DEMOGRAPHV@ Volume 15, Number 4 November 1978

POPULATION MOVEMENT AND CITY-SUBURBREDISTRIBUTION: AN ANALYTIC FRAMEWORK

William H. FreyCenter for Demography and Ecology, University of Wisconsin, Madison, Wisconsin 53706

Abstract-This paper introduces an analytic framework that can be used to assessthe relationships between individual movement differentials and place char­acteristics, on the one hand, and aggregate mobility levels and city-suburbpopulation change (in size or composition), on the other. Application of thisframework using census data for individual metropolitan areas allows theanalyst to decompose population changes due to net migration into contrib­uting mobility streams and their component rates which are subject to uniquecommunity and individual influences. The paper provides both theoreticaland empirical rationale for the framework, illustrates its use with 1970censusdata, and discusses its implications for empirical research on city-suburbpopulation redistribution.

INTRODUCTION

It is often difficult for demographers toexplain or project population redistribu­tion patterns in a specific context on thebasis of existing literature. The large bodyof empirical work on geographic mobilitythat exists ranges in scope from a descrip­tion of mobility differentials by individualcharacteristics, e.g., age, to a documenta­tion of net migration rates for geographicareas (U.S. Commission on PopulationGrowth and the American Future, 1972;Shaw, 1975; Ritchey, 1976). For the ana­lyst wading through this literature, itwould be useful to have an analytic frame­work to aid in piecing together the variousdifferentials, streams, and aggregate ratesin a manner that provides an explanationof the redistribution process under investi­gation.

A greater economy of effort might ac­crue from an orientation to movement re­search that stresses the integration ofanalysis levels along the lines proposed byLee (1966). Such an orientation would befurthered by the development of analyticframeworks that specify relationshipsamong those differentials, streams, and

explanatory factors that pertain to redis­tribution analyses in particular contexts.Each framework could then serve to ori­ent future data collection and empiricalresearch related to specific redistributionproblems.

We introduce in this paper an analyticframework which can be used to explaincity-suburb redistribution in a metropol­itan area, through place determinants andindividual differentials that are associatedwith contributing movement streams.Much of what is presently known aboutmovement which leads to population re­distribution in cities and suburbs has beenreported in a few empirical studies whichare based on the parallel subject reports ofthe 1960 and 1970 censuses on Mobilityfor Metropolitan Areas (U.S. Bureau ofthe Census, 1963, 1973) (see Taeuber andTaeuber, 1964;Tarver, 1969; Farley, 1970,1976; Olsen and Guest, 1977). While thisresearch has contributed significantly toexisting knowledge on the topic, it is pri­marily descriptive in nature. Like mostempirical work on migration, it fails toestablish linkages between determinantsof mobility at different levels of analysis

571

, .

572 DEMOGRAPHY, volume 15, number 4, November 1978

necessary to make inferences about theaggregate redistribution process. Theframework presented here, while using thesame data sources, will permit analyses ofmovement and redistribution patternswhich go beyond the descriptive ap­proach.

In the remainder of this paper, then, wepresent the parameters of the analyticframework along with its underlying ra­tionale and illustrate how these parame­ters can be related to the aggregate redis­tribution process in analyses of city­suburb population size and compositionalchange.

FRAMEWORK PARAMETERS

The analytic framework consists of anumber of parameters which are associ­ated with the following migration or mo­bility streams:

I. intrametropolitan-city-to-suburb orsuburb-to-city-mobility streams,

Chart l.-City-Suburb Movement Streams and As­sociated Parameters

I. Intrametropolitan City-to-suburb (or Suburb­to-city) Mobilitya. Mobility incidence rate ofresidents: ic (or i.)

The rate at which city (or suburb) residentsmove anywhere within the SMSA during aninterval

b. Destination propensity rate of movers: Pc~.(or P.~c)The rate at which city-origin (or suburb-ori­gin) movers relocate in a suburb (or city)destination during an interval

II. In-migration to City (or Suburbs) From Outsidethe SMSAa. Migration into the SMSA: Mo

Number of migrants into the SMSA duringan interval

b. Destination propensity rate of in-migrants:Po~c (or Po~.)The rate at which SMSA In-migrants relo­cate to a city (or suburb) destination duringan interval

III. Out-migration From the City (or Suburbs)a. Out-migration incidence rate of residents:

mc~o (or m.~o)The rate at which city (or suburb) residentsmigrate out of the SMSA during an interval

II. in-migration streams to city or sub­urbs from outside the SMSA, and

III. out-migration streams from cities orsuburbs to places outside the SMSA.

These streams represent all avenueswhereby the city or suburb population isaffected by mobility. Each represents ananalytically distinct type with respect toits geographic scope and the class of placeattributes which affect its size. A clear dif­ferentiation can be made between intra­metropolitan mobility streams (I), on theone hand, and in- and out-migrationstreams (II and III), on the other. Theformer streams can be characterized aslocal, residential mobility which takeplace within the single labor market areaof the SMSA. The latter are termed mi­gration streams because moves into andout of the SMSA are generally associatedwith a change in labor market areas.

Although a distinction is made betweenthree types of mobility /migrationstreams, the analytic framework presentedhere is based on more than three parame­ters. These parameters, with one ex­ception (discussed below), represent rateswhich are applied to various populationsat risk and are presented in Chart 1. Pre­vious research on individual mobility de­cision making and on aggregate move­ment patterns suggests that the threestreams cannot necessarily be viewed assingle-stage events. Rather, each can beanalyzed in terms of one or more stageswith different causal factors operating ateach stage. These considerations underliethe specification of multiple parametersassociated with streams I and II.

Stream I Parameters

Both theoretical and empirical work onthe individual residential mobility processindicates that mobility from an origin to adestination involves at least two major de­cisions: the resident's decision to move;and the mover's choice of a destination(Rossi, 1955; Butler, 1969; Brown andMoore, 1970; Speare et aI., 1975). Thisview of the individual mobility processimplies an analog for aggregate movement

, .

572 DEMOGRAPHY, volume 15, number 4, November 1978

necessary to make inferences about theaggregate redistribution process. Theframework presented here, while using thesame data sources, will permit analyses ofmovement and redistribution patternswhich go beyond the descriptive ap­proach.

In the remainder of this paper, then, wepresent the parameters of the analyticframework along with its underlying ra­tionale and illustrate how these parame­ters can be related to the aggregate redis­tribution process in analyses of city­suburb population size and compositionalchange.

FRAMEWORK PARAMETERS

The analytic framework consists of anumber of parameters which are associ­ated with the following migration or mo­bility streams:

I. intrametropolitan-city-to-suburb orsuburb-to-city-mobility streams,

Chart l.-City-Suburb Movement Streams and As­sociated Parameters

I. Intrametropolitan City-to-suburb (or Suburb­to-city) Mobilitya. Mobility incidence rate ofresidents: ic (or is)

The rate at which city (or suburb) residentsmove anywhere within the SMSA during aninterval

b. Destination propensity rate of movers: Pc~s(or Ps~c)The rate at which city-origin (or suburb-ori­gin) movers relocate in a suburb (or city)destination during an interval

II. In-migration to City (or Suburbs) From Outsidethe SMSAa. Migration into the SMSA: Mo

Number of migrants into the SMSA duringan interval

b. Destination propensity rate of in-migrants:Po~c (or Po~s)The rate at which SMSA In-migrants relo­cate to a city (or suburb) destination duringan interval

III. Out-migration From the City (or Suburbs)a. Out-migration incidence rate of residents:

mc~o (or ms~o)The rate at which city (or suburb) residentsmigrate out of the SMSA during an interval

II. in-migration streams to city or sub­urbs from outside the SMSA, and

III. out-migration streams from cities orsuburbs to places outside the SMSA.

These streams represent all avenueswhereby the city or suburb population isaffected by mobility. Each represents ananalytically distinct type with respect toits geographic scope and the class of placeattributes which affect its size. A clear dif­ferentiation can be made between intra­metropolitan mobility streams (I), on theone hand, and in- and out-migrationstreams (II and III), on the other. Theformer streams can be characterized aslocal, residential mobility which takeplace within the single labor market areaof the SMSA. The latter are termed mi­gration streams because moves into andout of the SMSA are generally associatedwith a change in labor market areas.

Although a distinction is made betweenthree types of mobility /migrationstreams, the analytic framework presentedhere is based on more than three parame­ters. These parameters, with one ex­ception (discussed below), represent rateswhich are applied to various populationsat risk and are presented in Chart 1. Pre­vious research on individual mobility de­cision making and on aggregate move­ment patterns suggests that the threestreams cannot necessarily be viewed assingle-stage events. Rather, each can beanalyzed in terms of one or more stageswith different causal factors operating ateach stage. These considerations underliethe specification of multiple parametersassociated with streams I and II.

Stream I Parameters

Both theoretical and empirical work onthe individual residential mobility processindicates that mobility from an origin to adestination involves at least two major de­cisions: the resident's decision to move;and the mover's choice of a destination(Rossi, 1955; Butler, 1969; Brown andMoore, 1970; Speare et aI., 1975). Thisview of the individual mobility processimplies an analog for aggregate movement

, .

Population Movement and City-Suburb Redistribution 573

which can be applied to the intra­metropolitan movement streams in I. Thefirst stage of the aggregate process as­sumes that, during a specified interval, aproportion of city residents and suburbresidents will move residentially (withinthe SMSA). In the second stage, a propor­tion of city movers will relocate in thesuburbs, and, correspondingly, a propor­tion of suburb movers will relocate in the

city. These latter groups of movers repre­sent the intra metropolitan mobilitystreams associated with the framework.

Accordingly, it is possible to designatestream rates of city-to-suburb mobility orsuburb-to-city mobility, that can be de­fined as the product of two componentrates, as follows:

city-to-suburb mobility rate = iePe~•• (1)

suburb-to-city mobility rate = i.P.~e' (2)

where ie, i., Pe~.' and p.~e are defined inChart 1. The first component rate of eachmobility rate represents the proportion ofan area's total population which movesresidentially during a given interval. In thepresent framework, "mobility incidencerates" will be applied to the resident citypopulation and the resident suburb popu­lation (parameters ie and i., respectively)at the beginning of an interval.

The second component of each mobilityrate indicates the proportion of an area'smovers that relocate in the opposite partof an SMSA during an interval. This ratehas been termed a "destination propensityrate" and can be used more generally torefer to the proportion of an at-risk popu­lation of movers which relocate in a speci­fied destination. The present frameworkidentifies the suburb propensity rate ofcity movers as Pe~. and the city propensityrates of suburb movers as P.~e'

The utility of representing each streammobility rate as the product of a mobilityincidence rate and a destination propen­sity rate rests on the assumption that dif­ferent sets of origin and/or destinationcausal factors operate on each com­ponent. A supporting illustration is pre-

sented in Figure I based on 1965-1970movement in 59 of the nation's 65 largestSMSAs. (Six SMSAs were omitted fromthe analysis due to large military popu­lations or substantial changes in centralcity boundaries.)

Panel A shows the component rates ofcity-to-suburb stream mobility: the mobil­ity incidence of city residents and the sub­urban propensity of city-origin moversplotted for 59 SMSAs in ascending orderaccording to the stream rate. It is clearfrom this graph that variation acrossSMSAs is dissimilar for each componentrate, suggesting that each rate responds todifferent sets of metropolitan-specific in­fluences or differentially to the same influ­ences. (An analogous plot for suburb-to­city stream mobility rates and associatedcomponent rates appears in panel B.)Moreover, analyses (not shown) establishthat the suburb destination propensityrate for city-origin movers uniquely ac­counts for 72 percent of intermetropolitanvariation in the city-to-suburb mobilityrate, and that the city destination propen­sity rate for suburb-origin movers ac­counts for 64 percent of intermetropolitanvariation in the suburb-to-city mobilityrate.

Stream II Parameters

If one turns now to the in-migrationstreams, two framework parameters areidentified for each stream: (a) the numberof SMSA in-migrants, and (b) the city (orsuburb) destination propensity rate of in­migrants:

number of SMSA in­

migrants to the city

and

number of SMSA in- - M 4migrants to the suburbs - oPo~., ()

where Mo, Po~e, and Po~. are defined inChart 1, and Po~. = (1.0 - Po~e)' Thisseparation of parameters is suggested byfindings which indicate that migrants areinitially attracted to labor market attrib­utes of the area in which the destination is

, .

U\...•.•'"

Panel A Panel B

Source: U.S. Bureau of the Census, 1973 (see Appendix).

Figure l.-SMSA-specific Values for Mobility Incidence Rates of Residents and Destination Propensity Rates of Movers, Leading to City-to­suburb Mobility Rates and Suburb-to-city Mobility Rates Based on 1965-1970 Movement, 59 SMSAs

SMSAs RANKED ACCORDING TO SUBURB-TO-CITY MOBILITY RATES

~:IC3rr~.'"Zo<CD

3rr~

Cm~oQ:lIJ):.."::r::

.-<<!!..c3CD

,,,I

(I,

~ '\,I ,I \

, I ~City Propensity Rate of II {'Suburb-origin \ I I \ IMovers (p )" 1\ , \ ~ •

s-+c ~ I I I" ... I'" tv 1I

I 1\, ·\1" I 1/.•.•' ~ \'/\1

~>H\'""'"-hoo "' ,000,' h." ••" ".'

.00

.60

.20

.40

.50

.10

'" . 30OJ

S

.20

.00

.~O

.40

.50

.60

'"OJ

~ .30

,I,

.0

...•.•

GO

Population Movement and City-Suburb Redistribution 575

= ms->o (6)

= mc->o (5)

located (Saben, 1964; Lansing and Muel­ler, 1967). The community of destinationwithin the labor market area (SMSA) thenbecomes a secondary consideration for in­migrants and will be chosen on the basisof a different set of factors.

For the in-migration streams only, theframework will identify numbers of in­migrants associated with each streamrather than rates of stream participationwhich can be applied to a resident popu­lation. Unlike the other streams whichcontribute to city-suburb redistribution inan SMSA, the appropriate at-risk residentpopulation for an SMSA in-migrationrate would necessitate including the totalpopulation that resides outside the bound­aries of the metropolitan area.

Stream III Parameters

Only one framework parameter is usedto specify the out-migration rates frommetropolitan cities or suburbs, since, forthe purpose of measuring intrametropol­itan population change, it is not necessaryto identify the destinations of thesestreams. Therefore,

out-migrationincidence rate

from the city

and

out-migrationincidence ratefrom the suburbs

where mc->o and ms->o are defined in Chart1.

To put all the framework parametersinto perspective, the redistribution ofmovers and in-migrants across a metro­politan area's city and suburbs can be seenas an allocation of three "pools" ofmovers-movers among city residents,those among suburb residents, and in-mi­grants to the SMSA. The magnitudes ofthe first two pools are determined by themobility incidence rates ic and i•• while thesize of the third pool is represented by theMo parameter. The movers in each poolare then allocated to city and suburb des-

tinations in accordance with destination

propensity rates: Pc->s for city-originmovers, Ps->c for suburb-origin movers,and Po->c, Po->s for SMSA in-migrants.Out-movement from cities and suburbsoccurs, first, through out-migration fromthe SMSA as determined by parametersmc->o and ms->o and, second, through theintrametropolitan allocation process de­scribed above.

Framework Parameter Differentials byIndividual Characteristics

It is possible to document individualdifferentials for other framework parame­ters in addition to the more convention­

ally reported mobility incidence differ­entials. Differentials in destination

propensity rates, for example, reflect se­lectivity among mover subgroups in bothpreferences and means to relocate in a cityversus a suburb destination. To illustrate,framework parameter differentials by cat­egories of race and years of school com­pleted are presented in Table 1. The pa­rameter values shown represent means forthe 59 SMSAs.

Differentials associated with the racial

categories demonstrate the point that dif­ferent patterns of selectivity can occuracross parameters. Although blacks tendto display higher mobility incidence ratesthan nonblacks, the suburban propensityof black city movers to relocate in thesuburbs is less than a quarter of that fornonblack city movers (.081 to .375). Citypropensity rates for both black suburbanmovers and black SMSA in-migrants arestrikingly higher than the correspondingrates for nonblacks. These destination

propensity differentials indicate a high de­gree of racial selectivity in the suburban­ward relocation of movers.

Different selectivity patterns are alsoevident by categories of education. Thereexists a generally positive relationship be­tween years of schooling, on the one hand,and levels of mobility incidence andSMSA in-migration, on the other. How­ever, as with race, destination propensityrates associated with years of schooling

, .

Table I.-Mean Values for Framework Parameters by Categories of Race and Years of School Completed Based on 1965-1970 Movement, 59SMSAs

UI...•0-

Suburb-to-CityMobility

bPo-+(!

Race and Years of

School Completed

City-to-Suburb

Mobility

ia Pa+sis Ps-+(!

In-Migrationto SMSA

Mao

Out-Migrationc

m ma-+o s-+o

a - Expressed in hundreds.b - Refers to proportion of SMSA in-migrants relocating in a city destination; mean value for

proportion relocating in a suburb destination equals 1.0 minus the value shown.

c - Because the desired rates of out-migration from city and suburb (ma~ and ms~) cannotbe obtained from the published census volumes, data refer to the rate of out-migrationfrom the SMSA as a whole.

.l,

1970 population aged 5+

Black

Nonblack

1970 population aged 25+

ElementaryLess than 8 years8 years

High school1-3 years

4 yearsCollege1-3 years4+ years

.413

.445

.410

.376

.318

.315

.376

.402

.424

.432

.315

.081

.375

.322

.210

.266

.286

.365

.392

.420

.301

.422

.298

.274

.240

.228

.270

.286

.304

.310

.156

.471

.146

.156

.175

.158

.163

.147

.169

.171

1709

1251584

891

5955

122274

150231

.370

.747

.340

.355

.455

.385

.388

.328

.338

.352

.140

.092

.148

.120

.064

.076

.089

.118

.172

.246

Cm~8:Ill:.'1lI:z:~~<oC3It

yo::IC3g-It..~".Zo<It3cr'It..

Source: U.S. Bureau of the Census, 1973 (see Appendix) •

..•-0...•CD

Population Movement and City-Suburb Redistribution 577

and

Po~. = suburb destination propensityrate of SMSA in-migrants be­tween t and t + n.

n,Pc~. = suburb destination propensity

rate of city-origin movers be­tween t and t + n,

P.~c = city destination propensity rateof suburb-origin movers be­tween t and t + n,

Mo = number of in-migrants to theSMSA between t and t + n whowere alive at time t,

Po~c = city destination propensity rateof SMSA in-migrants between tand t + n,

where

pc.t+n = city population aged n and overat time t + n,

p ••t+n = suburb population aged nandover at time t + n,

P/ = city population at time t,P/ = suburb population at time t,

s = survival rate specific to eachmover, migrant, or non moverpopulation,

mc~o = out-migration incidence rate ofcity residents between t and t +n,

m.~o = out-migration incidence rate ofsuburb residents between t and t

+ n,ic = mobility incidence rate of city

residents between t and t + n,i. = mobility incidence rate of sub­

urb residents between t and t +

exhibit contrasting patterns in selectiv­ity-positive for city movers, negative forin-migrants, and mixed for suburbmovers. It is apparent from these datathat the stream selectivity in race and so­cioeconomic status, which has been re­ported in previous research, may have re­sulted from selectivity in destinationpropensity, selectivity in mobility in­cidence, or both.

THE RELATIONSHIP BETWEENFRAMEWORK PARAMETERS AND CITY­

SUBURB REDISTRIBUTION

The framework parameters can be seenas basic elements in an analysis of mobil­ity-induced change in city-suburb distri­bution of population. As has been dis­cussed, each parameter is associated withan analytically distinct mobility /migra­tion stream and is subject to unique com­munity and individual influences. By em­ploying readily obtainable migrationtabulations, it is possible to examine therelationships between framework parame­ter values and the aggregate redistributionof population in cities and suburbs thatresults from the mobility/migrationstreams with which they are associated.These relationships, as they pertain tochanges in population size and composi­tion for cities and suburbs, are elaboratedbelow along with supporting illustrations.

Relating Parameter Values to Changes inthe Population Sizes of Cities and Suburbs

Given the sizes of an SMSA's city andsuburb populations at the beginning of amigration interval, one can reconstructstream movement levels and "net" popu­lation changes that are associated with aparticular set of framework parameters.The following equations specify theserelationships for the migration interval (t,t + n):

p •• t+n = sP.t - sP/m.~o

- s(P/ - P/m.~o)i.p.~c

+ s(Pct - P/mc~o)icpc~.

+ sMoPo~., (8)

Pct+n = sP/ - sP/mc~o

- s(Pct - Pctmc~o)icpc~.

+ s(P/ - P/m.~o)i.p.~c

+ sMoPo~c,

As they appear above, these equationsleave out the fertility component, since thepredicted populations at the end of themigration interval, pc.t+n and p ••t+n, do

(7) not include individuals who were born

, .

578 DEMOGRAPHY, volume 15, number 4, November 1978

during the interval. It is possible to specifythe total populations at the end of theinterval assuming that one is able to esti­mate the number of births occurring tovarious nonmover, mover stream, and mi­grant stream populations for the interval(t, t + n) along with appropriate survivalrates. Hence,

(9)

and

where

p/+n = city population at time t + n,p/+n = suburb population at time t +

n,Be = number of births occurring to

city non movers, suburb-to-citymovers, and SMSA in-mi­grants to the city between t andt + n, and

Bs = number of births occurring tosuburb non movers, city-to­suburb movers, and SMSA in­migrants to the suburbs be­tween t and t + n.

One should also note that equations (7)and (8) assume that the movementstreams occur in a specific sequence. Resi­dential movement within the SMSA dur­ing an interval is conditional on not mi­grating out of the SMSA during theinterval. This is predicated on the assump­tion that only one move can be registeredfor each resident during an interval andthat, from the resident's perspective, aresidential move is not considered a sub­stitute for a migratory move. Hence, thetotal city and suburb populations at time t

(P/ and P/) constitute the respective at­risk populations for out-migration param­eters me~o and ms~o, while the at-risk pop­ulations for ie and is are specified as (P/ ­Petme~o) and (P/ - P/ms~o), respectively.[Some analysts may not wish to make theassumption that residential movementduring the interval is conditional on notmigrating out of the SMSA. If this is not

assumed, equations (7) and (8) can be re­written, respectively, as

Pe.t+n = sP/ - sP/me~o - sP/iePe~s

+ sP/isPs~e + sMoPo~e'

and

Each term on the right-hand side ofequations (7) and (8) represents the (t, t +n) survivors of a resident, mover stream,or migrant stream population. The firstterm in each equation identifies the resi­dent city (or suburb) population at thebeginning of the migration interval. Sub­tracted from this is the out-migrantstream (in the second term) and thestream of movers to the opposite SMSApart (in the third term). The last two termsin each equation represent the in-moverstream from the opposite part of theSMSA and in-migrant stream from out­side the SMSA, respectively. Simple rela­tionships between the framework pa­rameters and aggregate populationredistribution can be observed. When ac­tual values are retained for all framework

parameters, the city and suburb popu­lation totals for the end of the interval

(Pe.t+n, Ps.t+n) will be computed, whereasthe substitution of zero values for all in­

cidence and in-migration parameters willyield only the first terms (sPet and sP/) asthe end-of-interval totals.

Illustration. In order to demonstratehow movement streams can be relatedthrough their framework parameters tochanges in the population size of citiesand suburbs, the above relationships willbe employed using appropriate data forthe Cleveland, Ohio, Atlanta, Georgia,and San Jose, California SMSAs. Thedata are taken from the 1970 census sub­

ject report, Mobility for MetropolitanAreas (U.S. Bureau of the Census, 1973)and pertain to the 1965-1970 migrationinterval. (Details regarding use of thisdata source in the estimation of frame-

, .

Population Movement and City-Suburb Redistribution 579

where, in the present application, t = 1965and n = 5.

Appropriate framework parameter andpopulation values for Cleveland, Atlanta,and San Jose appear in Table 2. An exam­ination of the parameters reveals two sig­nificant differences across SMSAs. First, acontrast can be made between the destina­tion propensity rates for San Jose, on theone hand, and Cleveland and Atlanta, on

work parameters and population valuesappear in the Appendix.)

This illustration focuses on the con­tributions to a redistribution of thepopulation of the four intrametropolitanand in-migration streams: city-to-suburbmovement, suburb-to-city movement,SMSA in-migration to the city, andSMSA in-migration to the suburbs. Thetwo SMSA out-migration streams wereexcluded from consideration for two rea­sons. The first is a practical one in thatdata from the published census report donot allow calculation of the mc~o andms~o, parameters which are necessary toexamine these two streams. Second, thelatter portion of this illustration evaluatesthe impact of destination propensity rateson the reallocation into cities and suburbsof movers in the three "mover pools."Only the four above-mentioned streamsare involved in this reallocation process.Because of this restricted focus, it is con­venient to rewrite demographic account­ing equations (7) and (8) as equations (11)and (12), respectively:

Pc.l+n = s(P/ - P/mc~o)

- s(Pcl - P/mc~o)icpc~s

+ s(P/ - P/ms~o)isps~c

+ sMoPo~c,

and

PS.I+n = s(P/ - P/ms~o)

- s(P/ - P/ms~o)isps~c

+ s(Pcl - P/mc~o)icpc~s

+ sMoPo~s,

(11)

(12)

the other. Destination propensity rates as­sociated with the former SMSA providefor a more balanced city-suburb alloca­tion of "mover pools" than those associ­ated with the latter areas. Specifically, SanJose's Pc~s rate represents the lowest forthe three metropolitan areas, while its Ps~cand Po~c rates are substantially higherthan those in either Cleveland or Atlanta.A second disparity among SMSAs occurswith the in-migration parameter (Mo) anddistinguishes Atlanta and San Jose fromCleveland, since the former metropolitanareas experience substantially higher lev­els of in-migration relative to their resi­dent populations. These contrasts shouldbe taken into account when reviewing theaggregate redistribution patterns.

The redistribution consequences of thefour movement streams can be ascer­tained from the measures of change pre­sented in the upper portion of Table 3(Rows Al through B4). These measuresrepresent the percent of change in the cityor suburb population that can be attrib­uted to each movement stream. They arebased on Pc.1970 and Ps•1970 values com­puted from equations (11) and (12) whichassume the presence or absence of desig­nated movement streams (discussed infootnote a to Table 3). The denominatorfor each change measure indicates the1970 city (or suburb) population thatwould have resulted if none of the fourmovement streams had occurred. Hence,the change measures for all cities and sub­urbs in Row Al are equal to O.

The measures of change in row Bl in­dicate that, in all three SMSAs, the city­to-suburb stream contributes sub­stantially to the growth of population inthe suburbs and loss of population in thecities. Its effect is smallest in San Jose,reflecting the lower Pc~s in that metropol­itan area. In contrast, the redistributionattributed to the suburb-to-city stream(row B2) is less significant, particularly inthe Cleveland and Atlanta SMSAs. This isa function of the generally lower is andPs~c rates in each metropolitan area.However, since the latter rate is highest in

, .

Table 2.-Framework Parameter and Population Values Utilized in the Analysis of the Redistribution of SMSA City and Suburb PopulationsDuring 1965-1970:Cleveland, Ohio, Atlanta, Georgia, and San Jose, California

UI011o

Framework Parameter and Population ValuesaSMSAs

ii'C

Pc-+8

i*8 P8-+0

8Mo Po-+cP0+8 8(p1965 _ p1965m )c C c-+O

8(p1965_ p1965m )8 8 8-+0

Source: U.S. Bureau of the Census, 1973 (see Appendix) •

a - The notation here is consistent with that required as input for text equations (11) and (12). The mobility

incidence rates are denoted as i~ and i~, since they include slight adjustments for the survival differential

between intrametropolitan movers and nonmovers (see Appendix).

San Jose, California .458

,II

Cleveland, Ohio

Atlanta, Georgia

.451

.493

.374

.369

.322

.285

.383

.347

.105

.124

.258

160,408

238,050

234,639

.278

.264

.418

.722

.736

.582

664,363

397,166

282,722

950,094

525,057

375,873

om~oCi):llII)0-.":J:~-<<oC3CD

~~c3IT~.~Zo<CD

3IT~oC).....•

011

,I,

Table 3.-Percent Change in 1970 City and Suburb Populations Aged 5 and Over in the Cleveland, Ohio, Atlanta, Georgia, and San Jose, CaliforniaSMSAs That Can Be Attributed to Actual 1965-1970 Mobility and In-migration Streams and to Hypothetical Streams Associated with Value

Changes for Parameters Pc~., P.~c, Po~c, and Po~.

Cleveland ~ OhioAtlanta. GeorgiaSan Jose. CaliforniaActual and Hypothetical

Percent Change in PopulationaPercent Change in PopulationPercent Change in PopulationMovementStreams, 1965-1970 City

Suburbs CitySuburbs CitySuburbs

Al. No intrametropolitan mobilityor in-migration

0.00.0 0.00.0 0.00.0Impact of actual movementstreams

b

BI. City-to-suburb mobility only-16.9+11.8 -18.2+13.8 -14.7+11.1

B2. Suburb-to-city mobility only+ 4.3- 3.0 + 6.3- 4.7 +11.7- 8.9

B3. In-migration to SMSAonly+ 6.7+12.2 +15.8+33.4 +34.7+36.4

B4. All mobility and migration- 5.9+21.0 + 3.9+42.5 "31.9+38.6

Impact of hypothetical movementstreamsassuming changes in destination pro-pensity valuescCL City-to-suburb mobility only

-20.3+14.2 -21. 8+16.5 -17.7+13.3C2. Suburb-to-city mobility only

+ 3.4- 2.4 + 5.0- 3.8 + 9.5- 7.2C3. In-migration to SMSAonly

+ 5.4+13.1 +12.7+35.8 +27.7+41.6C4. All mobility and migration

-11.5+24.9 - 4.1+48.5 +19.5+47.7

[_1970 (Based on assumption\ pl970(BaSed on assumption of no)]a - Percent change in city population is computed as ~* indicated in row ) - a* mobility or in-migration x 100

r-970 (BaSed on assumption of no)a. mobility or in-migration

[ 1970 (BaSed on assumPtiOn) -]970 (BaSed on as'sumption of no)]Percent change in suburb popula,tion is computed as p. i di d i - T. bi1i f i i i x 100,B n cate n row s mo ty 0 n-m grat on

p1970(BaSed on assumption of no)s* mobility or in-m1gration

1Ihere p;~70 and ~~70 are 1970city and suburb populations aged 5 and over as computed from equations (11) and (12) andbased on actual or

assumedvalues for parameters i a' Pc+s' is' Ps-+c' Mo' ~-+c' and P0+8. Whenno intrametropolitan mobility or migration takes place, each para­meter value above is assumed to equal O.

b - Bl assumes that i c and Pc-+8 ,.. actual values and that i 8~ Ps-+c.ll Mo~ Po-+cand Po-+s- O. B2 assumes that is and PS-+C - actual values andthat ic~ Pc....s.ll Mo~ Po...cJand P0.+8 - O. B3 assumes that Mo~ Po+c~ and P0-+8 - actual values and that ic~ Pc-+8~is and Ps-+c- o.

B4 assumes actual values for all mobility and migration parameters.

c - Cl through C4 assume the same mobility incidence and in-migration values as Bl through B4 but differ in destination propensity valuesassociated with the various streams. in each case reflecting an increased propensity out of the city, or decreased propensity intothe city. Hence: Cl assumes the same parameter values as Bl. except that Pc+s - 1.20 :x actual value. C2 assumes the same parameter

values as B2, except that P8+C - .80 x actual value. C3 assumes the same parameter values 8S B3,except that Po-+c- .80 x actual value.

C4 assumes actual parameter values for ic~ i8~ Mo~ Mo+O' and m/t+o and hypothetical propensity values in Cl, C2, and C3.

Source: Table 2.

."o."Cii'..o·::I

~o<•3•aell::III:L.

n~·encg-c..g-:Ill•II:L.

;;.....iTc..O·::I

VIClIt-

582 DEMOGRAPHY, volume 15, number 4, November 1978

San Jose, the suburb-to-city stream has agreater impact on redistribution there­almost countering the effect of the city-to­suburb stream.

The redistribution effects of in-migra­tion streams (row B3) are functions of theMo as well as Po->c, po->s parameters. Be­cause Atlanta and San Jose experiencehigh levels of in-migration (Mo), the im­pact of in-migration streams on both cityand suburb change is greater for theseSMSAs than for Cleveland. Yet the allo­cation of in-migrants to cities and suburbs(Po->c, Po->s) differs across SMSAs as well.Unlike Atlanta or Cleveland, a substantialportion of San Jose in-migrants locate in acity destination, thus resulting in almostequal percentage increases for the city andthe suburb populations in that SMSA.

The measures in row B4 represent thecumulative impact of all four movementstreams on city and suburb populationchange. The net negative change observedfor Cleveland's city population can be at­tributed to the large city-to-suburb move­ment stream which is countered by rela­tively little in-movement from the suburbsand in-migration from outside the SMSA.The net suburban gain can be accountedfor by both city-to-suburb and in-migra­tion streams. The pattern for Atlanta issimilar except that larger in-migrationstreams contribute to a small populationgain for the city and add substantially toincreases in the suburban populationwhich would result from the intra­metropolitan exchange alone.

The pattern changes somewhat for SanJose. Here, neither intra metropolitanstream dominates the other, and large netgains for the city and suburb populationsresult, for the most part, from in-migra­tion. In terms of framework parameters,the unique San Jose pattern can be attrib­uted to more balanced destination pro­pensity rates among mover groups,coupled with a high level of SMSA in­migration.

The above exercise illustrates how pop­ulation redistribution between cities andsuburbs can be evaluated as a product of

parameter values associated with theanalytic framework. Equations (11) and(12) can also be employed to assess cityand suburb population changes thatwould result from hypothetical or ex­pected changes in one or more of the pa­rameter values. These values could be se­lected arbitrarily, drawn from theexperience of another SMSA, or derivedfrom cross-sectional multivariate (regres­sion) analyses which estimate frameworkparameters as functions of metropolitan­specific determinants. In order to illus­trate this use of the framework, a re­analysis of redistribution of city andsuburb populations in three metropolitanareas will be performed.

In this reanalysis, all framework param­eters with the exception of the destinationpropensity rates will retain their actualvalues as shown in Table 2. The latter, it isassumed, will be arbitrarily changed toreflect the increased propensity for moversfrom all origins (city, suburbs, and outsidethe SMSA) to select a suburban destina­tion. In each metropolitan area, the sub­urb propensity rate of city-origin movers(Pc->s) will be increased by 20 percent,while the city propensity rates of subur­ban-origin movers (Ps->c) and in-migrants(Po->c) will be decreased by 20 percent. Al­though the magnitudes of these assumedchanges are purposefully exaggerated,their directions are consistent with whatmight be expected if, for example, sub­stantial increases in the construction ofsuburban dwelling units had taken place.

Under these hypothetical conditions,rows Cl through C4 represent parallelcomparisons to rows BI through B4. Acomparison of the simulated net effects(row C4) with those associated with theactual streams (row B4) reveals a sub­stantially greater redistribution of moversand migrants into the suburbs. The moststriking change occurs for San Jose, wherethe city population undergoes a hypothet­ical increase of only 19.5 percent as op­posed to 31.9 percent under actual condi­tions.

The preceding exercises indicate how

, .

Population Movement and City-Suburb Redistribution 583

the framework can be employed to mea­sure the redistribution consequences asso­ciated (a) with actual parameter values ofan SMSA, and (b) changes in those valuesthat may be determined by the analyst.They also point up the significance of thedestination propensity rate as a distinctparameter in the redistribution process. Inthe analysis of actual parameter values(rows BI through B4), the unique destina­tion propensity rates in San Jose effected asubstantially different redistribution ofcity and suburb populations in thatSMSA than occurred in Atlanta, whichmaintained similar values on other frame­work parameters. In the second exercise,hypothetical changes in destination pro­pensity rates resulted in extremely differ­ent city-suburb population redistributionpatterns than those which had occurred inthe first exercise.

Relating Parameter Values to Changesin the Population Compositions ofCities and Suburbs

Just as migration and mobility streamscontribute to changes in the populationsize of cities and suburbs, they contributeas well to change in the population com­position of each area. This compositionalchange can be attributed to the distinctmovement patterns of various racial, eth­nic, and socioeconomic subgroups in themetropolitan area which, in the aggregate,reflect the total redistribution process.Given appropriate data, the analyticframework can be applied separately topopulation subgroups disaggregated withrespect to such characteristics. In thismanner, one can determine stream contri­butions to changes in city-suburb popu­lation compositions which accompany ag­gregate changes in city-suburb populationsizes.

Illustration. The procedure for this dis­aggregated redistribution analysis is iden­tical to that used in the analysis of thetotal population (Tables 2 and 3), exceptthat, in this case, the procedure is appliedseparately to each population subgroup.To illustrate, a disaggregated analysis is

undertaken with data from the Cleveland,Ohio SMSA decomposed into six "yearsof school completed" subgroups. The sub­group-specific parameter and populationvalues for this analysis are displayed in theupper portion of Table 4. As a basis ofcomparison, analogous values for the ag­gregated population (age 25 and over) arealso presented in the table.

An examination of these values revealstwo fairly consistent patterns among themobility incidence and in-migration com­ponents (ie, i., Mo), on the one hand, andthe destination propensity components(Pe~.' P.~e, Po~e, Po~.), on the other. Theformer parameters-those which deter­mine the sizes of SMSA "mover pools"­display a generally positive relationshipwith years of school completed, indicatinga greater level of mobility and migrationamong individuals in the more highly edu­cated subgroups. Among the destinationpropensity rates, an even more consistentpositive association is indicated betweenyears of schooling completed and a subur­ban destination, among movers from allthree origins.

These parameter differentials imply thata substantial degree of subgroup selectiv­ity should take place among streams. Sub­group selectivity in city-to-suburb move­ment results from subgroup differentialsin the mobility incidence parameter (ie),which are further compounded by sub­group differentials in the suburb propen­sity parameter (Pe~.)' Because the streammobility rate is a product of the mobilityincidence and destination propensityrates, the proportion of city college gradu­ates moving to the suburbs equals .545multiplied by .696, or .379. I'n contrast,the proportion for city residents with lessthan eight years of schooling moving tothe suburbs equals .330 multiplied by.268, or .088. Similar reinforcing differen­tials lead to positive educational selectiv­ity for SMSA in-migration to the suburbs.

Changes in the educational back­grounds of the populations of cities andsuburbs resulting from all four streamscan be ascertained from subgroup-specific

:, .

UICD~

Table 4.-Percent Change in1970 City and Suburb Populations Aged 25 and Over That Can Be Attributed to Actual 1965-1970 Mobility and In­migration Streams and to Hypothetical Mobility and In-migration Streams, by Categories of Years of School Completed: SMSA of Cleveland, Ohio

Framework Parameters and Population ValuesYears of SchoolCompleted

1:a* Pc-+s i8* Ps~8M

o Po"", po.••• 8(~965 - p~965 me..,,) 8 (~965 _ p;965 m8..,,)

i:r~::~l;~~~~eaCity Suburbs

Assuming Actual Fl'Cl1I1ewopk Pcnwnetel' Values

Source: U.S. Bureau of the Census, 1973 (see Appendix).

a - The percent of change in city and suburbpopulations for specific categories are computedas described in Table 3.

b - ::~~:~e:~~tcategory-specific values for PC'"?s' PS-+C' Po-+c' and Po-+-s are equal to the values of the total population aged 25+ for these

Assuming Equal Destination Propensity Values Across Categoriesb

.414 .412 .272 .104 83,873 .251

Cm~8;lD:.­"V%~-<<oC3••

Y':IC3a-~~Zo<••3a-~

+17.3+26.9

+29.7+19.1

+21.8+16.0

+20.1

+22.6+36.5

+15.9+18.4

+15.5+14.2

+20.1

-12.1- 8.2

-10.5-10.0

- 7.9

- 7.9

-13.8-13.4

- 6.1-12.5

+ 0.5+25.3

- 2.2- 5.2

66,84674,321

46,01453,510

66,84674,321

579,710

46,01453,510

114,813224,206

114,813224,206

579,710

78,45455,997

24,72518,568

78,45455,997

112,543120,320

24,72518,568

410,607

410,607

112,543120,320

.749

.749

.749

.749

.749

.749

.749

.848

.873

.749

.424

.593

.576

.757

.080 11,663 .152

.055 24,666 .127

.202 5,437 .576

.135 4,651 .407

.104 83,873 .251

.148 11,359 .424

.090 26,097 .243

.330 .412 .228 .104 5,437 .251

.343 .412 .208 .104 4,651 .251

.419 .412 .245 .104 11,359 .251

.460 .412 .274 .104 26,097 .251

.490 .412 .313 .104 11,663 .251

.545 .412 .345 .104 24,666 .251

.490 .566 .313

.545 .696 .345

.419 .337 .245

.460 .487 .274

.330 .268 .228

.343 .330 .208

.414 .412 .272

Total populationaged 25 and over

ElementaryLess than 8 years8 years

High school1-3 years

4 yearsCollege1-3 years4+ years

Total populationaged 25 and over

ElementaryLess than 8 years8 years

High school1-3 years

4 yearsCollege

1-3 years4+ years

,I,

..•-0'ICD

Population Movement and City-Suburb Redistribution 585

measures of change entered in the finaltwo columns of Table 4. These subgroupmeasures are computed in the same man­ner as are those for the total population(Table 3) and are based on equations (II)and (12) utilizing subgroup-specific popu­lation parameters and values. The overallpattern of compositional change-in­creasing loss for the more highly educatedsubgroups in the city and increasing gainsfor their counterparts in the suburbs-ishardly surprising in light of the high de­gree of stream selectivity outlined above.

The parameter values in the upper por­tion of Table 4 suggest that differentials indestination propensity rates contribute inlarge measure to stream selectivity andchange in the population composition ofcities and suburbs. These differentials areassociated with movers in each of thethree "pools" and may be linked witheducation-specific preferences and/or theability to pay for a suburban location. Ifone uses equations (II) and (12) and ap­propriate parameter values, it is possibleto determine the aggregate impact of thesedifferentials on the educational composi­tions of city and suburb populations. Thisis accomplished by creating a hypotheticalredistribution in which uniform levels ofdestination propensity are assumed acrosssubgroups. The difference between thechange in the educational composition re­sulting from this hypothetical redistribu­tion and the actual change in the educa­tional composition of a city or suburbpopulation can then be attributed to sub­group differentials in destination propen­sity.

The framework parameter and popu­lation values used to generate the hypo­thetical redistribution are presented in thelower portion of Table 4. These values areidentical to those in the upper portion ofthe table for all parameters except the des­tination propensity parameters PC~8' P8~C'

Po~c, and PO~8' Among the latter, eachsubgroup takes on the value of the total(aggregated) population. As in the analy­sis of actual change in population compo­sition of cities and suburbs, subgroup-spe-

cific measures of population change havebeen computed and appear in the lowerportion of Table 4.

The hypothetical measures of changereveal a more moderate pattern of educa­tional selectivity. The most striking differ­ence between the two redistribution proc­esses occurs among city college graduates.In the absence of destination propensitydifferentials, this subgroup experiences apopulation increase of 25.3 percent as op­posed to the 13.4 percent loss observedunder actual conditions. Among subur­ban subgroups, there no longer exists apositive association between populationincrease and the educational status of thesubgroup. In sum, these data indicate thatthe actual pattern of selective city loss andsuburb gain among educational sub­groups in Cleveland can, in large part, beaccounted for by differentials in destina­tion propensity rates.

Oisaggregated redistribution analysessuch as those just illustrated can be per­formed on the basis of age, race, or anyother individual-level characteristic. Asthe latter exercise demonstrates, these­analyses can be particularly useful for ex­amining compositional changes thatmight be associated with a widening ornarrowing of subgroup movement differ­entials.

rMPLICA TrONS

Although the changing sizes and com­positions of central city populationswithin metropolitan areas has increasinglybecome a focus among urban specialists,existing demographic studies have fallenshort of providing a basis from which thesources for these changes can be identi­fied. The present framework can be usedto confront the most problematic demo­graphic component of these changes-netmigration-by examining separately thedeterminants of each contributing stream.Because the framework parameters asso­ciated with each stream are subject tounique sets of influences, they can be em­ployed to establish relationships betweenindividual mobility differentials and place

\ .

586 DEMOGRAPHY, volume 15, number 4, November 1978

characteristics, on the one hand, and ag­gregate movement levels and city-suburbpopulation change, on the other. [Appli­cation of this framework to substantive

issues appears in Frey (1977, 1978).]One purpose in advancing this frame­

work is to provide an orientation to fur­ther empirical research on city-suburbpopulation redistribution. The consid­erations which have been discussed sug­gest that the following areas be given pri­ority in this work.

1. Community Determinants of Intra­metropolitan Streams. Although a host ofdemographic studies have been devoted tometropolitan determinants of in-migra­tion and out-migration to the SMSA perse, there has been little systematic researchof community attributes which affectintra metropolitan streams leading to city­suburb redistribution. The framework

parameters related to these streams repre­sent analytically distinct stages of the mo­bility process which can be used as "de­pendent variables" in future evaluationsof community-level mobility determi­nants.

2. Relationships Between Subgroup Differentials and Stream Selectivity. It hasbeen clearly demonstrated that stream se­lectivity and city-suburb compositionalchange can result from subgroup differen­tials in all framework parameters and notonly those in conventionally reported mo­bility incidence rates. There is a need formore empirical research directed to identi­fying subgroup differentials for eachframework parameter relevant to city­suburb compositional change and to eval­uating the consistency of these differen­tials across metropolitan areas.

3. Research on Destination PropensityRates. The illustrations presented herehave underscored the importance of thedestination propensity rates of intra­metropolitan movers and in-migrants asdistinct parameters in the analysis of city­suburb redistribution. An empirical exam­ination of these rates indicates that theyrespond to community and individual in­fluences independently of other frame-

work parameters. Unfortunately, far lesswork has been amassed on the local desti­nation locations of movers than on theincidence of movement or the levels of in­migration to an area. Greater effortshould be devoted to documenting de­terminants, differentials, and geographicvariations in mover destination propen­sity rates as a separate area of inquiry inmigration research.

APPENDIX: USE OF 1970 CENSUS DATA INTEXT ILLUSTRATIONS

Source of Data

The data employed to illustrate variousaspects of the analytic framework in thetext are taken from the 1970 census sub­

ject report, Mobility for MetropolitanAreas (U.S. Bureau of the Census, 1973).This source cross-tabulates the 1965 city,suburb, and nonmetropolitan residencestatus of 1970 SMSA residents aged fiveand over in each of the 65 SMSAs. Fur­

ther detail is provided according to age,race, and measures of socioeconomicstatus. With this source, it is possible topartition residents of individual SMSAsinto appropriate categories for estimatingvalues for seven of the nine framework

parameters. These categories are given be­low.

1965-1970 nonmobile city population1965-1970 within-city movers1965-1970 suburb-to-city movers1965-1970 SMSA in-migrants to city1965-1970 nonmobile suburb popula-

tion1965-1970 within-suburb movers

1965-1970 city-to-suburb movers1965-1970 SMSA in-migrants to sub-

urbsThe two out-migration parameters, mc-oand ms_o, cannot be estimated from thissource, since 1965 city or suburb residencelocation for 1965-1970 SMSA out-mi­

grants has not been published.It should be emphasized that only one

move can be registered for each individualover the course of the interval. Hence, the"move" registered for a multiple mover

\ .

587

SMSA in-migrants to city

SMSA in-migrants to suburbs

SMSA in-migrants to city+SMSA in-migrants to suburbs

SMSA in-migrants to city+SMSA in-migrants to suburbs

and

It should be recognized that, by comput­ing these parameters among survivors atthe end of the migration interval, theirvalues (in comparison to those computedat the beginning of the interval or at thetime of the move) incorporate the differ­ential mortality experiences of non­movers, movers, and migrants over theinterval. Due to the relative youthfulnessof movers and migrants, the end of inter­val ie and is parameters slightly overstates,and the end of interval Mo slightly under­states, those values which would be mea­sured earlier in the interval. The analysesusing equations (11) and (12), in Tables 2,3, and 4, require that parameters be mea­sured at the time of the move. Hence, theincidence parameters in Tables 2 and 4 aredenoted as ie* and is* [where ie* = fie andis* = gis, such that f and g represent sur­vival rate adjustment factors, and ie and isrepresent the incidence parameters inequations (11) and (12)]. In like manner,SMSA in-migration is denoted as sMo inTables 2 and 4 to be consistent with equa­tions (11) and (12). Since there is no apriori reason to believe that those locatingin a city experience significantly differentmortality levels than those locating in asuburb, the destination propensity param­eters (Pe~s' Ps~e' Po~e, and Po~s) are as­sumed to be constant across the interval.

Finally, the following two terms arenecessary to perform calculations withequations (11) and (12): s(Pet - P/me~o)= 1965-1970 nonmobile city population+ 1965-1970 within-city movers + 1965­1970 city-to-suburb movers, and s(P/ ­P/ms~o) = 1965-1970 nonmobile suburbpopulation + 1965-1970 within-suburb

city-to-suburb movers

suburb-to-city movers

within-city movers+city-to-suburb movers

within-suburb movers

+suburb-to-city movers

nonmobile city population+within-city movers+city-to-suburb movers

nonmobile suburb population+within-suburb movers

+suburb-to-city movers

within-city movers+city-to-suburb movers

within-suburb movers

+suburb-to-city movers

ie =

M = SMSA in-migrants to cityo +SMSA in-migrants to suburbs,

Population Movement and City-Suburb Redistribution

over the interval is defined according tothe origin of the first move (reported 1965residence) and the destination of the lastmove. No "move" is registered for a resi-dent who may have moved during the in­terval but reported living in the samedwelling unit in 1965 and 1970.

Also, each individual's 1965 place ofresidence is based on recall, and a signifi­cant number of residents misreport or donot report their previous residence. In theMobility for Metropolitan Areas subjectreport, the latter individuals are includedin a category denoted as "residenceabroad or not reported." These individ­uals did not enter into the calculation of

parameters for the present analysis.

Computation of Framework Parameter andPopulation Values

The framework parameters were com­puted for the 1965-1970 period from theabove-mentioned census categories as fol­lows:

\ .

588 DEMOGRAPHY, volume 15, number 4, November 1978

movers + 1965-1970 suburb-to-citymovers.

ACKNOWLEDG MENTS

This research for this paper has beensupported by grant 1 ROL HD-10666-01,"Migration and Redistribution: SMSADeterminants," from the Center for Pop­ulation Research of the National Instituteof Child Health and Human Developmentand, in part, by the Institute for Researchon Poverty, University of Wisconsin. Theauthor is indebted to Alden Speare, Jr.,Basil Zimmer, Avery M. Guest, SamuelPreston, Carl Fredrickson, Karl Taeuber,and Franklin Wilson for providing com­ments and suggestions on this frameworkduring various stages of its development.

REFERENCES

Brown, L. A., and E. G. Moore. 1970. The Intra­Urban Migration Process: A Perspective. Geogra­fiska Annaler 52B:I-13.

Butler, Edgar W., F. S. Chapin, Jr., G. C. Hemmens,E. J. Kaiser, M. A. Stegman, and S. F. Weiss.1969. Moving Behavior and Residential Choice­A National Survey. National Cooperative High­way Research Program Report No. 81. Washing­ton, D.C.: Highway Research Board, NationalAcademy of Sciences.

Farley, R. 1970. The Changing Distribution of Ne­groes Within Metropolitan Areas: The Emergenceof Black Suburbs. American Journal of Sociology75:512-529.

--. 1976. Components of Suburban PopulationGrowth. Pp. 3-38 in Barry Schwartz (ed.), TheChanging Face of the Suburbs. Chicago: The Uni­versity of Chicago Press.

Frey, William H. 1977. White Flight and CentralCity Loss: Application of an Analytic MigrationFramework. Presented at the 1977 Annual Meet­ings of the American Statistical Association, Chi­cago. Center for Demography and EcologyWorking Paper 77-27. Madison: Center for De­mography and Ecology, University of Wisconsin.

---. 1978. "Black Movement to the Suburbs:Potentials and Prospects for Metropolitan-wideIntegration." Forthcoming in Frank D. Bean andW. Parker Frisbie (eds.), The Demography of Ra­cial and Ethnic Groups. New York: AcademicPress.

Lansing, John B., and E. Mueller. 1967. The Geo­graphic Mobility of Labor. Ann Arbor: Institutefor Social Research.

Lee, E. 1966. A Theory of Migration. Demography3:47-57.

Olsen, R. A., and A. M. Guest. 1977. Migration andCity-Suburb Status Differences. Urban AffairsQuarterly 12:523-532.

Ritchey, P. N. 1976. Explanations of Migration. An­nual Review of Sociology 2:96-137.

Rossi, Peter H. 1955. Why Families Move. NewYork: The Free Press.

Saben, Samuel. 1964. Geographic Mobility andEmployment Status, March 1962-March 1963.Monthly Labor Review 87:883-891.

Shaw, R. Paul. 1975. Migration Theory and Fact.Philadelphia: Regional Science Research Insti­tute.

Speare, Alden Jr., S. Goldstein, and W. H. Frey.1975. Residential Mobility, Migration and Metro­politan Change. Cambridge, Mass.: BallingerPublishing Company.

Taeuber, K. E., and A. F. Taeuber. 1964. WhiteMigration and Socioeconomic Diffel<:nces Be­tween Cities and Suburbs. American SociologicalReview 29:718-729.

Tarver, J. D. 1969. Migration Differentials in South­ern Cities and Suburbs. Social Science Quarterly50:298-324.

U.S. Bureau of the Census. 1963. Census of Popu­lation, 1960. Subject Reports, Final ReportPC(2)-2C. Mobility for Metropolitan Areas.Washington, D.C.: U.S. Government Printing Of­fice.

---. 1973. Census of Population: 1970. SubjectReports, Final Report PC(2)-2C. Mobility forMetropolitan Areas. Washington, D.C.: U.S.Government Printing Office.

U.S. Commission on Population Growth and theAmerican Future. 1972. Population Distributionand Policy, edited by Sara Mills Mazie. Vol. V ofthe Commission research reports. Washington,D.C.: U.S. Government Printing Office.

, .