You are on page 1of 43

Continuities in Transnational Migration: An Analysis of Nineteen Mexican Communities Author(s): Douglas S.

Massey, Luin Goldring, Jorge Durand Source: American Journal of Sociology, Vol. 99, No. 6 (May, 1994), pp. 1492-1533 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/2782582 . Accessed: 25/10/2011 17:40
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to American Journal of Sociology.

http://www.jstor.org

Continuities in Transnational Migration: An Analysis of Nineteen Mexican Communities1


Douglas S. Massey of University Chicago Luin Goldring of University Illinois at Chicago JorgeDurand Universidadde Guadalajara

in have observedboth working Mexican communities Researchers in migraand regularities inconsistencies theway thattransnational thataccounts a presents theory tiondevelopsovertime.This article fortheseuniformities discrepancies and and proposesa methodto It acrosscommunities. also argues comparetheprocessofmigration of for and control theprevalence migration thatstudiesmustreport showthat Data from Mexicancommunities 19 within communities. changesaccompany social, and economic demographic, predictable migraprevalence.Althoughinternational increasesin migratory socioecotion beginswithina narrowrange of each community's othersocial over timeit broadensto incorporate nomic structure, groups. in working Mexicobeganto uncover investigators Duringthe 1970s,field in to empiricalregularities the way that migration the United States were froma community developed over time. The earliestemigrants heads from family almostalways males of working age, usuallymarried structure. Typicallythey some identifiable niche in the socioeconomic came from middleofthelocal hierarchy-notso poorthattheycould the that mibut not afford costs and risksof migration, not so affluent the Withinthe United States theywent to a few grationwas unattractive. sectorsof the economy,such as locationsto work in particular specific They adopted strategiesof or railroads, agriculture, manufacturing.
1 Address correspondence Douglas S. Massey,Population Center, Unito Studies all 19104-6298. of 3178LocustWalk,Philadelphia, Pennsylvania versity Pennsylvania,

? 1994byThe University Chicago.All rights reserved. of 1.50 0002-9602/94/9906-0003$0

1492

AJS Volume 99 Number6 (May 1994): 1492-1533

Continuities Migration in movement appropriate the worktheydid, settling prolonged to for periodsto workin manufacturing moving or back and forth agricultural for labor. Over time the proportion people with U.S. experience of tended to increasewithinthe community. migratory As behaviorspread, foreign in accumulated thepopulation, ties multiexperience kinship to migrants plied, and the streamdiversified includeyounger to unmarried males, less women,and children. Migration becameprogressively class selective of and more representative the whole community. Withinthe United States the arrayof locations,occupations,and economicsectorswhere migrants workedexpanded,and the timing and durationof U.S. trips reflected thesechanges. Eventually,mostmen were drawnintothe migrant workforce and a largeplurality womenand children of began migrating well. In places as withlong and well-established histories migration, of information about jobs, housing,and lifein theUnitedStatesbecame widelydiffused, and nearlyeveryone was relatedto someonewho had been north the borof der. The social, economic,and demographic composition the outflow of In reacheda highdegreeof diversity. the United States the numberof branchcommunities consisting migrant of families who appearedto have settledabroad permanently began to grow. These trends werefirst notedbyJoshuaReichert thetownof Guafor dalupe, Michoac'an(see Reichert 1979, 1981, 1982; Reichert and Massey in 1979, 1980). Richard Mines uncoveredsimilarpatterns the town of Las Animas, Zacatecas (Mines 1981, 1984; Mines and Anzaldua 1982; of Mines and de Janvry 1982). A systematic comparison thesecommuniset ties by Mines and Massey (1985) showed that risingout-migration offstructural morelikely.The changesthat made additionalmigration of has self-feeding character migration been notedin othercommunities within Mexico (Wiest 1973; Massey et al. 1987; Alarcon1992) as well as othercountries (Baucic 1972;Fergany1982; Pessar 1982; Rhoades 1979). studieshave also found discrepancies Despite these commonalities, withrespectto key variablesin the migration among communities prothe of cess, such as the proportion people with migration experience, class background migrants, proportion documented of of the migrants, of the numberof migrant the women and children, importance settled and versus recurrent the strategies, numberand typesof destinations, the kind of U.S. occupationheld. Both Mines and Massey (1985) and to differences structhesecommunity-level Massey et al. (1987) attributed at turalfactors thatshaped the courseof migration each location. As Mexican community studies accumulatedduringthe 1970s and differences became moreand moresa1980s,however,intercommunity lientand beganto overshadow continuities investhe identified earlier by 1493

American Journal Sociology of tigators. Whereassome studiesfoundthatmigrants wereprimarily landless workers(see Cornelius 1976a; Stuart and Kearney 1981), others concludedthattheywere mainlylandowners (Lopez 1986) or bothlandownersand sharecroppers (Mines 1981). Althoughmarriedmen dominatedthemigrant in workforce sometowns(Wiest1973;Cornelius1976a, 1976b; Dinerman 1982), in othersthe participation women and chilof drencame closerto thatof men(Reichert 1979; Fernaindez 1988; Cornelius 1990). Findingsconcerning relative the importance different of migrant strategies became particularly confused.In some studies,migrants appeared to favora strategy temporary of migration, movingsporadically the to UnitedStatesforshortperiodsof work(Cornelius1978). In others, they engaged in recurrent migration, movingback and forthannually for seasonal wage labor (Reichert1979; Lopez 1986). In some settings, migrantsadopted a settledstrategy, establishing in themselves one place forlong periodsof time(see thetownof Santiagoanalyzedby Massey et al. [1987]), while in others,theyemployedseveral strategies once, at without to appearing favoranyone in particular thetownofChamit(see lan in Massey et al. [1987]). In orderto resolvethe tensionbetweenearlierfindings a common of migratory processand the growingevidenceof intercommunity differences, Durand and Massey (1992) reviewedstudiesof 25 Mexican communities. Theyfoundthat"apparently inconsistent about generalizations Mexico-U.S. migration not necessarily are when theyare contradictory examinedin comparative perspective. Rather,diverseoutcomes occurin various communities when commonprocessesof migration shaped are and differentiated structural by variables operatingat the community level" (p. 4). They arguedfor"a research designthatwould incorporate the studyof manydifferent communities a commonanalytic into framework" (p. 4). Goldring(1990, 1992b) reached similarconclusionsfrom her comparative analysisof two migrant communities. This commonanalyticframework ideally should involve more than the simple applicationof a standardsurveyinstrument a range of to different communities. suchan exercise Although would addresstheissue of sample generalizability, would not providea means of analytically it communities comparing withdifferent histories and levels of migration. Giventhecumulative natureofmigration fruitful processes, comparisons mustsomehowtake intoaccountpriormigration histories. People livingin communities wheremigration just begun,forexhas face significant ample, generally deterrents international to movement. Since the numberof migrants small, few nonmigrants is have friends and relativeswho have been abroad, and even if theydo, the migrants are likely have limited to aboutjobs, housing, knowledge and transporta1494

Continuities Migration in tionat destination sites.In contrast, peoplelivingin a community characterizedby a long history are and highprevalenceof out-migration very likelyto be connectedsociallyto people who have been abroad, and thesepeople tendto have considerable knowledgeabout conditions and In resources pointsof destination. communities at witha well-developed in migratory tradition, other words,nonmigrants have access to valuable social capital thatcan be used to facilitate movement. International migration a costly is and risky enterprise, thosewho and it undertake are usuallyselectedon demographic, social, economic,and role in psychological grounds.Social capital,however,plays a powerful thesecostsand risks,and its accumulation over timetendsto mitigating reducethe selectivity migration. of Variationin the amountand quality of social capital can, therefore, streams produceverydifferent migration over timeand acrosscommunities, makingmigration patterns appear to be discrepant the when,in fact,theyreflect same underlying process. The conceptualframework called forby Durand and Massey requires a techniquethat permitsdirectcomparisonamong communities with In different histories levelsofmigration. orderto satisfy methodand this a ological need, we introduce new analyticaltool: the migration prevain lence ratio. For any community any year,the prevalenceratiois definedas the numberof people with international migratory experience divided by the total numberof people alive. It can be calculatedretrofor spectively anyyearin therecent past givenjust twopiecesofinformation about everycommunity member: date of birthand the date of the his or herfirst foreign trip. This ratio, when calculatedfordifferent years withina community, of has providesa simpleindicator how widespreadmigratory experience become at any point in time. It serves as a proxyforthe extentof a in involvement themigratory community's processand allows us to comIn at pare communities verydifferent stagesof migratory development. this way, the prevalenceratiopartially controls the effect differfor of In of ences in the history and timing migration. essence,it standardizes the unitsof comparison. In thisarticle, employ ratiosto characterize underlythe we prevalence ing processof transnational migration it develops across a range of as Our data, drawnfrom community settings. representative samplesof 19 Mexican communities, compiledusingidenticalmethods are and instrumentsas part of a single,comprehensive study.Guided by the proposition that the natureof migration shifts it becomesmorewidespread as in a community, describethe demographic, we social, economic,and character international of as communities from geographic migration go low to highprevalence.Takingaccountofmigratory prevalence provides in a usefulway of resolvingapparentinconsistencies the literature by 1495

American Journal Sociology of revealing common patterns thedevelopment international in of migration across communities. Simplyknowing prevalence migration the of does not,of course,identify whichspecific structural conditions (e.g., immigration policies,political conditions, economictrends)mighthave influenced or at migration key historical junctures.Althoughwe recognizethat structural factors shape thelocal expression migratory of in processes different communities at different times,our purposehere is not to studythe effect these of factorsbut to characterize differentiating the basic processesacross a rangeof sites. We beginwitha reviewof recent research transnational on migration thathighlights empirical the commonalities differences and observedby earlierresearchers. thenelaborateour methodological We approachand describea data set that allows us to employit. Finallywe use the data to characterize migratory the processas itdevelopsacross19 communities thatdiffer withrespect economic to ethnic structure, ruralcomposition, urban status,and emigration history.
A CUMULATIVE THEORY OF MIGRATION

Priorempiricalworksuggests thattransnational in migration unfolds a consistent relatively way over time. It displaysa distinct to tendency become moreprevalent and to broadenits base of demographic, social, and economicrepresentation within community. the These trends follow from factthatmigration the theoretically affects individualmotivations and social structures ways thatencourageadditionalmigration. a in As result, transnational migration tendsto becomea self-reinforcing process that acquires an internal momentum its own. Over timeit becomes all increasingly independent the conditions of thatoriginally caused it. This theoretical logic predictsthe emergenceof commonempirical acrossdiversecommunities migration trends as becomesmoreprevalent. The seemingly diversearrayof migration that arises fromthe patterns variouscase studiesis explainedbythedifferent levelsofmigratory prevalence thateach community achieved. has Althoughwe drew our theoretical argument fromresearch primarily on Mexico-UnitedStates migration and to a lesser duringthiscentury, extent fromrecentresearch migration on from Caribbeanand Latin the America,we put it forward a generalconceptualmodel. It is meant as to apply to cases of transnational labor migration where host-country immigration policiesare relatively thosecases where open, particularly clandestine is migration feasible. Transnational labor migration for may originate a variety compleof reasons.Migrantsmay observewage differentials mentary betweenori1496

Continuities Migration in gin and destination areas and respondto expectedpositivereturns to foreign labor (Todaro and Maruszko 1987). Householdsmay seek to dirisksto theireconomicwell-being sendingfamily versify by members to workin different regionallabor markets, one of whichis foreign (Stark 1991). Migrants maybe recruited foreign by employers seeking import to for workers specific tasks (Piore 1979). People may be impelledto move in because structural transformations the local economy eliminate traditionalsourcesof sustenance (Sassen 1988)or because politicalupheavals cause people to fearfortheirphysical safety (Portesand Rumbaut 1990). how international from No matter migration begins,thefirst migrants a community likelyto experience as a verycostlyand risky are it enterprise,bothin monetary and psychological terms.They have littleor no in knowledgeof conditions the hostcountry and are ignorant its culof ture,language,and ways of life.In mostcases, theyincurthe expenses of the trip and absorb the opportunity while costs of income forgone movingand lookingforwork.They arrivehavingto pay offtheseoverhead expensesand are thusrelatively dependent theirfirst on employer. Giventheir lack of knowledge about prevailing wage rates,workhabits, legal conventions, social expectations, and theyare vulnerable exploito if tationand mistreatment, especially theyare undocumented do not and speak the languageof the hostcountry. usuGiven thesecostsand risks,thefirst transnational labor migrants the ally come not from bottom the socioeconomic of hierarchy from but the lower middleranges(Portes1979; Portesand Rumbaut 1990). Such to people have enoughresources absorb the costs and risksQfthe trip but are not so affluent thatforeign labor is unattractive. in followa patriarchal sexSince families low-wagecountries typically are ual divisionof labor within household, first the the migrants usually marriedmen of primelabor forceage who seek to maintaintheireconomic and genderroles throughmigration (Lindstrom1991; Pedraza 1991; Hondagneu-Sotelo1992; Alarcon 1992; Goldring1992c). Among menare seenas better ruraland working-class able thanwomen families, to absorb the physicalrisksof international movement (Reichert1979); in and given prevailing genderdifferentials wage rates(England 1992), men can be expectedto earn moreon average,than women. Thus, if a seeksto maximize one family foreign earnings sending worker by abroad, the logical choiceis the male householdhead or perhapsan olderson. This patternof male-ledmigration appears to hold well forsending areas throughout centraland westernstates of Mexico. Mexican the women in the northern crossed the states,however,have historically and industrial borderto workas domestics, serviceworkers, operatives Donato (1992)has shown (Ruiz and Tiano 1987;Taylor 1980).Moreover, thatthe gendercomposition migration shaped strongly historical of is by 1497

AmericanJournal Sociology of relationships betweennations,patterns social organization sending of in countries, and otherstructural factors. leave their Whethermale or female,however,the earliestmigrants familiesand friends behindand strike out forsolitary workin an alien land. Most transnational migrants begin as targetearners(Piore 1979), seeking earnas muchmoney possibleas quicklyas possiblein order to as to recoup theirinitialinvestment, attaina predetermined incomegoal, in and return hometo family friends. and Theyhave little interest permanentsettlement abroad. Once one or morepeoplehave comeand gonein thisfashion, however, in to the situation the sendingcommunity does notreturn the statusquo ante. Each act of migration generatesa set of irreversible changes in individualmotivations, social structures, culturalvalues that alter and the contextwithinwhich futuremigration decisionsare made. These thatmakeadditional changesaccumulateacrosstimeto createconditions morelikely.Massey (1990) has labeled thisself-generating migration process "the cumulativecausationof migration," following Myrdal (1957). and Reichert (1981) calls it the"migrant syndrome" Alarcon(1988, 1992) refers it as "northernization." to in At the individuallevel, participation a high-wage economy induces that turn people away fromtarget changes in tastes and motivations of earningand towardpersistent migration (Piore 1979). Satisfaction the led wantsthatoriginally to migration createsnew wants. Access to high wages and the goods theybuy createsnew standardsof materialwellin raise expectations being,and first-hand experience an affluent society and createnew ambitions upward mobility. migrants for As earn high and wages and altertheir consumption patterns, they adoptnew lifestyles local economicpursuitsbecome less attractive (Goldring1992a, 1992b, The first-hand experience gainedfrom migration makesthesatisfaction ofthesenew wantsincreasingly feasible.Once someonehas migrated and thatpersonhas direct of returned, knowledge employment opportunities, labor-market conditions, ways oflifein thedestination and country; they use theseunderstandings migrate to risksand coststhan again withfewer before (Massey 1986). Once it has been experienced, therefore, migration thatcan be embecomesa familiar and reliablesocioeconomic resource ployed again and again as new needs arise and motivations change (Reichert1979; Mines 1981). in Empiricalresearch Mexico showsconclusively thatonce a man has to migrated the United States,the odds are extremely highthathe will migrate again (Massey 1987b;Masseyet al. 1987).Indeed,theprobability of takingan additionaltriprises monotonically the numberof trips as increases(Massey 1985). The morea man migrates, morehe is likely the 1498
1992d).

Continuities Migration in to continue migrating, pattern a thathas provedto be remarkably persistentin theface of restrictive immigration policies(Donato, Durand, and Massey 1992). Given theirstatusas targetearners,duringthe first few tripsand in the earlyhistory migration of froma community, migrants tend to live underratherspartanconditions, sleepingin barracksor sharingapartmentswithothermen and sleepingin shifts save money.They work to longhoursand have littlesocial life.In some cases theyworktwo eightin hour shifts the same day (Durand 1992). Most of theirearningsare repatriated the formof savingsor remittances in (Massey et al. 1987). Migrants themselves members their see as of homecommunities not and as participants the hostsociety in (Piore 1979). As migrants spendincreasing timeabroad,however, thisform social of lifebecomesmoreand moreproblematic. staysabroad lengthen As and the numberof trips rises, pressurefromfamilymemberswantingto migrate grows(Hondagneu-Sotelo 1992;Alarcon1992).The first relatives to accompany married a migrant typically are unmarried sonsofworking age, since theyhave the greatest earnings potential after father the and theirmigration consistent is with prevailinggenderroles. Over time, however,unmarried working-age daughters, wives, and youngerchildrenare likelyto accompany himas well. Otherrelatives, such as nephAs ews, nieces, and cousins,eventually join experienced migrants. increasingnumbersof youngmen acquire migrant experience, theyalso as begin to travelnorthin groupsbased on friendship well as kinship (Lopez 1986). As a result,the demographic base of migration steadily widensand themean age of migration and Massey 1979; drops(Reichert Piore 1979; Massey et al. 1987). The act of migration onlyinduceschangeswithinindividualminot thatmake further it grants movement morelikely, also initiates changes in social structures spreadmigration that through community the (Mines is 1981; Massey et al. 1987). Each migrant inevitably linkedto a set of nonmigrants through variety social tiesthatcarryreciprocal a of obligationsforassistancebased on shared understandings kinship,friendof ship,and commoncommunity origin (Lomnitz1977). Giventheexpectations and practicesassociatedwith kinshipand friendship, each act of createsa set ofpeoplewithsocial tiesto thereceiving migration country. and Nonmigrants draw upon these ties to gain access to employment assistanceabroad, substantially the reducing costsand risksofmovement comparedto earliermigrants (Taylor 1986; Massey and Garcia Espania 1987). Every new migrant thusreducesthe costsand risksand increasesthe and relaattractiveness feasibility migration a set of friends of for and tives. Withtheseloweredcostsand risks,additionalpeople are induced 1499

American Journal Sociology of to migrate thefirst for time,whichfurther expandsthesetofpeoplewith ties abroad. This additionalmigration reducescostsand risksfora new set of people, causing some of themto migrate,and so on. Once the numberof networkconnections reaches a criticalthreshold, migration becomesself-perpetuating because each act of movement createsthe social structure to necessary sustainit (Massey 1990). Empiricalstudiesin Mexico clearlyshow that havingnetwork connections greatly increases the likelihood of international movement(Taylor 1986; Massey and Garcia Espafia 1987; Massey et al. 1987). As migrants make successive trips, theyaccumulate foreign experience and knowledge thatrender ties to themincreasingly valuable. As information about the destination country and its socioeconomic resources accumulatesin the population,the costs of migration steadilydrop to make the cost-benefit calculationpositiveforan increasingly large set of a people, while the risksof movement steadilyfall to rendermigration for feasiblerisk-diversification number households. of strategy a growing Over time,therefore, becomesprogressively selectiveand less migration morerepresentative the community a whole. as of Migration also changesthecultural context within whichdecisionsare made, and international movement becomesincreasingly attractive for reasonsthatare notpurely economic.Migrants evincea widely-admired lifestyle that othersare drawn to emulate.Although some of its attractiveness material-based on theability consumegoodsand purchase is to property-the lifestyle also acquires a strongnormativecomponent (Reichert 1979;Lopez 1986;Alarcon1992).In communities whereforeign into local values and expectawage labor has become fullyintegrated do tions, people contemplating entryinto the labor forceliterally not in consider otheroptions:theyexpectto migrate frequently the courseof theirlives and assume theycan go whenever theywish. As migration it assumes a greaterrole in the community, becomes increasingly important a rite of passage foryoung men, providing as an accepted means of demonstrating theirworthiness, ambition,and manhoodto others(Reichert1979; Alarcon 1992). Moreover,as women becomemoreintegrated within postindustrial society, theybeginto push that lead to formore egalitariangenderroles and encourageactivities in longerstays abroad, such as investing householdgoods and buying in property the destination country (Hondagneu-Sotelo 1992; Grasmuck and Pessar 1991; Goldring1992b, 1992c). communities Over timeand withextensive movement back and forth, of originand destination come to comprisetransnational increasingly circuits-social and geographicspaces that arise through the constant of circulation people,money, goods,and information (Rouse 1989, 1991, 1992). As thesecircuits develop, practicesand values thatonce demar1500

Continuities Migration in on cated distinct societiesbeginto have a transformative influence each other.Over time,migrant communities becomeculturally "transnationalized," incorporating ideologies,practices,expectations, and political claims fromboth societiesto create a "cultureof migration" that is distinct fromthe cultureof both the sendingand receiving nation (see Reichert1979; Rouse 1989, 1991, 1992; Georges1990; Goldring1992b, 1992c; Smith1992). As migration increasingly is takenforgranted, demographic the coma position and socioeconomic oftheplace oforigin role undergo dramatic In transformation. manyplaces,women,children, olderpeopledomand inatea reducedpopulation exceptduring fewweeksor months the when migrantsreturnfor holidays and celebrations(Reichert1979; Mines 1981). In economically marginalagricultural areas, farming and other traditional activities lose importance (Mines 1981). As theplace of origin becomesa siteof restand recreation, sharpcontrast the routine in to of work abroad, its social meaningundergoes changes(Rouse 1989, 1992; on Goldring1992a, 1992b, 1992d). Migrantsspend moneycollectively infrastructure othercommunity and projectsaimed at transforming the landscape intoa place ofleisure,a place wheremigrants theirfamiand lies can display their status and exercisepolitical claims and power and Smith1993). (Goldring1992a, 1992b, 1992d; Smith1992; Goldring The first migrants froma community typically to a specific go niche in the destination country's politicaleconomy,yieldinglittlediversity withrespect destination, to or of occupation, strategies movement. Early followthe path of the first migrants migrant because that is wherethe costsand risksof migration lowestand the chancesof successgreatare est. Once theyhave identified promising a migrant worker,moreover, labor recruiters and contractors tend to use themas vehiclesto recruit from additionalworkers theircircleof friends and relatives(Mines and in Anzaldua 1982).As experience thehostcountry accumulates, however, and as morepeople are drawnintotheprocess,somemigrants inevitably in seek out better opportunities new places and occupations.In thisway the diversity foreign of destinations, jobs, and strategies increases. As the migration processproceeds,however,typically someonefrom the sendingcommunity achievesa position responsibility enables of that himor herto channelemployment, to housing, and other resources fellow townspeople(Mines 1981; Massey et al. 1987). The positionmay be a in a crew boss in a railroad,a foreman a factory, union representative in a company,a majordomoin a restaurant, labor contractor a a for or it to grower, perhapseven a businessowner.Although is impossible whereor how it will occur,sooneror latersomeoneattainssuch predict a positionand beginsto recruit fellowtownspeople work. for At thispoint,themigration stream and beginsto focusmorenarrowly 1501

American Journal Sociology of the diversity jobs, destinations, of and strategies beginsto constrict, a process Jones (1982a, 1982b, 1984) has called "channelization."This concentration does not necessarily involvea singleforeign locationfora particular labor-exporting community(Mines 1981; Goldring 1990, 1992b),but the overallpattern earlydiversity of followedby increasing concentration one or moresitesis a generalfeature the processby in of whichmigrants in establishbranchcommunities receiving societies. As migrants make repeatedtripsand accumulatemoretime abroad, as wivesand children join themigrant workforce, morepeoplebecome as involvedin the migration process,and as stronger linksare formed with in specific employers particular locations,a growing number migrants of and familiessettlein the host society.They acquire informal to its ties inhabitants and establishformallinks with institutions such as banks, and government, schools.They learnthe host-country languageand become permanent legal residents. Empiricalstudiesshow thatthe probability settlement of risessteadily migrant as experience increases(Massey 1985, 1987b; Massey et al. 1987). As familiessettlearound specific places of employment, branchcommunities long-term permanent of and out-migrants These beginto form. communities anchorthe networks and further reducethe costsand risks of movementby providinga secure and familiarenvironment within whichnew migrants arrive,find can and employment, learn and housing the ropes in the receiving is country. Increasingly, migration channeled to thesecommunities the diversity destinations and of associatedwitha place of originis further reduced. As migrants in becomepartof established communities thehostcounto or try, theyadapt themselves thelocal setting. Whether nottheyhave legal documents, theysend theirchildren school,learn a minimum to of the host country's and social language, and use financialinstitutions services. Over time the local landscape of the receiving is community transformed (Goldring1992a, 1992b, 1992d). Whetheror not theyare immigrant entrepreneurs, migrants the contribute the creationand to growthof a marketfor specializedfoods, entertainment, cultural and products.The formation ethnicneighborhoods of a represents processof socioeconomic that adaptation and transformation permitsmany "forin eign" practicesto be maintained the new setting. If the processof migration continues long enough,networks reach a pointof numericalsaturation. Larger and largersharesof the transnational community residein the branchcommunities, more birthsoccur all are abroad, and virtually who remainin the home community connectedeitherto someonelivingabroad or to someonewith substantial reach thislevel of development, the When networks foreign experience. 1502

Continuities Migration in costsof migration stop falling witheach new entrant and the processof migration loses its dynamicmomentum growth. for As the processapproachesits limit,migratory experience becomesso diffused withinthe community thatthe stockof potential new migrants gets verysmall; increasingly is composedof youngchildrenand the it and local elderly. Labor shortages beginto occurin core sending regions coincides with wage ratesrise(Gregory 1986).If theprocessofsaturation a recessionary periodabroad, an oversupply immigrant of workers may result,leading to lower wages and makingit moredifficult experifor enced migrants fnd work forfriends to and relatives.These developmentsfurther dampen the pressures migration for and cause the rate of intothe migrant to and trailoff.The prevaentry workforce decelerate lenceof migration thestockofmigrant and experience thenapproachan upperasymptote. Observed in the aggregate, this asymptotic trendmay be difficult to are drawnintothemigradetect,because new communities continuously in tory process.As therateofout-migration decelerates communities with long-standing traditions, new communities drawnintotransnational are circuits and theirratesof out-migration beginto accelerate.As a result, thetotalflowofmigrants mayremainconstant steadily or increase.Only by studying data at the community level can we identify general the sequence of eventsthatoccurin the processof transnational migration. Thus, ourtheory positsa cumulative modeloftransnational migration. It outlineshow, once initiated, processbuilds upon a growing the base of knowledge, experience, social contacts, and otherforms social and of culturalcapital in self-reinforcing fashion.It arguesthatthe processof migration alterssendingand receiving localitiesin such a way thatfurthermigration encouraged.Subsequentmigration made to and from is is communities that are undergoing profound cultural,economic,social, and even physicalchanges. This theoretical argument elaboratesupon earlierstudies,synthesizes of the strands theory thathave appearedin diverse sources, and sharpens to underlying conceptual linkages.As constructed this point,the model in to appliesprimarily transnational migrant circuits arising nonmetropolitanlocations. The dynamics international of migration largemetropolifrom tan areas have not been well studied.This topicmerits further empirical in research and maynecessitate modifications thetheory. DATA Data forthisanalysiscomefrom simplerandomsamplesgathered during 1982-83 and 1987-91 in 19 communities located in the Mexican states 1503

American Journal Sociology of of Jalisco,Michoacan, Guanajuato, and Nayarit,areas thathave traditionallysent large numbersof migrants the United States (Gamio to 1930; Dagodag 1975; North and Houstoun 1976; Consejo Nacional de about the samples is summaPoblacion 1986; Jones1988). Information rizedin table 1. Withineach community, 100-200 households wereranthe of domlyselectedand interviewed during months Novemberthrough thatrangedfrom.029 to .699, deJanuary, yielding samplingfractions In of frame. all cases pending thenumber households thesampling on on In but one, the framecoveredthe entire community. San Franciscodel a Rincon,Guanajuato, we constructed framefora singleworking-class neighborhood sampledit instead. and These proceduresproduced a total sample size of 3,400 households across 19 communities, coveringa total populationof about 236,000 the not people. Refusalsweregenerally a problem: although ratereached in 15% in one case and 11% in another, 14 cases therefusal ratewas 5% or less, and overallthe rate was only4.8%. The higher refusalratesin fromlocal the two communities reflect generalizeddistruststemming rather than suspicionof our studyper se. politicalconditions The winter months generally besttimeto locate and interview are the returned migrants withinMexico, since most come back to spend the holidayswith theirfamilies.In the case of La Yerbabuena, however, initialfieldwork revealedthat large numbersof migrants also returned in July(because theyworkedin Florida's wintercitrusharvest),so we sent an interviewer duringthe summer complete survey.In gento the of eral, however,the Mexican community samples are representative dwellingsoccupied duringthe wintermonthsof 1982-82 and 1987-91 in 19 nonmetropolitan Mexican communities. These community data were supplemented withnonrandom samples of out-migrants located in the United States duringthe summersubsequent to each winter's survey.From the community samples,we deterwentand thensentinterviewminedwherein theUnitedStatesmigrants ersto thoseareas to survey households thathad permanently out-migrant settledabroad. Snowball samplingmethods (Goodman 1961) were used to compilethisU.S.-based sample. In mostof the communities, out20 householdswere interviewed, in Mineral de Pozos only 10 but migrant weresampled.In Tepec we wereunableto includeany U.S. households, because this sample was incorporated fromanotherstudyafterthe fact and no U.S. sample had been originally compiled(see Massey et al. 1987). Althoughthese U.S. samples are not representative all outof fromselecfor migrants, theydo providesome control biases stemming in tive emigration and settlement the UnitedStates. In choosingcommunities study,we soughtto include a range of for was not sizes and economicbases, but the prevalenceof U.S. migration 1504

0-

CI

C14

'c -

f)

En

r- Mt t- (7 r- w 0

-O

i o o o o

C)

C,

in

CN

D t t C m E-4 0 D
'I

I O0o

oo 0

O o
C1 m

O
1. in

C
ir

O
1r

Co
ON cl

CN

in

It

cr

cn

a ur
>

-|

E l~~~~~n <O Q >

in tt O

OQQ~ O-

cn

1-d if

Q -

00

r-

Qn

cn

O '

CN

O Q Q

0 H

cn 4-4

O O OQ

CO O Q

CO OCO COO

O O Q

0O O COD O O O O OO

O O t O O rQO O OO OO

a/)

CH
X

.J'
0

4)
CLQ

C\

r-

oN

-4

r-

It

0,d-

nS

O?q Seu roOato OO

0~~~~~~~~~~~~~~~~~~~~~~~~~~~.
0

~~ ~~~~ >
P: C\

a/)

C4C t~~~-0'.-~~~~~~OQ~~~OC140 QOON0 m C 00'a)0a)OO o000'C0CO00


-4-4d4 \o o -4

0a)C'
-4 -4 NXo \o o Xr --4

a)0 'lO0.-4 00 00 00 000 C


-4 -4 -4 -4

CN o

0 2)

::

Ct e" 00 00 c \o

cy

4 o

, rC

' 00 C Cc 4 c.

u) m)~')C'.-4 H

o\

R'>

0 0)

\0

-o 4 --4

cz c.-4

0-o

c.) C

r4

Q
0

'0 C0.e0
*.l

OQ

cO

1~~~~00 )
d ? ln

. oc 0 . 0
) n

-l I\ lz 74

0 q
u u

0 .?

o o

- No

'.'.

0(0 O DO

. 0) 0 .0

00

0' A

C3

0 Cl)

..

. ' tC

~~~~~~~~~~~~~

W~~~~~~~~~~~~~~~~~~~~~~~~~~C
(U U 0 )

:00c

,bi U D *0(0~~~~~~~
-

Cc5AAe o mQ~~~~~~~~c

U)

H
:z?

o04 = ;;S

American Journal Sociology of itselfa criterion inclusion.Althoughsamples were gatheredin five for large metropolitan areas withpopulationsover 100,000,thesedata are excludedfrom present data suggest the analysis.Limited thatU. S.-bound from migration Mexican urbanareas maybe rising (Cornelius1992),but we believe that the social dynamicsof migration frommajor citiesare sufficiently different fromthose of smallertownsand citiesto warrant separatestudy. The populationsof the communities understudyrangefrom1,000 to just over 50,000. The eightsmall citiesincludedin the data set have and include two industrial populationsin excess of 10,000 inhabitants cities(San Franciscodel Rinconand El Salto), severalcommercial cities serving agrarianhinterlands (Los Reyes,Ameca, San Felipe, Ixtlan,and Romita),and one coastal community engagedin fishing, tobacco growing, and sugar cane cultivation (Las Varas). All exceptthe last communityare seats of theirrespective municipios(the roughequivalentof a U.S. county). The six communities designatedas towns have populationsranging from of 2,500 (theofficial definition an urbanplace in Mexico)to 10,000; all are essentially Chavinda and Ario de Rayon agrariancommunities. are located in Michoacan's lush Zamora Valley, a regionof intense, highly capitalizedcommercial agriculture. Amacueca,locatedin southern of Jalisco,is a more traditional agrariancommunity small landholders and ejido farmers. Union de San Antonioand San Diego de Alejandria are located in the Los Altos regionof Jalisco,a dry,windsweptregion ofrain-fed and agriculture cattleranching. Finally,Nahuatzenis a Tarascan Indian community located in the highlands Michoacan. Except of forArio, all of thesecommunities municipioseats. are The fivesmallest communities ruralranchos, are political dependencies within their withpopulations under2,500 inhabitants. Three municipios of the communities (Santa Maria, La Yerbabuena,and Tepec) are small in are of settlements agrarian outlying regions; they composed poortenant farmers smalllandholders. and The smallranchoofLa Soledad is located divide just outsidethe cityof Irapuato, Guanajuato, and its inhabitants Mineralde Pozos is theirtimebetweenagrarianand industrial pursuits. a half-abandoned townwhose depositsgave out aroundthe turn mining of the century, it turning intoa poor farmtown. were interviewed Respondents livingin thesecommunities usingethnosurveymethods(Massey 1987a; Massey et al. 1987). Withineach and household gathered we basic information aboutthesocial,economic, characteristics thehead, thespouse,the head's children, of demographic which household and other household members.We also determined membershad ever been to the United States and fromthemgathered and first mostrecent and basic data about their trips:thedate, duration, 1506

Continuities Migration in as destination, well as the migrant's legal status,occupation,and wage earnedon thosetrips.Fromeach household head we collecteda detailed lifehistory thatincludeda labor history (including migration a history), a property history, maritalhistory, a fertility a and history.
THE PREVALENCE OF MIGRATION

The fundamental from earliertheoretical insight emerging our discussion is that migration unfoldsin a regular,predictableway over time. We argue that questionsabout migrants' characteristics-whether theyare predominantly male or female, youngor old, legal or illegal,richor poor, sojourners settlers-are misplacedbecause thesetraits notdescribe or do the migrant flowper se, but rather, phase in its development. a These outcomesmustbe understood qualitiesof the migration as streamthat evolve as partofa larger developmental process,notas fixed characteristicsof migrants from particular communities. therefore We comclassify munitiesaccordingto theirlevel of migratory prevalenceand examine in the characteristics migrants communities roughly same level of at the of migratory development.Our objective is to chartthe changes that occurin a community it movesfrom in as limited participation transnationalmigration a stateof mass involvement. to Trackingchangesin thisfashionallows us to overcomethe problems in inherent makinggeneralconclusions based on individualcase studies. It also providesa moretractableway of describing migratory processes A thatemerge overtimein a cumulative, nonlinear fashion. disadvantage ofthetechnique thatit tendsto dehistoricize is events migration: specific such as the Bracero Programof the 1940s or the economiccrisisof the levelsin different commuearly1980smay occurat different prevalence nities,a factthatshouldbe keptin mindwheninterpreting findings. our We definea community's stage in the migratory process based on prevalenceratios computedfor each year in each community. These ratiosare calculatedusingeveryrespondent's date of birthand the date of his or her first of U.S. trip.The denominator the ratiois the number of people 15 yearsold or olderalive in a givenyear,and the numerator is the numberof such people who have ever been to the United States we up to and includingthat year. Withineach community, computed prevalenceratiosforeach yearfrom1940to thesurvey date to createan annual estimateof the proportion adultswho have ever been to the of UnitedStates. is Since thiscomputation based on retrospective data, it assumesthat similarrates of mortality and migrantsand nonmigrants experienced in is internal out-migration the past. The mortality assumption likelyto be quite robust.Although smallnumbers survivors from of earlyperiods 1507

American Journal Sociology of in can producerandomfluctuations the ratios,thereare not likelyto be largemortality differentials based on migrant status.In orderto enhance of the stability estimates, ratios however,we do not consider prevalence foryearsbefore1940. To the extent that the communities experiencedpermanentoutwithinMexico, however,and to the degreethatthisinternal migration is for the migration a substitute international migration, ratioswill tend to overstate prevalence U.S. migration. the of This overestimation results because internalmigrants who were in the base populationin earlier yearshad leftby thetimeofthesurvey and werethusexcludedfrom the denominator, biasingthe ratioupwards. This bias tendsto be moresevere in earlierperiodsbecause the numberof permanent internaloutovertime.Since prevalence migrants accumulates ratiostendto riseover an time,however,the bias is conservative: upward bias in earlieryears acts to mitigate curve of risingprevalencethatis observedempirithe cally. Another sourceof potential bias stemsfrom permanent out-migration to the UnitedStates. To theextent thatwe have failedto includepeople who began migrating some pointin the past and thensettledin the at United States permanently, have excluded cases that contribute we to and proportionately to the denominator, strongly the numerator less ratiosdownward.Moreover, because setthereby biasingtheprevalence tlement tendsto occuramongthosewho have builtup considerable U.S. we experience, are mostlikelyto exclude people who lefton theirfirst the tripssometimeago, thereby exacerbating bias morein earlier periods thatis notconservative thanlaterones, and thusyielding pattern with a thatwe have respectto the empiricaltrendswe observe.To the extent of the capturedthe experience settlers through snowball samples comthisproblem. piled at U.S. destination sites,we have mitigated Althoughthese potentialbiases should be kept in mind, we believe for thatour conclusions robust.In orderto gauge the potential bias, are information about internal and international however,table 2 presents in columnshowstheyearof migration each of the 19 samples.The first to of thefirst U.S. tripin each community indicatetheroughbeginnings international migration, and the second columnreports prevalence the as columns of U.S. migration of the surveydate. The thirdand fourth report corresponding the data fortripswithinMexico, and the last colof umnshowsthepercentage adult respondents bornwithin commuthe of nity(an indicator the degreeof in-migration). The sample clearlyoffers wide rangeof U.S. migratory a experiences. such as La Yerbabuena,became heavilyinvolvedin Some communities, U.S. migration the earlyon and rapidlymovedtowardmass migration: in first left migrant this community 1923, and by the surveydate 60% 1508

Continuities Migration in
TABLE
PREVALENCE OF INTERNAL

2
MIGRATION

AND INTERNATIONAL

Community

Prevalence Year of of U.S. Year of in Earliest Earliest Migration U.S. SurveyYear Mexican (%) Migration Migration

% of Prevalence Population of Mexican 15 and Older Migrationin SurveyYear Born in (%) Municipio

Smallcities: El Salto................ ...... San Francisco Rinc6n del ........ Romita ...................... Los Reyes...................... San FelipeTorres Mochas ....... Ixtlan Rio ...................... del Ameca............. ......... Las Varas ...................... Towns: Chavinda ...................... Amacueca ...................... San Diegode Alejandria ......... Uni6n San Antonio de ............ Ario Ray6n...................... de ...................... Nahuatzen Ranchos: ...................... Tepec SantaMariadelValle ............. Mineralde Pozos ...................
La Yerbabuena ..................... La Soledad ...................... Average......................

1923 1920 1940 1943 1940 1941 1942 1943 1914 1920 1919 1925 1935 1940 1940 1923
1949 1923 1939 1932

21 21 17 30 30 27 31 29 34 34 43 23 39 18 18 25
9 60 28 28

1936 1918 1933 1923 1921 1928 1936 1944 1930 1927 1926 1926 1949 1936 1915 1928
1937 1928 1932 1930

14 5 8 21 21 16 18 11 9 17 23 23 11 28 16 13
23 10 6 15

62 84 78 57 78 45 76 55 79 82 78 83 76 95 85 78
87 83 95 77

15 NOTE.-Prevalence ratiosare calculatedformigrants yearsold and olderwhose first was for trip two months longer,fortripsmade in the period 1940-89. or

of the entireadult populationhad been to the UnitedStates. Similarly, in San Diego de Alejandria began sendingmigrants 1920, and by the timeof thesurvey 43% ofits adult population had acquiredU.S. experience. At theotherextreme, U.S. out-migration not beginin Mineral did de Pozos until1941,and only9% ofitsadultpopulation had migrated by thesurvey date. Acrossall communities, prevalence U.S. migration the of averaged 28% in the survey year. a sourceof bias in estimating Internalmigration constitutes potential thatitis permaprevalence ratiosforinternational migration theextent to If nent,substitutes U.S. migration, is prevalent. internal for and migrationis not permanent, are thenmigrants likelyto have theirexperience reflected thecomputed in to ratiosbecause theyreturned be interviewed. 1509

American Journal Sociology of If internaland international migration not substitutes, are thenpeople withand without S. experience equallylikely migrate U. are to internally and thuslittlebias enters computation prevalenceratios,because the of theabsenceofexperience affects numerator denominator the and equally. Even if internal out-migration permanent is and acts as a substitute for international migration, moreover, bias would be smallifit werenot the veryprevalent. Although have no way of knowingwhether we thereis muchpermanent out-migration Mexican destinations, overall prevalenceof to the internal migratory experience does notappear to be high,averaging only about 15% overall,about halftheprevalence international of migration. The ratiosrangefrom in San Franciscodel Rinconto 28% in Nahuat5% zen. In some communities, like Tepec and San Francisco del Rincon, internalmigration began in the earlyteensof this century, whereasin like Ario de Rayon and Las Varas, it began in the 1940s. There others, is littleevidence,however,thatU.S. and Mexican migration substiare tutesforeach other.If thatweretrue,we would expectan inversecorrein lation betweenthe two sets of prevalenceratiosacross communities; factthe correlation nearlyzero (.02). is In mostcases, thecommunities have experienced little in-migration as well. Overall, 77% ofthe adult respondents werebornin themunicipio, butthepercentage is variessomewhat size. The percentage exceptionby allyhighin smallranchos, whereit averages86%, butis somewhat lower in towns(78%) and cities(68%). The percentage locallybornresidents of is lowestin Ixtlan del Rio and Las Varas, bothin the stateof Nayarit, and the latterin a growing coastal area. In general,therefore, although internal migration cannotbe dismissed as a potential sourceof bias, the evidencemarshaledin table 2 suggests thatinternal thatit is unlikely be seriousin mostcases. To theextent to does bias the computation prevalenceratios,however,it of migration will tend to be conservative the observedempiricaltrendis one of if it risingprevalenceover time.Although is not practicalto show graphs in 1 of prevalenceratiosforall 19 communities the sample,figure plots trendsfor six cases forthe period 1940-89. Ratios are shown for all as community members, well as formen and womenseparately. To varying show a pattern risingprevaof degrees,the communities frominternal lence over time,which suggeststhat any bias stemming in is migration likelyto be conservative nature,workingagainst the of direction the apparenttrend.Despite the generalconsistency the of in differences the rate of change trend,however,thereare pronounced in over time.In general,trends U.S. migratory followone of prevalence threecharacteristic patterns. of La Yerbabuena and San Diego displaythe classic pattern rapidly 1510

0)

00~~~~~

O,~~~~ 0D

N0t

C~~~~~~C
ci ci ci ci ci

o*

c o

Oorcso-o

vN

o'

co

_ o e

ea

ci*

C0

ciO

a ,

American Journal Sociology of rising prevalencenotedfirst Reichert by (1979). (Indeed, La Yerbabuena is his community "Guadalupe," whichwe have resampled.)The top of curve shows prevalenceratiosformen, the bottomone is forwomen, in and the middlecurvecaptures trend prevalence thepopulation the for as a whole.In bothplaces, U. S. migration tookoff a cumulative really as processin 1942, when the United States initiated labor recruitment in Mexico underthe auspices of the Bracero Program(Galarza 1964; Samora 1971). Afterthis date, the prevalenceof migration among adult men rose veryrapidly,reaching50% in La Yerbabuena by 1950 and 80% by 1980. Women were largelyuninvolveduntil 1970, when their prevalencelevels began to rise,verysteeply La Yerbabuena, to reach in 40% by 1989. The lag betweenthe onsetof male and femalemovement is 20-30 yearsin bothcommunities. The second pattern,observed in Union de San Antonio and La Soledad, is that of a rapid risein the prevalenceof migration earlyon, followed a stagnation by and declineduringthe 1950sand 1960s,thena revivalof growth duringthe early1970s. A declinein the prevalenceof U.S. migration does not mean that new people were not entering the in migrant workforce. simply It meansthatthenumber new migrants of of anyyearwas less thanthenumber peopleturning causinga decline 15, in theaverageprevalence migratory of within population. experience the The last pattern, that of veryslow growth, expressedby Mineral de is Pozos and El Salto. In these communities, prevalenceratiosrose very slowlyand neverachievedhighlevels. in Our purposehereis not to explainintercommunity differences the timingand rate of growthin migratory prevalencebut to describethe characteristic flowsas prevalence changesthatoccurin migration moves fromlow to high. We accomplishthis task by classifying communities to each year, according their stagein theunderlying processofmigration in based on the estimated ratios.We inspected trends prevaprevalence in lence ratios for all 19 communities orderto discernwhetherrough break points could be definedto capturedistinct phases in the develof transnationalmigration,but in the end we opmental process simplycreated fourprogressive categoriesat evenlyspaced, arbitrary cutpoints. The first stage occurswhen under 10% of adult community members have been to the United States. At this stage, migration overwhelmis inglymale and may persistfor prolongedperiods withoutincreasing. The nextstage occurswhen 10%-19% of all community members have abroad. In thisphase oftheprocess,male migration acquiredexperience in continues apace but womenhave notyetbegunto migrate largenumbers. The thirdstageis reachedwhentheoverallprevalenceratiovaries between20% and 29%. Here, male migration decelerates womenbegin as 1512

Continuities Migration in to enterthe migrantflow in significant numbers,causing the overall prevalenceratioto reach a temporary plateau. The fourth stage occurs when prevalencevaries between30% and 39%. At this stage, female migration acceleratesand male movement continues, bringing level the of community participation highlevels. The fifth to stage constitutes a situation mass migration, of withoverallprevalence ratiosabove 40%as prevalence amongwomenapproaches30%, thatamongmalesreaches 80% or more. These five prevalence categoriescorrespondroughlyto successive In stagesin theprocessofmigration. theensuing section employ we them as independent variablesto studyqualitative changesin migrant streams as communities in move froman initial,tentativeparticipation U.S. migration a situation mass involvement. to of Specifically, examine we in the changesacross communities the stockof U.S. experience, demographicand socioeconomic characteristics migrants, of and the nature and destination U.S. tripsas the communities fromlow to high of go prevalence.

COMMONALITIES IN TRANSNATIONAL MIGRATION The Accumulation Social Capital of Table 3 examineschanges in U.S. experience that occur as migration becomesmoreprevalent. For each community yearfrom1940to the and presentwe computedprevalencelevels among men, women, and all thenwe calculatedthe totalstockof U.S. expericommunity members, ence and the percentage people withkinshiplinksto U.S. migrants. of We assignedeach community-yearone ofthefiveprevalence to categories and computed shown averagesby prevalence levelto derivethenumbers in thetable. Because communities weresampledat widely different rates, the averagescomputedin thisand all subsequenttables were estimated as usingthe inverseof the samplingfraction a case weight. In interpreting findings, is important remember our it to thatonlytwo communities reached the highestprevalencecategory.In contrast,18 communities achieved migration levels placing them into the second 15 themin thethird prevalence category, reachedlevelsputting category, All and eightreachedlevelsplacingthemin thefourth category. communities,of course,contribute yearsto the first prevalencecategory (since each community begins at a low prevalencelevel). This distribution across categoriessuggeststhat patterns change will be fairly of robust acrossthe first fourcategories thatthosein the fifth but should category be interpreted withsome caution,sincetheyare based on theexperience of just two communities, both agrarian. Heterogeneous patternsmay 1513

American Journalof Sociology


TABLE 3
PREVALENCE RATIO, RATE OF CHANGE IN PREVALENCE, AND CUMULATIVE

U.S. EXPERIENCE IN COMMUNITY


PREVALENCE MIGRATION COMMUNITY OF IN CHARACTERISTIC Prevalence ratio (%):* 0%-9% 10%-19% 20%-29% 30%-39% ?40%

Males ........................... All ............................


Females ...........................

14.3
.9

27.8
5.1

41.6
11.3

58.7
12.5

80.8
29.8

7.6

15.7

25.4

33.7

54.9

Change in prevalence ratio (mean absolute % change [t to t + 1]):* Males ........................... Females ........................... All ............................ Cumulative stock of U.S. experience (mean years per person):** Males ............... ............ Females ..................... ...... All ............................ Kinship links to U.S. migrants:** % with migrant parent ............... % with migrant grandparent ........ % with migrant sibling ............... % with no migrant relatives ......... Community-years (N) .................... No. of communities .......................

1.5 .3 .7

1.3 .6 .8

1.4 .6 .8

1.5 1.0 1.1

1.4 2.0 1.4

.1 .0 .1 13.8 1.8 22.6 65.6 228 19

1.1 .7 1.1 27.7 4.7 38.6 45.2 384 18

3.6 1.3 3.5 41.6 10.3 59.0 25.3 234 15

2.7 1.8 2.6 60.5 19.3 62.7 10.3 56 8

7.3 3.7 7.1 75.4 40.1 77.8 7.8 23 2

* Estimatedfrom in the sampleof all householdmembers 19 communities.

** Estimatedfrom sample of all householdheads in all 19 communities. the

the withrising reflect progressive out partially selecting of communities prevalence. The first as panel traces shiftsin the prevalenceof U.S. migration communities the pass through various phases of the migration process. from criteria the used Although totalprevalence the ratiosfollowdirectly to define fivestages,trends males and femalescapturethe interthe for play of sex-specific movements different at phases of the migration process. At theearliest few stagesofmigration, people,male or female,have been to the UnitedStates:only14% of menand under1% of women.In theensuing phases,however, migration spreadsprogressively throughout the adult male population,and by the timemass migration achieved is eightout of ten men have been abroad. The prevalenceof U.S. migration among femaleslags behindthat of males at all phases of the migration process,but the differential grows becomesmoreprevalent. low levels At progressively smalleras migration 1514

Continuities Migration in of migration, male prevalencelevels exceed thoseof femalesby a factor ofnearly16: 1. Movingthrough nexttwolevels,thedifferential the drops to 5.5: 1 and thento 3.7: 1. In thefourth category, ratioclimbsminithe mally,to 4.7: 1, but in thehighest prevalence category, when30% of all adult womenhave been to theUnitedStates,thesex differential dropsto 2. 7: 1. As Reichert (1979), Mines (1981), and others noted,transnational womenas well. migration beginsamongmenbut ultimately incorporates The nextpanel ofthetable showshow rapidly U.S. migration spreads withincommunity populationsat different phases of the migration process. We estimatethe instantaneous rate of change in prevalenceby computing averageabsolutepercentage the changein prevalence between timest - 1 and t + 1, wheret standsforthetarget year,whichis then in classifiedby prevalencecategoryfor presentation the table. These computations revealthatmale migratory behaviorspreadsat a relatively constantrate. At the lowest prevalencelevel, the absolute percentage changeformalesis 1.5%, a figure thatfluctuates only0. 1%-0.2% across prevalencecategories. In contrast, rate of changeof migration womenis muchlower the for in thefirst prevalence category, it nearlydoublesin thesecondstage, but remainsconstantin the third,climbsagain in the fourth, and doubles again in thefifth. Thus, transnational migration appearsto spreadamong menat a fairly constant rateirrespective thedegreeofmigratory of prevalencethathas been achieved,butthespreadofmigratory behavioracceleratesrapidly a amongwomenas prevalence rises,yielding steadyaccelerationin migration the community at level. The last two panels oftable 3 show how U.S. migrant experience and networkconnections accumulateas migratory behaviorbecomes more in diffused communities. Special questionsput to householdheads allow us to computethe totalamountof timepeople have spentmigrating to or the United States and whether not certainrelativeshad gone to the UnitedStatesbefore them. As thesefigures about qualitaindicate,thespreadofmigration brings tivechangesthatalterthedecision-making context actorsat different for Potential depointsin the developmental processof migration. migrants from community a low prevalence a or with cidingwhether notto migrate of migration have littleaccess to information about potential generally in jobs and opportunities the United States. At low prevalencelevels, the typicalhouseholdhead has accumulated only0.1 yearsof experience in the United States,only 14% have a parentwithmigrant expe-rience, 22% have a siblingwith U.S. experience, and just 2% have a migrant grandparent. As U.S. migration spreads withinthe community, however,kinship connections the United States proliferate to and migratory experience 1515

American Journalof Sociology accumulatesto the pointwherenonmigrants contemplating move can a draw on substantial social capitalto reducethe costsand risksof a U.S. trip.Movingthrough second,third, the and fourth prevalence levels,the average years of U.S. experience per persongrow from1.1 to 2.6 and the percentage people with a migrant of parentincreasesfrom28% to 60%. Acrossthe same categories, percentage the witha migrant sibling goes from39% to 63% and the percentage witha migrant grandparent growsfrom5% to 19%. Once a level of mass migration been achieved,potential has migrants considering triphave a vast storeof experience a and kin connections theycan use to gain access to jobs, housing,and otherresources the in UnitedStates.At thehighest prevalence levelthetypical household head has accumulatedan average of seven yearsof experience the United in States, 75% have a parentwho has been to the UnitedStates,and 78% have a siblingwho has been there.Indeed, thedepthof kin connections to the United States is such that 40% have a grandparent with U.S. migrant experience. This growth the stockof migratory of knowledge and experience and the proliferation networkconnections the United States are both of to causes and effects the spread of migratory of behaviorthroughout the community. They are effects because each new migrant adds to thestock of experience and expands the rangeof network connections. They are causes because connections experienced to a migrants constitute valuable form social capital (Coleman 1988) thatpeople who have notyetmiof gratedcan employto improvetheirodds of obtaining job and income a in the UnitedStates. IncreasingDemographicDiversity to According the theory outlinedabove, as migration we becomesmore in prevalent a community demographic its base progressively broadens. This hypothesis generally is confirmed the data presented table 4, in by whichexamines demographic the of background migrants leavingsample communities different at stages in the migration process.In this table, are migrants classified the prevalencelevel of the place theywere in by when theymade theirfirst trip. The first panel showsthattheshareoffemales risesas migration develops and expands. Although thereis a slightdrop betweenthe thirdand fourth the in prevalence categories, rateclimbssharply thefifth category. Whereas only 6.7% of U.S. migrants leaving on theirfirst tripare female in the lowest prevalencecategory, the time mass prevalence by is achieved44% of new migrants women.Thus, theshareofwomen are risesas migration moves from beinga rareto a mass phenomenon. 1516

in Continuities Migration
TABLE 4
DEMOGRAPHIC CHARACTERISTICS OF MIGRANTS ON THEIR FIRST U.S. TRIP
PREVALENCE OF MIGRATION IN COMMUNITY CHARACTERISTIC 0%-9% 10%-19% 20%-29% 30%-39% ?40%

Sex: Female (%) .............................. 6.7 Age of male migrants (years): 0-14 (%) .............................. 10.3 15-19 (%).............................. 20.7 20-34(%) .............................. 57.6 ?35 (%) .............................. 11.4 Mean age .............................. 23.7 Diversity (above groups,n = 4) ..... 57 Diversity (5-yeargroups,n = 11) ... 52 Age of femalemigrants (years): 0-14 (%) .............................. 54.4 15-19 ............%.................. 2.9 20-34 .............................. 27.9 ?35 (%) .............................. 14.8 13.7 Mean age .............................. Diversity (above groups,n = 4) ..... 20 Diversity (5-yeargroups,n = 11) ... 16 Household position: Head (%) .............................. 86.0 3.4 Spouse (%) .............................. Son (%) .............................. 7.3 3.0 Daughter(%) ............................. Other(%) .............................. .3 18 Diversity(n = 5) ........................ No. of migrants 227 (unweighted)...........

25.6 13.9 23.2 53.7 9.2 21.8 74 67 43.9 13.0 32.2 10.9 16.4 69 50 59.7 15.2 15.1 8.6 1.4 53 1,148

32.6 19.5 29.8 44.2 6.6 19.4 83 68 42.7 16.5 27.0 13.8 16.3 83 63 43.3 12.4 25.7 16.6 1.9 71 1,480

27.5 27.8 33.7 40.6 8.0 19.9 83 64 39.6 14.5 32.2 13.6 17.7 81 66 41.5 10.6 30.7 14.1 3.1 66 635

44.3 42.6 39.8 13.6 4.0 14.5 78 51 33.8 24.8 24.3 17.1 20.1 87 72 19.2 23.2 38.5 18.7 .5 78 343

The representation different groups of age likewise broadensovertime. of 35 Amongmentheproportion migrants yearsold or olderfallssteadily as communities the of proceedthrough fivecategories migratory prevalence, goingfrom11% and 9% at low prevalencelevels at 4% at mass levels.The percentage maleswho are 20-34 yearsold likewisedeclines of across stages, going from58% to 14%. In contrast, the percentageof males who are 15-19 yearsold steadilyrisesfrom21% to 40% and the a mean age correspondingly 23.7 yearsto 15 years,yielding dropsfrom of flow. progressive "greening" the male migration In general, lowerprevalence levelscorrespond moreto earlier historical levels. For thosein the lowestprevaperiodsthan do higher prevalence lence category, mean year of migration people leaving on their the for 1517

American Journalof Sociology first U.S. tripwas 1953; it was 1964, 1971, and 1974, respectively, for thosein thesecond,third, and fourth categories. Amongmigrants the in highestcategory,the mean year of migration was 1981. Thus, older migrants who began migrating whencommunities werecharacterized by low prevalence levelsare less likely thanthoseleavingfrom communities withhighlevels to have survivedto the survey date to report theirtrip. The selectivemortality oldermigrants of from earlierperiodsconstitutes a conservativebias, however,and the drop in the mean age of first is thanindicatedby our data. migration probably even morepronounced These figures suggest thattransnational migration beginsamongmales in theirpeak labor forceyears and spreads progressively otherage to groups.In orderto measurethisincreasein age heterogeneity moresucwe indices(Shannon 1948; Theil 1972; White cinctly, computedentropy 1986) to measurediversity each level of prevalence.These indicesare at in reported table 4. The entropy index (henceforth called the diversity index)is defined the formula: by
n

Diversity=

Pi x log(pi) l() x 100,

(1)

wheren is the numberof categories this case, age groups)and pi is (in the proportion people in category of i. The indexvariesbetween0 and 100. Minimumdiversity occurswhen in all people are concentrated one category and the index equals zero. Maximumdiversity contains same proporthe occurswheneach category tionofpeople,yielding indexof 100. We computed an indicesto measure then within each community diversity prevalence by category separately, averaged themto obtain the indicesshown in the table. The resulting indifigures indicateaveragewithin-community diversity. computed We ces forthefourbroad age groupsshownin thetable, as well as formore detailedfive-year groups(11 categories). age Both setsofindicesshow an increasein age diversity occurring among male migrants subsequentto theinitialstageoftransnational migration, at level of prevalence. althoughthereis a drop in diversity the highest to According calculationsbased on the five-year intervals, age diversity is limitedat first, with an index value of 52. It thenrisesto 68 by the third prevalence phase beforeedging downward to 51 in the massto returns a level thatis close migration phase. Thus, the concentration in are to the originalvalue, but migrants now concentrated different, younger, age groups. The nextpanel oftable4 showstheage composition female of migrants 1518

Continuities Migration in by prevalencelevel. Moving from low to mass prevalencecategories, it is clear that developmental trendsin diversity roughly parallel thoseof men, movingtowardsteadilygreater heterogeneity respectto age. with In the case of women,however,the trendcontinues unabated through themass migration category. The diversity indexcomputed five-year for from in thelowestprevalence age intervals increases 16 category, through values of 50, 63, and 66 in the intermediate categories, end up at 72 to underconditions mass migration of (comparedto a value of 51 among men). increases Unlike men, however,the average age of femalemigrants the moves through successivestages. In general, steadilyas migration the percentage womenyounger of than 15 drops,while the percentages of those who are 20-34 years old and of those35 and over rise. Thus, whereas among men migration begins among older marriedhousehold heads and thenspreadsto younger heads,oldersons,and finally young to boys,femalemigration appears to beginamongyoungand working-age daughtersand then spreads to wives and older women. The order of precedencein migration thus appears to be fathers, older sons, older and children, and thenolderwomen. daughters, youngmothers in This conclusion consistent is withtheinformation presented thelast panel of table 3, whichshows householdpositionby stagein the migrationprocess.These data mustbe interpreted cautionbecause housewith hold positionis measuredat the timeof the survey,not at the timethat of left number the housemigrants on theirfirst trip.Thus, a significant hold heads shown in the panel are likelyto have been sons when they actuallymade theirfirst trip. to As a community moves froma state of low to intermediate mass prevalence levels,theshareof household heads amongmigrants progressivelyfalls while the proportion sons, spouses, and daughtersrises. of Accordingly, diversity householdpositions in beginswitha low value of 18 at thelowestprevalence level and ends up at a highvalue of 78 under conditions mass migration. of Sons begintheirupward trendat an initiallevel of participation that is quite low comparedto heads but is overtwo timesthatof spousesand come daughters (7% at thelowestprevalence levels),and theyeventually to dominate outflow the (38% at thehighest prevalence levels).Daughters rapidlyincreasetheirparticipation (from at the lowestlevel to 19% 3% at the highest).Spouses are onlya small part of the flowat the lowest substanlevel of prevalence(3%), but theyincreasetheirrepresentation, tiallymatchingsons, in the second prevalencecategory.The share of spouses then drops, but rises again in the highestcategory, surpassing of the proportion daughters (23% vs. 19%). 1519

American Journalof Sociology The orderof precedenceis consistent withthe prevailing divisionof labor withinthe householdand withnormsabout how men and women shouldoccupyand move through space. It also reflects changesin norms thatoccur as wives negotiate withhusbandsin orderto securea larger economicroleforthemselves and through migration as theyseek to join members abroad. The prevailing family view ofMexico-U. S. migration, whichsees womenas an undifferentiated their groupof wives following husbandsabroad, thusneeds modification. Risingand Falling Diversity Trip Characteristics of According the developmental to model outlinedabove, the diversity of should first rise and then fall as migration unfolds trip characteristics overtime.This curvilinear is are pattern evidentwhenU. S. destinations an examinedby the prevalenceof migration, analysisthatis carriedout in the top panel of table 5. a Althoughsouthern Californiaattracted majority all migrants of at the everyprevalencelevel exceptthe first, share goingto Los Angeles rose and thenfell,whilethe portion goingto VenturaCountyincreased the steadily.Those goingto theinlandvalleysrepresented highest initial share, but theirproportion fell,rose again, and thendropped.The perroseacrossdevelcentagegoingto the San FranciscoBay area generally opmentalstages,the proportion goingto Texas and Illinoisfell,and the share going to otherlocationsdropped, except in the last prevalence category. The trends observedin thehighest prevalence category largely are due to the particular communities here. People fromLa Yerbarepresented buena and San Diego de Alejandria,the two places thatachievedmass work in the involved in agricultural migration,are overwhelmingly in to United States, which is reflected the destinations which theytravel. in difficult see thecurvilinear to shift diversity from the It is somewhat distributions shownin thetable, because theyare aggregated geographic and intobroad zones. The pattern changof acrosscommunities grouped is ing diversity more clearlydetectedusing indicescomputedfromdein tailedgeographic categories (notreported thisarticle;no. of categories = 67). At thelowestlevel ofprevalence, small migrants to a relatively go = numberof U.S. locations(diversity 35), but the rangeof destinations increasesmarkedly when prevalencereachesthe secondcategory (diverbetween and 46 across 43 and sity= 46). It thenstopsincreasing remains the remaining stages. A similarpattern of the typifies distribution migrants tripduration, by is shown in the second panel of table 5. Once again diversity relatively 1520

Continuities Migration in low in the first prevalencelevel, thenit increasesas prevalencemoves in throughthe second and thirdphases, decreases slightly the fourth category, thenfallsfurther and whenmigration reachesa mass stage. At low levels of prevalence, migrants take either longor shorttrips.In this category 54% of the tripslasted underone year and 20% forfiveyears or more. This bimodal distribution yields a mean trip lengthof 3.4 years. As migration spreadsand becomesmoreprevalent, shareofpeople the stayingover five years at first rises but then progressively falls. The proportion takingmoderate-length trips(one-fiveyears) rises but then shortens. The greatest levelsof diverdrops,and the averagetriplength sity(91 and 90) are reachedin thethird and fourth prevalence categories, but diversity fallsback to 87 duringthe mass phase. At thisprevalence level, only24% of people stayedover fiveyears,49% stayedunderone year,and 27% stayedone-fiveyears. Althoughthereare no strongdevelopmental trendsin the aggregate of falldistribution migrants legal status,thereis a pattern rising, of by ing, and then risingdiversity withincommunities. low prevalence At levelsthe diversity indexis only43, it risesto a peak of 57 in the second prevalencestage, and by the timemass migration been achieved it has has fallen back to 54. In the initial phases of migration, 46% of all migrants undocumented, are 16% are legal, and 38% are braceros,but at the finalstage braceroshave droppedto zero and the share of legal has migrants increasedto 64%. The foregoing resultssuggestan orderlyshiftin migrantstrategies of across developmental stages. At the beginning the migration process, or tendeither adopt a settled to of residence migrants strategy long-term a short-term of movement. strategy back-and-forth Duringintermediate as withvariations thesestrategies theytry on stagesmigrants experiment out different locationsand different jobs. As the developmental process which are turntoward settledstrategies, proceeds,however,migrants in made moreattractive theformation stemcommunities theUnited of by which are enabled by the emerStates, or toward recurrent strategies, In networks. some cases a community gence of well-developed migrant for may "specialize"in morethanone strategy, depending, example,on whether have achieveda toeholdin morethanone destination migrants or occupation(Goldring1992 b). In orderto studymigrant the strategies, employed criteria we used by first the Massey et al. (1987) to characterize period betweenmigrants' made theirfirst and mostrecenttrips.New migrants tripto the United either Statesduringthe threeyearspriorto the survey.Settledmigrants stayedin the United States forthreeyearson theirmost recenttripor had an average tripdurationof at least threeyears betweentheirfirst 1521

ID

CN rl

Al

1-

in

cl

'IC

,I

? 1

~~

-~
("1

~~rt 00

C,
mNNo e

..I

O0~QQ~-C~O0~

~Q

e
o0ro x 0 N-

00

00X r

I;

Nt

(O

z
o s N _ e N o oo t-\O

H
0

~O

t-

~~~
I

~~~~~~~~~~~~~~~~~~~~~~~~0
t 0 r 6 ; in 0

w0
H
M

Z0
i

lflf

e0X~-f

v 08 > m t m t N oO

~~~~~~~
> U U~~~~~~~~~~~C +. ..Cl o$... v ..... t ..t............. ? vN . .M e t - t oo v
O

...... .

7v

E-1

m m N 00 ? m ? m m

>

.cn1

1522

'O
uo e e N

O
ur

i :n

I ot
N q:

ro

ON

0) -4

m n M It m "t

~~00

00-4O

ONt

'O

1Cy

r-

(N

-e

t-

ON

'

cn

Oo in

00 oo t

1.r

C\

00

C)

00

"t

C t

0-t0

"d

tn

"t

cdq n

Cd
bi

Go' bD

0~~~~~~~~
n

cl,
oo m N oO m e ? t O

C-d

m t N cr o ? e+ t

bD

bD

-- ,

Cd~

n ln o X N m oo t o Not u: tO O O t W In m t N ? e et e 0 U

bOCd

bD~~~~~~ ~~~~~~.

g~~b

Q~~~~~~~b
~~~~~io
bi tQ

sz~~~~

m~~~~~~~
.

cd

~ ~~ ~ ~ ~ ~ ~ ~ ~ 0~~~ ~ ~
4 *
. . . .

. .
n .

*. . . .
C .

0~~~~~~~ .
. .
. . .

. . . . . . . . 'e~~~~~~~~~~~~4.

Cd

--

*4

: w
.

:3 :

.
*

: .
.

: .n
.

: .4

:-

(:$

:
.

.
. .

. w>. . *

.
.

.
n. .

.
. .

.
.

. . .

.d

(3.)

.n

bb.h

C-d :n

:.

:1 :

:7

:--

:~
.

.~~~~~~~~~~l
:
. + . .

:J

!V

.4

n~~~~

. 1.52..3

American Journal Sociology of and last trip.Recurrent took at least threetripsand averaged migrants one tripeverytwo yearsor spentat least halfof the timebetweentheir first and last tripin the UnitedStates. Finally,temporary migrants, the residualcategory, took fewerthan threetrips,averagedfewerthan one tripeverytwo years,or spentless than halfof theirtimebetweentrips in the United States. We used the midpoint the periodbetweenthe of first last tripas theyearby whichto classify and strategy prevalence. by For people who had made onlyone trip,we used its date to define the relevantprevalencecategory. As themigration in processunfolds stagesand transnational movement becomesmoreprevalent, shift migrants generally away from settled and and temporary strategies increasingly favora strategy recurrent of movement.Acrossthe fivestages,the relativenumber migrants of employing a temporary falls from14% to 5%, the share using a settled strategy strategy drops from34% to 20%, but the proportion using a recurrent strategy increasesfrom50% to 73%. At the same time,the diversity of risesfrom at thelowestprevalence 46 strategies level to 73 at thefourth prevalence categorybut drops to 61 when mass migration has been reached. Thus, the generaltendency toward greaterdiversity strategies is in over time, despite some curvilinearity. When migration becomes extremely prevalent, thereis a tendency however, towardrenewedspecialization, focusing on The growingpreprincipally a recurrent strategy. dominance of recurrent is migration enabled by the accumulationof social capitalin theform network of connections community and migrant whichenable anyone,even new migrants, experience, quicklyto adopt a patternof recurrent movement back and forth regularperiodsof for paid labor abroad. The last panel of table 5 shows shifts the U.S. occupations in held by at migrants different stagesin themigration is process.Once again, there a pattern rising of and falling increases diversity. Occupationalvariation between first third the sharply and prevalence levels,thendeclinesacross laterphases. There is, however,an apparentreversalof the shift away fromagricultural betweenthe thirdand fifth employment prevalence levels. Whereasthe proportion migrants of in working agriculture drops from 83% at the lowestprevalence level to 42% in thethirdcategory, it risesagain to 71% and 84% during last two phases. the The generaltendency probablytowardgreater is in concentration urban jobs as prevalencerises;the apparentrespecialization agriculture in withinthelast two categories reflects natureofthecommunities the that in achieved high levels of prevalence.Whetherspecialization U.S. agricultural labor is itselfa factorthat promotes the emergence mass of 1524

in Continuities Migration
TABLE 6
SOCIOECONOMIC CHARACTERISTICS OF MIGRANT HOUSEHOLD HEADS PRIOR TO LEAVING ON THEIR FIRST U.S. TRIP IN PREVALENCE MIGRATION COMMUNITY OF CHARACTERISTIC 0%-9% 10%-19% 20%-29% 30%-39% 240%

Education: None (%) .............................. 1-5 years (%).............................. 6+ years(%).............................. Mean yearsof schooling.................
Diversity (n = 20) ........................

23.2
54.5

23.5
36.8 39.7

16.4 29.9

22.4 2.7
36 6.1 9.3 61.2 29.4

3.7
53 8.1

53.7 5.2
55 4.9

21.6 50.0 28.4

6.4 65.3 28.3

2.9
51

4.3
63

Land ownership: Landowners(%) ........................... Businessownership: Businessowners(%) ...................... Mexican occupation: Agriculture (%) ............................ Nonagriculture (%)* ...................... Professional-manager-owner ... (%) Technical-sales-clerical .......... (%) Skilledmanual (%) ..................... Unskilledmanual (%) ................. Services(%) .............................. Not in workforce ..................... (%)
Diversity (above groups, n = 7) ...... Diversity (original groups, n = 64) ...

5.1
3.9 70.0 15.5

.5
8.1 57.5 10.8

5.8
48.7 38.7

7.4
38.0 46.9

.0
5.8 7.2 9.6 6.9

1.0
7.2 12.8 11.3 6.5

3.2
6.0 18.1 10.7 8.8

.0
3.0 6.5 2.8 3.2

.0
2.0 6.2 2.0 .8

9.4
32 27

12.6
50 45

15.1
59 48

14.5
47 48

31.7
41 38

No. migrant householdheads .............

172

670

674

190

57

of NOTE.-The diversity for figures "original groups"are based on a breakdown theMexicanoccupain tionsinto64 categories morespecific thanthe seven generalcategories reported thisarticle. * Categoriesof nonagricultural total due to rounding employment may not sum to nonagricultural error.

or we thosecommunities selected migration whether just happenedto be from small agrariantownscannotbe determined thesedata. of IncreasingSocioeconomic Heterogeneity Migration The gradualaccumulation network of connections migratory and knowlan comedge across developmental stagesmakes migration increasingly monsocial and economicpractice and lowersthecostsand risksofmovea ment,makingmigration less selectiveprocess.Table 6 examinesthe in of socioeconomic selectivity migration termsof education,property and Mexican occupations.These variablesare measuredfor ownership, in U.S. trip. migrants the year before theytake theirfirst 1525

American Journal Sociology of Changingeducationaldistributions somewhat are difficult interpret to historical because thestagesofmigration tendto occurat different times. Overall levels of education have been risingin Mexico and the local has availability postprimary of education expandedovertime.In general, botheducationallevelsand diversity increaseas migration becomesmore prevalent,indicatingthat the educationalselectivity migration of decreases. Mean educationincreasesfrom to 4.3 yearsfrom lowest 2.7 the to the mass prevalencelevel, and the diversity index increasesfrom36 in to 63, afterremaining constant the second through fourth relatively prevalencecategories. Distributions property of ownershipalso suggestthat migration becomes less socioeconomically selectiveas migration spreads throughout In the community, althoughthereis some curvilinearity. the first two prevalencecategories, and 8% of the migrants landowners, 6% are but thepercentage and fourth dropsto about 5% in thethird stagesand ends Trends in businessownerup at 0.5% in the mass prevalencecategory. are ship are less clear. In theinitialstage,9% of migrants businessownthe ers,but as migration becomesmoreprevalent, proportion falls,rises, fallsagain, thenrises. It is possiblethatsome of thesefluctuations reflect processin which a use remittances acquire businessesor property to that early migrants in latermigrants the family report beingtheirown priorto theirfirst as That is, theaccumulation property of overtimebyindividuals departure. who pass it on to family members may explainsome of the fluctuations. It may also reflect particular the kindsof communities achieve high that levels of migratory prevalence. is The broadeningof socioeconomic representation suggestedby the Mexican occupationaldata, shown in the last panel of table 6. At the are from lowestlevel of prevalence, migrants drawnlargely agricultural an while61% report occupations:29% come from nonagrarian pursuits, are agrarianoccupation.Amongnonagricultural occupations,migrants distributedrelativelyevenly among four categories: technical-salesclerical workersmake up 6%, skilled manual workers7%, unskilled manual workers10%, and serviceworkers7%. As migration proceeds the thirdprevalencecategory, through however,thereis a clear shift of migrants' of originsto include a higherproportion nonagricultural are backgrounds. the timethisstageis reached,38% of all migrants By have whilenonagricultural workers engagedin agricultural occupations, risenslightly a 47% share.The increaseis particularly to largeforskilled workers. The trendtoward lower shares of agricultural backgroundsamong in first-time reverses thefourth wherethe migrants prevalencecategory, share jumps sharplyto 70%, beforefallingback down to 58% in the 1526

Continuities Migration in last category. This increasing predominance agricultural of backgrounds probablyagain reflects nature of the communities the achievinghigh levels of prevalencein our sample, but it mightalso reflect Mexico's sharply worsening economicconditions during early1980s. The "crithe sis" sentmanypeople north, did thepossibility legalization as of through the Immigration Reform and ControlAct. The occupationaldata in table 6 also show that the proportion of migrants reporting Mexican occupationpriorto leavingon theirfirst no In tripincreasesacrossprevalence categories. theinitialprevalencecategory9% of migrants report havingworkedbefore not leavingMexico, a percentage thatrisessteadily through fourth the phase and jumps to 32% in the highestprevalencecategory.This patternreflects decline in the themean age ofmigration notedin table4, whichyieldsa steadyincrease in thepercentage migrants of outsideofthelaborforce.This pattern also suggestsa phenomenon "northernization," of wherebyU.S. migration gainsforceas a social and cultural phenomenon and people increasingly first. migrate abroad without gaininglocal occupationalexperience As migration movesfrom low to mass prevalence, degreeof diverthe sityin migrant's occupationalbackgrounds follows curvilinear a pattern. When calculatedusingthe broad occupationalgroupsshownin table 6, it growsfrom27 to 48 and thenfallsback to 38. When based on more detailed categories(55 categories, not reportedin this article),it rises from to 59 and thenfallsback to 41. Although 32 thereis sometendency foroccupationaldiversity constrict to betweenthe thirdand the mass phases in curvilinear fashion,the overall tendency toward broader is representation less socioeconomic and selectivity the streamof outin migrants. CONCLUSION in Field investigators working Mexico duringthe 1970s and early 1980s in uncovereda varietyof empiricalcontinuities the way that transnationalmigration communities. to developedwithin Migration theUnited Statesgenerally of began witha smallnumber migrants leavingthecomand demographic niche. munityfroma rathernarrow socioeconomic Over time,however, number migrants the of tendedto growand eventuall ally came to incorporate virtually groups and classes in the community. In this articlewe outlineda cumulativetheory migration of that accountsforempirical observedby earlier regularities investigators. Migration tends to increasein prevalenceand become more diversebecause transnational movement causes relatively permanent changesin individual motivations, and thesechanges social structures, cultural and milieus, 1527

American Journal Sociology of cumulateover timeto changethe context withinwhichsubsequentmigrationdecisionsare made. As information about migration growsand network connections the UnitedStates ramify, costs and risksof to the international movement dropand migration becomesmoreattractive. As and network morepeople are inducedto migrate, knowledge connections and so on. Withtime, morepeople to migrate, expand further, inducing becomesa generalized migration social and economicpractice. No previousstudy examined has developmental processes migration of usingsucha broad sampleofcommunities surveyed usingidentical methods and procedures. Drawing on data from19 Mexican settlements, we defined fivebasic stagesin the social processof migration based on the in of We overallprevalence migration thecommunity. showedthatwhen high prevalencelevels are achieved, the vast majorityof townspeople are relatedto someonewho has been to the United States and average U.S. experience accumulatedto veryhighlevels. has of This accumulation social capital qualitatively changesthedecisionfor making environment potentialmigrantsand makes transnational members.Although movement accessible and attractive community to thefirst international tendto be married male householdheads migrants ofprimelabor force a and age, usuallyfrom nonagricultural background a becomesmoreprevalent often from property-owning as migration class, and social capital accumulates,thisprofile changes. Migrationspreads among males at a relatively constantpace, but rate. As a result, among femalesits prevalencegrowsat an increasing of cohorts the representation females increasesmarkedly amongmigrant as the level of prevalencerises. The range of ages steadilybroadens among both men and women. Amongthe former, migration generally spreads fromfathersto older sons and then to young boys. Among and youngwives, then women,migration beginsamongolderdaughters moves to older wives and younggirls.As migration develops froman it isolatedsetofeventsto a mass phenomenon, also becomesless selective in class terms.Educationaland occupationalbackgrounds becomemore workers falls. diverseas the percentage landowners of and agricultural At early stages in the developmental processof migration, migrants tendto go to a rather narrowgeographic and occupafrom community a in tional niche in the United States, largelyfollowing the footsteps of As the first migrants. networks develop and migrants acquire greater in experienceabroad, however,theyseek out new opportunities new of locations.Eventually, someoneachievesa position authority however, thatallows himto distribute to employment thepeoplein his community whichcauses the diversity destinations stop rising. of to network, was low in Our empiricalanalysesshowed that geographic diversity 1528

Continuities Migration in the initialstagesof migratory increaseddramatically development, during the intermediate stages,and thenstayedconstant fellslightly or as a mass level of migration was achieved. Movementstrategies and trip durations displayeda pattern rising of and thenfalling diversity comas munities moved from lowestto thehighest the prevalencecategory. Ocin cupationsand legal statusesformigrants the UnitedStatesdisplayed a generaltrendtowardgreater diversity across developmental stages. Thus, usinga broad sampleofMexicancommunities, find we evidence of commondevelopmental of thatare in line withthe patterns migration observations earlyinvestigators thedevelopmental of and empirical theoutlinedabove. Our resultsreinforce Durand oryof network migration and Massey's (1992) conclusionthatcommonmigration processesoccur across a wide rangeof Mexican communities, even thoughtheirexpression may be shaped by factors level. operating the community at We also underscore Durand and Massey's caveat that care must be taken when attempting generalizefromisolated community studies. to As we have shown,depending whether selects community on one a where is transnational migration incipient well established, "nature"of or the in transnational migration may be characterized verydifferently social, and economictermsand the patterns movement demographic, of may In varyconsiderably. orderto aid in future comparative work,investigatorsundertaking case studiesof migrant communities shouldreport the degreeof migratory prevalenceso thatothers can determine what phase thecommunity achievedin thedevelopmental has processand can avoid communities markedly at different comparing stages. Our contribution thuslies in synthesizing available materialand sugregesting conceptualmodel thatcan reconcile a seemingdiscrepancies are portedin individualcase studies.If identifiable patterns associated withthe expansionof migratory behaviorand accumulatedexperience, thenthesedifferences shouldbe the resultof diversehistories and levels of migration of experience. Simple crosstabulations data from multiple sitesoffer little intotheprocessof migration unlesstheyare staninsight dardized for purposesof comparison.We introduce the prevalenceof as migration a conceptualand empiricalmeasurethat can capturethe cumulative as processof migration it unfolds. in a Having discussedcommonalities migration from broad sample of and Mexican communities, nextneed to studythe structural factors we thatshape thespreadofmigration within communipopulation processes tiesto understand attaina stateof mass migrawhysome places rapidly tionwhile othersdevelop moreslowlyand achieve onlymodestratesof We out-migration. hope to address these more complicatedmultilevel research. analysesin future 1529

American Journalof Sociology


REFERENCES Impactode la Migraci6n Alarc6n,Rafael. 1988. "El Processode 'Nortefnizaci6n': de Internacional Chavinda, en Michoacan." 337-58 inMovimientos Poblaci6n Pp. en el Occidente Mexico,edited ThomasCalvo and GustavoL6pez. Zamora: de by El Colegiode MichoacanPress. from MexicanTown." a Self-Perpetuating Migration . 1992. "Nortefiizaci6n: editedby Pp. 302-18 in U.S.-MexicoRelations:Labor MarketInterdependence, Stanford, Calif.: Stanford Jorge Bustamante, Hinojosa, and Clark Reynolds. R. University Press. and theProblems of Baucic, Ivo. 1972. The Effects Emigration of fromYugoslavia Nijhoff. Returning Emigration Workers. Hague: Martinus The of Coleman, James 1988."SocialCapitalin theCreation HumanCapital."AmeriS. can Journal Sociology of 94:S95-S120. a Consejo Nacional de Poblaci6n(CONAPO). 1986. Encuestaen la Frontera Trade Devueltospor las Autoridades los Estados UnibajadoresIndocumentados de dos de America:Diciembre 1984. Mexico City:Consejo Nacional de Poblato States:The View from Cornelius, WayneA. 1976a.MexicanMigration theUnited Monograph C/76-12. Rural SendingCommunities. Migration and Development Mass.: MIT Center International for Studies. Cambridge, . 1976b."Outmigration RuralMexicanCommunities." Interdisciplinary from Communications Program OccasionalMonograph Series5 (2): 1-39. Washington, D.C.: Smithsonian Institution. and to . 1978.MexicanMigration the UnitedStates:Causes,Consequences, U.S. Responses.Migration and Development MonographC/78-9.Cambridge, Mass.: MIT Center International for Studies. and . 1990. "Labor Migration the UnitedStates:Development Outcomes to for of Alternatives Mexican in Sending Communities." Commission theStudy InterWorking Paper no. national Migration and Cooperative EconomicDevelopment 38. Commission theStudy International Economic and for of Migration Cooperative Development, Washington, D.C. Profile MexicanLabor of . 1992."FromSojourners Settlers: Changing to The in Relations: Labor Migration California the1980s."Pp. 155-95in U.S.-Mexico to MarketInterdependence, R. editedby Jorge Bustamante, Hinojosa, and Clark Reynolds. Stanford, Calif.:Stanford University Press. Immiof Dagodag,W. Tim. 1975."Source Regions Composition IllegalMexican and gration California." to International Migration Review9:499-511. StudyofRural Dinerman, R. 1982.Migrants Stay-at-Homes: Comparative Ina and A in Mexico.Monographs U.S.-Mexican no. Studies, 5. Migration from Michoacdn, in La Jolla,Calif.:Program United University Califorof States-Mexican Studies, nia, San Diego. M. Why Donato,Katharine 1992."Understanding Immigration: SomeCountries U.S. Ground: Send Womenand Others Men." Pp. 159-84in Seeking Common Women to Conn.: Immigrants the UnitedStates,editedby Donna Gabbacia. Westport, Greenwood Press. the Donato,Katharine Jorge M., Durand,and DouglasS. Massey.1992."Stemming Tide?Assessing Deterrent Reform Control and Act." the Effects theImmigration of Demography 29:139-58. en de Durand,Jorge. 1992."Paso del Norte:Patrones Migrantes el Occidente M6xand of ico." Ph.D. dissertation. University Toulouse,Department Geography of Management. to Durand,Jorge, and Douglas S. Massey. 1992. "MexicanMigration the United Review27:3-42. States:A Critical Review."LatinAmerican Research 1530
ci6n.

in Continuities Migration
England, Paula. 1992.Comparable Worth: Theories Evidence.Hawthorne, and N.Y.: Aldinede Gruyter. Fergany, Nader. 1982."The Impactof Emigration NationalDevelopment the on in Arab Region:The Case of the Yemen Arab Republic."International Migration Review16:757-80. Fernandez, Celestino. 1988. "Migraci6n hacia los EstadosUnidos:Caso Santa Ines, Michoacan. Pp. 113-24inMigraci6n el Occidente Mexico,edited Gustavo en de by L6pez and SergioPardo. Zamora:El Colegiode Michoacan. Galarza,Ernest.1964.Merchants Labor: TheMexicanBracero of Story.Santa Barbara, Calif.:McNally& Loftin. to Gamio,Manuel. 1930.MexicanImmigration theUnited States.Chicago:University ChicagoPress. of DeGeorges, Eugenia. 1990.TheMaking a Transnational of Community: Migration, velopment, Cultural and Changein theDominican Republic.New York: Oxford University Press. A Luin. 1990."Development Migration: Comparative and of Goldring, Analysis Two MexicanMigrant Circuits." Commission theStudyofInternational for Migration and Cooperative Economic Development Working Paperno. 37. Commission for the Studyof International Migration and Cooperative EconomicDevelopment, D.C. Washington, in and Social Transformation . 1992a. "BlurringBorders: Community on Mexico-U.S.Migration." Paperpresented theConference New Perspectives at on Mexico-U.S.Migration. of 23-24. University Chicago,October in A . 1992b."Diversity, Community Transnational and Migration: ComparaPh.D. dissertation. tive Studyof Two Mexico-U.S.MigrantCircuits." Cornell University, Department RuralSociology. of . 1992c."Gendered Memory: ReconstructionstheVillage MexicanTransof by national Migrants." Paperpresented theeighth at WorldCongress RuralSociolof ogy,University Park,Pa., August. 1992d. "La migracion Mexico-EUAy la Transnacionalizaci6n Espacio del Politico Social: Perspectivas y desde el Mexico Rural."EstudiosSociol6gicos10 (29): 315-40. and C. Goldring, Luin, and Robert Smith.1993."Substantive Citizenship TransnaDimensions Mexico-U.S.Transnational of tionalCommunities: Sociolegal Migration."Manuscript. of Research Center. University Chicago,Population AnnalsofMathematical Statistics 32: Leo. 1961. "SnowballSampling." Goodman, and PatriciaR. Pessar. 1991.BetweenTwo Islands: Dominican Grasmuck, Sherri, and of Press. International Migration. Berkeley Los Angeles: University California Failure:Employment the Peter.1986.TheMyth Market and LaborMarket Gregory, of in Mexico. Baltimore: Johns HopkinsUniversity Press. Patriarchal Constraints: ReconThe Hondagneu-Sotelo, Pierette. 1992."Overcoming Women struction Gender of Relations MexicanImmigrant and Men." Genamong derand Society6:393-415. to of MexicanMigrants Jones,RichardC. 1982a. "Channelization Undocumented theUnitedStates."EconomicGeography 58:156-76. . 1982b."Undocumented Migration from Mexico:Some Geographical QuesAssociation American 72:77-87. tions."Annals, of Geographers Na. 1988. "MicroSourceRegionsof MexicanUndocumented Migration." tionalGeographic Research 4:11-22. in Role Networks Individual David P. 1991."The Differential ofFamily Lindstrom, of at Decisions."Paper presented the annual meeting the Population Migration Association America, of Washington, D.C., March21-23. and New York:Academic Press. Larissa. 1977.Networks Marginality. Lomnitz, 1531
117-51.

American Journal Sociology of


L6pez, Gustavo. 1986.La Casa Dividida: Un Estudio de Caso SobreMigraci6n a Estados Unidosen un PuebloMichoacano.Zamora:El Colegiode Michoacan. Process Massey,Douglas S. 1985."The Settlement amongMexicanMigrants the to United States:New Methods Findings." 255-92in Immigration and Pp. Statistics: A Story Neglect, of edited Daniel B. Levine,Kenneth by Hill, and Robert Warren. D.C.: NationalAcademy Press. Washington, . 1986. "The Settlement ProcessamongMexican Migrantsto the United States."American Sociological Review51:670-85. in . 1987a. "The Ethnosurvey Theory and Practice." International Migration Review21:1498-522. . 1987b."Understanding MexicanMigration theUnitedStates."American to Journal Sociology of 92:1372-403. . 1990."SocialStructure, Household Strategies, theCumulative and Causation of Migration." Population Index 56:3-26. Massey,Douglas S., RafaelAlarc6n, Jorge Durand,and Humberto Gonzalez.1987. Return Aztlan: to TheSocial Process International of Migration from Western Mexico. Berkeley Los Angeles: and University California of Press. Massey,DouglasS., and FelipeGarciaEspafia.1987."The SocialProcess Internaof tionalMigration." Science 237:733-38. A Mines, Richard. 1981. Developing Community a Tradition Migration: Field of Studyin Rural Zacatecas,Mexico and California Settlement Areas.Monographs inU. S -Mexican Studies, 3. La Jolla, no. in Calif.:Program United States-Mexican of Studies, University California, Diego. San . 1984."Network A Migration MexicanRuralDevelopment: Case Study." and Pp. 136-55inPatterns Undocumented of Migration: Mexicoand theUnited States, editedby Richard Jones. C. Totowa,N.J.: Rowman& Allanheld. and Anzaldua.1982.NewMigrants OldMigrants: vs. AlterMines,Richard, Ricardo nativeLaborMarket in in Structures theCalifornia Citrus Industry. Monographs in U.S.-MexicanStudies, 9. La Jolla,Calif.:Program UnitedStates-Mexican no. of Studies, University California, Diego. San 1982. "Migration the UnitedStatesand to Mines, Richard,and Alain de Janvry. A MexicanRural Development: Case Study."American JournalofAgricultural Economics 64:444-54. and DouglasS. Massey.1985."Patterns Migration theUnited to Mines,Richard, of Statesfrom Two MexicanCommunities." Latin American ResearchReview 20: 104-24. Gunnar.1957.Rich Lands and Poor. New York:Harper& Row. Myrdal, and North,David S., and MarionF. Houstoun.1976. The Characteristics Role of An IllegalAliensin theU.S. LaborMarket: Exploratory D.C.: Study.Washington, Linton. Pedraza,Silvia. 1991."Women Migration: SocialConsequences Gender." and The of Annual ReviewofSociology 17:303-25. in and Pessar,PatriciaR. 1982. "The Role of Households International Migration the Case of U.S.-BoundMigration from Dominican the Republic."International Review16:342-64. Migration Piore, MichaelJ. 1979.BirdsofPassage:Migrant Laborin Industrial Societies.New York: Cambridge University Press. and Lessons Portes, Alejandro.1979."IllegalImmigration theInternational System: from RecentLegal Immigrants Mexico."Social Problems from 26:425-38. A America: Portrait. and RubenC. Rumbaut.1990. Immigrant Portes, Alejandro, and of Press. Berkeley Los Angeles: University California An of Reichert, JoshuaS. 1979. "The Migrant Syndrome: Analysis U.S. Migration and Its Impacton a RuralMexicanTown." Ph.D. dissertation. Princeton Univerof sity, Department Anthropology. 1532

in Continuities Migration
Syndrome: SeasonalU.S. WageLaborand RuralDevel. 1981."The Migrant in opment Central Mexico."HumanOrganization 40:56-66. in . 1982. "A Town Divided:EconomicStratification Social Relations a and MexicanMigrant Community." Social Problems 29:411-23. Reichert, Joshua, and DouglasS. Massey.1979."Patterns U.S. Migration S., of from A a MexicanSendingCommunity: Comparison Legal and Illegal Migrants." of International Migration Review13:599-623. . 1980."History Trends U.S. BoundMigration and in from MexicanTown." a International Migration Review14:475-91. Rhoades, Robert.1979. "From Caves to Main Street:Return and the Migration Transformation a SpanishVillage."Papersin Anthropology of 20:57-74. Rouse,RogerC. 1989."Mexican Migration theUnited to in States:Family Relations theDevelopment a Transnational of Migrant Circuit." Ph.D. dissertation. Stanford of University, Department Anthropology. . 1991. "MexicanMigration the Social Space of Postmodernism." and Diaspora 1:8-23. . 1992."Making SenseofSettlement: Class Transformation, Cultural Struggle, and Transnationalism amongMexicanMigrants theUnitedStates."Annalsof in theNew York Academy Sciences645:25-52. of Ruiz, VickiL., and Susan Tiano, eds. 1987. Women the U.S.-MexicoBorder: on Responses Change.Boston:Allen& Unwin. to NotreDame, Ind.: UniverSamora,Julian.1971.Los Mojados: The Wetback Story. of sity NotreDame Press. Sassen,Saskia. 1988. The Mobility Labor ar I Capital:A Studyin International of Investment LaborFlow. Cambridge: and Press. Cambridge University Bell Shannon,C. E. 1948. "A Mathematical Theoryof Communication." System Technical Journal 27:379-422,623-56. Robert 1992."'Los Ausentes C. Presentes': Imagining, The Smith, Siempre Making, and Politics a Transnational of between New YorkCityand Ticuani, Community Puebla." Paperpresented theconference theLatinAmerican at of Studies Association,Los Angeles, September. Mass.: Basil Blackwell. Stark,Oded. 1991.TheMigration Labor. Cambridge, of and 1981."Causesand Effects Agricultural of Labor Stuart, James, MichaelKearney. Migrationfrom the Mixteca of Oaxaca to California."WorkingPaper in in U.S.-Mexican no. States-Mexican Studies, 28. La Jolla,Calif.:Program United of Studies, University California, Diego. San Taylor, J. Edward. 1986. "Differential Migration, Networks, Information, and Risk."Researchin HumanCapitaland Development: HumanCapital, Migration, and Development 4:147-71. Paul S. 1980."Mexican in Taylor, Women Los Angeles 1. Industry." Aztlan11:99-13 Theil,Henri.1972.Statistical Decomposition Analysis. Amsterdam: North Holland. Todaro,MichaelP., and LydiaMaruszko.1987."IllegalMigration U.S. Immiand A gration Reform: Conceptual Framework." Population Development and Review 13:101-14. MichaelJ. 1986."Segregation Diversity: in and Measures Population DistribuWhite, tion."Population Index 52:198-221. in E. and Wiest,Raymond 1973."Wage-Labor Migration theHousehold a Mexican Town." Journal Anthropological of Research29:108-209.

1533

You might also like