Addiction (1999) 94(10), 1477± 1487

RESEARCH REPORT

Heavy alcohol consumption and maritalstatus: disentangling the relationship in anational study of young adults

CHRIS POWER,1 BRYAN RODGERS2 & STEVEN HOPE1

1Epidemiology and Public Health, Institute of Child Health, London, UK & 2NHMRCPsychiatric Epidemiology Research Centre, The Australian National University, Canberra,Australia.

Abstract

Aims. To investigate why alcohol consumption varies by marital status, assessing (i) differences in heavyconsumption prior to changes in marital status (indicating selection) and increases or decreases in heavy

consumption associated with changes in marital status (indicating causation), (ii) whether such increases ordecreases are transient, and (iii) the possible mediating effect of parental status. Design. Longitudinal

cohort. Setting. Great Britain. Participants. Data from the 23- and 33-year surveys of the 1958 British

birth cohort (all born in England, Wales and Scotland, 3± 9 March 1958). Measurements. Heavydrinking, de® ned as more than 35 (men) and 20 (women) units/week; changes between ages 23 and 33 in

consumption and marital status. Findings. The divorced had the highest consumption levels at both ages,

the married had the lowest. Selection effects were minimal in both sexes. Overall, heavy drinking declinedbetween ages 23 and 33 (21.4± 13.0% in men, 6.4± 3.4% in women), but increased among individuals who

divorced, compared to the continuously married (adjusted OR 5 2.05, 95% CI 5 1.49,2.83 for men;

OR 5 2.61, 95% CI 5 1.67,4.09 for women), most strikingly for recent divorces (adjusted OR 5 4.97, 95%CI 5 2.86,8.57 and OR 5 5.25, 95% CI 5 2.60,10.65). High rates of heavy drinking persisted for never

married men (19.1%) and women (5.2%). Conclusions. The heavy drinking level of divorced young

adults was not due to selection. Marital separation was accompanied by increases in heavy drinking, withpronounced short-term effects. Adverse alcohol-related health consequences may occur in the immediate period

around divorce. Individuals who never marry appear to have a chronic heavy consumption pattern that may

contribute to their increased mortality.

Introduction

One of the most consistent ® ndings in the litera-ture on alcohol use is that patterns of consump-tion vary by marital status, with heavier drinkinglevels among divorced or single people thanamong those who are married (Wilson, 1980;

Johnson, 1982; Wilsnack et al., 1984; Goddard,1991; Thomas et al., 1994; Leino et al., 1995;Single, 1995; Luoto et al., 1998). However, theprocesses leading to these associations remainunclear. It may be, for example, that excessalcohol consumption among divorced people is

Correspondence to: Chris Power, Epidemiology and Public Health, Institute of Child Health, 30 Guilford Street,London WC1N 1EH, UK.

Submitted 13th November 1998; initial review completed 12th February 1999; ® nal version accepted 26th March1999.

ISSN 0965± 2140 print/ISSN 1360-0443 online/99/101477± 11 Ó Society for the Study of Addiction to Alcohol and Other Drugs

Carfax Publishing, Taylor & Francis Limited

1478 Chris Power et al.

due to elevated rates of divorce in heavydrinkers. A selective process of this sort is indi-cated by the presence of heavy drinking beforemarital breakdown occurs. Alternatively, divorceand its concomitants, such as ® nancial hardshipor poor social support, may play a causal role asindicated by an increase in consumption accom-panying marital breakdown. Excess alcohol con-sumption among single people could similarly bedue to selection, with heavy drinkers being lesslikely to marry; or, alternatively, marriage may beaccompanied by a decrease in consumption. Fewstudies have attempted to disentangle the associ-ation between marital status and alcohol con-sumption to help determine underlying causalrelationships. Some evidence exists, however,suggesting that selective processes are operating,whereby problem drinkers are less likely thanmore moderate drinkers to commence and sus-tain a partnership (Power & Estaugh, 1990;Wilsnack et al., 1991; Chilcoat & Breslau, 1996).Evidence exists also for decreases in consump-tion following marriage (Power & Estaugh, 1990;Hajema & Knibbe, 1998) and increases in con-sumption accompanying divorce (Doherty et al.,1989; Power & Estaugh, 1990; Miller-Tutzaueret al., 1991; Romelsjo et al., 1991; Temple et al.,1991; Leino et al., 1995; Chilcoat & Breslau,1996; Horwitz et al., 1996; Bachman et al., 1997;Hajema & Knibbe, 1998). Increased drinkingmay also be a short-term response to the distressof marital breakdown (Bloom et al., 1978; Hor-witz et al., 1996), in which case stronger effectswill be evident in studies with shorter follow-up.The relative contributions of selection and cau-sation will therefore appear to vary with durationof follow-up. A further possibility is that drinkinglevels are in¯ uenced primarily by parentalresponsibility, or anticipation of parenthood(Bachman et al., 1997). It may be parentalresponsibility that is implicated in the lowerconsumption of married people.

We investigated the contributions of selectiveand causal processes in early adulthood to theassociation between marital status and alcoholconsumption, using data from the 23- and 33-year follow-ups of 1958 British birth cohort. Ouraim was to establish whether differences in heavyconsumption exist prior to changes in maritalstatus (indicating selection), or whether increasesor decreases in heavy consumption accompany achange in marital status (indicating causation).We also examined whether changes in heavy

consumption related to marital status are transi-ent or long-term, and whether effects on con-sumption associated with marital status re¯ ectparental status.

Methods

Study sampleThe 1958 birth cohort study includes all childrenborn in England, Scotland and Wales during oneweek, 3± 9 March 1958. The study methodologyhas been described in detail elsewhere (Power,1992; Ferri, 1993). In brief, data were collectedat birth, and at ages 7, 11, 16, 23 and 33.Immigrants to Britain born in the same weekwere added to the study sample at ages 7, 11 and16. At age 23, 12 537 subjects were interviewed(76% of the target population), and at age 33,11 405 subjects (69% of the target). Sampleattrition has been associated with under-repre-sentation of those with the most disadvantagedbackgrounds. Subjects with lower social class atbirth (Power et al., 1996), whose parents sepa-rated during their childhood (Hope et al., 1998),or who experienced psychological distress at age23 (Hope et al., 1999) were less likely to respondat age 33. However, such biases tend to be smalland the sample is considered to be generallyrepresentative of the original birth cohort (Ferri,1993). For example, 19.4% of men respondingat age 33 had been born into classes IV and Vcompared with 21.1% in the original sample, forwomen the ® gures are 20.5% and 21.4%, re-spectively (Power et al., 1996). Information atage 23 and 33 was obtained through personalinterview with the study subject.

MeasuresAt ages 23 and 33 individuals reported theamount of beer, lager, spirits, wine and otheralcoholic drinks they had consumed in the pre-vious week. Units of alcohol were estimated,with 1 unit equivalent to half a pint of beer, onemeasure of spirits or one glass of wine (1unit 5 8 gms of alcohol). Heavy drinking wasde® ned as more than 20 units a week for womenand more than 35 units for men. Those drinkingless than these cut-offs, including abstainers,formed a comparison group.

Marital status ascertained at ages 23 and 33was grouped as: single (i.e. never married andnot presently cohabiting), married (or presently

Alcohol and marital status 1479

cohabiting), divorced or remarried. We com-bined cohabitees with the married or remarriedsince they were self-de® ned as ª living as mar-riedº . Cohabitees were more likely than the mar-ried to drink heavily, especially at age 23 (21.8%compared with 12.6%, respectively, for men,10.0% and 2.9% for women). However, only202 men and 269 women were cohabiting at age23 and subdivision by marital status at age 33(remaining cohabiting, marrying, separating andremarrying) resulted in too few observations foranalysis. A small number of widow(er)s wereexcluded from analyses (10 at age 23, 25 at age33). Change and continuity in marital statusbetween 23 and 33 years were identi® ed.Suf® cient numbers of subjects were available for:continuously single, single to married, continu-ously married, married to divorced, and marriedto remarried. Partnership histories, including thetiming of marriage and divorce, were reported atage 33. We identi® ed the age when subjectsstopped living with a partner rather than theirage of legal divorce. Data obtained at age 33were also used to ascertain whether the respon-dent had dependent children aged 16 or youngerliving in their household at this age, whether astheir own child, adopted, fostered or step-child.

Statistical analyses

Chi-square was used to test (i) the cross-sectional association between marital status andheavy drinking at ages 23 and 33, and (ii) selec-tive processes, as indicated by comparison ofheavy drinking at age 23 in stable and changingmarital status groups (e.g. continuously single vs.single to married). To examine causal processes,as indicated by changes in drinking level betweenthe two ages, we estimated odds ratios for heavydrinking at age 33 for groups who had changedtheir marital status relative to those remainingstable (e.g. divorced vs. continuously married),conditional upon drinking level at age 23 (i.e.heavy drinking vs. not heavy drinking). Theseconditional odds ratios, obtained from logisticregression models, were tested for homogeneityby including interactions between drinking levelat age 23 and marital status change and stability,but we found no signi® cant effects indicative ofheterogeneity (Agresti, 1996). We also examinedwhether relationships between marital status andheavy drinking differed for men and women, byincluding interaction terms with gender, and

found no signi® cant interactions. Results arepresented separately for men and women so thatthe sex differences in heavy drinking remainapparent at each stage. Finally, the analyses wererepeated excluding cohabitees from the categoryof ª marriedº and the main ® ndings were foundto be unaffected (data not presented), hence theresults shown here are for the combined married/cohabiting group.

Results

Marital status and alcohol consumption at 23 and33 years

Table 1 shows the proportion of heavy drinkersand mean alcohol consumption according tomarital status at ages 23 and 33. At both ages,alcohol consumption was greater in men thanwomen (p , 0.001) and, in both sexes, consump-tion was greater at age 23 than age 33(p , 0.001). The divorced had the highest con-sumption levels, and the married had the lowest.The single and remarried had intermediateconsumption levels. Heavy drinking and meanconsumption showed similar relationships withmarital status. To assess whether sample attri-tion between ages 23 and 33 introduced bias, wecompared alcohol consumption of those withdata at age 23 but not at age 33. The proportionof heavy drinkers at age 23 was similar amongsubsequent non-responders and those followed-up at age 33 (22.8% and 21.4%, respectively formen; 6.2% and 6.4% for women). These differ-ences were not statistically signi® cant. Further-more, the association between heavy drinkingand marital status at age 23 was signi® cant andshowed a similar pattern for those followed upand those not.

Selective processes

Table 2 shows heavy drinking at ages 23 and 33in relation to change in marital status. Selectiveprocesses would be indicated by differences inheavy drinking before any change in maritalstatus occurred, that is at age 23. Three com-parisons were made separately for men andwomen: (i) continuously single vs. single to mar-ried, (ii) continuously married vs. married todivorced or remarried, and (iii) married to di-vorced vs. married to remarried. Only one of thesix analyses gave a signi® cant ® nding: heavydrinking was less common in continuously single

1480 Chris Power et al.

Table 1. Heavy drinking (%) and mean alcohol consumption (units/week) and current marital status at ages 23 and 33

Men Women

N % Heavy drinkers Mean units/week N % Heavy drinkers Mean units/week

Alcohol consumption at age 23Marital status

Single 2630 26.1 25.22 1676 10.5 8.58Married 1779 13.6 17.26 2911 3.5 4.71Divorced 66 43.9 34.32 156 14.7 9.35Remarried 23 21.7 17.26 99 8.1 6.52All 4498 21.4 22.17 4842 6.4 6.23

c2 (df) 118.6 (3)* 105.4 (3)*

Alcohol consumption at age 33Marital status

Single 686 19.0 19.48 485 5.2 5.90Married 3160 10.3 15.11 3367 2.6 4.49Divorced 223 24.7 25.14 373 5.9 6.13Remarried 429 17.2 18.31 617 5.0 6.14All 4498 13.0 16.58 4842 3.4 4.97

c2 (df) 74.7 (3) * 22.7 (3)*

* p , 0.001.

Table 2. Heavy drinking (%) at ages 23 and 33 and change in marital status

Heavy drinking

Change in marital status age 23 age 33(ages 23± 33) N (%) (%)

MenContinuously single³ 665 22.9 , 19.1Single to married 1766 26.8 , 10.0Continuously married³ 1392 12.9 10.8Married to divorced/remarried³ 367 15.5 19.9

Married to divorced² 116 14.7 27.6Married to remarried² 251 15.9 16.3

WomenContinuously single³ 459 9.4 5.2Single to married 1095 11.2 3.6Continuously married³ 2264 3.2 2.2Married to divorced/remarried³ 621 4.5 5.6

Married to divorced² 249 4.0 6.0Married to remarried² 372 4.8 5.4

² Subgroups of those divorcing between ages 23 and 33; , comparison

c2 5 4.0 (1 df), p 5 0.045; ³ small discrepancies between numbers in Tables

1 and 2 are due to differences in cohabitation as reported contemporaneouslyat age 23, or retrospectively at age 33; 21 men and 26 women reported thatthey were single± never married at age 33 but were cohabiting at age 23.

compared with single to married men. Expressedin terms of drinking status at age 23, heavydrinking men were signi® cantly more likely thanothers to marry by age 33 (76% vs. 72%). Thedirection of this association was the opposite tothat required to produce lower levels of heavy

drinking among the married through a selectiveprocess. The non-signi® cant ® ndings includedthe small differences between the continuouslymarried and those who divorced between age 23and 33. Expressing this in terms of drinkingstatus at the younger age, heavy drinkers were

0

% h

eavy

drin

king

23

SM

SC

33 23

MDR

MC

33 23

MRMD

33

5

Age

10

15

20

25

30

0

% h

eavy

drin

king

23

SM

SC

33 23

MDR

MC

33 23

MRMD

33

2

Age

4

6

88

10

12

Alcohol and marital status 1481

not signi® cantly more likely to divorce than thosewho did not drink heavily.

Causal processes(i)Transition from single. Figures 1 and 2 show

consumption changes for marital status groups.Causal processes associated with change in mari-tal status are indicated by changes in heavydrinking between ages 23 and 33. These changes

need to be viewed within the context of thegeneral decline in heavy drinking of 21.4% to13.0% in men and 6.4% to 3.4% in women(Table 1). As Figs 1 and 2 illustrate, there arelarge differences in the magnitude of decline formarital status groups. The greatest decline wasin those who moved from single to married(SM), and this differed from the smaller declinein the continuously single (SC). The single whosubsequently married (SM) had a lower risk of

Figure 1. Change in heavy drinking according to transitions (men). SM, single-married; MDR, married-divorced/re-married; MR, married-remarried; SC, continuously single; MC, continuously married; MD. married-divorced.

Figure 2. Change in heavy drinking according to transitions (women). SM, single-married; MDR, married-divorced/re-married; MR, married-remarried; SC, continuously single; MC, continuously married; MD. married-divorced.

0

% h

eavy

drin

king

23 33

5

Age

Continuously married(N =1392)

Not recent divorce(N =283)

Recent divorce(N =73)

10

15

20

25

30

35

40

1482 Chris Power et al.

heavy drinking at age 33 than the continuouslysingle (SC), after adjustment for 23-year drink-ing level (adjusted OR 5 0.41, 95%CI 5 0.31,0.53 for men; OR 5 0.61, 95%CI 5 0.36,1.04 for women). These analyses wererepeated, strati® ed by parental status at age 33,to establish whether responsibility for children,rather than marriage, was the major factor in thedecline of heavy drinking. For men, the primaryrole of marriage is suggested by similar levels ofheavy drinking among married men irrespectiveof whether they had children (11.7% of menwithout children and 9.2% of fathers). In con-trast, the reduction in heavy drinking amongwomen who married between ages 23 and 33was associated primarily with parental status:8.2% of married women without children wereheavy drinkers compared with only 1.3% ofmarried mothers.

(ii) Transition from married. Changes in levelsof heavy drinking associated with the transitionfrom marriage are also shown in Figs 1 and 2.Small declines are evident for the continuouslymarried (MC), whereas heavy drinking increasedamong individuals who divorced or remarried(MDR). The risk of heavy drinking at age 33among the divorced group (MDR) wassigni® cantly greater than that for the continu-ously married (MC), following adjustment for

drinking level at age 23 (adjusted OR 5 2.05,95% CI 5 1.49,2.83 for men; OR 5 2.61, 95%CI 5 1.67,4.09 for women). Among divorcedmen, the increase in heavy drinking was onlyevident in those who did not remarry by age 33.This is re¯ ected in a greater risk of heavy drink-ing among divorced men who remained single(MD) in comparison with remarried men (MR)(adjusted OR 5 2.13, 95% CI 5 1.22,3.71).

To establish whether increased consumptionre¯ ected the acute distress of recent break-up,we examined separations before and after age 31(Figs 3 and 4). Recently divorced men andwomen had a higher risk of heavy drinking at age33 than those who were continuously married(adjusted OR 5 4.97, 95% CI 5 2.86,8.57 andOR 5 5.25, 95% CI 5 2.60,10.65). The excessrisk was less pronounced for those who had notdivorced recently compared with thecontinuously married (adjusted OR 5 1.46,95% CI 5 1.00,2.14 for men; OR 5 2.16, 95%CI 5 1.30,3.58 for women). These analyses onthe timing of separations also showed that thesmall increases in heavy drinking for remarriedmen and women (Table 2) re¯ ected the lowerproportions of recent separations in these groups(9.8% and 7.7%, respectively, for men andwomen).

Finally, we examined the role of parentalstatus on the relationship between heavy drink-

Figure 3. Change in heavy drinking and timing of divorce (men). Recent divorces occurred after age 31; not recentdivorces occurred at age 31 or earlier.

0

% h

eavy

drin

king

23 33

2

Age

Continuously married(N =2264)

Not recent divorce(N =505)

Recent divorce(N =104)

4

6

8

10

12

Alcohol and marital status 1483

Figure 4. Change in heavy drinking and timing of divorce (women). Recent divorces occurred after age 31; not recent divorcesoccurred at age 31 or earlier.

ing and divorce. For women, the excess riskassociated with divorce relative to the continu-ously married was observed irrespective ofparental status at age 33. The adjusted oddsratio for heavy drinking at age 33 was 2.54 (95%CI 1.49,4.30) for divorced women with childrenand 1.71 (95% CI 0.74,3.99) for divorcedwomen without children. For men, levels ofheavy drinking at age 33 were signi® cantly higheramong the married who subsequently divorced,relative to the continuously married, for thosenot living with children at age 33 (adjustedOR 5 2.17, 95% CI 1.07,4.37), but to a lesserextent for men living with children (adjustedOR 5 1.51, 95% CI 0.92,2.46). However, mostdivorced men living with children (91%) hadseparated before age 31, and it was this earliertiming of divorce that appeared to account fortheir lower risk of heavy drinking.

Discussion

Levels of heavy drinking within any particularsociety are widely thought to be related to macrosocio-economic and cultural factors thatin¯ uence average alcohol consumption (Edwardset al., 1994). These macro factors include re-ligious and social sanctions, and availability andpricing of alcohol (Skog, 1980; Johnson, 1982;

Kendell et al., 1983; Fillmore, 1990). However,within a particular society other factors are asso-ciated with alcohol consumption including, forexample, employment and marital status (Tem-ple et al., 1991; Leino et al., 1995). In focusingon marital status, our study deals with an indi-vidual rather than a macro in¯ uence, althoughmacro cultural in¯ uences may well affect thesocial roles attached to marital status in differentpopulations. With respect to alcohol consump-tion, we used a de® nition of heavy drinking(more than 35 and 20 units in men and women,respectively); that is, above medically acceptedsensible limits (British Medical Association,1995) in order to identify individuals who wouldhave high mortality risk within this age group(Thun et al., 1997). Although some mis-classi® cation of heavy drinkers may haveoccurred, such biases tend, in general, to lead toan under-estimation of effects.

The strength of this study is the availability ofprospectively collected data on both drinkingand marital status at a stage in the life-coursewhen major changes in behaviour occur. Alcoholconsumption tends to decline after a peak inearly adulthood, as found in our study and alsoin other UK studies (Bennett et al., 1996;Hedges 1996) and elsewhere (Fillmore, 1987a,b;Single et al., 1995). Changes in marital status are

1484 Chris Power et al.

also evident over this period, as illustrated by thedecline shown here in the percentage of singlesubjects at the two time points: from 58% to15% of men and from 35% to 10% of women.However, a limitation of our study is that we useinformation on current status at ages 23 and 33to represent marital status stability and change,thereby ignoring marital status in the interveningperiod. The distinction between cohabitationand single status has become less clear with thesocial trend away from marriage. We combinedcohabitees with the married on the basis thatthey were both ª living as marriedº , recognizingthat this is an imperfect solution. Results for themarried held, none the less, regardless of in-clusion or exclusion of cohabitees. Increasingly,research will need to represent the more diverseand ¯ uid relationship histories that are emergingwithin the UK and elsewhere.

We found that married men and women havea low risk of heavy drinking, while the divorcedhave a high risk, which accords with previousstudies (Rosengren et al., 1989; Miller-Tutzaueret al., 1991; Temple et al., 1991; Ebrahim et al.,1995; Joung et al., 1995). The contribution ofselection to this relationship was minimal,whereas the evidence was consistent with causalprocesses associated with marital status change.Over the 10 years between ages 23 and 33, heavydrinking increased in men and women who di-vorced. This increase was particularly striking forrecent divorce. More than a third of recentlydivorced men reported drinking more than 35units in the week preceding interview; and one in10 recently divorced women had an intakegreater than 20 units. Men and women marryingfor the ® rst time showed a dramatic decline inheavy drinking over the same period (adjustedOR 5 0.41 for men, 0.61 for women, comparedwith the continuously single).

Although numerous studies examine maritalstatus in relation to alcohol consumption, typi-cally they do not identify recent separations. Onelongitudinal study with a relatively short follow-up of 3.5 years reported a substantial risk ofincident problem drinking associated with div-orce over that period (Chilcoat & Breslau,1996). Studies with longer follow-ups may fail todetect such changes if increases in alcohol con-sumption dissipate with time (Romelsjo et al.,1991). Our ® ndings on timing of separation sug-gest that the heavy consumption for divorcedgroups in cross-sectional studies and increased

consumption associated with divorce in longi-tudinal studies (Temple et al., 1991) re¯ ect apronounced short-term response in the subgroupof recent divorces. It has not been recognizedpreviously that this short-term response couldaccount for much of the excess consumption ofthe divorced evident in cross-sectional studies.

This study also shows the well-known elevatedalcohol consumption levels of single men andwomen compared to the married. It was notablein longitudinal analysis that high rates of heavydrinking for the never married persisted through-out early adulthood, with 19.1% and 5.2% ofcontinuously single men and women, respect-ively, drinking heavily at age 33. These rateswere even greater than those for the divorced,after excluding recent separations. The heavydrinking levels of those who never marry areevident in other studies that focus, for example,on women (Wilsnack et al., 1991) and on olderage groups (Ebrahim et al., 1995). Hence, thenever married are likely to include individualswho have drunk heavily throughout adulthood,and this chronic consumption may contribute toincreased mortality in this marital status category(Ebrahim et al., 1995).

With respect to the transition from single tomarried, the reduction in heavy drinking with® rst marriage appears to be suggesting a protec-tive effect of marriage, at least among men.However, after an initial decline in heavy drink-ing associated with ® rst marriage, there was nofurther reduction for those remaining married inthis study population. Moreover, an apparentprotective effect of remarriage on alcohol con-sumption is misleading, as this was found to bedue to confounding with recency of separation.(Remarried individuals tended not to have sepa-rated recently and hence were less likely thanindividuals who remained divorced to bedrinking heavily.)

In these data, the difference between the singleand the married did not appear to be due toheavy drinkers being selectively less likely tomarry. Rather, those who were to marry subse-quently were more likely to drink heavily. This isin marked contrast to our ® ndings from parallelanalyses of psychological distress, in which thehigher symptom levels of the single appeared tobe entirely attributable to selection (Hope et al.,1999). This discrepancy in the role of selectionfor the two outcomes may relate to the life stageexamined and at which the transition to marriage

Alcohol and marital status 1485

tends to occur. At age 23, heavy drinking may bemore normative than at later ages, and wouldnot necessarily be regarded as disorder in thesame way as elevated psychological symptomsare at the same age. High drinking levels in thepopulation at age 23 could therefore precludethe detection of (or ª maskº ) selective processesin relation to remaining single. In other words,the continuing heavy drinking levels of thisgroup cannot be detected at the start of theperiod of follow-up, at age 23. However, withrespect to changes accompanying divorce, thepattern reported here of an increase in alcoholconsumption mirrored that observed for psycho-logical distress, although the former followed amore transient course. Findings for psychologi-cal distress emphasized changes for women, onaccount of their higher baseline symptoms(Hope et al., 1999), whereas changes in alcoholconsumption achieve more prominence in menbecause of their higher drinking levels. Collec-tively, the results for psychological distress andheavy alcohol consumption are consistent with acausal effect of divorce in early adulthood. Thesimilarity in the patterns for these two outcomessuggests that they are related, but with a strongin¯ uence of gender directing the socially ac-cepted outcome which is likely to manifest.

This paper has also examined whetherparental responsibilities could account forchanges in alcohol consumption associated withmarital status. We have attempted to disentanglemarital status and parenthood effects because ithas been argued that the responsibilities imposedby parenthood may affect drinking levels(Chilcoat & Breslau, 1996; Hajema & Knibbe,1998). Few other studies have considered therole of marital and parental status simul-taneously but some evidence exists showingstronger effects for marital status than for parent-hood in the transition from single to married andfrom married to divorced (Chilcoat & Breslau,1996; Horwitz et al., 1996; Bachman et al.,1997). The results shown here for the 1958cohort are consistent with this previous work insuggesting a primary in¯ uence of marital statusin respect of the transition from married to di-vorced, and also among men, for the transitionfrom single to married. Among women our studysuggests, however, that parental responsibilitymay underlie the reduction of heavy drinkingassociated with the transition from single tomarried.

Finally, two major public health implicationsof our ® ndings can be identi® ed. First, divorcemay have more relevance for public health withrespect to acute increases in consumption in theimmediate period around separation than it hasin relation to chronic heavy drinking. Adversehealth consequences of the former could includealcohol-related injury both to the individual con-cerned and to others, and increased sicknessabsence (Andreasson et al., 1988; Romelsjo,1995). This applies both to currently separatedmen and women and to a group who re-partnersoon after separation who also have heavy drink-ing levels. An important caveat to these ® ndingsis that they apply to early adulthood and theirgeneralizability to other ages cannot be assumed.The initial measure of alcohol consumptionwas obtained at age 23 when heavy drinking wasrelatively common. At later ages the same levelof consumption would be more atypical.Stronger selective effects of heavy drinking couldemerge in older groups (Ebrahim et al., 1995), asimplied by reports that alcohol abuse contributesto many divorces (Santamaria, 1972; Bloom et

al., 1985). Second, the study showed particularlyadverse trends in consumption for young adultswho remained single, and it is likely that thisgroup will have the greatest risk of alcohol-related mortality and morbidity.

Acknowledgements

The authors are grateful for the ® nancial supportfrom the Department of Health for England andWales (Chris Power and Steven Hope) and theNational Health and Medical Research Councilof Australia (Bryan Rodgers through Unit GrantNo. 973302). Chris Power is also supported bythe Canadian Institute for Advanced Research asa Weston Fellow. For the data we thank theCentre for Longitudinal Studies, the Institute ofEducation, National Child Development Studycomposite ® le, including selected Perinatal Dataand sweeps one to ® ve (computer ® le). NationalBirthday Trust Fund, National Children’sBureau, the Centre for Longitudinal Studies(original data producers), Colchester, Essex,UK: The Data Archive (distributor), 21 June1994. SN: 3148.

ReferencesAGRESTI, A. (1996) An Introduction to Categorical Data

Analysis (New York, John Wiley and Sons).

1486 Chris Power et al.

ANDREASSON, S., ALLEBECK, P. & ROMELSJO, A. (1988)Alcohol and mortality among young men: longitudi-nal study of Swedish conscripts, British MedicalJournal, 296, 1021± 1025.

BACHMAN, J. G., WADSWORTH, K. N., O’MALLEY, P.M., SCHULENBERG, J. & JOHNSTON, L. D. (1997)Marriage, divorce, and parenthood during the tran-sition to young adulthood: impacts on drug use andabuse, in: SCHULENBERG, J., MAGGS, J. L. & HURREL-

MANN, K. (Eds) Health Risks and DevelopmentalTransitions During Adolescence, pp. 246± 279(Cambridge, Cambridge University Press).

BENNETT, N., JARVIS, L., ROWLANDS, O., SINGLETON,N. & HASELDEN, L. (1996) Living in Britain: resultsfrom the 1994 General Household Survey, pp. 113± 135(London, HMSO).

BLOOM, B. L., ASHER, S. J. & WHITE, S. W. (1978)Marital disruption as a stressor: a review andanalysis, Psychological Bulletin, 85, 867± 894.

BLOOM, B. L., NILES, R. L. & TATCHER, A. M. (1985)Sources of marital dissatisfaction among newly sepa-rated persons, Journal of Family Issues, 6, 359± 373.

BRITISH MEDICAL ASSOCIATION (BMA) (1995) Alcohol:guidelines on sensible drinking (London, BMA).

CHILCOAT, H. D. & BRESLAU, N. (1996) Alcohol disor-ders in young adulthood: effects of transitions intoadult roles, Journal of Health and Social Behavior, 37,339± 349.

DOHERTY, W. J., SU, S. & NEEDLE, R. (1989) Maritaldisruption and psychological well-being, Journal ofFamily Issues, 10, 72± 85.

EBRAHIM, S., WANNAMETHEE, G., MCCALLUM, A.,WALKER, M. & SHAPER, A. G. (1995) Marital status,change in marital status, and mortality in middle-aged British men, American Journal of Epidemiology,142, 834± 842.

EDWARDS, G., ANDERSON, P., BABOR, T. F. et al.(1994) Alcohol Policy and the Public Good (Oxford,Oxford University Press).

FERRI, E. (Ed.) (1993) Life at 33: the ® fth follow-up ofthe National Child Development Study (London,National Children’ s Bureau and City University).

FILLMORE, K. M. (1987a) Prevalence, incidence andchronicity of drinking patterns and problems amongmen as a function of age: a longitudinal and cohortanalysis, British Journal of Addiction, 82, 77± 83.

FILLMORE, K. M. (1987b) Women’ s drinking acrossthe adult life course as compared to men’s, BritishJournal of Addiction, 82, 801± 811.

FILLMORE, K. M. (1990) Critical explanationsÐbiological, psychological, and socialÐ of drinkingpatterns and problems from the alcohol-relatedlongitudinal literature, Research Advances in Alcoholand Drug Problems, 10, 15± 38.

GODDARD, E. (1991) Drinking in England and Wales inthe late 1980s: an enquiry carried out by the SocialSurvey Division of OPCS on behalf of the Department ofHealth in association with the Home Of® ce (London,HMSO).

HAJEMA, K.-J. & KNIBBE, R. A. (1998) Changes insocial roles as predictors of changes in drinkingbehaviour, Addiction, 93, 1717± 1727.

HEDGES, B. (1996) Alcohol consumption, in: COL-

HOUN H. & PRESCOTT-CLARKE P. (Eds) Health Sur-

vey for England 1994, pp. 337± 368 (London,HMSO).

HOPE, S., POWER, C. & RODGERS, B. (1998) The rela-tionship between parental separation in childhoodand problem drinking in adulthood, Addiction, 93,505± 514.

HOPE, S., RODGERS, B. & POWER, C. (1999) Maritalstatus transitions and psychological distress: longi-tudinal evidence from a national population sample.Psychological Medicine, 29, 381± 389.

HORWITZ, A. V., WHITE, H. R. & HOWELL-WHITE, S.(1996) The use of multiple outcomes in stressresearch: a case study of gender differences in re-sponses to marital dissolution, Journal of Health andSocial Behavior, 37, 278± 291.

JOHNSON, P. B. (1982) Sex differences, women’ s rolesand alcohol use: preliminary national data, Journal ofSocial Issues, 38, 93± 116.

JOUNG, I. M. A., STRONKS, K., VAN DE MHEEN, H. &MACKENBACH, J. P. (1995) Health behaviours ex-plain part of the differences in self-reported healthassociated with partner/marital status in The Nether-lands, Journal of Epidemiology and Community Health,49, 482± 488.

KENDELL, R. E., ROUMANIE, M. D. & RITSON, E. B.(1983) Effect of economic changes on Scottishdrinking habits 1978± 82, British Journal of Addiction,78, 365± 379.

LEINO, E. V., AGER, C. R., FILLMORE, K. M. & JOHN-

STONE, B. M. (1995) A meta-analysis of multiplelongitudinal studies from the Collaborative Alcohol-Related Longitudinal project, American Journal onAddictions , 4, 141± 149.

LUOTO, R., POIKOLAINEN, K. & UUTELA, A. (1998) Un-employment, sociodemographic background andconsumption of alcohol before and during the econ-omic recession of the 1990s in Finland, InternationalJournal of Epidemiology, 27, 623± 629.

MILLER-TUTZAUER, C., LEONARD, K. E. & WINDLE, M.(1991) Marriage and alcohol use: a longitudinalstudy of ª maturing outº , Journal of Studies onAlcohol, 52, 434± 440.

POWER, C. & ESTAUGH, V. (1990) The role of familyformation and dissolution in shaping drinking behav-iour in early adulthood, British Journal of Addiction,85, 521± 530.

POWER, C. (1992) A review of child health in the 1958birth cohort: national child development study,Paediatric and Perinatal Epidemiology, 6, 81± 110.

POWER, C., MATTHEWS, S. & MANOR, O. (1996) In-equalities in self-rated health in the 1958 birth co-hort: life time social circumstances or socialmobility? British Medical Journal, 313, 449± 453.

ROMELSJO, A., LAZARUS, N. B., KAPLAN, G. A. & CO-

HEN, R. D. (1991) The relationship between stressfullife situations and changes in alcohol consumption ina general population sample, British Journal ofAddiction, 86, 157± 169.

ROMELSJO, A. (1995) Alcohol consumption and unin-tentional injury, suicide, violence, work perform-ance, and inter-generational effects, in: HOLDER,H.D. & EDWARDS, G. (Eds) Alcohol and Public Policy:evidence and issues, pp. 114± 142. (Oxford, OxfordMedical Publications).

Alcohol and marital status 1487

ROSENGREN, A., WEDEL, H. and WILHELMSEN, L.(1989) Marital status and mortality in middle-agedSwedish men, American Journal of Epidemiology, 129,54± 64.

Santamaria, J. N. (1972) The social implications ofalcoholism, Medical Journal of Australia, 2, 523± 528.

SINGLE, E. W., BREWSTER, J. M., MACNEIL, P.,HATCHER, J. & TRAINER, C. (1995) The 1993General Social Survey I: alcohol use in Canada,Canadian Journal of Public Health, 86, 397± 401.

SKOG, O.-J. (1980) Social interaction and the distri-bution of alcohol consumption, Journal of DrugIssues, 10, 71± 92.

TEMPLE, M. T., FILLMORE, K. M., HARTKE, E., JOHN-

STONE, B., LEINO, E. V. & MOTOYOSHI, M. (1991) Ameta-analysis of change in marital and employmentstatus as predictors of alcohol consumption on atypical occasion, British Journal of Addiction, 86,1269± 1281.

THOMAS, M., GODDARD, E., HICKMAN, M. & HUNTER,P. (1994) Drinking, in: General Household Survey1992, pp. 79± 93 (London, HMSO).

THUN, M. J., PETO, R., LOPEZ, A. D. et al. (1997)Alcohol consumption and mortality among middle-aged and elderly U.S. adults, New England Journal ofMedicine, 337, 1705± 1714.

WILSNACK, R. W., WILSNACK, S. C. & KLASSEN, A. D.(1984) Women’ s drinking and drinking problems:patterns from a 1981 national survey, American Jour-nal of Public Health, 74,1231± 1238.

WILSNACK, S. C., KLASSEN, A. D., SCHUR, B. E.&WILSNACK, R. W. (1991) Predicting onset andchronicity of women’ s problem drinking: a ® ve-yearlongitudinal analysis, American Journal of PublicHealth, 81, 305± 318.

WILSON, P. (1980) Drinking in England and Wales(London, HMSO).