 0085

Dummy 5 1 if born inU SGermany 0896 0123 0894 0123Ireland 1376Dagger 0094 1374Dagger 0094Great Britain 1399Dagger 0120 1400Dagger 0119Other 1205dagger 0093 1204dagger 0093

Age at enlistment 0989Dagger 0003 0989Dagger 0003Dummy 5 1 if married 1287Dagger 0092 1286Dagger 0092Log (total household personal

property) 18600964Dagger 0013 0964Dagger 0013

Dummy 5 1 if illiterate 1582Dagger 0192 1586Dagger 0192Company-level measures

Birthplace fragmentation 1612 0462 1619 0464Occupational fragmentation 2239dagger 0844 2245dagger 0844Coefcient of variation for

age 3 1001027dagger 0014 1028dagger 0014

Log (population) cityenlistment

1048dagger 0023 1048dagger 0023

Dummy 5 1 if mustered in18611862 1287Dagger 0131 1309Dagger 01331863 1649Dagger 0264 1702Dagger 02741864 1291dagger 0136 1330Dagger 01431865 2089Dagger 0308 2060Dagger 0305

Dummy 5 1 if volunteer 0751Dagger 0085 0752Dagger 0085Percent in county of enlistment

voting for Lincoln0993Dagger 0002 0993Dagger 0002

Percent in company dying 3100 (time-varying)

1036Dagger 0009 1035Dagger 0009

Fraction Union victories (time-varying)

0737Dagger 0070

Duration dependenceparameter

0752 0028 0782 0030

x2(32)x2(33) forSignicance of all coefcients 75249 75259

Days until rst desertion arrest or AWOL are measured from rst mustering in The rst instance ofeither is an event Standard errors are clustered on the company The symbols dagger and Dagger indicate that thecoefcient is signicantly different from 1 at the 10 5 and 1 percent level respectively Signicance of allcoefcients is for equality of all coefcients to 1 Men who died became POWs were discharged were missingin action or changed companies before the rst desertion arrest or AWOL are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

535COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

TABLE IVCOMBINED COMPETING RISK HAZARD MODEL FOR DESERTION ARREST OR AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1240Dagger 0065 1236Dagger 0065Professionalproprietor 1267Dagger 0082 1261Dagger 0082Laborer 1374Dagger 0085 1372Dagger 0085

Dummy 5 1 if born inU SGermany 0896 0123 0894 0123Ireland 1376Dagger 0094 1374Dagger 0094Great Britain 1399Dagger 0120 1400Dagger 0119Other 1205dagger 0093 1204dagger 0093

Age at enlistment 0989Dagger 0003 0989Dagger 0003Dummy 5 1 if married 1287Dagger 0092 1286Dagger 0092Log (total household personal

property) 18600964Dagger 0013 0964Dagger 0013

Dummy 5 1 if illiterate 1582Dagger 0192 1586Dagger 0192Company-level measures

Birthplace fragmentation 1612 0462 1619 0464Occupational fragmentation 2239dagger 0844 2245dagger 0844Coefcient of variation for

age 3 1001027dagger 0014 1028dagger 0014

Log (population) cityenlistment

1048dagger 0023 1048dagger 0023

Dummy 5 1 if mustered in18611862 1287Dagger 0131 1309Dagger 01331863 1649Dagger 0264 1702Dagger 02741864 1291dagger 0136 1330Dagger 01431865 2089Dagger 0308 2060Dagger 0305

Dummy 5 1 if volunteer 0751Dagger 0085 0752Dagger 0085Percent in county of enlistment

voting for Lincoln0993Dagger 0002 0993Dagger 0002

Percent in company dying 3100 (time-varying)

1036Dagger 0009 1035Dagger 0009

Fraction Union victories (time-varying)

0737Dagger 0070

Duration dependenceparameter

0752 0028 0782 0030

x2(32)x2(33) forSignicance of all coefcients 75249 75259

Days until rst desertion arrest or AWOL are measured from rst mustering in The rst instance ofeither is an event Standard errors are clustered on the company The symbols dagger and Dagger indicate that thecoefcient is signicantly different from 1 at the 10 5 and 1 percent level respectively Signicance of allcoefcients is for equality of all coefcients to 1 Men who died became POWs were discharged were missingin action or changed companies before the rst desertion arrest or AWOL are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

535COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

TABLE VSEPARATE COMPETING RISK HAZARD MODELS FOR DESERTION ARREST AND AWOL

Desertion Arrest AWOL

Hazardratio

Stderr

Hazardratio

Stderr

Hazardratio

Stderr

Dummy 5 1 if occupationFarmerArtisan 1435Dagger 0093 0925 0115 0910 0088Professionalproprietor 1359Dagger 0105 1132 0195 1045 0146Laborer 1572Dagger 0121 1063 0136 1043 0121

Dummy 5 1 if born inU SGermany 0884 0146 0918 0164 0857 0143Ireland 1310Dagger 0103 2007Dagger 0287 1181 0152Great Britain 1396Dagger 0148 1691Dagger 0280 1247 0219Other 1245dagger 0120 1100 0170 0935 0148

Age at enlistment 0985Dagger 0003 0985dagger 0006 1004 0005Dummy 5 1 if married 1382Dagger 0128 1141 0218 1211 0147Log (total household personal

property) 18600950Dagger 0017 0987 0027 0968 0021

Dummy 5 1 if illiterate 1601Dagger 0243 1076 0314 1551 0464Company-level measures

Birthplace fragmentation 1405 0496 3001Dagger 1230 2593Dagger 1007Occupational fragmentation 3428dagger 1682 2983dagger 1451 0759 0376Coefcient of variation for

age 3 1001032 0017 0993 0025 1014 0022

Log (population) cityenlistment

1058dagger 0028 1006 0037 1027 0036

Dummy 5 1 if mustered in18611862 1632Dagger 0200 1390dagger 0216 0749dagger 00991863 2338Dagger 0437 1748Dagger 0322 0729 01541864 1472Dagger 0196 2505Dagger 0400 1326 02321865 2628Dagger 0437 1921dagger 0503 1191 0333

Dummy 5 1 if volunteer 0749dagger 0100 0854 0144 0651dagger 0113Percent in county of

enlistment voting forLincoln

0995dagger 0003 0994 0004 0990Dagger 0003

Percent in company dying(time-varying)

1036Dagger 0011 0990 0019 1068Dagger 0014

Fraction Union victories(time-varying)

0610Dagger 0075 0599dagger 0128 0605 0105

Duration dependenceparameter

0682 0027 1325 0072 1298 0050

x33 for signicance of allcoefcients

78432 34934 21706

Days until desertion arrest or AWOL are measured from rst mustering in In addition for arrest andAWOL men who deserted are treated as censored Standard errors are clustered on the company Thesymbols dagger and Dagger indicate that the coefcient is signicantly different from 1 at the 10 5 and 1 percent levelrespectivelySignicance of all coefcients is for equality of all coefcients to 1 Men who died became POWswere discharged were missing in action or changed companies before rst desertion are treated as censoredCovariates include height in inches and dummy variables indicating missing information for occupation the1860 census literacy and county voting Included region xed effects are for Middle Atlantic East NorthCentral West North Central Border and West (New England is the omitted category)

536 QUARTERLY JOURNAL OF ECONOMICS

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

predictor when we dropped occupational fragmentation from theregression Men in companies in which birthplace and occupa-tional diversity was high were signicantly more likely to bearrested The only company socioeconomic and demographiccharacteristic that signicantly predicted AWOL was birthplacediversity When we included the company Gini coefcient for bothpersonal and property wealth calculated from the 1860 censuswe found that while men in companies where inequality was highwere more likely to desert the effect was statisticallyinsignicant

We tested whether our birthplace and occupation fragmen-tation measures proxy for average skills within the companycharacteristics instead of company heterogeneity The percent ofthe company of foreign birth or of a given occupation and themean age of the company had no predictive power We also testedwhether our company fragmentation measures perform betterthan county-level fragmentation measures for the male popula-tion of military age Higher birthplace fragmentation in county ofenlistment increased desertion rates but the effect was not sta-tistically signicant19 Finally we tested whether unobservedsorting on ability is driving our results by excluding large coun-ties where men had more companies to choose from in enlistingIn the case of desertion occupational fragmentation became aninsignicant predictor as variation in this variable fell but birth-place fragmentation became a statistically signicant predictor

We investigated using alternative measures of birthplaceand occupational diversity such as percent of own nativity oroccupation and concentration ratios Concentration ratios forbirthplace and occupation were collinear but individually ahigher concentration ratio signicantly decreased the probabilityof desertion Measures such as percent of own nativity or occu-pation are not suited to the Union Army data because there wasno dominant ethnic group However we did nd some evidence ofethnic favoritism when we investigated whether there was anyinteraction between own ethnicity and that of an ofcer for thelimited set of companies for which we know something about theofcers because they rose from the ranks In the case of AWOLthe Irish were signicantly more likely to be AWOL if the com-

19 We used the 1860 census of population and created fragmentation mea-sures for men age 16 to 39 in counties with at least 25 such men We found noeffect at all of county-level occupational fragmentation

537COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

pany had an Irish ofcer but we could not determine whetherpunishments for AWOL were lower in these companies Howeverboth the Irish and the British were more likely to be arrested ifthe company contained an Irish or British ofcer and the Britishwere signicantly less likely to desert if the company contained aBritish ofcer We also investigated whether the interactionsbetween own occupation and the proportion of men in the com-pany in that occupation and own birthplace and the proportion ofmen of that ethnicity were at all signicant We only obtainedsignicant results for laborers They were more likely to desertand to be arrested if the proportion of laborers in the companywas high

We have the opportunity to study peer groups for brothersfathers and sons among men linked to the 1860 census Thesemen might either be more likely to shirk because collusion iseasier or be less likely to shirk because of loyalty We nd thathaving close kin in the same company increased the probability ofdesertion but the coefcient was not statistically signicant Itsignicantly decreased the odds of going AWOL and did not affectarrests

We have not tested whether there was a contagion effectleading to increased individual probabilities of desertion whencompany desertion rates rose However because of the nonlinear-ity of our estimation equation (3) this endogenous interaction canbe estimated off of the functional form [Manski 1993 2000 Brockand Durlauf 2001] We therefore included a time-varying mea-sure of the fraction in the company deserting in our desertionspecication We found that this measure signicantly increaseddesertion rates but that the company death rate became aninsignicant predictor of desertion and that the signicance of thecoefcient on the proportion of Union victories fell from 1 to 10percent All other coefcients were unaffected

Were there any individual benets to soldiers of being in ahomogeneous company Because the ghting unit sent to battlewas the regiment but because regiments contained both homoge-neous and heterogeneous companies we can examine time untildeath on the battleeld as a function of company characteristicsindividual characteristics and regimental xed effects to controlfor battleeld experience Our results are mixed Mortality waslower among men in companies with high birthplace fragmenta-tion (hazard ratio 5 0420 s 5 0196) but it was higher amongmen from large cities (hazard ratio 5 1071 s 5 0039) and higher

538 QUARTERLY JOURNAL OF ECONOMICS

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

among men in companies with high occupational fragmentation(hazard ratio 5 2382 s 5 1180)

Our ideology proxies predicted desertion arrest and AWOLMen who enlisted in 1861 were less likely to desert or to bearrested Surprisingly soldiers mustered in 1862 and 1863 wereless likely to be AWOL than soldiers mustered in 1861 Men whovolunteered and men from pro-Lincoln counties were less likely todesert or to be AWOL Using the percentage of the county votingfor Fremont in the 1856 presidential election as an alternativemeasure of ideology yielded virtually similar results to using thepercentage of the county voting for Lincoln in 1860 We ndmixed evidence that soldiers became more committed to the causethe longer they remained in the army Although desertion haz-ards decrease with time arrest and AWOL hazards increase withtime

Lastly our morale proxies were predictors of all of our mea-sures of cowardice and of heroism Men were more likely to desertwhen company mortality was high and when the Union waslosing Arrest rates were higher when the Union was losing Ahigh company mortality rate signicantly reduced time untilAWOL

We experimented with different outcome variables as robust-ness checks We investigated what predicted reenlistment foranother three-year term among men who enlisted in 1861 andwho had already served a three-year term Approximately half ofreenlistees in the sample received a bounty upon reenlistmentGenerally men reenlisted as regiments or companies [Hess 1997p 89] Older men men from large cities and Germans were lesslikely to reenlist and men who received a bounty for reenlistingwere more likely to reenlist but these were the only characteris-tics that predicted reenlistment20 We also examined the deter-minants of promotion from the ranks to ofcer nding that suchindividual characteristics as social status (being a professionalproprietor or artisan rather than a farmer or laborer) beingnative-born and being tall increased the likelihood of promotion

We performed further robustness tests by experimentingwith state xed effects for all regressions One of the difculties

20 There was no dishonor in not reenlisting Newton Scott a private in the36th Iowa Infantry Company A wrote to Hannah Cone ldquoI think it the Duty ofEvery Able Bodied man If Necessary to Help Defend His country But I think 3years Sufcient long for one man to Serve while they all take there [sic] turns rdquohttpwwwcivilwarletterscom

539COWARDS AND HEROES

we faced is that when the number of companies within a statewas small correlation between birthplace and occupational frag-mentation was high In the case of AWOL and arrest the coef-cients on company socioeconomic and demographic characteris-tics the percentage of the county voting for Lincoln and thecompany death rate remained unchanged For desertion bothbirthplace and occupational fragmentation were statistically sig-nicant predictors of desertion but the proportion of the countyvoting for Lincoln (a measure that varies more across states thanwithin states) became an insignicant predictor We also experi-mented with regimental xed effects In the case of arrest andAWOL our results were very similar to those presented in ourtables In the case of desertion the size of the coefcients onbirthplace and occupational fragmentation increased markedlyand both were statistically signicant in all specications but thecoefcient on age diversity became statistically insignicant

VII IMPLICATIONS

Our results show that the same types of variables that pre-dict commitment to organizations in civilian life today predictedloyalty to the Union Army in the past21 Group loyalty requiresinteractions with fellow workers or community members butcommitment to interacting with others varies by demographicgroup [Glaeser Laibson and Sacerdote 2000] Previous studieshave found that community heterogeneity lowers public expendi-tures [Luttmer 2001 Poterba 1997 Alesina Baqir and Easterly1999 Goldin and Katz 1999] and reduces time allocation andorganizational membership [Alesina and La Ferrara 2000 Costaand Kahn 2003] Studies of rms have found that heterogeneityin age education tenure race and sex is positively related toturnover but which of these heterogeneity measures is moreimportant depends upon the organization studied22

Table VI shows the relative importance of community socio-economic and demographic characteristics and of our morale andideology proxies for the predicted probability of desertion arrest

21 Our results stand in contrast to Bearman [1991] who argues that amongConfederate soldiers from North Carolina local homogeneity led to high desertionrates However heterogeneity within the state may have led to high desertionrates Weitz [2000] nds that among Georgia soldiers men most likely to desertwere from the subsistence farming areas of the Upcountry whose families facedstarvation without them and who had nothing to gain from secession

22 See Pfeffer [1997 pp 83ndash85] for a review

540 QUARTERLY JOURNAL OF ECONOMICS

and AWOL Predictions are based upon the predicted survivalfunction from the time of muster calculated for every individualand then averaged over the whole sample In the case of oursummary measure of disloyalty and of desertion the single mostimportant variables were age and occupational diversity withinthe company In the case of arrests birthplace and occupationalfragmentation the fraction of Union victories the percentage ofthe county voting for Lincoln and year of muster were the singlemost important predictors Birthplace diversity age diversityand the fraction of the county voting for Lincoln were the mostimportant predictors of AWOL On the whole company socioeco-nomic and demographic characteristics were the most important

TABLE VIPREDICTED PROBABILITIES OF DESERTION ARREST AND AWOL BY COMPANY

CHARACTERISTICS MORALE AND IDEOLOGY

DesertionArrest or

AWOL Desertion Arrest AWOL

Using true variable values 0131 0094 0020 0033Community characteristics

If birthplace fragmentation 5 0 0101 0078 0010 0019If occupational fragmentation

5 00085 0047 0010 0039

If coefcient of variation forage 5 0

0065 0042 0023 0023

If all of above 0031 0016 0006 0016If population in city of

enlistment 5 25000125 0088 0020 0032

If all of above 0030 0015 0006 0015Morale

If company death rate 5 0 0118 0084 0020 0027If fraction Union victories 5 1 0108 0067 0015 0034If both 0097 0060 0022 0028

IdeologyIf volunteer 0128 0091 0020 0032If 866 county voted for

Lincoln0103 0079 0015 0022

If mustered in 1861 0107 0066 0015 0039If all of above 0082 0054 0011 0025

Our summary measure (the rst desertion arrest or AWOL) is predicted from the second specicationin Table IV All other probabilities are predicted using the specications in Table V Predictions are basedupon the predicted survival function from the time of muster using the actual data and averaging over thewhole sample In this sample the largest share of the vote Lincoln received in a county was 866 percentCities with a population of less than 2500 were not even listed in the census and are therefore consideredsmall towns

541COWARDS AND HEROES

predictors of desertion arrest and AWOL and our ideology prox-ies were relatively more important than our morale proxies23 Wealso examined predicted probabilities for our summary measureof all recruits being literate native-born 30-year old single farm-ers with personal property wealth equal to the mean plus a halfstandard deviation Being literate and a being a farmer were thesingle most important individual predictors of loyalty and all ofour individual characteristics combined were better predictors ofloyalty than our ideology or morale proxies

Why does the Army today not make greater use of socialcapital by creating socioeconomic and demographically homoge-neous ghting units Desertion on the modern battleeld is nowharder because the battleeld is larger more congested in therear with administrative soldiers and also more leveled by bomb-ing [Keegan 1976 p 316] Two additional reasons include diver-sication and human capital specialization in the modern armyHighly publicized losses to communities during World War IIended any practice of locally based companies In the modernarmy soldiers perform a myriad of tasks requiring differenttraining In contrast during the Civil War the job of a soldier wasunskilled largely consisting of learning the movement of linearformations of obeying orders without hesitation and of master-ing the nine steps of loading a musket and ring in the directionof an enemy hidden by the smoke of the battleeld [Hess 1997 pp18ndash19 137] Worker skills were perfect substitutes

VIII CONCLUSION

Why do ordinary soldiers ght when pay is low and whendesertion is a choice that many have the opportunity to exerciseIs it the attributes of the person is it loyalty to a small group ofindividuals ideology or morale Most sociologists and psycholo-gists have emphasized the importance of loyalty to a small groupof individuals Military historians however have reminded us ofthe importance of a moral crusade in motivating democratic sol-diers and of morale in keeping an army from faltering We haveshown that among Union Army soldiers in the Civil War individ-ual socioeconomic and demographic characteristics that proxy for

23 If the Union Army had had the power to construct companies at randomit could not have reduced shirking by following this policy because birthplaceoccupational and age fragmentation would have increased for many companies

542 QUARTERLY JOURNAL OF ECONOMICS

ghting ability company socioeconomic and demographic charac-teristics ideological commitment and morale were all importantdeterminants of group loyalty Company socioeconomic and de-mographic characteristics were particularly important evenmore so than ideological commitment and morale in one of ourcountryrsquos more ideological wars Heterogeneity is an importantdeterminant of participation in organizations today [Alesina andLa Ferrara 2000 Costa and Kahn 2003] and it was also animportant predictor in a very different environment in the past

DATA APPENDIX

This appendix describes the construction of our dependentvariables our demographic and socioeconomic variables our com-munity variables our ideology variables and our morale vari-ables All data on Union Army recruits are obtained from Aging ofVeterans of the Union Army Robert W Fogel Principal Investi-gator httpwwwcpeuchicagoedu

Dependent Variables

We calculated days from muster until desertion arrest orAWOL We allowed for censoring by also calculating days frommuster until death discharge changing company becoming pris-oner of war or missing in action In examining time until arrestor AWOL individually we treated men who deserted as censoredFirst arrests therefore exclude those for desertion but could befor AWOL insubordination theft sleeping on picket duty drunk-enness or other infractions Men who were AWOL illegally ex-tended their leaves or straggled from the company after a battleor during a march A military court convened in the eld deter-mined whether a man deserted Those determined by the court tohave deserted are listed as having deserted from when they wererst missing

Socioeconomic and Demographic Characteristics

1 Occupation Dummy variables indicating whether atenlistment the recruit reported his occupation as farmerartisan professional or proprietor or laborer Farmersrsquosons who were not yet farmers in their own right wouldgenerally report themselves as farmers

2 Birthplace Dummy variables indicating whether at en-

543COWARDS AND HEROES

and AWOL Predictions are based upon the predicted survivalfunction from the time of muster calculated for every individualand then averaged over the whole sample In the case of oursummary measure of disloyalty and of desertion the single mostimportant variables were age and occupational diversity withinthe company In the case of arrests birthplace and occupationalfragmentation the fraction of Union victories the percentage ofthe county voting for Lincoln and year of muster were the singlemost important predictors Birthplace diversity age diversityand the fraction of the county voting for Lincoln were the mostimportant predictors of AWOL On the whole company socioeco-nomic and demographic characteristics were the most important

TABLE VIPREDICTED PROBABILITIES OF DESERTION ARREST AND AWOL BY COMPANY

CHARACTERISTICS MORALE AND IDEOLOGY

DesertionArrest or

AWOL Desertion Arrest AWOL

Using true variable values 0131 0094 0020 0033Community characteristics

If birthplace fragmentation 5 0 0101 0078 0010 0019If occupational fragmentation

5 00085 0047 0010 0039

If coefcient of variation forage 5 0

0065 0042 0023 0023

If all of above 0031 0016 0006 0016If population in city of

enlistment 5 25000125 0088 0020 0032

If all of above 0030 0015 0006 0015Morale

If company death rate 5 0 0118 0084 0020 0027If fraction Union victories 5 1 0108 0067 0015 0034If both 0097 0060 0022 0028

IdeologyIf volunteer 0128 0091 0020 0032If 866 county voted for

Lincoln0103 0079 0015 0022

If mustered in 1861 0107 0066 0015 0039If all of above 0082 0054 0011 0025

Our summary measure (the rst desertion arrest or AWOL) is predicted from the second specicationin Table IV All other probabilities are predicted using the specications in Table V Predictions are basedupon the predicted survival function from the time of muster using the actual data and averaging over thewhole sample In this sample the largest share of the vote Lincoln received in a county was 866 percentCities with a population of less than 2500 were not even listed in the census and are therefore consideredsmall towns

541COWARDS AND HEROES

predictors of desertion arrest and AWOL and our ideology prox-ies were relatively more important than our morale proxies23 Wealso examined predicted probabilities for our summary measureof all recruits being literate native-born 30-year old single farm-ers with personal property wealth equal to the mean plus a halfstandard deviation Being literate and a being a farmer were thesingle most important individual predictors of loyalty and all ofour individual characteristics combined were better predictors ofloyalty than our ideology or morale proxies

Why does the Army today not make greater use of socialcapital by creating socioeconomic and demographically homoge-neous ghting units Desertion on the modern battleeld is nowharder because the battleeld is larger more congested in therear with administrative soldiers and also more leveled by bomb-ing [Keegan 1976 p 316] Two additional reasons include diver-sication and human capital specialization in the modern armyHighly publicized losses to communities during World War IIended any practice of locally based companies In the modernarmy soldiers perform a myriad of tasks requiring differenttraining In contrast during the Civil War the job of a soldier wasunskilled largely consisting of learning the movement of linearformations of obeying orders without hesitation and of master-ing the nine steps of loading a musket and ring in the directionof an enemy hidden by the smoke of the battleeld [Hess 1997 pp18ndash19 137] Worker skills were perfect substitutes

VIII CONCLUSION

Why do ordinary soldiers ght when pay is low and whendesertion is a choice that many have the opportunity to exerciseIs it the attributes of the person is it loyalty to a small group ofindividuals ideology or morale Most sociologists and psycholo-gists have emphasized the importance of loyalty to a small groupof individuals Military historians however have reminded us ofthe importance of a moral crusade in motivating democratic sol-diers and of morale in keeping an army from faltering We haveshown that among Union Army soldiers in the Civil War individ-ual socioeconomic and demographic characteristics that proxy for

23 If the Union Army had had the power to construct companies at randomit could not have reduced shirking by following this policy because birthplaceoccupational and age fragmentation would have increased for many companies

542 QUARTERLY JOURNAL OF ECONOMICS

ghting ability company socioeconomic and demographic charac-teristics ideological commitment and morale were all importantdeterminants of group loyalty Company socioeconomic and de-mographic characteristics were particularly important evenmore so than ideological commitment and morale in one of ourcountryrsquos more ideological wars Heterogeneity is an importantdeterminant of participation in organizations today [Alesina andLa Ferrara 2000 Costa and Kahn 2003] and it was also animportant predictor in a very different environment in the past

DATA APPENDIX

This appendix describes the construction of our dependentvariables our demographic and socioeconomic variables our com-munity variables our ideology variables and our morale vari-ables All data on Union Army recruits are obtained from Aging ofVeterans of the Union Army Robert W Fogel Principal Investi-gator httpwwwcpeuchicagoedu

Dependent Variables

We calculated days from muster until desertion arrest orAWOL We allowed for censoring by also calculating days frommuster until death discharge changing company becoming pris-oner of war or missing in action In examining time until arrestor AWOL individually we treated men who deserted as censoredFirst arrests therefore exclude those for desertion but could befor AWOL insubordination theft sleeping on picket duty drunk-enness or other infractions Men who were AWOL illegally ex-tended their leaves or straggled from the company after a battleor during a march A military court convened in the eld deter-mined whether a man deserted Those determined by the court tohave deserted are listed as having deserted from when they wererst missing

Socioeconomic and Demographic Characteristics

1 Occupation Dummy variables indicating whether atenlistment the recruit reported his occupation as farmerartisan professional or proprietor or laborer Farmersrsquosons who were not yet farmers in their own right wouldgenerally report themselves as farmers

2 Birthplace Dummy variables indicating whether at en-

543COWARDS AND HEROES

predictors of desertion arrest and AWOL and our ideology prox-ies were relatively more important than our morale proxies23 Wealso examined predicted probabilities for our summary measureof all recruits being literate native-born 30-year old single farm-ers with personal property wealth equal to the mean plus a halfstandard deviation Being literate and a being a farmer were thesingle most important individual predictors of loyalty and all ofour individual characteristics combined were better predictors ofloyalty than our ideology or morale proxies

Why does the Army today not make greater use of socialcapital by creating socioeconomic and demographically homoge-neous ghting units Desertion on the modern battleeld is nowharder because the battleeld is larger more congested in therear with administrative soldiers and also more leveled by bomb-ing [Keegan 1976 p 316] Two additional reasons include diver-sication and human capital specialization in the modern armyHighly publicized losses to communities during World War IIended any practice of locally based companies In the modernarmy soldiers perform a myriad of tasks requiring differenttraining In contrast during the Civil War the job of a soldier wasunskilled largely consisting of learning the movement of linearformations of obeying orders without hesitation and of master-ing the nine steps of loading a musket and ring in the directionof an enemy hidden by the smoke of the battleeld [Hess 1997 pp18ndash19 137] Worker skills were perfect substitutes

VIII CONCLUSION

Why do ordinary soldiers ght when pay is low and whendesertion is a choice that many have the opportunity to exerciseIs it the attributes of the person is it loyalty to a small group ofindividuals ideology or morale Most sociologists and psycholo-gists have emphasized the importance of loyalty to a small groupof individuals Military historians however have reminded us ofthe importance of a moral crusade in motivating democratic sol-diers and of morale in keeping an army from faltering We haveshown that among Union Army soldiers in the Civil War individ-ual socioeconomic and demographic characteristics that proxy for

23 If the Union Army had had the power to construct companies at randomit could not have reduced shirking by following this policy because birthplaceoccupational and age fragmentation would have increased for many companies

542 QUARTERLY JOURNAL OF ECONOMICS

ghting ability company socioeconomic and demographic charac-teristics ideological commitment and morale were all importantdeterminants of group loyalty Company socioeconomic and de-mographic characteristics were particularly important evenmore so than ideological commitment and morale in one of ourcountryrsquos more ideological wars Heterogeneity is an importantdeterminant of participation in organizations today [Alesina andLa Ferrara 2000 Costa and Kahn 2003] and it was also animportant predictor in a very different environment in the past

DATA APPENDIX

This appendix describes the construction of our dependentvariables our demographic and socioeconomic variables our com-munity variables our ideology variables and our morale vari-ables All data on Union Army recruits are obtained from Aging ofVeterans of the Union Army Robert W Fogel Principal Investi-gator httpwwwcpeuchicagoedu

Dependent Variables

We calculated days from muster until desertion arrest orAWOL We allowed for censoring by also calculating days frommuster until death discharge changing company becoming pris-oner of war or missing in action In examining time until arrestor AWOL individually we treated men who deserted as censoredFirst arrests therefore exclude those for desertion but could befor AWOL insubordination theft sleeping on picket duty drunk-enness or other infractions Men who were AWOL illegally ex-tended their leaves or straggled from the company after a battleor during a march A military court convened in the eld deter-mined whether a man deserted Those determined by the court tohave deserted are listed as having deserted from when they wererst missing

Socioeconomic and Demographic Characteristics

1 Occupation Dummy variables indicating whether atenlistment the recruit reported his occupation as farmerartisan professional or proprietor or laborer Farmersrsquosons who were not yet farmers in their own right wouldgenerally report themselves as farmers

2 Birthplace Dummy variables indicating whether at en-

543COWARDS AND HEROES

ghting ability company socioeconomic and demographic charac-teristics ideological commitment and morale were all importantdeterminants of group loyalty Company socioeconomic and de-mographic characteristics were particularly important evenmore so than ideological commitment and morale in one of ourcountryrsquos more ideological wars Heterogeneity is an importantdeterminant of participation in organizations today [Alesina andLa Ferrara 2000 Costa and Kahn 2003] and it was also animportant predictor in a very different environment in the past

DATA APPENDIX

This appendix describes the construction of our dependentvariables our demographic and socioeconomic variables our com-munity variables our ideology variables and our morale vari-ables All data on Union Army recruits are obtained from Aging ofVeterans of the Union Army Robert W Fogel Principal Investi-gator httpwwwcpeuchicagoedu

Dependent Variables

We calculated days from muster until desertion arrest orAWOL We allowed for censoring by also calculating days frommuster until death discharge changing company becoming pris-oner of war or missing in action In examining time until arrestor AWOL individually we treated men who deserted as censoredFirst arrests therefore exclude those for desertion but could befor AWOL insubordination theft sleeping on picket duty drunk-enness or other infractions Men who were AWOL illegally ex-tended their leaves or straggled from the company after a battleor during a march A military court convened in the eld deter-mined whether a man deserted Those determined by the court tohave deserted are listed as having deserted from when they wererst missing

Socioeconomic and Demographic Characteristics

1 Occupation Dummy variables indicating whether atenlistment the recruit reported his occupation as farmerartisan professional or proprietor or laborer Farmersrsquosons who were not yet farmers in their own right wouldgenerally report themselves as farmers

2 Birthplace Dummy variables indicating whether at en-

543COWARDS AND HEROES

listment the recruit reported his birthplace as the UnitedStates Germany Ireland Great Britain or other

3 Age at enlistment Age at rst enlistment4 Height in inches Height in inches at rst enlistment5 Married in 1860 This variable is inferred from family

member order and age in the 1860 census This variablewas set equal to zero if the recruit was not linked to the1860 census

6 Log (total household personal property) in 1860This variable is the sum of personal property wealth ofeveryone in the recruitsrsquo 1860 household This variable isset equal to zero if the recruit was not linked to the 1860census

7 Missing census information A dummy equal to one ifthe recruit was not linked to the 1860 census Linkagerates from the military service records to the 1860 censuswere 57 percent The main characteristic that predictedlinkage failure was foreign birth

8 Illiterate This variable is from the 1860 census andprovides illiteracy information only for those age 20 andolder

9 Missing illiteracy information A dummy equal to oneif we do not know whether the recruits were illiterateeither because he was not linked to the 1860 census orbecause he was less than age 20 in 1860

10 Region effects Our region dummies are New EnglandMiddle Atlantic East North Central West North Cen-tral Border and West

Community Characteristics

1 Birthplace fragmentation We calculated by com-pany the fraction of individuals born in the UnitedStates in New England in the Middle Atlantic in theEast North Central in the West North Central the Bor-der states the south and the west and born abroad inGermany Ireland Canada Great Britain Scandinavianorthwestern Europe (France Belgium Luxembourgthe Netherlands) other areas of Europe and other areasof the world Our birthplace fragmentation index fi isthen

fi 5 1 2 Ok

ski2

544 QUARTERLY JOURNAL OF ECONOMICS

where k represents the categories and where sk i is theshare of men of born in place k in company i

2 Occupational fragmentation We calculated by com-pany the fraction of individuals who were farmershigher class professionals and proprietors lower classprofessionals and proprietors artisans higher class la-borers lower class laborers and unknown Our occupa-tional fragmentation index is then calculated similarly toour birthplace fragmentation index

3 Coefcient of variation for age We calculated bycompany the coefcient of variation for age atenlistment

4 Population in city of enlistment We obtained popu-lation in city of enlistment from Union Army Recruits inWhite Regiments in the United States 1861ndash1865 (ICPSR9425) Robert W Fogel Stanley L Engerman ClaynePope and Larry Wimmer Principal Investigators Citiesthat could not be identied were assumed to be cities ofpopulation less than 2500

Ideology Variables

1 Year of muster Dummy variables indicating the yearthat the soldier was rst mustered in

2 Volunteer A dummy equal to one if the recruit was avolunteer instead of a draftee or a substitute

3 Percent of vote in 1860 Presidential election Weobtained by county of enlistment the fraction of the votecast for Lincoln and for other candidates from ElectoralData for Counties in the United States Presidential andCongressional Races 1840 ndash1972 (ICPSR 8611) JeromeM Clubb William H Flanigan and Nancy H ZingalePrincipal Investigators Because we cannot attribute acounty to each recruit our categories are percent incounty of enlistment voting for Lincoln other candidateand unknown

Morale Variables

1 Fraction in company dying We calculated by com-pany the fraction dying overall and the fraction dying(among all men at risk to die) within all half-years thateach recruit served Our means present the fraction dy-ing overall Our regression results use the time-varying

545COWARDS AND HEROES

covariate fraction of men at risk dying during all half-years that each recruit served

2 Fraction of major Union victories This is a time-varying variable that indicates for each half-year thatthe recruit was in the service the fraction of major Unionvictories to all major battles in that half-year It takes thevalue zero if there were no major battles

MASSACHUSETTS INSTITUTE OF TECHNOLOGY AND NATIONAL BUREAU OF ECONOMIC

RESEARCH

THE FLETCHER SCHOOL TUFTS UNIVERSITY

REFERENCES

Alesina Alberto Reza Baqir and William Easterly ldquoPublic Goods and EthnicDivisionsrdquo Quarterly Journal of Economics CXIV (1999) 1243ndash1284

Alesina Alberto and Eliana La Ferrara ldquoParticipation in Heterogeneous Com-munitiesrdquo Quarterly Journal of Economics CXV (2000) 847ndash904

Bearman Peter S ldquoDesertion as Localism Army Unit Solidarity and GroupNorms in the U S Civil Warrdquo Social Forces LXX (1991) 321ndash342

Beevor Antony Stalingrad (New York Viking 1998)Berman Eli ldquoSect Subsidy and Sacrice An Economistrsquos View of Ultra-Ortho-

dox Jewsrdquo Quarterly Journal of Economics CXV (2000) 905ndash954Brock William A and Steven N Durlauf ldquoDiscrete Choice with Social Interac-

tionsrdquo Review of Economic Studies LXVIII (2001) 235ndash260Cai T L J Wei and M Wilcox ldquoSemiparametric Regression Analysis for

Clustered Failure Time Datardquo Biometrika LXXXVII (2000) 867ndash 878Coleman James Foundations of Social Theory (Cambridge MA Harvard Uni-

versity Press 1990)Costa Dora L and Matthew E Kahn ldquoUnderstanding the American Decline in

Social Capital 1952ndash1998rdquo Kyklos LVI (2003)Dollard John Fear in Battle (New Haven CT The Institute of Human Relations

Yale University 1943)Fogel Robert W Without Consent or Contract The Rise and Fall of American

Slavery (New York WW Norton and Company 1989)mdashmdash ldquoEarly Indicators of Later Work Levels Disease and Deathrdquo Grant submit-

ted to NIH February 1 2001Gibbons Robert ldquoIncentives in Organizationsrdquo Journal of Economic Perspectives

XII (1998) 115ndash132Glaeser Edward L David Laibson and Bruce Sacerdote ldquoThe Economic Ap-

proach to Social Capitalrdquo National Bureau of Economic Research WorkingPaper No 7728 2000

Goldin Claudia and Lawrence F Katz ldquoHuman Capital and Social Capital TheRise of Secondary Schooling in America 1910 to 1940rdquo Journal of Interdis-ciplinary History XXIX (1999) 683ndash723

Gould Benjamin Apthrop Investigations in the Military and AnthropologicalStatistics of American Soldiers (New York Published for the United StatesSanitary Commission by Hurd and Houghton 1869)

Hanson Victor Davis The Soul of Battle From Ancient Times to the Present DayHow Three Great Liberators Vanquished Tyranny (New York Anchor Books1999)

mdashmdash Carnage and Culture Landmark Battles in the Rise of Western Power (NewYork Doubleday 2001)

Hattaway Herman M ldquoThe Civil War Armies Creation Mobilization and De-velopmentrdquo On the Road to Total War The American Civil War and theGerman Wars of Unication 1861ndash1871 Stig Foster and Jorg Nagler eds

546 QUARTERLY JOURNAL OF ECONOMICS

where k represents the categories and where sk i is theshare of men of born in place k in company i

2 Occupational fragmentation We calculated by com-pany the fraction of individuals who were farmershigher class professionals and proprietors lower classprofessionals and proprietors artisans higher class la-borers lower class laborers and unknown Our occupa-tional fragmentation index is then calculated similarly toour birthplace fragmentation index

3 Coefcient of variation for age We calculated bycompany the coefcient of variation for age atenlistment

4 Population in city of enlistment We obtained popu-lation in city of enlistment from Union Army Recruits inWhite Regiments in the United States 1861ndash1865 (ICPSR9425) Robert W Fogel Stanley L Engerman ClaynePope and Larry Wimmer Principal Investigators Citiesthat could not be identied were assumed to be cities ofpopulation less than 2500

Ideology Variables

1 Year of muster Dummy variables indicating the yearthat the soldier was rst mustered in

2 Volunteer A dummy equal to one if the recruit was avolunteer instead of a draftee or a substitute

3 Percent of vote in 1860 Presidential election Weobtained by county of enlistment the fraction of the votecast for Lincoln and for other candidates from ElectoralData for Counties in the United States Presidential andCongressional Races 1840 ndash1972 (ICPSR 8611) JeromeM Clubb William H Flanigan and Nancy H ZingalePrincipal Investigators Because we cannot attribute acounty to each recruit our categories are percent incounty of enlistment voting for Lincoln other candidateand unknown

Morale Variables

1 Fraction in company dying We calculated by com-pany the fraction dying overall and the fraction dying(among all men at risk to die) within all half-years thateach recruit served Our means present the fraction dy-ing overall Our regression results use the time-varying

545COWARDS AND HEROES

covariate fraction of men at risk dying during all half-years that each recruit served

2 Fraction of major Union victories This is a time-varying variable that indicates for each half-year thatthe recruit was in the service the fraction of major Unionvictories to all major battles in that half-year It takes thevalue zero if there were no major battles

MASSACHUSETTS INSTITUTE OF TECHNOLOGY AND NATIONAL BUREAU OF ECONOMIC

RESEARCH

THE FLETCHER SCHOOL TUFTS UNIVERSITY

REFERENCES

Alesina Alberto Reza Baqir and William Easterly ldquoPublic Goods and EthnicDivisionsrdquo Quarterly Journal of Economics CXIV (1999) 1243ndash1284

Alesina Alberto and Eliana La Ferrara ldquoParticipation in Heterogeneous Com-munitiesrdquo Quarterly Journal of Economics CXV (2000) 847ndash904

Bearman Peter S ldquoDesertion as Localism Army Unit Solidarity and GroupNorms in the U S Civil Warrdquo Social Forces LXX (1991) 321ndash342

Beevor Antony Stalingrad (New York Viking 1998)Berman Eli ldquoSect Subsidy and Sacrice An Economistrsquos View of Ultra-Ortho-

dox Jewsrdquo Quarterly Journal of Economics CXV (2000) 905ndash954Brock William A and Steven N Durlauf ldquoDiscrete Choice with Social Interac-

tionsrdquo Review of Economic Studies LXVIII (2001) 235ndash260Cai T L J Wei and M Wilcox ldquoSemiparametric Regression Analysis for

Clustered Failure Time Datardquo Biometrika LXXXVII (2000) 867ndash 878Coleman James Foundations of Social Theory (Cambridge MA Harvard Uni-

versity Press 1990)Costa Dora L and Matthew E Kahn ldquoUnderstanding the American Decline in

Social Capital 1952ndash1998rdquo Kyklos LVI (2003)Dollard John Fear in Battle (New Haven CT The Institute of Human Relations

Yale University 1943)Fogel Robert W Without Consent or Contract The Rise and Fall of American

Slavery (New York WW Norton and Company 1989)mdashmdash ldquoEarly Indicators of Later Work Levels Disease and Deathrdquo Grant submit-

ted to NIH February 1 2001Gibbons Robert ldquoIncentives in Organizationsrdquo Journal of Economic Perspectives

XII (1998) 115ndash132Glaeser Edward L David Laibson and Bruce Sacerdote ldquoThe Economic Ap-

proach to Social Capitalrdquo National Bureau of Economic Research WorkingPaper No 7728 2000

Goldin Claudia and Lawrence F Katz ldquoHuman Capital and Social Capital TheRise of Secondary Schooling in America 1910 to 1940rdquo Journal of Interdis-ciplinary History XXIX (1999) 683ndash723

Gould Benjamin Apthrop Investigations in the Military and AnthropologicalStatistics of American Soldiers (New York Published for the United StatesSanitary Commission by Hurd and Houghton 1869)

Hanson Victor Davis The Soul of Battle From Ancient Times to the Present DayHow Three Great Liberators Vanquished Tyranny (New York Anchor Books1999)

mdashmdash Carnage and Culture Landmark Battles in the Rise of Western Power (NewYork Doubleday 2001)

Hattaway Herman M ldquoThe Civil War Armies Creation Mobilization and De-velopmentrdquo On the Road to Total War The American Civil War and theGerman Wars of Unication 1861ndash1871 Stig Foster and Jorg Nagler eds

546 QUARTERLY JOURNAL OF ECONOMICS

covariate fraction of men at risk dying during all half-years that each recruit served

2 Fraction of major Union victories This is a time-varying variable that indicates for each half-year thatthe recruit was in the service the fraction of major Unionvictories to all major battles in that half-year It takes thevalue zero if there were no major battles

MASSACHUSETTS INSTITUTE OF TECHNOLOGY AND NATIONAL BUREAU OF ECONOMIC

RESEARCH

THE FLETCHER SCHOOL TUFTS UNIVERSITY

REFERENCES

Alesina Alberto Reza Baqir and William Easterly ldquoPublic Goods and EthnicDivisionsrdquo Quarterly Journal of Economics CXIV (1999) 1243ndash1284

Alesina Alberto and Eliana La Ferrara ldquoParticipation in Heterogeneous Com-munitiesrdquo Quarterly Journal of Economics CXV (2000) 847ndash904

Bearman Peter S ldquoDesertion as Localism Army Unit Solidarity and GroupNorms in the U S Civil Warrdquo Social Forces LXX (1991) 321ndash342

Beevor Antony Stalingrad (New York Viking 1998)Berman Eli ldquoSect Subsidy and Sacrice An Economistrsquos View of Ultra-Ortho-

dox Jewsrdquo Quarterly Journal of Economics CXV (2000) 905ndash954Brock William A and Steven N Durlauf ldquoDiscrete Choice with Social Interac-

tionsrdquo Review of Economic Studies LXVIII (2001) 235ndash260Cai T L J Wei and M Wilcox ldquoSemiparametric Regression Analysis for

Clustered Failure Time Datardquo Biometrika LXXXVII (2000) 867ndash 878Coleman James Foundations of Social Theory (Cambridge MA Harvard Uni-

versity Press 1990)Costa Dora L and Matthew E Kahn ldquoUnderstanding the American Decline in

Social Capital 1952ndash1998rdquo Kyklos LVI (2003)Dollard John Fear in Battle (New Haven CT The Institute of Human Relations

Yale University 1943)Fogel Robert W Without Consent or Contract The Rise and Fall of American

Slavery (New York WW Norton and Company 1989)mdashmdash ldquoEarly Indicators of Later Work Levels Disease and Deathrdquo Grant submit-

ted to NIH February 1 2001Gibbons Robert ldquoIncentives in Organizationsrdquo Journal of Economic Perspectives

XII (1998) 115ndash132Glaeser Edward L David Laibson and Bruce Sacerdote ldquoThe Economic Ap-

proach to Social Capitalrdquo National Bureau of Economic Research WorkingPaper No 7728 2000

Goldin Claudia and Lawrence F Katz ldquoHuman Capital and Social Capital TheRise of Secondary Schooling in America 1910 to 1940rdquo Journal of Interdis-ciplinary History XXIX (1999) 683ndash723

Gould Benjamin Apthrop Investigations in the Military and AnthropologicalStatistics of American Soldiers (New York Published for the United StatesSanitary Commission by Hurd and Houghton 1869)

Hanson Victor Davis The Soul of Battle From Ancient Times to the Present DayHow Three Great Liberators Vanquished Tyranny (New York Anchor Books1999)

mdashmdash Carnage and Culture Landmark Battles in the Rise of Western Power (NewYork Doubleday 2001)

Hattaway Herman M ldquoThe Civil War Armies Creation Mobilization and De-velopmentrdquo On the Road to Total War The American Civil War and theGerman Wars of Unication 1861ndash1871 Stig Foster and Jorg Nagler eds

546 QUARTERLY JOURNAL OF ECONOMICS

(Cambridge-New York German Historical Institute and Cambridge Univer-sity Press 1997)

Hess Earl J The Union Soldier in Battle Enduring the Ordeal of Combat(Lawrence KS University Press of Kansas 1997)

Holmstrom Bengt ldquoMoral Hazard in Teamsrdquo Bell Journal of Economics XIII(1982) 324ndash340

Ichino Andrea and Giovanni Maggi ldquoWork Environment and Individual Back-ground Explaining Regional Shirking Differentials in a Large Italian FirmrdquoQuarterly Journal of Economics CXV (2000) 1057ndash1090

Kandel Eugene and Edward P Lazear ldquoPeer Pressure and Partnershipsrdquo Jour-nal of Political Economy C (1992) 801ndash817

Keegan John The Face of Battle (Hartmondsworth Middlesex UK PenguinBooks 1976)

Kemp Thomas R ldquoCommunity and War The Civil War Experience of Two NewHampshire Townsrdquo Toward a Social History of the American Civil WarExploratory Essays Maris A Vinovskis ed (New York-Cambridge Cam-bridge University Press 1990)

Knippschild Dieter ldquoDeserteure im Zweiten Weltkrieg Der Stand der DebatterdquoArmeen und ihre Deserteure Vernachlassigte Kapitel einer Militargeschichteder Neuzeit Ulrich Brockling and Michael Sikora eds (Gottingen GermanyVandenhoeck amp Ruprecht 1998)

Lazear Edward P ldquoWhy Is There Mandatory Retirementrdquo Journal of PoliticalEconomy LXXXVII (1979) 1261ndash1284

Lee W W L J Wei and D A Amato ldquoCox-Type Regression Analysis for LargeNumbers of Small Groups of Correlated Failure Time Observationsrdquo SurvivalAnalysis State of the Art J P Klein and P K Goel eds (Dordrecht Neth-erlands Kluwer 1992)

Levitt Steven D and Sudhir Alladi Venkatesh ldquoAn Economic Analysis of aDrug-Selling Gangrsquos Financesrdquo Quarterly Journal of Economics CXV (2000)755ndash790

Linderman Gerald F Embattled Courage The Experience of Combat in theAmerican Civil War (New York The Free Press 1987)

Luttmer Erzo F P ldquoGroup Loyalty and the Taste for Redistributionrdquo Journal ofPolitical Economy CIX (2001) 500ndash528

Manski Charles F ldquoIdentication of Endogenous Social Effects The ReectionProblemrdquo Review of Economic Studies LX (1993) 531ndash542

mdashmdash ldquoEconomic Analysis of Social Interactionsrdquo Journal of Economic Perspec-tives XIV (2000) 115ndash136

Margo Robert A and Richard H Steckel ldquoHeights of Native-Born Whites Duringthe Antebellum Periodrdquo Journal of Economic History XLIII (1983) 167ndash174

McPherson James M For Cause and Comrades Why Men Fought in the CivilWar (Oxford-New York Oxford University Press 1997)

Menand Louis The Metaphysical Club (New York Farrar Straus and Giroux2001)

Mitchell Reid ldquoThe Northern Soldier and His Communityrdquo Toward a SocialHistory of the American Civil War Exploratory Essays Maris A Vinovskised (New York-Cambridge Cambridge University Press 1990)

OrsquoReilly C A D F Caldwell and W P Barnett ldquoWork Group DemographySocial Integration and Turnoverrdquo Administrative Science Quarterly XXXIV(1989) 21ndash37

Pfeffer Jeffrey New Directions for Organization Theory (New York-Oxford Ox-ford University Press 1997)

Poterba James M ldquoDemographic Structure and the Political Economy of PublicEducationrdquo Journal of Policy Analysis and Management XVI (1997) 48ndash 66

Sikora Michael ldquoDas 18 Jahrhundert Die Zeit der Deserteurerdquo Armeen und ihreDeserteure Vernachlassigte Kapitel einer Militargeschichte der Neuzeit Ul-rich Brockling and Michael Sikora eds (Gottingen Germany Vandenhoeckamp Ruprecht 1998)

Stouffer Samuel A et al The American Soldier Combat and its AftermathVolume II (Princeton NJ Princeton University Press 1949)

United States Provost Marshall General Final Report United States House of

547COWARDS AND HEROES

Representatives Executive Document No 1 39th Congress 1st SessionSeries Numbers 1251 1252 1866

Vinovskis Maris A ldquoHave Social Historians Lost the Civil War Some Prelimi-nary Demographic Speculationsrdquo Toward a Social History of the AmericanCivil War Exploratory Essays Maris A Vinovskis ed (New York-Cam-bridge Cambridge University Press 1990)

Weitz Mark A A Higher Duty Desertion Among Georgia Troops During the CivilWar (Lincoln NB and London University of Nebraska Press 2000)

548 QUARTERLY JOURNAL OF ECONOMICS