Benefit or Burden? Social Capital, Gender, and the Economic Adaptation of Refugees

Benefit or Burden? Social Capital,Gender, and the Economic Adaptationof Refugees 1

Ryan AllenUniversity of Minnesota

This paper examines the role of co-ethnic social capital on the earn-ings of refugees, using a unique data set for adult refugees who reset-tled in Portland, Maine, between 1998 and 2004. Multiple regressionmodels test the effect of access to co-ethnic social capital on the logearnings of refugees in their first and most recent years of work.Results show that over time access to co-ethnic social capital uponarrival decreased earnings for female refugees. The paper concludeswith a discussion of the implications of the findings for social capitaland immigration research.

INTRODUCTION

Though disagreements about the exact nature of social capital still exist inthe academic community (Portes and Landolt, 1996; Putnam, 2000; Lin,2001a; Newton, 2001), most researchers define social capital as the abilityof an individual who is a part of a social network to access variousresources that reside within that social network. In other words, socialcapital is a collection of benefits belonging to a network of individuals(Bourdieu, 1985). Existing research on the importance of co-ethnic socialcapital, or access to resources controlled by a social network of individualsof the same ethnicity, suggests that it helps to shape a variety of individualoutcomes (Stack, 1974; Portes and Sensenbrenner, 1993). Still, there islingering debate over whether co-ethnic social capital has a positive or neg-ative effect on labor market outcomes, particularly for low-skill, low-wageimmigrant workers (Sanders, Nee, and Sernau, 2002; Tsuda, Valdez, andCornelius, 2003). This debate has mostly focused on economic immigrants

1The author would like to acknowledge the financial support of the U.S. Department ofHousing and Urban Development and the Maine Department of Labor, which made the

research upon which this article is based possible.

� 2009 by the Center for Migration Studies of New York. All rights reserved.DOI: 10.1111/j.1747-7379.2009.00767.x

332 IMR Volume 43 Number 2 (Summer 2009):332–365

instead of refugees, or immigrants who resettle in a receiving country afterleaving their countries of origin because of persecution (Black, 2001).

Since refugees arrive in the United States with few materialresources, government assistance that is limited in time, uneven access toco-ethnic social capital, and ambitious expectations regarding their abilityto become economically self-sufficient, understanding how co-ethnic socialcapital shapes refugee economic adaptation is vital (Haines, 1996). Accessto co-ethnic social capital could give refugees much needed informationregarding available jobs and the support necessary to work in one of thesejobs, producing positive outcomes (Gold, 1992). At the same time, accessto co-ethnic social capital that helps refugees find employment could alsoconstrain their labor market activity because of reciprocal obligations andadherence to social norms that accompany the use of social capital (Bachand Carroll-Seguin, 1986; Portes and Sensenbrenner, 1993; Read, 2004).

Several changes in the admission of refugees to the United Statescompound the importance of understanding the effect of co-ethnic socialcapital on the economic outcomes of refugees. First, compared to the highpoint of refugee admissions in 1980, the United States now admits farfewer refugees, and those refugees who are admitted comprise a muchsmaller percentage of the total number of immigrants admitted to thecountry. According to the Office of Refugee Resettlement and the Depart-ment of Homeland Security, in fiscal year 1980 the United States admit-ted over 200,000 refugees, representing almost 40 percent of the totalnumber of immigrants admitted in that year (DHS, 2005; ORR, 2005).In contrast, the same sources indicate that by fiscal year 2004 the UnitedStates admitted about 74,000 refugees, representing only 8 percent of thetotal number of immigrants admitted in that year. Similar to the declinein the number of refugees admitted, the federally funded benefits for refu-gees have also waned during this period, falling from 36 months of cashassistance to 8 months of cash assistance. The smaller number of refugeesadmitted to the country with less financial support from the governmentmay mean that co-ethnic social capital plays a different role in the resettle-ment of refugees today compared to the role it played in the early 1980s.

Second, the diversification of refugees resettling in the United Statesmay mean that existing research on co-ethnic social capital and refugeesin the United States is less applicable to the current situation faced by ref-ugees. In the early 1980s most refugees in the United States came fromSouth-East Asia, but currently 38 percent come from a country in Africaand 40 percent come from the former Soviet Union or Eastern Europe

Social Capital, Gender, and the Economic Adaptation of Refugees 333

(ORR, 2004). Early refugee research focused on South-East Asian refugeeswho arrived with unique cultural traditions and one set of resettlementpolicies, but we know comparatively little about more contemporarygroups of refugees who arrive in the United States with different culturaltraditions and must respond to a different set of resettlement policies.

Third, compared to the early years of the refugee resettlement pro-gram in the United States, refugees are now more likely to resettle insmall cities and even rural areas that have less experience with refugeesspecifically and diversity in general (Fennelly and Leitner, 2002; Singerand Wilson, 2006). With different economic conditions and institutionalenvironments present in smaller cities and rural areas compared to largecities in the United States, it stands to reason that refugees in these areasmay have experiences with co-ethnic social capital that differ significantlyfrom the experiences of previous waves of refugees (Hein, 2006).Together, these factors make the role of co-ethnic social capital on theeconomic adaptation of refugees living in small cities an important topicto consider.

This paper responds to the question, does access to co-ethnic socialcapital upon arrival in the city of resettlement give refugees an advantagein the labor market compared to those who do not have access to co-ethnic social capital upon arrival? Using a mixed-methods approach itanswers this question through an analysis of the effect of sponsorship, ameasure of access to co-ethnic social capital, on the earnings of a diversegroup of adult refugees that recently resettled in Portland, Maine, a smallNew England city with a relatively large and growing refugee population.Research results suggest that co-ethnic sponsorship has no effect on theinitial earnings of female and male refugees, but over time a negativeeffect on the earnings of female refugees.

This paper proceeds in six sections. The first section explores rele-vant prior scholarship on this topic and places this current study withinthe context of this work. The second section describes the data used inthis analysis, as well as an overview of the demographic characteristics andeconomic conditions in Portland, Maine, compared to other U.S. metro-politan areas where refugees comprise a significant portion of the foreignborn population. The third section describes qualitative data collected inPortland and how these data informed the research hypotheses tested inthis study. The fourth section describes a quantitative methodology usedto test the relationship between access to co-ethnic social capital andrefugee earnings. The fifth section presents results from the quantitative

334 International Migration Review

analysis. Finally, the sixth section identifies research findings and discussestheir theoretical and practical implications.

THE IMPORTANCE OF SOCIAL CAPITAL ON THEECONOMIC ADAPTATION OF REFUGEES

Originally, scholars focused on refugee economic adaptation because ofthe large numbers of Cuban and Vietnamese refugees arriving in theUnited States in the 1960s and 1970s, and unease among politicians thatrefugees would become a burden on state welfare systems (Kennedy,1981; Gold, 1992; Haines, 1996). In fact, political concerns were onlythinly veiled in the 1980 Refugee Act, which explicitly states that employ-able refugees should be placed in jobs as soon as possible. At least partlyin response to this policy stance, a substantial portion of the existingresearch on refugees in the United States has focused on refugee laborforce participation, rather than longer-term outcomes, such as earningsover time (Strand, 1984; Bach and Carroll-Seguin, 1986).

Studies that focus on the employment status and earnings of refu-gees have found that English language skills are an important determinantof employment status, but differ in whether education level upon arrival,gender, and age are important predictors (Strand, 1984; Chiswick, 1993;Waxman, 2001; Mamgain and Collins, 2003; Cortes, 2004). Some analy-ses have demonstrated that refugees start at a distinct disadvantagecompared to economic migrants but achieve parity in the probability ofemployment over time (Wooden, 1991). Very few studies have focusedon both refugee employment status and earnings as dependent variablesin the same study. One study that did focus simultaneously on these twodependent variables estimated earnings according to the mean earnings ofthe occupation of the refugee’s most recent job (Potocky-Tripodi, 2003),a method that almost certainly measures refugee earnings unreliably.

Social Capital in the Economic Lives of Refugees

Access to social capital helps to determine how refugees find employmentand the quality of the jobs that they find. This argument depends uponthe ability of social capital to act as a conduit that connects refugees toemployment. Scholars in the broader immigration literature argue thatestablished immigrants use their social networks to efficiently transferinformation about employment opportunities to more recent immigrants


(Portes, 1995). These established immigrants reduce the costs associatedwith recruiting labor for employers and reduce informational asymmetriesexperienced by recent immigrants, producing benefits for both parties(Bailey and Waldinger, 1991; Waldinger, 1994). As a result, immigrantswith access to social capital are likely to find a job faster and earn morethan immigrants without access to social capital.

Attempts to extend this finding to studies of the economic outcomesof refugees have been met with mixed success. Studies from Canada andthe United States that considered social capital as an independent variablefound that it had little effect on economic outcomes. For example,Montgomery (1996) estimated the effect of involvement in an ethnicsocial network on economic adaptation for refugees in Canada, aftercontrolling for typical demographic characteristics. His results failed tosupport the idea that participation in an ethnic network had any influenceon economic adaptation, though these results should be viewed withcaution given the non-random sample used in the analysis. Similarly,Potocky-Tripodi (2004) tested whether or not assistance of a social net-work helped refugees find employment and what effect this assistance hadon earnings after controlling for background characteristics. She foundthat assistance from a social network has a small, positive, and statisticallysignificant effect on employment status, but no significant effect on earn-ings. In contrast, other studies have found that the use of social capitalhas a positive effect on the economic adaptation of refugees. A study ofrecent refugees in Canada by Lamba (2003) found that refugees whoactively used co-ethnic social capital to help them find a job were signifi-cantly more likely to have higher quality employment compared to refu-gees who relied exclusively on their own efforts to find a job.

The divergent findings of these studies may be due to how eachresearcher measured social capital. For example, Montgomery measuredsocial capital as the extent of an individual’s involvement in an ethnicsocial network, rather than whether or not an individual used hisethnic social network for assistance in the labor market. Potocky-Tripodimeasured social capital by whether or not an individual used friends,family members, or compatriots to help him find his first job in theUnited States and used this dichotomous variable to estimate currentmonthly earnings, despite the fact that 98 percent of the sample had livedin the United States for longer than 5 years and had potentially workedin many jobs since their first entry point into the labor market. Finally,Lamba measured social capital by how refugees found their current jobs


and used this measure as an independent variable to estimate current jobquality. Of the three approaches, Lamba’s methods were likely to measuresocial capital and use it to estimate economic outcomes most effectively.

Earlier studies of the effect of social capital on the employmentstatus of refugees distinguish between co-ethnic social capital and socialcapital outside of the refugee group through refugee sponsorship status.Sponsorship was the original method of refugee resettlement in theUnited States, with refugees receiving one of two types of sponsorship:family ⁄ co-ethnic sponsorship or sponsorship outside of the refugee com-munity (Tran, 1991). Sponsors traditionally shouldered much of theburden involved with early resettlement challenges, including helping therefugee locate housing and employment (Lanphier, 1983). Studies thatused sponsorship status (or another variable approximating sponsorshipstatus) as an independent variable to predict economic adaptation foundthat refugees with non-refugee sponsors fared better than refugees withrefugee sponsors (Bach and Carroll-Seguin, 1986; Tran, 1991; Majka andMullan, 2002). Based on data from a national survey of South-East Asianrefugees in the early 1980s, Bach and Carroll-Seguin (1986) found thatfemale refugees who arrived after 1980 received a significant benefit totheir labor force participation rate from having non-refugee sponsors.They reasoned that co-ethnically sponsored female refugees may have beenunder pressure from an established refugee community to adhere to cul-tural norms and stay at home instead of work, resulting in lower laborforce participation. Analysis of the results from a national survey of Indo-chinese refugees in 1982 similarly found that refugees with non-refugeesponsorship experienced better economic adaptation than refugees withsponsorship from fellow refugees (Tran, 1991). A third study focused onrefugees from multiple countries of origin who resettled in Chicago andfound that by the late 1980s male and female refugees who had access tomainstream sources of support experienced better economic adaptationoutcomes than those who did not (Majka and Mullan, 2002).

Findings from all three of these studies suggest that ‘‘weak’’ ties out-side of the co-ethnic community provide an economic advantage for refu-gees. One interpretation of these studies supports traditional findings inthe literature that networks of weak social ties allow individuals to accesssuperior resources for attainment than networks of strong social ties (Lin,2001b). Another possible interpretation of these studies, perhaps mostconsistent with findings from Bach and Carroll-Seguin, is that the advan-tages provided by networks of strong, co-ethnic ties are offset by increased


reciprocal obligations and adherence to community norms that may con-strain involvement in the labor market.

Other qualitative studies that focus on how immigrants use socialcapital support this latter interpretation. For example, a study of Salvador-ans in San Francisco found that using co-ethnic social capital led to a vari-ety of outcomes, some positive and others negative, and was experienceddifferently by male and female immigrants (Menjivar, 2000). Specifically,research findings suggest that social capital may have positive effects onthe economic outcomes of male refugees and negative effects on the eco-nomic outcomes of female refugees, because of different expectationsregarding reciprocal obligations for men and women and because of com-munity norms that regulate the behavior of men and women differently.

The Current Study

This study reassesses the importance of sponsorship on refugee economicoutcomes. This reassessment is necessary because the notion of refugeesponsorship has changed dramatically since the 1980s. In the early years ofthe refugee resettlement program in the U.S. native born families,churches with non-Hispanic white congregations, and refugee resettlementorganizations sponsored refugees in addition to refugee families (Montero,1979). Traditionally, voluntary agencies (VOLAGS), which are the non-profit organizations that the U.S. government funds to assist refugeesduring the resettlement process, matched refugees who came to the UnitedStates with a sponsor who would be responsible for many of the economicand social needs initially experienced by a refugee (Holman, 1996). Asthe early studies reviewed in this paper have shown, native born familiesand non-Hispanic white congregations that sponsored refugees gave therefugees access to considerable advantages because of their personal andinstitutional networks that could connect refugees to employment.

Compared to the 1980s the resettlement system today is different intwo important ways. First, instead of native born families and faith-basedgroups, refugees already resettled in the United States sponsor the vastmajority of other refugees. Refugees who come to the United Statesbecause they are sponsored by a friend or family member are known asreunification cases. When reunification cases arrive in a resettlement city,they become a part of the established social network of their friend orfamily member, giving them immediate access to social capital. Thoserefugees who arrive without a friend or family member as a sponsor are


called free cases. Free case refugees have few if any existing social ties inthe refugee community of their resettlement city and therefore lack imme-diate access to social capital.

Second, in contrast to the earlier era when sponsors shoulderedmuch of the resettlement burden, VOLAGS provide extensive resettlementassistance for all reunification and free case refugees. The initial assistanceprovided by the VOLAG to recent refugees arriving in the United Statesis intensive and constant, regardless of whether the refugee is a reunifica-tion or a free case. Each newly arrived refugee or refugee household ispaired with a case manager at the VOLAG who assists them with theirinitial navigation of the city, including assigning them to temporary hous-ing, filling out appropriate paperwork, arranging a medical checkup, andbeginning to search for employment. Many VOLAGS have separateemployment specialists who match recently arrived refugees to availablejobs. Some VOLAGS may also use institutional networks with churchesor mentor networks to connect newly resettled refugees to the native borncommunity. In the case of Portland, Maine, where this research occurred,the VOLAG had disbanded its volunteer mentorship program and didnot use church-based networks since so many of the refugees who reset-tled in Portland were Muslim.

At the same time, reunification case refugees also rely on theirco-ethnic sponsors for assistance during resettlement. The decision tosponsor a friend or family member as a reunification case is significantfor refugees who are already established in the United States because itinvolves a large time commitment to go through the sponsorship pro-cess. In addition, refugees who sponsor reunification cases are expectedto assist their friend or family member as they adjust to life in theUnited States. Typically the VOLAG requests that a sponsor of a reuni-fication case assist with specific tasks, such as offering transportation andlooking for permanent housing or a job, all clear examples of social cap-ital. Reunification case sponsors are not legally required to provide anyof this assistance, but most help refugees in numerous ways. In otherwords, reunification case refugees actually have two sources of support:co-ethnic social capital from their sponsor and resources and servicesfrom a VOLAG. Free case refugees only have the support of theVOLAG. Without immediate access to co-ethnic social capital whenthey arrive in the United States, free case refugees may be missing animportant source of social, and potentially economic, support (Portesand Rumbaut, 2006).


In contrast to existing studies that use sponsorship status as a predic-tor of employment status for refugees, this study uses qualitative and quan-titative data to determine the effect of sponsorship status on the earningsof male and female refugees. I paired findings from prior studies with myown insights from data from interviews with refugees to develop a set ofresearch hypotheses regarding the effect of sponsorship status on the earn-ings of male and female refugees. I then used a unique data set to modelthe earnings of male and female refugees and test these hypotheses. Giventhe kinds of informal support that co-ethnic sponsors can provide recentlyarrived refugees, I use sponsorship status as a proxy for access to co-ethnicsocial capital in these models.

QUALITATIVE DATA AND RESEARCH HYPOTHESES

Given the significant changes to the refugee resettlement system in theUnited States that have occurred since most of the studies that examinedthe role of sponsorship on employment outcomes, I relied on qualitativedata that I collected in Portland to help generate research hypotheses.Due to time and resource constraints, I limited the qualitative portion ofmy study to interviews of Sudanese and Somali refugees living in Port-land. When choosing which refugees to interview, I snowball sampledfrom multiple entry points, including religious institutions, English lan-guage classes, and social service agencies. Snowball sampling increasesaccess to hard-to-reach populations and individuals who are unlikely torespond positively to a ‘‘cold call.’’ This sampling method resulted in asample of 42 refugees that varied by country of origin (38 percent Somaliand 62 percent Sudanese), gender (33 percent female), and sponsorshipstatus (43 percent free cases and 57 percent reunification cases).

Part of my interviews focused on how respondents used social capi-tal in their lives. Interviews revealed that reunification case refugees weresubjected to the social norms and reciprocity that help to define socialcapital to a greater degree than free case refugees (Portes and Sensenbren-ner, 1993). For example, Somali women in Portland tend to wear thetraditional long and flowing clothing common in Somalia (McMichael,2002). Two of the largest and most successful manufacturing businessesin Portland pay relatively high wages and hire large numbers of refugees,but have strict guidelines that require production employees to wearclothing that will not become entangled in equipment, potentially causingpersonal injury to the employee. To be eligible to work at these jobs,


Somali women would have to reject their traditional clothing and adoptWestern-style dress that reveals the contours of their bodies, somethingthat is forbidden in a public setting by some interpretations of theQur’an.

In addition to the personal question of faith and adherence to reli-gious principles that adopting Western-style dress represents for Somaliwomen, they also face social pressure from the Portland Somali commu-nity to maintain their traditional dress. Since, by definition, reunificationcase Somali female refugees have more social ties and likely receive moreassistance from social ties in the Somali community than their free casecounterparts, they experience more of this pressure and stand to lose morefrom the stigma and ostracism that might follow a decision to changetheir dress and work at one of these businesses. I observed that, comparedto free case Somali female refugees, reunification case Somali female refu-gees were less likely to stray from important customs and had less accessto well-paying jobs as a result.

Co-ethnic social ties frequently expected Somali and Sudanese reuni-fication case female refugees who received assistance when they arrived inPortland to reciprocate this assistance. Many of the female refugees whoreceived assistance reciprocated by assuming family or extended familyresponsibilities that limited their ability to work in the labor market(Portes and Sensenbrenner, 1993). Reunification case female refugeesoften helped close friends and extended family members with informaland temporary child care and a variety of household chores, includingcooking and cleaning. They provided this help because their social tiesrequested their assistance, but also as a part of an unspoken traditionimported from their countries of origin. In formal interviews and informaldiscussions, these reunification case female refugees discussed how thistype of labor pooling was a common way for women in Somalia andSudan to ease the burden of their tasks and socialize simultaneously. Incomparison with reunification case female refugees, free case female refu-gees had fewer reciprocal obligations because they had fewer social tieswhen they arrived in Portland. Data from interviews suggest that, overtime, free case female refugees increased the size of their social networksin Portland, but their social networks rarely approached the size or inti-macy of the social networks of reunification case female refugees. There-fore, reunification case female refugees felt these reciprocal obligationsmore acutely than their free case counterparts and spent more time out-side of the labor market as a result.


In contrast to female refugees, male refugees face a different set ofsocial norms and reciprocal obligations. For example, social norms withinthe Somali and Sudanese refugee communities in Portland frown on mennot participating in the labor market. My interviews with male refugeesindicate that they expect and feel social pressure to arrange financial sup-port for their families and extended families to a greater extent thanfemales. There was widespread agreement that women should enter thelabor force since it was nearly impossible for households to get by in Port-land on the earnings of males alone. However, social pressure limitedsome women’s work to the minimum amount necessary to make endsmeet.

In addition, the co-ethnics who assisted male refugees when theyarrive in Portland sometimes expected reciprocal obligations that involvedfinancial support, including pooling financial resources to pay bills, invest-ing in start-up businesses, or remitting money to friends and relatives stillliving abroad. These norms and reciprocal obligations have the effect ofencouraging labor market participation for male refugees, but, because oftheir initial connections to co-ethnics when they arrive in Portland, reuni-fication case male refugees seemed to feel pressure to conform to culturalnorms more than free case male refugees. Similarly, reunification casemale refugees may have more reciprocal obligations than their freecase counterparts. As a result, compared to free case males, reunificationcase male refugees felt a greater need to conform to cultural norms andrise to obligations that ultimately encouraged labor market activity.Though my qualitative data did not include interviews of Eastern Euro-pean refugees, a group that has resettled in Portland in relatively largenumbers, I take these findings from the qualitative portion of my studyto be suggestive of how social capital has different consequences forfemale and male refugees.

This analysis of qualitative data suggests three research hypothesesregarding the effect of sponsorship status on the earnings of resettled refu-gees. Each hypothesis treats sponsorship status as a proxy for access toco-ethnic social capital upon arrival.

Hypothesis 1: Male and female reunification case refugees will earn morein their first year of work than male and female free case refugees.Hypothesis 2: In their most recent year of work, male reunification caserefugees will earn more than free case male refugees.


Hypothesis 3: In their most recent year of work, female reunification caserefugees will earn less than free case female refugees.

The first hypothesis suggests that, regardless of gender, reunificationcase refugees have an initial advantage in the labor market over free caserefugees. Initial support from co-ethnics combined with the support fromthe VOLAG is likely to help reunification case refugees find better payingwork faster than free case refugees who rely solely on the VOLAG forassistance. The next two hypotheses are consistent with the idea that overtime male and female refugees experience co-ethnic social capital differ-ently, with females experiencing a negative effect and males experiencing apositive effect. Female reunification case refugees will experience moresocial constraints on their labor market activity and earn less than femalefree case refugees. Compared to male free case refugees, male reunificationcase refugees will experience more intense social pressure to engage inlabor market activity and they will earn more than male free caserefugees.

RESETTLEMENT CONTEXT

Portland is one of many metropolitan areas where resettled refugees arean important component of the foreign born population. Table 1 lists 10metropolitan areas where refugees account for more than 40 percent ofthe recently arrived foreign born population. The metropolitan areas onthis list are strikingly similar in size and geographic location. With theexception of Spokane, WA, none of the metropolitan areas have morethan 300,000 people and all are located in the Midwest or Northeast.Among this group of metropolitan areas, the economic conditions inPortland make it stand out from the rest. In 2000 only one metropolitanarea had a lower unemployment rate or higher median household earn-ings. Portland’s natural deep water port and thriving tourism industryprobably account for much of its economic health. According to the U.S.Army Corps of Engineers, in 2006 the Port of Portland was the largesttonnage port in New England and the largest oil port on the East Coast.Port officials estimate that the economic activity associated with the port(including a large number of cruise ships that bring tourists to thearea) contributes half a billion dollars to the Maine economy each year(Monroe, 2004), with much of this sum undoubtedly concentrated in thePortland metropolitan area economy.


Portland’s healthy economy is one factor that prompted theNational Conference of Catholic Bishops, one of the eight activeVOLAGS in the United States, to begin resettling refugees there in theearly 1980s. Figures from the federal Office of Refugee Resettlementshow that almost 4,000 refugees resettled in Portland between 1983 and2004. As Table 1 indicates, of the foreign born population arriving inPortland between 1990 and 2000, almost half came to the United Statesas a refugee. None of these figures reflect a significant population of sec-ondary migrants, those refugees who originally resettled in a differentcity in the United States and then decided to move to Portland. Reset-tlement officials in Portland estimate that the recent secondary migrantpopulation in Portland contributes at least 2,000 additional refugees tothe total number recognized by official data. Therefore, refugees proba-bly represent close to 10 percent of the 64,000 people living in Portlandtoday.

QUANTITATIVE DATA

The sample of refugees for this analysis came from the administrativefiles of Catholic Charities Maine Refugee and Immigrant Services(CCMRIS) in Portland, Maine. While most states have several

TABLE 1DEMOGRAPHIC CHARACTERISTICS AND ECONOMIC CONDITIONS FOR SELECT U.S. METROPOLITAN AREAS

WITH SIGNIFICANT REFUGEE POPULATIONS

Metro AreaPopulation

(2000)

Foreign-BornPopulation

Arrived1990–2000

No.RefugeesResettled

1990–2000UnemploymentRate (%) (2000)

MedianHouseholdEarnings

($) (1999)

Utica-Rome, NY 299,896 7,013 6,084 6.3 35,292Fargo-Moorhead,ND-MN

174,367 3,572 2,718 4.5 38,069

Erie, PA 280,843 3,992 2,969 5.8 36,627Sioux Falls, SD 172,412 4,391 2,684 2.8 43,387Binghamton, NY 252,320 4,760 2,601 5.3 36,374Spokane, WA 417,939 9,131 4,466 7.9 37,308Portland, ME 243,544 3,888 1,871 3.6 44,707Lincoln, NE 250,291 9,398 4,131 3.6 41,850Waterloo-CedarFalls, IA

128,012 3,307 1,397 4.8 37,266

Burlington, VT 169,391 4,559 1,803 3.9 46,732

Note: Adapted from Appendix A in ‘‘From ‘There’ to ‘Here’: Refugee Resettlement in Metropolitan America,’’ byAudrey Singer and Jill Wilson (2006), Brookings Institution Living Cities Census Series, and Census 2000(Summary File 3).


organizations that resettle refugees, CCMRIS is the only functioningrefugee resettlement organization in the state of Maine. Those includedin the sample are all refugee clients who used the services at CCMRISbetween January 1, 1998, and December 31, 2004, and were at least18 years old as of September 1, 2005. Since the cultural contexts of acountry of origin may influence how refugees use social capital in theUnited States, the sample was limited to individuals from EasternEurope (former Soviet Republics or the former Yugoslavia), Somalia,and Sudan to create a more focused comparison. There were 938 refu-gees who met these selection criteria. This sample included refugeeswho were resettled in Portland by CCMRIS as well as secondarymigrants.

Employment and earnings data came from the Maine Departmentof Labor (MDOL). Each quarter, most employers in Maine submit a listof their employees along with the pre-tax quarterly earnings of eachemployee to the Bureau of Unemployment Compensation’s tax division.These individual-level data from this source span a time period betweenthe first quarter of 1998 and the third quarter of 2005 (the last quarter ofdata available at the time of this analysis). These data do not includeinformation about self-employment or informal employment, thoughavailable evidence from interviews with refugees suggests that these formsof employment are not major sources of employment for refugees in Port-land. On the other hand, it is possible that refugees were hesitant to men-tion earnings derived from the informal economy in interviews if theybelieved it might cause trouble for them should the resettlement agencyor Internal Revenue Service find out.

The MDOL data do indicate, for each quarter in this time period,whether a refugee formally worked in Maine (an employment occur-rence), how many employment occurrences a refugee had in eachquarter, the corresponding industry of each employment occurrence, andhow much a refugee earned at each employment occurrence. Whetheran employment occurrence was part-time or full-time work and thenumber of hours worked per week are not available from this datasource. Connecting the data from CCMRIS and MDOL using thesocial security numbers of those in the sample yielded a data set withrich demographic information on the adult refugee population in


Portland, as well as a comprehensive work history in Maine for eachadult refugee who worked in the state.2

Dependent Variables

This study considers two dependent variables. The first dependent variableis the total, inflation-adjusted earnings of a refugee in his first year ofworking in Maine. I calculated this variable by adding earnings from allof the jobs that a refugee held in Maine during the four quarters immedi-ately following his or her first quarter of work in Maine. The seconddependent variable is the total inflation-adjusted earnings of a refugee inhis most recent year of working in Maine. I calculated this variable byadding a refugee’s earnings in his or her most recent quarter of work tohis or her earnings in the previous three quarters. Both variables were cal-culated in constant 2005 dollars. Calculating the variables in this mannerensured that each refugee’s earnings were based on an equal amount oftime in the labor force.

Independent Variables

Independent variables included in this study fall into one of five areas:demographic characteristics, human capital characteristics, macroeconomicfactors, work experience, and access to social capital. Table 2 lists anddefines the independent variables used in this analysis.

Overview of the Sample

A significant number of observations in the data had missing values forlevel of English (19 percent of cases had missing values), educationalattainment (12 percent of cases had missing values), and family size (46percent of cases had missing values). I used the multiple imputation (MI)

2I received approvals from the Institutional Review Boards (IRBs) at my university and

CCMRIS as a part of this research project. Part of this approval process resulted in CCM-RIS granting me a waiver of informed consent for the use of client Social Security Num-bers (SSNs) in this research. After using the SSNs to match each client to their labor

market histories in Maine, I deleted all identifying information, including SSNs, from thedata set. The confidentiality protocols in place ensured that the risks for any refugeeincluded in this data set were minimal, probably no greater than the risks faced in every-

day life.


technique to create 10 complete data sets with imputed values for themissing data (King et al., 2001). Given the amount of missing data, MIappears to be the best statistical technique to ensure a useable data set.Other methods of dealing with the missing data (e.g., list-wise deletion orimputing the mean) either throw away information or create bias in thedata set (Little and Rubin, 1987). MI works by replacing missing observa-tions with a value selected at random from a set of plausible values. Thistechnique creates multiple complete sets of data, analyzes the data usingstandard statistical procedures, and then combines the results from theseanalyses. The MI strategy for imputing missing data represents the uncer-tainty represented by the missing data more accurately than other imputa-tion techniques, but still provides efficient and robust results (King et al.,

TABLE 2LIST AND DEFINITION OF INDEPENDENT VARIABLES USED IN THE ANALYSIS

Variable Definition

DemographicAge Refugee’s age in years (September 1, 2005, minus refugee’s date of

birth)Age_Sq Refugee’s age squaredMinor_ME 1 = refugee was a minor upon arrival in Portland, 0 = otherwiseOrigin Dummy variable for region of origin: Eastern Europe (omitted),

Sudan, SomaliaHuman capital

English Dummy variable for English speaking proficiency upon arrival inPortland: Yes (Good or Fair English), No (Poor or No English).No is omitted

Education Dummy variable for educational attainment upon arrival in Portland:High (Higher or Secondary education), Low (Primary or No

education). High is omittedMacroeconomic

First_U Maine average statewide unemployment rate during a refugee’s firstyear of work

Recent_U Maine average statewide unemployment rate during a refugee’s mostrecent year of work

ExperienceQtrs_Work Total number of quarters in Maine that a refugee workedQtrs_ME Total number of quarters that a refugee lived in Maine

Social capitalMigrant 1 = refugee was resettled in Portland originally, 0 = refugee moved to

Portland as a secondary migrant or an asyleeSponsor 1 = reunification case, 0 = free case

OtherRelation 1 = primary applicant for refugee resettlement, 0 = otherwiseWork_First 1 = had earnings in the four quarters after the first quarter of working

in Maine, 0 = otherwiseWork_Recent 1 = had earnings in the three quarters prior to the most recent quarter

worked in Maine, 0 = otherwise


2001). Some research suggests that as few as five imputed data sets areenough to ensure efficient and robust results (Horton and Kleinman,2007), but I created 10 imputed data sets to ensure the validity of myresults. Table 3 describes key characteristics of the sample based on asummary of the imputed data sets.

Economic Experiences of Refugees in Portland

In their most recent job in Portland, the majority of refugees in the sam-ple worked in one of five industries. By far the most popular industry wasthe administrative services industry, which employed about 20 percent of

TABLE 3CHARACTERISTICS OF THE IMPUTED SAMPLE, BY GENDER

Total Male Female

N 938 526 412Age (mean, SD) 36.0, 13.0 35.5, 12.7 36.6, 13.3Minor_ME 71 (8%) 42 (8%) 29 (7%)Gender

Male 526 (56%) – –Female 412 (44%) – –

Family_Size (mean, SD) 2.6, 1.7 2.4, 1.7 2.7, 1.8Origin

Eastern Europe 389 (41%) 208 (40%) 181 (44%)Sudan 315 (34%) 181 (34%) 134 (33%)Somalia 234 (25%) 137 (26%) 97 (24%)

Educationa

High (Higher or Secondary) 554 (59%) 347 (66%) 207 (50%)Low (Primary or None) 384 (41%) 179 (34%) 205 (50%)

Englisha

Yes (Good or Fair) 393 (42%) 255 (48%) 138 (33%)No (Poor or None) 545 (58%) 271 (52%) 274 (66%)

Relationa

Principal Applicant 512 (55%) 338 (64%) 174 (42%)Not Principal Applicant 426 (45%) 188 (36%) 238 (58%)

Sponsora

Reunification Case 673 (72%) 362 (69%) 311 (75%)Free Case 265 (28%) 164 (31%) 101 (25%)

Migranta

General 631 (85%) 402 (76%) 344 (83%)Secondary Migrant or Asylee 112 (15%) 124 (24%) 68 (17%)

Qtrs_ME (mean, SD) 19.2, 7.5 19.3, 7.3 19.2, 7.7Qtrs_Work (mean, SD) 15.7, 7.2 15.5, 7.2 16.0, 7.3Work_Firsta 664 (71%) 395 (75%) 269 (65%)Work_Recenta 679 (72%) 406 (77%) 273 (66%)Earn_First (mean, SD)a $19,190, $11,736 $20,287, $12,150 $17,580, $10,925Earn_Recent (mean, SD)a $21,231, $14,484 $22,143, $15,232 $19,875, $13,204

Notes: Administrative data from Catholic Charities Maine and the Maine Department of Labor.aStatistically significant difference between males and females at the 95% confidence level.


refugees in their most recent job. Most of the refugees working in thisindustry were employed by a temporary help services business, whichplaces individuals in a variety of industries for temporary periods of time.The refugee employment concentration in this industry was followed bysocial assistance, hospitals, and educational services (14 percent); manufac-turing (13 percent); accommodation and food services (11 percent); andwholesale trade (6 percent). Male refugees were disproportionately repre-sented in the manufacturing industry, while female refugees were dispro-portionately represented in the accommodation and food servicesindustry. Concentrations of these magnitudes suggest the emergence ofethnic niches, where individuals from a particular ethnic group becomeoverrepresented in a given industry due to network hiring practices(Waldinger, 1994). In comparison, in 2004 almost half of the total workforce in Portland worked in a service industry. Significant portions ofPortland’s work force were concentrated in ‘‘white collar’’ industries, suchas finance, insurance and real estate, and professional services, where fewrefugees held jobs.

Refugees earned significantly less than typical workers in Portland,but their earnings improved dramatically over time. Among refugees inthe sample who worked in Maine, the median refugee worked in his firstjob in 2001 and in his most recent job in 2004. The average, inflation-adjusted monthly earnings (2005 dollars) of a refugee during his first yearof work was $1,570, compared to $1,790 in his most recent year of work(an increase of 14 percent). In contrast, the Maine Department of Laborestimates that the average, inflation-adjusted monthly earnings (2005 dol-lars) of a typical worker in Portland in 2001 was $3,146, compared to$3,195 in 2004 (an increase of almost 2 percent) (LMIS, 2006). Thoughrefugees experienced impressive earnings gains compared to typicalworkers in Portland, they still earned about 45 percent less than typicalworkers.

QUANTITATIVE METHODS

This paper estimates two models, both of which follow the same essentialform: a refugee’s earnings in Portland are a function of demographic char-acteristics, human capital characteristics, macroeconomic conditions, workexperience, and access to social capital. Estimating this model through theordinary least squares method requires a randomly selected sample. Unfor-tunately, the sample under consideration is not randomly selected.


Instead, only those refugees who entered the labor market have earningsand are observed. Since it is not possible to observe a counterfactual inthis case (what the earnings of a refugee who did not enter the labor mar-ket would be had he entered the labor market), it is likely that an estima-tion of refugee earnings using ordinary least squares may unwittinglyinclude unobservable characteristics of the refugees who worked. In otherwords, there is a case of ‘‘selection on the unobservables’’ (Winship andMorgan, 1999).

It is possible to correct for the sample-selection bias problem in thiscase with a two-step estimation procedure developed by economist JamesHeckman (1979). This technique uses a derived term called the inverseMill’s ratio to correct for sample-selection bias when estimating earnings.Specifically, the inverse Mill’s ratio is ‘‘a monotone decreasing function ofthe probability that an observation is selected into the sample’’ (Heckman,1979:156). The first step of the technique uses a probit model to estimatethe probability that a refugee worked in Maine and derives the inverseMill’s ratio that reflects this probability. The second step of the techniqueestimates refugee earnings, using the inverse Mill’s ratio as one of manyindependent variables. The general form of the model used in the secondstep of this analysis is:

lnðYijÞ ¼ aþ bXij þ binvMillsij þ eij

where ln(Yij) is the natural log of earnings for individual i of sex j oversome period of time, a is constant term, b represents the coefficients tobe estimated, Xij is a vector of independent variables for individual i ofsex j, invMillsij is the inverse Mill’s ratio for individual i of sex j, and eij

is a randomly distributed error term.This paper specifies and estimates two series of models, correspond-

ing to the natural log of earnings in the first and most recent years ofwork for refugees. The base model in each series uses demographic andhuman capital variables to estimate the dependent variable. Subsequentmodels add a variable that controls for macroeconomic conditions in thestate, work experience variables (when appropriate), and social capital vari-ables. Each model includes the inverse Mill’s ratio as a predictor. Maleand female refugees are analyzed separately in each series of models, sinceoverwhelming evidence suggests that gender is an important determinantof earnings for individuals in the U.S. labor market.


Only summarized regression coefficients, standard errors, and good-ness of fit statistics from the imputed data sets are presented in this paper(King et al., 2001). The results from the estimation of the probit modelsused to derive the inverse Mill’s ratio are included in Appendix A. Theseprobit models estimate the probability that male and female refugeesworked in Maine as a function of age and age squared; the year of arrivalin Maine; whether or not the refugee was the primary applicant for refu-gee status; and family size upon arrival to Maine. Though it expands onthe methods and data used, the basis of this methodology is similar tothat used by Mamgain and Collins (2003) in their study of refugeeeconomic outcomes in Portland, Maine.

Multicollinearity is a concern when using the Heckman correctiontechnique, but typically only if the probit and the estimation regressionmodels use identical or nearly identical sets of independent variables.When the models share some, but not all, of the same independent vari-ables, multicollinearity is less of a concern. A review of variance inflationfactors (VIF) for all variables used in the models (including the inverseMills ratio) confirms this position. With the exception of age and age-squared, none of the variables have a VIF over 3.5 (a conservative rule ofthumb is a VIF value of 4 or higher reflects a multicollinearity problem).Predictably, age and age-squared have very high VIF values, but in thiscase the value of capturing the nonlinear relationship between age andearnings outweighs the multicollinearity problem.

The real challenge to using the Heckman correction technique effec-tively is to identify a variable that belongs to the probit model, but not tothe estimation regression model. When the probit model predicts entry tothe labor force and the regression model predicts earnings, one typicalindependent variable to use in the probit model is family size. Family sizeshould help to determine whether someone enters the labor force and, ifthe earnings are pre-tax earnings, should not be a significant predictor ofearnings (Puhani, 2000).

QUANTITATIVE RESULTS

Refugee Earnings in the First Year of Work

Table 4 presents results from a series of models that estimate the naturallog of earnings during the first year of work for female and male refugeesin the sample. Since this represents the first work experience of the


TA

BL

E4

MU

LT

IPL

ER

EG

RE

SSIO

NR

ESU

LT

SF

OR

FIR

ST-Y

EA

RE

AR

NIN

GS

M1

M2

M3

M4

M5

M6

M7

M8

Inte

rcep

t6

.17

02

**(0

.75

54

)6

.40

70

**(0

.82

32

)6

.32

31

**(0

.83

01

)5

.91

19

**(0

.70

79

)6

.50

64

**(0

.49

85

)6

.52

63

**(0

.60

43

)6

.44

20

**(0

.61

17

)6

.29

34

**(0

.48

01

)A

ge0

.16

87

**(0

.03

75

)0

.17

07

**(0

.03

78

)0

.17

13

**(0

.03

78

)0

.18

03

**(0

.03

59

)0

.14

86

**(0

.02

40

)0

.14

85

**(0

.02

40

)0

.14

89

**(0

.02

41

)0

.15

61

**(0

.02

30

)A

ge_

sq)

0.0

02

0**

(0.0

00

5)

)0

.00

21

**(0

.00

05

))

0.0

02

1**

(0.0

00

5)

)0

.00

22

**(0

.00

05

))

0.0

01

6**

(0.0

00

3)

)0

.00

16

**(0

.00

03

))

0.0

01

6**

(0.0

00

3)

)0

.00

17

**(0

.00

03

)M

inor_

ME

)1

.10

80

**(0

.22

80

))

1.1

01

**(0

.22

78

))

1.0

91

1**

(0.2

28

1)

)1

.01

39

**(0

.22

31

))

1.0

28

9**

(0.1

70

7)

)1

.02

95

**(0

.17

09

))

1.0

19

2**

(0.1

70

9)

)1

.00

47

**(0

.15

99

)E

ngl

ish

.Yes

0.1

68

0(0

.12

15

)0

.16

61

(0.1

21

7)

0.1

75

7(0

.12

24

))

0.0

19

8(0

.09

88

))

0.0

19

9(0

.09

31

))

0.0

06

1(0

.09

47

)E

du

cati

on

.Low

)0

.04

03

(0.1

50

2)

)0

.04

94

(0.1

51

5)

)0

.06

16

(0.1

56

5)

)0

.06

04

(0.0

98

8)

)0

.06

12

(0.0

99

6)

)0

.08

36

(0.1

02

0)

Ori

gin

.Som

alia

)0

.19

63

(0.1

40

3)

)0

.20

00

(0.1

40

2)

)0

.17

30

(0.1

58

9)

)0

.18

02

�

(0.1

05

8)

)0

.18

07�

(0.1

06

1)

)0

.15

09

(0.1

11

7)

Ori

gin

.Su

dan

)0

.18

25

(0.1

31

6)

)0

.18

16

(0.1

31

6)

)0

.17

69

(0.1

33

3)

)0

.02

95

(0.1

01

0)

)0

.02

97

(0.1

01

1)

)0

.02

35

(0.1

01

9)

Fir

st_

U)

0.0

70

6(0

.09

89

))

0.0

77

0(0

.10

53

))

0.0

04

2(0

.07

51

))

0.0

11

0(0

.07

58

)M

igra

nt

0.0

55

9(0

.14

63

)0

.08

46

(0.1

00

1)

Sp

onso

r0

.05

90

(0.1

10

0)

0.0

33

0(0

.07

79

)in

vMil

lsR

atio

0.6

68

2**

(0.2

15

7)

0.7

03

9**

(0.2

21

9)

0.6

92

0**

(0.2

27

7)

0.5

59

2**

(0.2

11

0)

0.6

66

5**

(0.2

35

9)

0.6

66

5**

(0.2

35

9)

0.6

88

9**

(0.2

39

8)

0.6

02

4**

(0.2

21

6)

R2

0.3

13

50

.31

50

0.3

16

30

.29

35

0.4

08

50

.40

85

0.4

10

00

.40

07

Ad

j-R

20

.29

24

0.2

91

20

.28

71

0.2

82

80

.39

63

0.3

94

70

.39

31

0.3

94

5

Not

es:

Par

amet

eres

tim

ates

,ap

pro

xim

ate

p-v

alu

es,

and

asso

ciat

edgo

odn

ess-

of-fi

tst

atis

tics

for

an

este

dta

xon

omy

ofO

LS-

fitt

edre

gres

sion

mod

els

that

des

crib

eth

ere

lati

on

ship

bet

wee

nlo

gof

fem

ale

(N=

269)

and

mal

e(N

=395)

refu

gee

earn

ings

inth

eir

firs

tye

arof

wor

kin

Por

tlan

d,

Mai

ne,

and

dem

ogra

ph

ican

dh

um

anca

pit

alch

arac

teri

stic

s;m

acro

econ

omic

con

dit

ion

sin

Mai

ne;

and

acce

ssto

co-e

thn

icso

cial

cap

ital

.M

odel

s1–4

corr

esp

ond

tofe

mal

ere

fuge

esan

dM

odel

s5–8

corr

esp

ond

tom

ale

refu

gees

.�p

<0.1

0;

*p<

0.0

5;

**p

<0.0

1(s

tan

dar

der

rors

are

inp

aren

thes

es).


refugees in Maine, independent variables that control for work experiencein Maine are not used as predictors in these models. Overall, a refugee’sage played the most significant role in determining his or her earnings inthe first year of work in Maine. Interpreting age and age squared together,an additional year in age was associated with a 17.6 percent increase inearnings for female refugees (Model 4) and a 15.3 percent increase inearnings for male refugees (Model 8). Since those refugees who arrivedwhile they were still minors probably still lack the education and workexperience to make much of a difference in the labor market, arriving inPortland as a minor was associated with substantially lower earnings inthe first year of work for both female and male refugees. The ability tospeak English and educational attainment were not statistically significantpredictors of first year earnings for either female or male refugees. Spon-sorship did not have a statistically significant effect on the first year earn-ings of female or male refugees.

Refugee Earnings in the Most Recent Year of Work

Table 5 summarizes findings from a series of models that estimate thenatural log of earnings during the most recent year of work for femaleand male refugees in the sample. In contrast to their first year of work inMaine, by their most recent year of work in Maine, refugees have usuallyaccumulated experiences in the local community and economy that mayhelp to determine earnings. Therefore, some of the models in this tableinclude independent variables that control for work experience and timespent in Portland. The age of refugees played an important role in deter-mining their earnings in their most recent year of work, but work experi-ence was also an important factor in determining earnings in thesemodels. The ability to speak English and educational attainment were notstatistically significant predictors of most recent year earnings for eitherfemale or male refugees. Sponsorship had a negative effect on female refu-gee earnings in their most recent year of work in Maine. Though it wasnot statistically significant at the 0.05 level, sponsorship had a positiveeffect on male refugee earnings in their most recent year of work inMaine.

The inclusion of temporal variables in Model 11 resulted in achange to the sign of the coefficient and statistical significance of the aver-age statewide unemployment rate in a female refugee’s most recent year ofwork. The change in level of statistical significance and the sign and size


TA

BL

E5

MU

LT

IPL

ER

EG

RE

SSIO

NR

ESU

LT

SF

OR

MO

ST

RE

CE

NT

YE

AR

EA

RN

ING

S

M9

M10

M11

M12

M13

M14

M15

M16

M17

M18

Inte

rcep

t6.0

298**

(0.8

661)

5.2

437**

(1.3

267)

10.4

015**

(1.4

237)

10.8

248**

(1.4

425)

10.6

618**

(1.3

850)

7.5

416**

(0.6

025)

6.8

913**

(0.9

397)

11.9

689**

(0.9

868)

11.4

639**

(0.9

874)

11.2

877**

(0.9

840)

Age

0.1

891**

(0.0

428)

0.1

890**

(0.0

428)

0.1

262**

(0.0

388)

0.1

256**

(0.0

387)

0.1

188**

(0.0

347)

0.1

133**

(0.0

293)

0.1

131**

(0.0

293)

0.0

630*

(0.0

261)

0.0

655*

(0.0

258)

0.0

836**

(0.0

232)

Age

_sq

)0.0

023**

(0.0

005)

)0.0

023**

(0.0

005)

)0.0

015**

(0.0

005)

)0.0

015**

(0.0

005)

)0.0

015**

(0.0

004)

)0.0

012**

(0.0

004)

)0.0

012**

(0.0

004)

)0.0

006

�

(0.0

003)

)0.0

006*

(0.0

003)

)0.0

009*

(0.0

003)

Min

or_M

E)

0.2

975

(0.2

830)

)0.3

100

(0.2

832)

0.2

460

(0.2

627)

0.2

803

(0.2

612)

)0.7

590**

(0.2

201)

)0.7

678**

(0.2

200)

)0.2

314

(0.2

049)

)0.2

023

(0.2

029)

En

glis

h.Y

es)

0.0

001

(0.1

454)

)0.0

045

(0.1

454)

)0.0

022

(0.1

300)

)0.0

098

(0.1

317)

)0.0

223

(0.1

166)

)0.0

204

(0.1

161)

)0.0

017

(0.1

033)

0.0

499

(0.1

037)

Ed

uca

tion

.Low

0.0

404

(0.1

693)

0.0

433

(0.1

689)

)0.0

473

(0.1

469)

)0.0

496

(0.1

481)

)0.0

367

(0.1

317)

)0.0

316

(0.1

315)

0.0

029

(0.1

206)

)0.0

836

(0.1

232)

Ori

gin

.Som

alia

)0.3

557*

(0.2

097)

)0.3

621*

(0.1

730)

)0.0

916

(0.1

541)

)0.0

813

(0.1

734)

)0.6

106**

(0.1

412)

)0.6

202**

(0.1

415)

)0.3

911**

(0.1

296)

)0.2

655�

(0.1

361)

)0.2

673*

(0.1

244)

Ori

gin

.Su

dan

)0.2

244

(0.1

591)

)0.2

368

(0.1

596)

0.0

483

(0.1

441)

0.0

428

(0.1

444)

)0.3

122*

(0.1

293)

)0.3

284**

(0.1

304)

)0.2

907*

(0.1

214)

)0.2

395*

(0.1

213)

)0.2

326*

(0.1

100)

Rec

ent_

U0.1

743

(0.2

231)

)0.7

795**

(0.2

215)

)0.8

022**

(0.2

206)

)0.7

536**

(0.2

153)

0.1

440

(0.1

608)

)0.6

929**

(0.1

632)

)0.7

184**

(0.1

620)

)0.7

457**

(0.1

603)

Qtr

s_M

E)

0.0

656**

(0.0

142)

)0.0

735**

(0.0

147)

)0.0

674**

(0.0

135)

)0.0

874**

(0.0

106)

)0.0

811**

(0.0

107)

)0.0

855**

(0.0

101)

Tim

e_W

ork

0.0

995**

(0.0

107)

0.1

011**

(0.0

106)

0.0

988**

(0.0

101)

0.0

944**

(0.0

092)

0.0

940**

(0.0

091)

0.0

967**

(0.0

089)

Mig

ran

t0.0

289

(0.1

575)

0.3

104**

(0.1

171)

0.2

737**

(0.1

108)

Spon

sor

)0.2

493*

(0.1

221)

)0.2

394*

(0.1

218)

0.1

793�

(0.0

942)

0.1

680�

(0.0

940)

invM

ills

Rat

io0.4

220

(0.2

643)

0.4

102

(0.2

643)

0.4

550

(0.2

897)

0.4

619

(0.2

913)

0.4

772

(0.2

872)

0.2

875

(0.2

801)

0.2

691

(0.2

799)

)0.0

538

(0.2

662)

0.1

086

(0.2

691)

0.1

304

(0.2

701)

R2

0.1

543

0.1

562

0.3

626

0.3

724

0.3

655

0.2

358

0.2

373

0.3

994

0.4

156

0.4

117

Ad

j-R

20.1

287

0.1

273

0.3

358

0.3

409

0.3

488

0.2

203

0.2

199

0.3

826

0.3

963

0.3

969

Not

es:

Par

amet

eres

tim

ates

,ap

pro

xim

ate

p-v

alu

es,

and

asso

ciat

edgo

odn

ess-

of-fi

tst

atis

tics

for

an

este

dta

xon

omy

ofO

LS-

fitt

edre

gres

sion

mod

els

that

des

crib

eth

ere

lati

onsh

ipbet

wee

nth

elo

gof

fem

ale

(N=

273)

and

mal

e(N

=406)

refu

gee

earn

ings

inth

eir

most

rece

nt

year

ofw

ork

inP

ortl

and

,M

ain

ean

dth

eir

dem

ogra

ph

ican

dh

um

anca

pit

alch

arac

teri

stic

s;m

acro

econ

om

icco

nd

itio

ns

inM

ain

e;w

ork

exp

erie

nce

;an

dac

cess

toco

-eth

nic

soci

alca

pit

al.

Mod

els

9–13

corr

esp

ond

tofe

mal

ere

fuge

esan

dM

odel

s14–18

corr

esp

ond

tom

ale

refu

gees

.�p

<0.1

0;

*p<

0.0

5;

**p

<0.0

1(s

tan

dar

der

rors

are

inp

aren

thes

es).


of the coefficient suggests an unstable relationship between this controlfor macroeconomic conditions and the temporal variables introduced inModel 11. This kind of instability usually suggests a multicollinearityproblem, but as mentioned above the VIFs in this model indicate thatmulticollinearity is not a problem in this model. Given the theoreticalimportance of controlling for economic conditions when estimating earn-ings, I elected to leave the unemployment rate in subsequent models.Female refugee work experience had a positive relationship with earningsin the most recent year of work. Sponsorship had a negative effect onearnings during the most recent year of work for female refugees. In fact,Model 13 predicts that, all else equal, sponsored female refugees earnedabout 24 percent less in their most recent year of work in Maine thannon-sponsored female refugees.

Table 5 also shows that, similar to female refugees, temporal vari-ables had an unstable relationship with macroeconomic conditions duringthe most recent year of work for male refugees (again, multicollinearity isnot a problem in this model). In addition, age and work experience wereimportant predictors of most recent year earnings for male refugees, butin contrast to female refugees, country of origin was also important. Malerefugees from Somalia and Sudan earned about 25 percent less than theirEastern European counterparts in their most recent year of work. Theability to speak English and educational attainment were not statisticallysignificant predictors in any of the models for male refugees. Results inModel 18 indicate that access to co-ethnic social capital had a positiveeffect on the earnings of male refugees in their most recent year of workin Maine. Male refugees who were originally resettled in Portland earnedabout 27 percent more than male refugees who elected to move fromtheir original resettlement city to Portland. Though it is not significant atthe 0.05 level of significance, results indicate that sponsored male refugeesearned about 17 percent more than their non-sponsored counterparts.

DISCUSSION

Prior to a discussion of these research results, it is important to under-stand the limitations of this study. First, the data come from a cross-section of refugees from a single city in the United States, so it is notpossible to generalize these findings for refugees overall. A more ambitiouslongitudinal research design that compared the economic outcomes of ref-ugees in small and large cities over a longer period of time would add


considerably to the explanatory power of any research findings. Second, itwas not possible to include data regarding households in this analysis,since the data were maintained in a manner which made it very difficultto determine which refugees in the data set lived in the same household.Therefore, the possible effects of single versus two-parent headed house-holds on refugee economic outcomes are missing from this analysis. Thisis a weakness of this study since substantial evidence on immigrants sug-gests that households are a preferred unit of analysis. Third, though quali-tative evidence indicates that sponsorship is an appropriate measure ofaccess to co-ethnic social capital upon arrival to Portland, it does notactually measure the extent to which refugees used co-ethnic social capitalto help them in the labor market or in other aspects of life related to thelabor market. Instead, it makes a plausible assumption regarding the dif-ferent degree of co-ethnic social capital that reunification and free caserefugees could access when they arrived in Maine.

The results from this research partially support the hypotheses laidout earlier in this paper. Specifically, my results failed to support Hypoth-esis 1, since sponsorship was not a statistically significant predictor ofearnings in the first year of work in Maine for male or female refugees.The research results failed to support Hypothesis 2 since sponsorship hada positive effect on male refugee earnings in their most recent year ofwork, but this result fell short of meeting the 0.05 level of statistical sig-nificance common in social science. Finally, the results supported Hypoth-esis 3 since sponsorship had a negative and statistically significant effecton female refugee earnings in their most recent year of work. The researchresults suggest two minor findings and one major finding that warrantattention.

Minor Findings

The first minor finding is that the ability to speak English and educa-tional attainment were not significant factors in determining the earningsof refugees in their first or most recent years of work. These human capi-tal characteristics may have been insignificant for several reasons. In mod-els estimating the first year of earnings, the lack of statistical significancefor the English and educational attainment predictors suggests that, atleast initially, male and female refugees are an undifferentiated pool ofunskilled workers. Reflecting the devaluation of refugee educational cre-dentials outside of their country of origin (Colic-Peisker and Walker,


2003) and an urgency to find work, refugees initially compete for thesame pool of available low-skill jobs, rather than sorting into jobs thatmatch their educational attainment level or English fluency.

My own observations in Portland suggest that, even by their mostrecent year of work, jobs that are available to the bulk of refugees in Port-land depend less on their ability to communicate well in English andmore on their ability to learn by example. The significant proportion ofrefugees working in the manufacturing industry, which requires little com-munication with the public, for their most recent job in Portland supportsmy observations. Also, for models estimating the most recent year of earn-ings for refugees, these results could be a function of measurement error.A refugee’s ability to speak English and educational attainment were mea-sured upon arrival to Portland and may be poor measures of a refugees’ability to speak English and educational attainment after several years ofattending English classes and other training while living in Portland. Thethriving English as a Second Language (ESL) courses offered in Portlandsuggest that many adult refugees do improve their English skills overtime. However, most ESL classes for recently resettled refugees focus on‘‘survival English,’’ which may improve a refugee’s ability to function intheir host society, but not necessarily provide a significant advantage inthe labor market.

The second minor finding is that in their most recent year of earn-ings, male refugees who resettled in Portland earned more than male refu-gees who moved to Portland as secondary migrants. Even thoughsecondary migrants in this sample used the services of CCMRIS, my ownobservations suggest that secondary migrants lacked familiarity with theorganization and had only a weakly developed relationship with a casemanager. The combination of these two factors may have dissuaded themfrom using the services intensely, resulting in their getting less supportfrom CCMRIS than other refugees. About 63 percent of the secondarymigrants in the data set were also reunification cases, meaning that theymoved to Portland to reunite with friends or family members. Thissuggests that relying primarily on the informal assistance of co-ethnics,instead of drawing on assistance from co-ethnics and a case manager atthe VOLAG, may have had a detrimental effect on the earnings of malesecondary migrants. Male secondary migrants had spent slightly less timein Portland than male refugees who were originally resettled there (18.1quarters and 19.7 quarters, respectively), suggesting that secondarymigrants had less knowledge about the local labor market. Finally, it is


also possible that secondary migrants move to a new city because theyhave been unsuccessful finding a job in the original resettlement contextbecause of lack of job opportunities or because they possess few labormarket skills. If the second case is true, then the negative self-selection ofsecondary migrants could help to explain the earnings differential. Moreinformation about the secondary migrant population, including their rea-sons for moving, would help to shed light on this result.

Major Finding

The major finding in this research is the negative effect of sponsorshipstatus on the earnings of female refugees in their most recent year ofwork. In their most recent year of work, female reunification case refugeesearned less than female free case refugees. Informed by my observations inPortland, I believe that the different ways that reunification and free caserefugees experienced social capital help to explain this result.

My observations indicate that female reunification case refugees notonly experience real costs associated with using social capital, but alsocould take advantage of non-pecuniary benefits unavailable to female freecase refugees. For example, female reunification case refugees frequentlyacknowledged expectations from their social ties regarding responsibilityfor child care and help with household chores for extended family mem-bers and friends that their free case counterparts did not. While theseobligations constrained the amount of time that reunification case femalescould participate in the labor market, it also resulted in a more satisfyingsocial life and allowed them to request reciprocation from their social tieswhen they needed child care and other types of assistance in their house-holds. In contrast, free case females were typically on their own whenarranging and paying for childcare, and performing household chores.How female reunification case refugees valued reciprocal contributionsfrom social ties and how researchers should factor these contributions intoanalyses of the effects of social capital on the lives of refugees are openquestions that a more comprehensive study must address.

My research results indicate that co-ethnic social capital can nega-tively affect female refugee earnings. This finding supports earlier researchfindings by Bach and Carroll-Seguin (1986) that indicated a negativeeffect of co-ethnic sponsorship on the labor force participation of femaleSouth-East Asian refugees. One important difference is that in Bach andCarroll-Seguin’s findings this effect was observed in recently arrived


refugees, but not for well-established refugees. They hypothesize that thelack of knowledge about employment opportunities and cultural con-straints on females entering the workforce explain their research finding.In contrast, my findings indicate that access to co-ethnic social capital hasno initial effect on the earnings of female refugees but has a negativeeffect on the earnings of female refugees over time. I hypothesize thatreciprocal obligations and social norms may impinge upon the earningspossibilities of female reunification case refugees but not on the earningspossibilities of female free case refugees.

It is important to mention two differences between the Bach andCarroll-Seguin study and this study to provide a contextual basis for com-paring each study’s research findings. First, and most importantly, Bachand Carroll-Seguin examine labor market participation while I focus onearnings. This difference is important because I am measuring the effectof co-ethnic social capital on the quality of employment, rather than theability to find employment. With this in mind, it is plausible and consis-tent with Bach and Carroll-Seguin’s findings that over time female refu-gees learn more about the local labor market and enter the labor forcethrough their own devices rather than relying on the assistance of a spon-sor as they did when they first arrived in the United States. It is equallyplausible and consistent with my own findings that, over time, female ref-ugees who have received assistant from a sponsor are called upon to recip-rocate or feel more subject to social norms that frown upon themworking too much, limiting how much they earn rather than whether ornot they are part of the labor force. Together, these ideas paint a morecomplete picture of the effect of co-ethnic social capital on economicoutcomes for female refugees.

Second, Bach and Carroll-Seguin compared the effects of co-ethnicsponsorship and sponsorship outside of the refugee community (primarilyfrom Anglo families or church groups) on the labor force participation ofrefugees. In the early 1980s, almost 70 percent of resettled refugees reliedon an Anglo family or church group for sponsorship (Bach and Carroll-Seguin, 1986). Bach and Carroll-Seguin argued in their study that thesocial and institutional networks for Anglo families and church groups aresuperior to those of refugee sponsors, resulting in an advantage in thelabor market for refugees with sponsors outside of the refugee community.This argument is undoubtedly still true today but is largely a moot pointsince Anglo families and church groups now play such a small role in ref-ugee resettlement. Instead, refugees who have co-ethnic sponsors have a


layer of support that may be a benefit or burden in the labor market com-pared to refugees who rely exclusively on a VOLAG for resettlementassistance.

The findings from this study expand our understanding of theconcept of embeddedness (Portes and Sensenbrenner, 1993), or howsocial relationships shape economic behavior. Female reunification caserefugees appear to be overly embedded, at least from the standpoint oftheir earnings prospects, through excessive obligations and stricter adher-ence to social norms that can become barriers to working. With this inmind, a robust understanding of social capital must pay attention tohow the obligations and claims on resources—or what an individualowes to other members of a social network—and the aid that an indi-vidual receives from other members of a social network differ for menand women.

APPENDIX A

Results of the Probit Analysis

There were two dependent variables in the probit analysis. Work_First isa dichotomous variable coded 1 if a refugee had earnings from employ-ment in his first year of work in Maine and 0 otherwise. Work_Recent isanother dichotomous variable coded 1 if a refugee had earnings fromemployment in his most recent year of work in Maine and 0 otherwise.The probit equations used five variables to estimate whether or not a refu-gee worked in Maine:

• Age reflects life experience and should have a positive effect on workingin Maine.

• Age squared should have a negative effect on working in Maine.• A set of dummy variables used to control for year of arrival in Portland

(1998–2004, 1998 was dropped). Refugees who arrived in earliercohorts should be more likely to work in Maine than refugees whoarrive in later cohorts.

• Relation is a dichotomous variable coded 1 if the refugee was the prin-cipal applicant in the refugee resettlement process and 0 otherwise. Theprincipal applicant in the refugee resettlement process was the personfrom a household who worked most closely with the resettlementbureaucracy. Refugees who are the primary applicants in the


resettlement process may face more pressure to provide financially.Therefore, primary applicant status should have a positive effect onworking in Maine.

• Family Size is the total number of individuals who arrived with a refu-gee in Portland (including the refugee). It should have a negative effecton working in Maine reflecting higher child care costs associated withworking.

TABLE A1PROBIT ANALYSIS FOR THE PROBABILITY OF FEMALE REFUGEES WORKING IN PORTLAND,

MAINE (WORK_FIRST)

Maximum Likelihood Estimates z-Statistic

Intercept )1.3388 )2.0011Age 0.1601 4.4794Age_sq )0.0021 )4.7960Cohort.1999 )0.6404 )2.7064Cohort.2000 )0.2059 )0.7829Cohort.2001 )0.8164 )3.3083Cohort.2002 )1.1052 )3.7030Cohort.2003 )0.6465 )2.0924Cohort.2004 )2.4133 )7.0886Relation )0.2781 )1.7746Family_Size )0.0547 )0.9230

Notes: Dependent variable = 1 if person experienced employment and 0 otherwise. v2(10) = 117.07; 412 observa-tions (143 = 0; 269 = 1).

TABLE A2PROBIT ANALYSIS FOR THE PROBABILITY OF MALE REFUGEES WORKING IN PORTLAND,

MAINE (WORK_FIRST)


Intercept )0.8095 )1.3744Age 0.1102 3.5116Age_sq )0.0014 )3.7600Cohort.1999 )0.6773 )3.0635Cohort.2000 )0.3094 )1.3610Cohort.2001 )0.4966 )2.166Cohort.2002 )0.2900 )1.0709Cohort.2003 )0.3747 )1.266Cohort.2004 )1.6316 )5.9068Relation )0.3529 )2.0932Family_Size 0.1460 2.6395

Notes: Dependent variable = 1 if person experienced employment and 0 otherwise v2(10) = 73.90; 526 Observa-tions (131 = 0; 395 = 1).


TABLE A3PROBIT ANALYSIS FOR THE PROBABILITY OF FEMALE REFUGEES WORKING IN PORTLAND,

MAINE (WORK_RECENT)


Intercept )0.9005 )1.3431Age 0.1396 3.9568Age_sq )0.0018 )4.3198Cohort.1999 )0.7786 )3.1924Cohort.2000 )0.3577 )1.3382Cohort.2001 )0.9054 )3.5742Cohort.2002 )1.2108 )3.9961Cohort.2003 )0.6855 )2.1504Cohort.2004 )2.5525 )7.4109Relation )0.2577 )1.6450Family_Size )0.0316 )0.5088


TABLE A4PROBIT ANALYSIS FOR THE PROBABILITY OF MALE REFUGEES WORKING IN PORTLAND, MAINE

(WORK_RECENT)


Intercept )0.3219 )0.5297Age 0.0930 2.8756Age_sq )0.0012 )3.1782Cohort.1999 )0.6837 )2.9487Cohort.2000 )0.3024 )1.2611Cohort.2001 )0.6051 )2.5352Cohort.2002 )0.4069 )1.4654Cohort.2003 )0.4139 )1.3518Cohort.2004 )1.7558 )6.1680Relation )0.4018 )2.2190Family_Size 0.1508 2.5300


REFERENCES

Bach, R. L., and R. Carroll-Seguin1986 ‘‘Labor Force Participation, Household Composition and Sponsorship among

Southeast Asian Refugees.’’ International Migration Review 20:381–404.

Bailey, T., and R. Waldinger1991 ‘‘Primary, Secondary, and Enclave Labor Markets: A Training System Approach.’’

American Sociological Review 56:432–445.

Black, R.2001 ‘‘Fifty Years of Refugee Studies: From Theory to Policy.’’ International Migration

Review 35:57–78.

Bourdieu, P.1985 ‘‘The Forms of Capital.’’ In Handbook of Theory and Research for the Sociology of

Education. Ed. J. G. Richardson. New York: Greenwood. Pp. 241–258.


Chiswick, B. R.1993 ‘‘Soviet Jews in the United States: An Analysis of Their Linguistic and Economic

Adjustment.’’ International Migration Review 27:260–285.

Colic-Peisker, V., and I. Walker2003 ‘‘Human Capital, Acculturation and Social Identity: Bosnian Refugees in Australia.’’

Journal of Community & Applied Social Psychology 13:337–360.

Cortes, K. E.2004 ‘‘Are Refugees Different From Economic Immigrants? Some Empirical Evidence

on the Heterogeneity of Immigrant Groups in the United States.’’ The Review ofEconomics and Statistics 86:465–480.

DHS2005 2004 Yearbook of Immigration Statistics. Washington: Department of Homeland

Security, Office of Immigration Statistics.

Fennelly, K., and H. Leitner2002 ‘‘Working Paper #59: How the Food Processing Industry is Diversifying Rural

Minnesota.’’ In JSRI Working Paper Series. East Lansing, MI: Julian SamoraResearch Institute, Michigan State University.

Gold, S. J.1992 Refugee Communities: A Comparative Field Study. Newbury Park, CA: Sage

Publications.

Haines, D. W.1996 ‘‘Patterns in Refugee Resettlement and Adaptation.’’ In Refugees in America in the

1990s. Ed. D. W. Haines. Westport, CT: Greenwood Press. Pp. 28–62.

Heckman, J. J.1979 ‘‘Sample Selection Bias as a Specification Error.’’ Econometrica 47:153–161.

Hein, J.2006 Ethnic Origins: The Adaptation of Cambodian and Hmong Refugees in Four American

Cities. New York: Russell Sage Foundation.

Holman, P. A.1996 ‘‘Refugee Resettlement in the United States.’’ In Refugees in America in the 1990s.

Ed. D. Haines. Westport, CT: Greenwood Press. Pp. 3–27.

Horton, N., and K. Kleinman2007 ‘‘Much Ado About Nothing: A Comparison of Missing Data Methods and Soft-

ware to Fit Incomplete Data Regression Models.’’ The American Statistician 61:79–90.

Kennedy, E. M.1981 ‘‘The Refugee Act of 1980.’’ International Migration Review 15:141–156.

King, G., J. Honaker, A. Joseph, and K. Scheve2001 ‘‘Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multi-

ple Imputation.’’ American Political Science Review 95:49–69.

Lamba, N. K.2003 ‘‘The Employment Experiences of Canadian Refugees: Measuring the Impact of

Human and Social Capital on Quality of Employment.’’ Canadian Review of Sociol-ogy and Anthropology [Revue Canadienne de Sociologie et D’ Anthropologie] 40:45–64.

Lanphier, C. M.1983 ‘‘Refugee Resettlement: Models in Action.’’ International Migration Review 17:4–33.


Lin, N.2001a ‘‘Building a Network Theory of Social Capital.’’ In Social Capital: Theory and

Research. Ed. N. Lin, K. Cook, and R. Burt. New York: Aldine de Gruyter. Pp. 3–30.

———2001b Social Capital: A Theory of Social Structure and Action. Cambridge: Cambridge

University Press.

Little, R., and D. Rubin1987 Statistical Analysis with Missing Data. Hoboken: John Wiley & Sons.

LMIS2006 ‘‘Local Employment Dynamics (LED).’’ <http://www.state.me.us/labor/lmis/lehd1.

html>. Accessed on March 18, 2009.

Majka, L., and B. Mullan2002 ‘‘Ethnic Communities and Ethnic Organizations Reconsidered: South-East Asians

and Eastern Europeans in Chicago.’’ International Migration 40:71–92.

Mamgain, V., and K. Collins2003 ‘‘Off the Boat, Now Off to Work: Refugees in the Labour Market in Portland,

Maine.’’ Journal of Refugee Studies 16:113–146.

McMichael, C.2002 ‘‘‘Everywhere is Allah’s Place’: Islam and the Everyday Life of Somali Women in

Melbourne, Australia.’’ Journal of Refugee Studies 15:171–188.

Menjivar, C.2000 Fragmented Ties. Berkeley, CA: University of California Press.

Monroe, J. W.2004 Seeking the Port in Portland. Portland, ME: Department of Transportation and

Waterfront.

Montero, D.1979 Vietnamese Americans: Patterns of Resettlement and Socioeconomic Integration in the

United States. Boulder, CO: Westview Press.

Montgomery, J. R.1996 ‘‘Components of Refugee Adaptation.’’ International Migration Review 30:679–702.

Newton, K.2001 ‘‘Trust, Social Capital, Civil Society and Democracy.’’ International Political Science

Review 22:201–214.

ORR2004 Report to Congress FY2003. Washington, DC: Office of Refugee Resettlement.

———2005 Report to Congress FY2004. Washington, DC: Office of Refugee Resettlement.

Portes, A.1995 ‘‘Economic Sociology and the Sociology of Immigration: A Conceptual Overview.’’

In The Economic Sociology of Immigration. Ed. A. Portes. New York: Russell SageFoundation. Pp. 1–41.

———, and P. Landolt1996 ‘‘The Downside of Social Capital,’’ The American Prospect 26:18–22.

———, and R. Rumbaut2006 Immigrant America: A Portrait. Berkeley, CA: University of California Press.


———, and J. Sensenbrenner1993 ‘‘Embeddedness and Immigration: Notes on the Social Determinants of Economic

Action.’’ American Journal of Sociology 98:1320–1350.

Potocky-Tripodi, M.2003 ‘‘Refugee Economic Adaptation: Theory, Evidence, and Implications for Policy and

Practice.’’ Journal of Social Service Research 30:63–91.

———2004 ‘‘The Role of Social Capital in Immigrant and Refugee Economic Adaptation.’’

Journal of Social Service Research 31:59–91.

Puhani, P.2000 ‘‘The Heckman Correction for Sample Selection and Its Critique.’’ Journal of

Economic Surveys 14:53–68.

Putnam, R. D.2000 Bowling Alone: The Collapse and Revival of American Community. New York: Simon

& Schuster.

Read, J. G.2004 ‘‘Cultural Influences on Immigrant Women’s Labor Force Participation: The

Arab-American Case.’’ International Migration Review 38:52–77.

Sanders, J., V. Nee, and S. Sernau2002 ‘‘Asian Immigrants’ Reliance on Social Ties in a Multiethnic Labor Market.’’ Social

Forces 81:281–314.

Singer, A., and J. H. Wilson2006 ‘‘From ‘There’ to ‘Here’: Refugee Resettlement in Metropolitan America.’’ In The

Living Cities Census Series. Washington, DC: Brookings Institution. Pp. 1–32.

Stack, C. B.1974 All Our Kin: Strategies for Survival in a Black Community. New York: Harper & Row.

Strand, P. J.1984 ‘‘Employment Predictors among Indochinese Refugees.’’ International Migration

Review 18:50–64.

Tran, T. V.1991 ‘‘Sponsorship and Employment Status among Indochinese Refugees in the United

States.’’ International Migration Review 25:536–550.

Tsuda, T., Z. Valdez, and W. Cornelius2003 ‘‘Human Versus Social Capital: Immigrant Wages and Labor Market Incorporation

in Japan and the United States.’’ In Host Societies and the Reception of Immigrants.Ed. J. G. Reitz. La Jolla, CA: University of California Press. Pp. 215–251.

Waldinger, R.1994 ‘‘The Making of an Immigrant Niche.’’ International Migration Review 28:3–30.

Waxman, P.2001 ‘‘The Economic Adjustment of Recently Arrived Bosnian, Afghan and Iraqi

Refugees in Sydney, Australia.’’ International Migration Review 35:472–505.

Winship, C., and S. L. Morgan1999 ‘‘The Estimation of Causal Effects from Observational Data.’’ Annual Review of

Sociology 25:659–706.

Wooden, M.1991 ‘‘The Experience of Refugees in the Australian Labor Market.’’ International Migra-

tion Review 25(3):514–535.


Documents

Benefit or Burden? Social Capital, Gender, and the Economic Adaptation of Refugees