econstor Make Your Publications Visible.

Transcription

1 econstor Make Your Publications Visible. A Service of Wirtscaft Centre zbwleibniz-informationszentrum Economics Oreffice, Sonia Working Paper Culture and Houseold Decision Making: Balance of Power and Labor Supply Coices of US-born and Foreign-born Couples IZA Discussion Paper, No Provided in Cooperation wit: IZA Institute of Labor Economics Suggested Citation: Oreffice, Sonia (2014) : Culture and Houseold Decision Making: Balance of Power and Labor Supply Coices of US-born and Foreign-born Couples, IZA Discussion Paper, No. 7997, Institute for te Study of Labor (IZA), Bonn Tis Version is available at: ttp://dl.andle.net/10419/93316 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenscaftlicen Zwecken und zum Privatgebrauc gespeicert und kopiert werden. Sie dürfen die Dokumente nict für öffentlice oder kommerzielle Zwecke vervielfältigen, öffentlic ausstellen, öffentlic zugänglic macen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt aben sollten, gelten abweicend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewärten Nutzungsrecte. Terms of use: Documents in EconStor may be saved and copied for your personal and scolarly purposes. You are not to copy documents for public or commercial purposes, to exibit te documents publicly, to make tem publicly available on te internet, or to distribute or oterwise use te documents in public. If te documents ave been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise furter usage rigts as specified in te indicated licence.

2 DISCUSSION PAPER SERIES IZA DP No Culture and Houseold Decision Making: Balance of Power and Labor Supply Coices of US-born and Foreign-born Couples Sonia Oreffice February 2014 Forscungsinstitut zur Zukunft der Arbeit Institute for te Study of Labor

3 Culture and Houseold Decision Making: Balance of Power and Labor Supply Coices of US-born and Foreign-born Couples Sonia Oreffice University of Surrey and IZA Discussion Paper No February 2014 IZA P.O. Box Bonn Germany Pone: Fax: Any opinions expressed ere are tose of te autor(s) and not tose of IZA. Researc publised in tis series may include views on policy, but te institute itself takes no institutional policy positions. Te IZA researc network is committed to te IZA Guiding Principles of Researc Integrity. Te Institute for te Study of Labor (IZA) in Bonn is a local and virtual international researc center and a place of communication between science, politics and business. IZA is an independent nonprofit organization supported by Deutsce Post Foundation. Te center is associated wit te University of Bonn and offers a stimulating researc environment troug its international network, worksops and conferences, data service, project support, researc visits and doctoral program. IZA engages in (i) original and internationally competitive researc in all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of researc results and concepts to te interested public. IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of suc a paper sould account for its provisional caracter. A revised version may be available directly from te autor.

4 IZA Discussion Paper No February 2014 ABSTRACT Culture and Houseold Decision Making: Balance of Power and Labor Supply Coices of US-born and Foreign-born Couples * Tis study investigates ow spouses cultural backgrounds mediate te role of intraouseold bargaining in te labor supply decisions of foreign-born and US-born couples, in a collective-ouseold framework. Using data from te 2000 US Census, I sow tat te ours worked by US-born couples, and by tose foreign-born coming from countries wit gender roles similar to te US, are significantly related to common bargaining power forces suc as differences between spouses in age and non-labor income, controlling for bot spouses demograpic and socioeconomic caracteristics. Houseolds wose culture of origin supports strict and unequal gender roles do not exibit any association of tese power factors wit teir labor supply decisions. Tis cultural asymmetry suggests tat spousal attributes are assessed differently across couples witin te US, and tat ow spouses make use of teir outside opportunities and economic and institutional environment may depend on teir etnicities. JEL Classification: D1, J15, J22 Keywords: culture, gender roles, ouseold bargaining power, labor supply Corresponding autor: Sonia Oreffice Scool of Economics University of Surrey Guildford, Surrey, GU2 7XH United Kingdom s.oreffice@surrey.ac.uk * Financial support from te Spanis Ministry of Science and Innovation (ECO ) and IVIE is acknowledged. I would like to tank te editor, two anonymous referees, and seminar participants at RES 2013 and Alicante for teir elpful comments. Any errors are mine.

5 1. Introduction Tis paper examines te labor supply coices of foreign-born and US-born couples in order to explore te role of spousal cultural background on te extent to wic bargaining power forces matter in ouseold decisions, using data from te 2000 US Census. Specifically, it investigates weter spouses wo live in te US but are from different etnicities relate eterogeneously to bargaining power forces, as measured by discrepancies in spouses ages and non-labor income, witin a collective ouseold labor supply framework. Te penomena of immigrants labor market outcomes and te influence of teir cultural background on one side, and intra-ouseold bargaining power and ouseold decision-making on te oter, ave been widely studied in te literature. A large body of teoretical and empirical literature sows tat te intra-ouseold distribution of power influences ouseolds outcomes in bot developing and developed countries (Ciappori, Fortin, Lacroix, 2002; Grossbard- Sectman, 1993; Lundberg and Pollak, 1996; Tomas, 1990). In particular, te collective ouseold beavior model predicts tat ouseold members make Pareto-efficient decisions according to teir respective bargaining power positions, wic in turn depend on outside opportunities and social and legal factors, suc as members relative sare of non-labor income, teir age differences, teir body mass index, abortion and divorce laws, and sex ratios (Browning, Bourguignon, Ciappori, Lecene, 1994; Ciappori et al., 2002; Negrusa and Oreffice, 2010; Oreffice, 2007; Oreffice, 2011; Oreffice and Quintana-Domeque, 2012). Tere is empirical evidence of tese bargaining power effects concerning various countries, from different cultures and parts of te world (Grossbard and Amuedo-Dorantes, 2007; Lundberg, Pollak, Wales, 1996; Rangel, 2006; Scultz, 1990; Tomas, 1990). Wile all tese empirical studies ave focused on a single country, Datta Gupta and Stratton (2010) present a two-country analysis on US and Denmark, and argue tat different social norms across countries can affect intra-ouseold bargaining power, mitigating or reinforcing te bargaining effect on couples decisions concerning leisure and labor supply. Still, no empasis as been devoted to differences across etnicities witin a country, and te recent immigration waves trougout Europe and te US, along wit te sizable presence in te US of immigrants from a variety of cultural backgrounds, prompt to explore te role of intra-ouseold bargaining power by culture and spouses gender role beliefs. 1

6 A recent strand of literature specifically empasizes te role of culture and family experience on economic outcomes suc as female labor force participation and fertility coices, and on te marriage market. Fernandez and Fogli (2006, 2009), Fernandez (2007), and Blau, Kan, and Papps (2011) sow tat culture, as measured by female labor force participation rate and total fertility in country of origin, explains fertility and labor market coices of American women living in te US but born from foreign parents, wile male labor supply is not affected by te country of origin. Using te same US Census data and te same sources of variation for cultural differences suc as country of origin, lengt of stay in te US, and linguistic distance, related literature studies te determinants of interetnic marriages and living arrangements (e.g., Angrist, 2002; Ciswick and Housewort, 2010; Giuliano, 2007). In particular, Giuliano (2007) empasizes te importance of analyzing te role of culture in a neutral environment, using samples drawn from te population of only one country. Finally, culture and immigration ave also been linked to labor market activity and te extent of wage and employment consequences for native and foreign-born workers. All tese studies igligt tat te main callenge in tis line of researc is to disentangle culture from institutional and traditional economic variables suc as prices and income. Indeed, tis paper addresses tis feature using one large data set suc as te US Census, wit US- and foreign-born individuals from a large variety of countries at different stages of development and assessment of gender roles in society, wo live and work in te US, facing te same institutions and bargaining power measures. Specifically, te cultural mecanism troug wic intraouseold bargaining power and couples labor supplies may be related is analyzed ere using te five-percent national random sample of te 2000 US Census data on married men and women. Te US Census data provide te largest sample of ouseolds wit foreign-born spouses, teir detailed etnic, demograpic, labor and income information, along wit standard samples of US-born individuals. Te degree of cultural differences from te US mainstream is captured by te country of birt of eac spouse, to wic te information on te corresponding measures of gender roles is associated. Following te definition of culture offered by Fernandez (2007), I consider culture as a set of beliefs and preferences, wic are important determinants of beavior. Te cultural proxies of bot te usband and te wife are analyzed, as te usband s culture may be important in driving work decisions and ouseold responsiveness to bargaining 2

7 power forces (Fernandez and Fogli, 2009). I ten follow a procedure in line wit Guiso et al. (2008) to assess countries cultural similarity to te US in terms of gender roles. Te focus is on first-generation immigrants, wo are less likely to ave assimilated and mitigated teir culture and beliefs to te US mainstream, to better capture te corresponding bargaining power disparities at stake. Individuals in te sample face te same markets and institutions by construction, so tat only te belief and preference components (te cultural components) are potentially relevant. Te main assumption ere is tat immigrants bring wit tem some of te attitudes of teir country of origin (Carroll, Ree, and Ree, 1994). Ten, since tey differ by teir own cultural eritage, it is still possible to estimate te impact of cultural differences on a common set of variables, as suc eliminating unobserved eterogeneity and comparability concerns wic would result from cross-country estimation from separate data sets. As additional evidence, second- and iger-generation immigrants are considered using te information on primary ancestry, examining US-born spouses wose culture may differ by ancestry rater tan by country of birt. Te analysis ere considers te differences in age and non-labor income ownersip between spouses, two commonly used indicators of intra-ouseold bargaining power. According to Ciappori et al. (2002), wen a spouse as a relatively better attribute (relatively older or ricer), te distribution of gains from te relationsip would sift in is/er favor, generating opposite income effects on te spouses, wic are testable on teir labor supplies. Consequently, te spouse wit a more favorable bargaining position would decrease is/er ours worked, wile is/er mate would increase is/ers. My identification strategy consists of estimating te associations of intra-ouseold age and non-labor income gaps wit te labor supplies of bot spouses, comparing teir ours of work cross-sectionally among US-born and foreign-born couples wit various extents of cultural disparities to te US in terms of gender roles. Specifically, I test weter tere is a stronger association for tose married individuals wose cultural background supports more egalitarian gender roles, and no association at all for couples wit very different gender role norms from te US. Terefore, tis study considers individuals wo are already married to men and women of teir same cultural background, leaving aside te interesting patterns of intermarriage or sorting by etnicity, wic ave recently been empirically analyzed in Furtado (2012), for instance. 3

8 Te empirical analysis sows tat te ours worked by US-born spouses are more responsive to te differences between spouses in age and non-labor income, tan tose by foreign-born couples wose country of origin supports a traditional role of women. In US-born couples, and in tose coming from countries wit gender roles similar to te US 1, a relatively older, or ricer spouse supplies less labor, te opposite olding for is/er mate, controlling for bot spouses demograpic and socioeconomic caracteristics. Tis evidence on bot spouses labor supplies, and te signs of te estimated coefficients, is consistent wit te ouseold bargaining power interpretation. Interestingly, ouseolds wose culture of origin is quite or somewat different on gender role grounds tan te US do not exibit any relationsip between tis measured balance of power and teir labor supply decisions. A 5-year difference implies working about 12 ours less, and 5,000 dollars more of nonlabor income relative to teir wives implies about 9 ours less. For wives, te corresponding figures are 8 and 50 ours more. Wile te foreign-born couples culturally distant from te US do not exibit any significant association wit tese bargaining forces, tose immigrant ouseolds wit very similar culture to te US sow a significant association wit tese bargaining power factors for bot spouses: for usbands, 19 and 5 ours less, and for teir wives 35 and 170 ours more, respectively. Te estimated correlation of female labor supply and age difference is iger in magnitude tan in te case of US-born couples. In addition, joining te sample of US- and foreign-born spouses and interacting te bargaining power variables wit te gender role index associated wit teir country of origin provides evidence consistent wit te bargaining power interpretation of te estimates on te subsamples of couples by cultural disparity from te US. I also present estimates controlling for and interacting wit te number of years spent in te US by foreign-born spouses, and obtain virtually identical results. Tese findings represent te first empirical evidence suggesting tat spousal attributes and empowerment are assessed differently across couples witin te US, and tat couples responsiveness to a common economic and institutional environment may depend on teir culture of origin. Te importance of tis novel approac rests in providing an empirical assessment and cross-cultural comparisons wic may be useful to devise public policies targeting immigrant ouseolds and women in particular. Tey may be te ones least likely to 1 Te similarity of institutions is defined according to te country of origin, and categorized in te tree groups very similar, somewat similar, and different from te US culture, as explained in Section 3. 4

9 respond and take advantage of an empowerment policy, because of teir cultural background constraining tem to ignore outside opportunities and welfare enancement measures. Te paper is organized as follows. Section 2 describes te teoretical framework and Section 3 te empirical specification and te data. Section 4 presents te empirical results, and discusses te sensitivity analysis and potential alternative explanations. Finally, Section 5 concludes te paper. 2. Teoretical framework Te collective ouseold labor supply model wit distribution factors (Ciappori et al., 2002) is applied to US-born and foreign-born couples wit different degrees of cultural disparities wit respect to te US, and tus wit a potentially different responsiveness to bargaining power forces. A ouseold is composed of two decision makers, usband and wife, eac aving a distinct utility function on consumption and leisure, and making Pareto-efficient decisions. Preferences are egoistic, in tat one mate s utility does not depend on te oter s consumption or leisure, altoug te model can be extended to allow for caring preferences and public goods. Let i and C i for i =, w denote member i s labor supply and consumption of a private composite good (wose price is normalized to unity), y te ouseold non-labor income, w i te wage rate of spouse i, and z i possible preference parameters and cultural background of eac spouse, suc as education, race, or country of birt. Finally, let s 1 and s 2 represent te two distribution factors (bargaining power forces) under analysis: te differences between spouses in i i i non-labor income and age. Te utility function of member i =, w is U (1, C ), were U is strictly quasi-concave, increasing, and continuously differentiable, wile, following convention, te utility from companionsip is assumed to be separable and not to influence te trade-off between leisure and consumption. Te optimal allocations of labor supply of eac spouse are determined by te following program: max, (1 U C subject to w, C y C ϕ ( w, w,, 1, 2,, w ) + s s ) z z w 5

10 were te spouse faces a symmetric problem. ϕ w, w, y, s, s, z, z ) represents te usband s ( w 1 2 w sare of ouseold non-labor income y, wile te wife receives y - ϕ( w, ww, y, s1, s2, z, zw ), so tat te stronger te usband s bargaining power, te iger is sare of ouseold non-labor income and te lower is wife s. Te saring rule ϕ w, w, y, s, s, z, z ) is a function of ( w 1 2 w prices (ere normalized to unity), spouses wages, ouseold non-labor income, distribution factors (ere te non-labor income gap and age gap) 2, and oter observable caracteristics z. Te couple s Pareto-efficient decisions yield te following equilibrium labor supply functions of te two spouses: w = [ w, ϕ( w, ww, y, s, s2, z, z =, y ϕ( w, w 1 w, y, s, s, z )], z w [ ww w 1 2 w Te derivatives of eac labor supply function wit respect to te second arguments are expected to be negative, reflecting a pure income effect (leisure is commonly assumed to be a normal good). Hence, factors tat strengten te usband s bargaining power reduce te labor supplied by te usband and increase te labor supplied by te wife, ceteris paribus, in particular controlling for own wage and te couples total non-labor income y. Te question of weter foreign-born couples ours of work are related to suc factors in te direction predicted by te teory and already estimated for US couples in general is investigated by testing teir relationsip wit te labor supplies of foreign-born couples from different cultural backgrounds, and comparing it to te corresponding evidence on US-born spouses. Te decisions of spouses wose country of origin supports strict gender roles and traditional family institutions relative to te US may be very weakly related to bargaining power forces. Teir ouseold beavior may reflect a setting were spousal personal caracteristics and outside opportunities do not influence teir ouseold decision-making. Te non-labor income difference and age differences represent relevant monetary and demograpic personal attributes enancing spouses outside opportunities, wic ave been extensively estimated to be sources of bargaining power (e.g., Browning, Ciappori, Weiss, 2014). If tey did not significantly enter )] 2 Te sex ratio, divorce laws, abortion legalization, alimony, and cild benefits laws, are oter examples of distribution factors tat ave been studied in te literature (Ciappori et al, 2002; Lundberg and Pollak, 1996; Oreffice, 2007). 6

11 an immigrant couple s decision process, tis would suggest tat te cultural background can inibit tis mecanism and any actual intra-ouseold bargaining, so tat te labor supplies of tis type of couples would not depend on tese factors. Te following empirical analysis specifically tests te predictions of Ciappori et al. (2002) wic were developed for married working couples. Witin tis framework, tis study as well focuses on tose couples were bot individuals are working (estimation including individuals wo do not work troug a Heckman s sample selection model is available upon request). 3. Empirical Specification and Data Description Te sample under analysis consists of married couples wit bot spouses between 18 and 65 years of age. US-born and foreign-born spouses are identified using information on te individual place of birt, furter categorizing foreign-born couples according to teir specific country of origin and to weter usband and wife sare te same cultural background. Spouses are defined to sare te same cultural background if tey are from te same country of origin, or from different countries of origin provided tat tese are caracterized by similar prevailing beliefs in terms of gender roles. To establis tese comparisons, I follow a similar procedure to Carroll et al. (1994) and Antecol (2000), wo identify sets of countries were te prevailing beliefs are comparable to te US ones, and oters were tey are different, assuming tat tere are cultural similarities among te countries of eac of tese groups. More refined groups of countries are used ere tan in Carroll et al. (1994) and more measures of gender roles are considered tan in Antecol (2000) or Carroll et al. (1994), in line wit Guiso et al. (2008). Specifically, I first considered four variables used in te literature as possible measures of gender roles across countries: te gender gap index, te political empowerment index, te female economic activity rate suc as te female labor force participation rate and te ratio of female to male labor force participation rate (age 15 and above), and te uman development index 3. To create a variable tat successfully captures variation in te underlying gender roles across countries, I ten confined te analysis to te gender gap variable and te 3 Te first tree variables are explicitly used by Guiso et al. (2008) and tey all come from te same sources: te Global Gender Gap report by te World Economic Forum (for te first two), and te International Labour Organization (ILO) (for te oter tree measures). I use te year 2006 for all variables since te first two are not available in earlier years. Te first two are defined as ordinal rankings across countries, and te last tree ave been translated into rankings to make tem comparable to te oter variables. 7

12 political index, making tese variables te preferred components of te gender role index constructed by principal component analysis. Tis approac of reducing te dimensionality of tese measures into teir common component is supported by te large Cronbac s alpa reliability coefficient 4. Tis constructed gender role variable captures te underlying common component of gender roles across countries, and is used to divide te observations of married couples into tree groups: tose wose country of origin ave gender roles very similar to te US, i.e., tose wit very similar cultural background to te US, tose wit somewat similar gender roles; tose wit very different cultural background from te US. Tere are 115 countries for wic te World Economic Forum data on gender role measures were collected in 2006 and for wic te above gender role variable as been created, and for 89 of tem te US Census records te detailed birtplace country code in te year Tese are divided as follows to create te tree groups of couples by country of origin in terms of te constructed gender role index. Countries wit culture very different from te US: Saudi Arabia, Yemen, Egypt, Iran, Algeria, Turkey, Jordan, Morocco, Kuwait, Nigeria, Pakistan, Malaysia, Italy, Greece, Sout Korea, Japan, Cile, Albania, Hungary, Cyprus, Guatemala, Uruguay, India, Nepal, Mexico, Honduras, Cameroon, Ukraine, Russia. Countries wit culture somewat different from te US: Indonesia, Romania, France, Kenya, Singapore, Uzbekistan, Cambodia, Slovakia, Brazil, Czec Republic, Ecuador, Poland, Bolivia, Venezuela, Nicaragua, Dominican Republic, Etiopia, Georgia, Moldova, Gana, Macedonia, Tailand, Paraguay, El Salvador, Banglades, Panama, Bulgaria, Trinidad and Tobago, Zimbabwe, Israel, Argentina, Costa Rica, Cina. Countries wit very similar culture to te US: Sweden, Norway, Finland, New Zealand, Denmark, Tanzania, Neterlands, United Kingdom, Germany, Canada, Ireland, Uganda, Australia, Switzerland, Pilippines, Austria, Spain, Latvia, Colombia, Sout Africa, Portugal, Croatia, Sri Lanka, Lituania, Belgium, Jamaica, Peru. Tis full dummy controls approac does not impose any restrictions on te influence of te country of birt, wereas specifications wit female labor force participation or total fertility rates would, so tat te former metod is preferred ere (e.g., 4 A test scale of 0.83 is acieved. Guiso et al. (2008) use tese measures separately and alternatively, and do sow tat te labor force participation measure as a lower correlation to te oters. Tey focus on 40 countries, 30 of wic are OECD countries, no African or Middle Eastern country, and very few from Central and Latin America, wereas tis analysis considers countries in all te five continents, in wic female labor force participation rate can be ig also wit uneven gender roles (several African sub-saaran countries, for instance). Fortin (2005) also uses only 30 countries, none of wic from Africa, Asia, or Sout America. 8

13 Antecol, 2000 for a discussion of tis advantage). Couples of US-born spouses are restricted to non-hispanic wites, to old a uniform reference group representing US culture. Te focus is on first-generation immigrants: on one and, a differential impact of bargaining power forces sould be more pronounced for tose wo actually were born and spent some time in a country different from te US, since recent immigrants are less likely to ave assimilated and mitigated teir culture and beliefs to te US mainstream (Antecol, 2000). On te oter and, it is not possible after 1970 to exactly identify eiter parent s country of birt in te US Census, and terefore te information of primary ancestry is used, tus making it impossible to distinguis second-generation from iger-generation immigrants in te US Census (Fernandez and Fogli, 2009). In addition, several observations ave a non-reported or unreadable ancestry, and several ancestries tat would ave a corresponding value for te gender role index are instead not available in te 2000 US Census. Terefore, a direct comparison in terms of assimilation using te evidence from first- and second and above-generation immigrants cannot be completely establised due to te different origins and composition of te present and past immigration 5. It is crucial to te empirical analysis tat te individuals in te sample face te same markets and institutions, and tat it is possible to estimate te impact of cultural differences on a common set of variables, since tey may differ by teir own and/or teir spouses cultural eritage, as suc eliminating unobserved eterogeneity and comparability concerns wic would result from cross-country estimation from separate data sets (Fortin, 2005). Many cross-country differences in te economic environment are difficult to capture wit aggregate statistics, so tat cross-country analyses may fail because of mis-measuring or omitting important variables. Te advantage of using US Census ouseold level data from a single country is tat all te immigrants face te same economic environment and institutions. Also, witin-country studies provide better controls for uman capital and labor factors, suc as education (Antecol, 2000). Estimation is carried out on te US Census data for te year 2000, specifically its fivepercent sample 5% IPUMS data (1-in-20 national random sample of te population), wic provides te largest sample of ouseolds wit foreign-born spouses, teir detailed etnic, demograpic, labor and income information, along wit standard samples of US-born 5 Evidence on te association of bargaining power and ours of work using spouses primary ancestry instead of country of birt to define cultural background will be presented in Section

14 individuals. Tese data allow to identify te country of birt of ouseold members ( birtplace detailed code), teir ancestry, along wit te number of years already spent living in te US. Country of birt is considered a more robust measure tan ancestry (e.g., Ciswick and Housewort, 2011), and is necessary to identify first-generation immigrants in te US Census Specifically, a random sample (50 percent) of married eads and spouses were extracted from te Census using te variables relationsip to ouseold ead and marital status. Records in tese files were ten matced on te ouseold identification code serial to create a single observation for eac couple. Using te variable sex, couples wit te ead being te usband were ten identified wit a dummy variable index, and individuals sorted into usbands and wives. Te Census defines te ead as te individual wo owns te ousing unit or signs te rental contract, and te partner/spouse is te individual wo identifies imself/erself as suc. Individuals wit imputed values for sex, marital status, relationsip to ouseold ead and country of origin were excluded from te main samples (about 1% of observations are dropped). Tis metod prevents couples from being misclassified, in particular in terms of cultural background. Only couples were bot te ead and te spouse are actually present are considered, wile I exclude ouseolds were tere are multiple spouses, or more tan two adults. Moreover, excluded from te sample are all individuals in scool, in te military, in farm ouseolds, or not working. Individual weigts are used to make te sample representative of te US population and economy. Te following labor supply equations are estimated for usbands and wives, and run separately on eac type of couples, US-born, foreign-born, and foreign-born divided into te tree above categories reflecting different degrees of cultural similarity to te US: w = w w y y diff age _ diff w α 1 ln + α 2 ln + α 3 + γ 1 _ + γ 2 + δ + = w w y y diff age _ diff w β 1 ln + β 2 ln + β 3 + λ1 _ + λ2 + ψ + were te dependent variable is total annual ours worked in te previous year, and y_diff and age_diff are te two bargaining power factors under consideration. Te former is defined as te usband s total non-labor income in dollars minus te wife s total non-labor income in dollars, wile te latter as te usband s age in years minus te wife s age in years. Bot te individual non-labor income and age variables do not ave any missing values and teir differences can be X X ε ε w 10

15 eiter positive or negative, or zero 6. In te Census, te age and all te income questions are asked to eac adult in te ouseold, so tat teir measures are self-reported, rater tan reported by a proxy respondent. Te identification strategy of tese bargaining power factors consists of estimatingγ 1 and γ 2 for usbands, and λ 1 and λ 2 for wives and comparing tese estimates across couples wit different cultural backgrounds. Te role of te non-labor income difference on te labor supply of usbands and wives is captured byγ 1 and λ 1. According to te collective labor supply framework, if a spouse is relatively ricer, ten is/er labor supply sould be lower and te labor supply of is/er mate sould be iger tan in oter ouseolds witout tis income disparity. Hence, γ 1 sould be negative, wile λ 1 sould be positive. Te corresponding coefficients for age difference areγ 2 and λ 2. If being relatively older is a favorable attribute associated to iger intra-ouseold bargaining power, so tat te older spouse s labor supply sould be lower and te labor supply of is/er mate sould be iger tan in te absence of tis age gap, ten γ 2 sould be negative and λ 2 positive. Te oter regressors are te logaritm of te ourly wage rate i w of eac spouse i =, w, te couple s total non-labor income y 7, and a vector of covariates X. X includes education of eac spouse (number of completed years of scooling), number of eac spouse s own cildren living in te ouseold, and only own age of spouse i, so tat te impact of age_diff can be identified. Dummy variables for racial profiles are also included at te individual level. Taking into account education can eliminate te indirect effect tat culture may ave on spouses labor supply coices, as tose individuals (especially women) wo work in te labor market due to teir cultural background are also more likely to be more educated. Tis aspect sould not interfere wit te bargaining power variables. 6 Te ratio of non-labor incomes and te ratio of ages were used as alternative distribution factors. However, te former is not defined for te several couples wit no non-labor income, and tey bot introduce non-linearities in te labor supply equations. 7 All wage and income variables refer to te previous year I consider individuals wo are not self-employed, so tat earned income coincides wit wage income, and non-labor income indeed represents non-earned income sources. Results are robust to te inclusion of te self-employed (less tan 10 percent of my samples). 11

16 Te vector X also includes state fixed effects, wic sould capture te different labor market opportunities and social and legal attitudes toward immigrants tat exist across states. 8 Robust standard errors clustered by state are used to allow for correlation of ouseold observations witin state. I alternatively clustered by metropolitan area (or microdata area puma or conspuma ) 9. Tese specifications do not use a differences-in-differences estimator: usbands and wives regressions are estimated separately, across types of couples. As suc, tey sould not suffer from te understated standard errors igligted by Bertrand, Duflo, and Mullainatan (2004). At any rate, clustering by state (metropolitan area) sould rectify any suc underestimation. [Table 1 about ere] Table 1 presents te descriptive statistics for te usbands and wives main variables, by type of couple. On average, married women are as educated as married men, but are younger, earn a lower wage and work fewer ours tan teir spouses, regardless of teir US or foreignborn status. In te US-born samples, spouses are more similar in terms of age and education, wile te more dissimilar to US culture, te less educated couples are. On average, te age difference is about 1.98 for US-born and 2.73 years for foreign-born ones overall, wile te nonlabor income difference is around $ 2,002 and $ 1,334, respectively. Bot of tese differences exibit a sizable variation in all tese samples, te standard deviations being several times larger tan teir means. 4. Results 4.1 Main evidence Tables 2 and 3 present te results of several regressions were te dependent variable is te usbands or te wives annual ours of work, separately for couples of different cultural backgrounds. [Tables 2 and 3 about ere] 8 Alternatively, I include te state unemployment rate, te state total labor force participation and female labor force participation (retrieved from te Bureau of Labor Statistics), to control for te level of economic activity in a state, and especially for employment opportunities. 9 Te Census reports tat many metropolitan areas ave only been partially identified in 2000, and tat users sould not assume tat te identified portion of a partly-identified metropolitan area is a representative sample of te entire metropolitan area. Terefore, te main specifications are clustered by state. 12

17 Columns 1-4 and sow a negative significant relationsip of te age and income differences on te ours worked by married men, wereas for married women te coefficients are positive significant, controlling for bot spouses demograpic and socioeconomic caracteristics. Te older or ricer usbands are, te lower teir labor supply and te iger te labor supply of teir wives, in te overall sample, in US-born couples, and in foreign-born couples wose cultural background is very similar to te US. Te older (ricer) spouse may old a more favorable bargaining position and works less, wile is/er mate works more. Specifically, in US-born couples, a 5 year difference is associated to working about 12 ours less, and a 5,000 dollars more non-labor income to about 9 ours less. For wives, te corresponding figures are 8 and 50 ours more. Wile te foreign-born wit cultural distance from te US do not exibit any significant association wit tese differences (columns 7-10), tose immigrant ouseolds wit a cultural background very similar to te US sow a significant association to tese bargaining power canges: for usbands, 19 and 5 ours less, and for teir wives 35 and 170 ours more. Te circumstance tat tese married women seem more responsive to bargaining power tan US-born spouses (te coefficients are statistically different), may reflect te fact tat tese type of immigrant couples react to canges in te environment and in outside opportunities, as sown by teir decision to migrate to anoter country. It is important to note tat tere is no teoretical prediction on te significant coefficients being similar or not across groups or between spouses, and we actually observe a more similar ours response to te income difference variable tan to te age difference wen comparing married men to married women, and US-born to foreign-borns wit similar gender role models. Tis could be due to different elasticities of female and male labor supplies, and/or to a different assessment of te relevance of te two bargaining power forces by foreign-born and US-born spouses. As reported in columns 7-10, in couples wose etnicity as a more traditional view on gender roles, spouses do not tend to work more or fewer ours according to te balance of power measures of age and income differences. Te lack of significance cannot be simply due to small sample size since tese traditional couples are muc more numerous tan te foreign-born couples in te oter groups. Tis independence of teir labor supply may strengten te ouseold-bargaining interpretation of te labor supply responses of te oter groups of couples, insofar as tese two specific factors are present in te ouseold power balance, as extensively found in te literature (e.g., Browning, et al., 2014). More traditional gender role attitudes may 13

18 prevent individuals from taking into account of or responding to balance of power incentives. Only immigrant couples wose cultural background is very similar to te US seem to be related wit bargaining power forces in te direction predicted and found for US-born couples. It may also be te case tat a strong disparity in cultural background is associated to more rigid labor supplies due to job type or commitment to work of tese immigrants, so tat te power factor measured ere would not play any role along te labor supply dimension (results are robust to controlling for occupation categories). Couples significant relationsips between tese forces and labor supply are sizable, corresponding to several days of work a year. Te concurrent significance on bot spouses, and wit opposite outcomes, is remarkable given te acknowledged rigidities in te labor supplies, and te frequency of te reported labor supply peaking around 40 ours of work per week. Traditional analyses do not empasize canges by bot spouses, let alone teir labor supply responding to bargaining power forces. No study finds tat ouseolds labor supply decisions are related to differences in nonlabor income ownersip and age according to teir cultural background, suggesting tat tey may reflect te collective ouseold beavior of US-born couples. Moreover, te intraouseold decision process does not appear to vary by foreign status per se, but it depends on te specific cultural views on gender roles. Tese findings point to married US-born and US-similar couples valuing being relatively old, controlling for wages and education of eac spouse, and for individual age. Tis evidence is consistent wit wat is found in te literature, were te spouses age difference is considered a traditional measure of bargaining power, and te older spouse as a favorable position (e.g., Browning et al., 1994; Oreffice, 2011). Tese findings also sow tat income pooling does not old for eiter US-born or immigrant couples wose culture is very similar to te US. So far, te income pooling ypotesis and te prediction tat bargaining power forces are irrelevant to intra-ouseold decisions ad been empirically rejected for several countries and time periods (Browning et al., 1994; Browning et al., 2014; Lundberg, Pollak, Wales, 1996; Scultz, 1990; Tomas, 1990). Instead, te evidence on tose foreign-born culturally different from te US is consistent wit income pooling, wic may indicate a unitary decision-making or inefficiencies, as found occasionally in developing countries (Udry, 1996). 14

19 As to te oter covariates in te labor supply equations, most parameter estimates for all couples are comparable to te literature. In particular, te spouses own wage response is negative significant, as is te cross-wage effects between spouses labor supplies. Te couple s total non-labor income as a negative effect on labor supply in most regressions, wile education as a positive impact, altoug te coefficients are not always precisely estimated. Cildren living in te ouseold are associated wit fewer ours of work for wives, but te opposite olds for usbands, for wom cildren ave a positive effect on labor supply. Tese estimates are in line wit te findings in te ouseold labor supply literature, reported for instance in Blundell and MaCurdy (1999). Adding own age squared and non-labor income squared to te main regressions does not alter te findings and interpretation of te bargaining power forces. Results are also robust to including te education difference between spouses along wit own education (te coefficient of te education difference is not significant, as in Browning et al., 1994), to interacting education wit te bargaining power variables, or to controlling for te number of years spent in te US. Adding to te main specification suc a control for te accumulation of US-specific uman capital, along wit its interactions wit te bargaining variables, does not affect te pattern of significance of te bargaining factors for te group of couples wit similar culture to te US, wit te oter types of foreign-born couples exibiting no significant association wit tese factors. As to te interaction between tese factors and te amount of time in te US, te estimation yields coefficients tat are basically not significant, apart from few coefficients for wives tat are positive for te age difference and negative for te income differences, wereas te estimated coefficient of te number of years spent in te US is always positive and almost always significant. Tis is in line wit te outcome variable being ours worked: better integration means also better knowledge of te caracteristics of te US labor market. [Table 4 about ere] It is important to notice tat assimilation for te group of couples from similar countries to te US is not necessary for bargaining power forces: a variety of empirical studies as already documented te presence of bargaining power responses and evidence consistent wit te collective model in several countries outside te US, suc as Brazil, Denmark, France, te UK, etc. (see for instance Datta Gupta and Stratton, 2010). Tis means tat te response of ours worked to bargaining power forces may not necessarily cange wit time spent in te US. If 15

20 tere were selection into aving migrated versus remaining in country of origin, ten tis sould make it less likely to find differences among te tree groups. Migration is not necessarily a confounding factor: if migrating to te US means a stimulus toward gender parity, ten people from Sweden, for instance, sould be muc less stimulated tan people say from Banglades after teir arrival to te US, te latter would feel muc more liberated, and may respond to bargaining power forces, but tis is not wat tis evidence sows. Finally, allowing for te nonlinearity of te age and non-labor income differences using dummy variables constructed from te terciles of teir distributions leads to estimates consistent wit te main evidence sown above (estimates available upon request). 4.2 Additional findings Next, I estimate te labor supply regressions on te joint sample of US- and foreign-born spouses, including te constructed gender role variable and its interactions wit te bargaining forces as additional controls. Results are presented in Table 5. Table 5 clearly sows tat te interaction terms exibit te expected estimated coefficients: for te wife s regression, bot interactions are negative significant, wile te bargaining variables are positive significant (almost all coefficients significant at te 1% level). Te iger te gender role variable by country of origin, te farter away te country of origin is from te US culture, so tat tese negative coefficients are consistent wit te bargaining power interpretation of tese regressions: a iger value makes te coefficients of age and income differences go toward zero, since for tese cultures te bargaining power impact on labor supply sould be negligible. For te usband s regression, bot interactions are positive, wile te bargaining variables are negative significant (almost all coefficients significant at te 1% level), wic is consistent wit te role of tis gender variable and te bargaining power interpretation. [Table 5 about ere] Tese findings support te contention tat bargaining power forces are associated wit te intra-ouseold allocation of resources and te ours worked by couples wit no cultural disparity from te US in terms of gender roles. However, tey sould be interpreted wit caution since tese interactions assume tat te gender role variable as a continuous linear relationsip wit cultural similarity to te US and wit bargaining power, wic is not warranted. Te measure of gender roles as no cardinal meaning, but ordinal, and ere te empirical exercise is 16

21 trying to capture bargaining power forces and labor supplies. In addition, te interacted gender role variable refers to eiter te usband s or te wife s country of origin (results are robust to using eiter of tose, or to control for bot), so tat tis specification wit interactions is not entirely equivalent to te main estimation by subsamples, since spouses are defined to sare te same cultural background wen tey are from countries wit similar gender roles wit respect to te US (see Section 3). Furtermore, as an alternative to first-generation immigrants born outside te US, second- and iger-generation immigrants are considered ere using te information on primary ancestry of US-born spouses, so tat now culture of origin differs by ancestry rater tan by country of birt. Following, for instance, Antecol (2000) and Fernandez and Fogli (2009), te variable primary ancestry ( ancestr1 ) is used to recover te information on second- and igergeneration immigrants. Since it is not possible after 1970 to exactly identify eiter parent s country of birt in te US Census, te information on ancestry is used, making it impossible to distinguis second-generation from iger-generation immigrants in te US Census (Fernandez and Fogli, 2009). In addition, using ancestry to determine te tree groups reduces te sample size given tat about 17% of te observations ave a non-reported or unreadable ancestry, and several ancestries tat would ave a corresponding value for te gender role variable are instead not available in te 2000 US Census. However, te following subsamples can be constructed: US ancestry is now defined as tose Americans witout foreign ancestry, and foreign ancestry as tose wit it, wo are ten divided into te tree groups according to cultural distance of teir ancestry to te US by te gender role variable. [Table 6 about ere] Table 6 reports te estimated bargaining factors coefficients of te labor supply regressions on te various groups of couples. Te table sows te significance of te bargaining power coefficients for bot te US ancestry and te foreign ancestry groups: now, actually, te foreign group is muc more numerous since it contains US-born, not foreign-borns as wit first-generation immigrants. It is interesting to note tat te very different and somewat different groups are very small and exibit non-significant coefficients (altoug te income coefficient is significant for usbands), wile te group wit similar culture to te US exibits significant coefficients and is muc more numerous tan in te main analysis wit firstgeneration immigrants (foreign-borns), reflecting on one and te large size of te American 17

22 population and on te oter te different origins and composition of te present and past immigration. Altoug te results in Table 6 matc te main evidence presented above, tey sould be interpreted wit caution since a direct comparison in terms of assimilation using firstand second and above-generation immigrants cannot be completely establised due to te abovementioned differences. Te estimated opposite labor supply patterns on bot spouses and for bot factors, and te additional empirical evidence described above are unlikely to old unless te bargaining power explanation and te collective-ouseold approac are correct and applicable to culture affecting ouseold decisions. In addition, te important influence of culture on ouseold production, marriage market matcing, or divorce/remarriages cannot consistently explain te body of results of tis study, as te following discussion will argue. Cultural background and gender role models are associated wit differential productivity in ouseold production and intra-ouseold division of labor. Te labor supplies of foreignborn couples wit strict gender roles may be associated wit stronger ouseold specialization and terefore be muc less responsive to any bargaining power force: wives work more ours and in worse jobs to allow for te usband to make te main uman capital investment (Ciswick and Housewort, 2010) or wives work few ours in te market sector because tey are devoted to ouseold production. Also, immigrants or older US-born individuals may work less in te market because of poor local economic opportunities, or unfriendly attitudes toward immigrants. However, tis study includes individuals wages and education, own age, and state fixed effects (and standard errors are clustered by state or metropolitan area) in te labor supply regressions, wic account for te variation in labor market opportunities and attitudes. Te findings are also robust to adding individual controls for occupation categories. Culture and gender role norms also influence family formation and terefore te matcing patterns of men and women in te marriage market. Te literature on marital selection by differently-aged spouses sows tat men and women in couples wit a large age disparity are negatively selected in earnings (Mansour and McKinnis, 2014). In particular, te estimated negative coefficient on te age difference in te usbands labor supply regressions is consistent wit men being negatively selected into marriage te older tey are wit respect to teir wives, wile teir wives positive coefficient may reflect teir increased labor supply in response to a matc wit a low earning type. It may also be te case tat tis marital selection is not present in 18