Culture and Divorce: Evidence from European Immigrants to the US

Transcription

1 Culture and Divorce: Evidence from European Immigrants to the US Delia Furtado, University of Connecticut and IZA Miriam Marcén, Universidad de Zaragoza Almudena Sevilla-Sanz, University of Oxford December 4, 2009 Abstract: This paper explores the role of culture in determining divorce decisions by examining differences in divorce rates by country of origin of immigrants in the United States. Because immigrants who arrived in the US at a young age all share a common set of American laws and institutions, we interpret cross-ancestry differences in divorce rates as evidence of the effect of culture. Using this basic epidemiological approach, we find that culture has quantitatively significant effects on divorce decisions, more so for women than men. Supplemental analyses suggest the effect of culture is especially strong for immigrants from low divorce countries that reside amidst a large number of co-ethnics. Given the importance of divorce as a determinant of later outcomes in life, our findings suggest that cultural effects should be taken into consideration when formulating family policies. JEL: J12, Z13 Keywords: Divorce, Culture, Immigrants 1

2 1 Introduction Researchers have found that divorce rates increase with more permissive divorce laws (see Wolfers 2006 for US, and González and Viitanen 2009 for Europe). However, despite a fairly homogeneous divorce law regime across Europe at present (e.g., González and Viitanen 2009), divorce rates vary substantially across European countries, ranging from 4.28 divorces per 1,000 inhabitants in Russia to 0.65 in Italy in the year Other determinants of divorce that have been suggested are the economic status of women (Bedard and Deschênes 2005), unemployment rates (Jensen and Smith 1990), female labor force participation rates (Johnson and Skinner 1986, Trent and Scott 1989, Allen 1998 ), gender differences in family roles (Kalmijn et al. 2004), sex ratios (Trent and Scott 1989), public transfers and tax laws (Dickert-Colin 1999), property distribution within marriage (Gray 1998), and laws concerning alimony payments, child support, and child custody (Kiernan 2004). In this paper, we argue that culture also plays an important role. Following Fernandez (2007), we conceptualize culture as a set of beliefs and preferences that vary across time, space, or social groups. The interrelationship between institutions and norms makes it difficult to rigorously disentangle the two. For example, countries in which inhabitants have more liberal attitudes toward divorce enact liberal divorce policies. At the same time, more liberal divorce policies can generate attitudes which are more accepting of divorce. To separate the effect of culture from institutions on an individual s probability of divorce we examine divorce patterns of immigrants who arrived to the US at or under the age of 5. Immigrants in our sample have lived under the laws, institutions, and markets of the United States. However, the attitudes of these immigrants are likely to reflect the attitudes of their parents and ethnic communities, implying that differences in divorce rates by 1 By 2003, almost all countries implicitly or explicitly allowed for a spouse to divorce unilaterally after a required separation period, which was considered proof of the irretrievable breakdown of the marriage. The only two countries that are not categorized as unilateral regimes even in 2003 are Ireland and Italy. In these two countries, even after the separation requirement is fulfilled, a divorce is not granted automatically if one of the spouses opposes to it. 2

3 country of origin may be interpreted as evidence of the importance of culture, and thus that cultural variation might be behind differences in divorce rates observed across European countries. Our results contribute to a growing literature on the effect of culture on economic outcomes (See Fernandez and Fogli 2005 and Guiso, et al for a review of this literature). Using methodologies very similar to ours, fairly recent studies have examined the effect of culture on savings rates (Carroll et al. 1994), female labor force participation (Antecol 2000), fertility and female labor force participation (Fernandez and Fogli 2006, Fernández 2007 and, Fernández and Fogli 2009), living arrangements (Giuliano 2007), unemployment rates (Brügger et al. 2009), and most recently preference for a child s sex (Almond et al. 2009). We add to this work by presenting evidence of the importance of culture on divorce decisions. In our empirical analysis, we use the 2000 US Census to estimate the probability that an European immigrant who arrived in the US as a young child is divorced based on the person s home country crude divorce rate (CDR), defined as the number of divorces in a year per 1,000 inhabitants. Our results suggest that culture plays an important role in explaining divorce even after controlling for an individual s socio-economic characteristics. We find that when the CDR increases by one, the probability that an immigrant in the US is divorced increases by almost three percentage points. In other words, immigrants from Russia, with the highest CDR of 4.28 in our sample, are 10 percentage points more likely to be divorced than immigrants from Italy facing a CDR of 0.65, the lowest among the countries considered. Consistent with the literature on identity (e.g., Akerlof and Kranton, 2000), and findings in Fernandez and Fogli (2009) that husband s culture has a larger effect than own culture on women s work decisions, we also find that women appear to be more sensitive to divorce culture than men. We also explore the role played by other determinants of divorce which vary by country of origin, but have little to do with culture such as household income, religiosity, and marriage quality. This ethnic-specific unobserved heterogeneity might be either embodied in the individual or in her ethnic network (see Fernández and Fogli, 2009). Although our cultural proxy decreases in 3

4 magnitude after including these variables, it remains statistically significant. We further look at the relationship between country of origin divorce rates and the probability of being single to test whether individuals in countries with high divorce rates enter bad marriages to start because they know there are few social sanctions to exiting the marriage. We find however no statistically significant impact of country of origin divorce rates on the probability of being single, which suggests that divorce culture operates through differential tendencies to divorce conditional on marriage quality as opposed to differential tendencies to enter marriages of low quality. Our findings are robust to using various measures of divorce culture such as total divorce rates (TDR) and attitudes about divorce, and a specification including country of origin fixed effects. Evidence seems to suggest that countries where divorce is more common also have higher cohabitation and lower marriage rates. Although the coefficient on the TDR (which measures the number of divorcees per 1,000 married individuals) is half the size that of the CDR, the fact that it is economically and statistically significant suggests that it is actually divorce tendencies, as opposed to marriage tendencies, that are driving our results. Concerned that our results may be driven by abnormally high divorce rates among late universal divorce law adopters, we also use data from the World Values Survey to examine a more direct measure of attitudes about divorce captured by the percentage of the country s population believing that divorce is never justifiable. Our results are not sensitive to our choice of cultural proxy. We also include a specification that relaxes the linearity assumption about the cultural effect by directly including country of origin fixed effects, and results are qualitatively the same. We further explore not only whether culture matters, but also how culture matters and is transmitted. We find that an increase in the concentration of individuals with the same ancestry leads to a smaller decrease in the probability of being divorced for immigrants from countries with relatively higher crude divorce rates. This finding suggests that culture is transmitted not only from parent to child but also within communities. Moreover, because we find evidence of culture even in this specification which includes country of origin fixed effects, we feel more confident that our 4

5 analyses is identifying the role of culture as opposed to unobserved individual characteristics which happen to be correlated within ethnic groups. The paper is organized as follows. Section 2 discusses the empirical strategy. Section 3 describes the data and provides the main evidence of the effect of culture on an individual s divorce probability. Section 4 presents evidence of peer effects in the transmission of culture and divorce. Section 5 concludes. 2 Empirical Strategy Our empirical approach makes use of the fact that all European immigrants who arrived in the US at a young age are, and have been, exposed to the same American markets and institutions. This observation implies that, if only markets and institutions matter, home country divorce rates should have no effect on the divorce decisions of these immigrants. Conversely, if attitudes about divorce get passed down from generation to generation, then home country divorce rates will have a positive effect on divorce probabilities of immigrants and their children. Thus, differences in divorce rates across European countries might be interpreted as evidence of culture. The following equation forms the empirical framework of this analysis: D = DR j + β 2 β 1 X + δ + ε (1) k where D is an indicator variable for whether an individual i of cultural origin j who lives in metropolitan area k reports being divorced. Standard errors are clustered at the country of origin level to account for any within-ethnicity correlation in the error terms. Our measure of culture, DR j, is the crude divorce rate in country j in the year If culture matters, then individuals originating from countries with more liberal cultures regarding divorce should have a higher probability of divorce than individuals from more traditional backgrounds. Thus we expect β 1 to be positive as 5

6 higher crude divorce rates would be associated with more liberal cultural attitudes regarding divorce. The vector X, includes individual characteristics, such as age, education, and gender, which might affect divorce rates for reasons unrelated to culture. A full set of metropolitan statistical area (MSA) fixed effects, denoted δ k, is included because of regional differences in US divorce rates (see, for instance, Gruber 2004, and Friedberg 1998 for an examination of how divorce rates differ by state). Not controlling for MSA fixed effects would be problematic if immigrants from countries with high divorce rates tend to settle in cities with high divorce rates, and might lead to a bias in the culture coefficient as the cultural proxy might be capturing the effect of US divorce laws, rather than the effect of culture. Figure 1 uses data from the World Values Survey and from the UN Demographic Yearbook to show that the percentage of married individuals who report having ever been divorced is higher in those countries which have a higher divorce rate. This positive correlation suggests that in countries where divorce is more common remarriage is also more common, and so the probability that an individual is currently divorced would be lower in a country with higher remarriage rates. This regularity implies that all of our estimates of the effect of culture shown below may be interpreted as lower bounds of the cultural effect on the probability of ever being divorced. 2 3 Data: The five percent Public-Use Microdata Sample (PUMS) 2000 US Census To conduct the main analysis, we use data from the five percent Public-Use Microdata Sample (PUMS) of the 2000 US census. Our sample consists of those immigrants from Europe who arrived in the US when they were less than five years of age. Although they are foreign born, these 2 Arguably a more interesting question to ask would be the effect of culture on the probability that a person ever divorces (conditional on having ever been married). Unfortunately, there is no recent data set that contains information on ever being divorced that also meets all of our variable and sample size requirements. We have estimated equation (1) using the 1980 Census. Results, shown in Appendix A, are qualitatively very similar and may suggest that the bias induced by the different remarriage probabilities is not empirically important. As explained below however, the analysis in this paper uses a much recent Census data to avoid measurement error in our divorce cultural proxy. 6

7 immigrants all grew up under US laws, institutions, and markets. However, because the attitudes of these immigrants are likely to reflect the attitudes of their ethnic communities, we argue that differences in divorce rates by country of origin can be interpreted as evidence of the importance of culture. Given that marriage is a prerequisite of divorce and we do not aim to study marriage decisions, we further restrict the sample to those who are either married or divorced. 3 Finally, we keep only those immigrants whose first reported ancestry corresponds to their country of birth in order to establish a stronger link between culture and place of birth than if we had only used country of birth to define cultural origin. For the main analysis, we use the crude divorce rate (CDR) in 2000 in the immigrant s country of origin as our cultural proxy. The CDR is defined as the number of divorces per thousand midyear inhabitants. The U.N. Demographic Yearbook, our source of data on crude divorce rates, defines divorce as the final legal dissolution of marriage, conferring on both involved parties the right to remarry as defined by the laws of each country. Choosing the same year for the cultural proxy as in our main data set reflects the idea that an immigrant s behaviour is best characterized by the current individual behaviour in his or her country of origin. Another approach would have been to use previous CDRs as our cultural proxies. Given that the average person in our sample is 46 years old, the parents of marriageable age foreign-born individuals in our sample in 2000 most probably arrived in the US in the 1950s. It could thus be argued that it is the divorce rates in the country of origin in the 1950s that should drive divorce probabilities. In Appendix B we report estimates of Equation (1) using crude divorce rates in different years and results remain robust, probably because culture evolves slowly (see Fernandez, 2007). Previous literature using the epidemiological approach to identify the effect of culture typically uses second-generation immigrants instead of first generation childhood arrivers. 3 See Gray (1988) for a discussion of how limiting the sample in this way takes marriage selection into account. 7

8 Unfortunately, after 1970, the US Census stopped asking for parents country of birth. Given our focus on understanding the role of culture in explaining European divorce rates, there is reason to favour a sample of childhood arrivers over the native-born children of immigrants. Certainly, home country culture is more important for a first generation immigrant than a native with one foreign born parent who might have arrived as a young child. 4 Thus, our estimates of the role of culture will more closely resemble the effect of culture in Europe than if we had used a sample of secondgeneration immigrants. There is another compelling reason to use more recent Census data in our main analysis which is particularly relevant for the study of divorce. Because divorce patterns in Europe and in the US have changed so dramatically since 1970, using older Census data is problematic as it increases the error in our cultural proxy, in the sense that cross-country differences in divorce rates as captured by the CDR proxy are less likely to represent differences in underlying beliefs about divorce. For example, many countries experienced a legalization of divorce during this period. Thus, a CDR of zero in a given year might not capture the evolving underlying preferences and attitudes toward divorce that were taking place. Similarly unilateral divorce laws have been found to be usually followed by temporary spikes in the crude divorce rate (see Wolfers 2006 for the US and Gonzalez and Viitanen 2009 for Europe) which presumably do not measure actual changes in attitudes about divorce since culture evolves slowly over time. We have nonetheless run the baseline Equation (1) using these alternative data sets and results are qualitatively the same (see Appendix C). Our final sample consists of 12,076 immigrants from 24 different countries. 5 Table 1 presents summary statistics of the relevant variables by country of origin. Ethnicities are ordered 4 Antecol (2000) finds that the effect of home country female labour force participation rates is stronger for first generation immigrants than for second-and-higher generations. This may be because natives spend more of their adult lives in the host country than immigrants, making them more heavily influenced by the host country s culture than home country culture. 5 United Kingdom includes England, Scotland, Wales and United Kingdom not specified. Czechoslovakia includes the Czech Republic. 8

9 from highest to lowest 2000 CDR in the country of origin. Column (1) shows large CDR variation across countries: from 4.28 divorces per 1,000 inhabitants in Russia to 0.65 in Italy. The other columns describe our PUMS sample. Overall, 15% of individuals are divorced, but Southern Europeans, Poles, Irish, and individuals from Romania or Bulgaria are significantly less likely to be divorced than the average. Immigrants from Portugal, Spain, Greece and Germany tend to be younger than other groups suggesting a relatively more recent arrival of these groups to the US. About 30% of European immigrants have at least a college degree, although education levels range widely across ancestries with Portugal having the lowest proportion of immigrants with at least a college degree (14%) and Switzerland the highest (48%). 4 Results: Culture and Divorce 4.1 Baseline Results Table 2 reports the estimates for the main specification. In all the specifications we group together immigrants from the Russian Federation, Latvia, and Lithuania (as these immigrants all belonged to the USSR at the time period being analyzed). We also group immigrants from Romania and Bulgaria together because they are culturally similar (i.e., they have similar divorce rates over time) and we have few observations for each of them separately. As can be seen in the first column an increase in the crude divorce rate in an immigrant s country of origin is associated with an increase in the probability that that immigrant is divorced. Overall, our estimates indicate that when the number of divorces per thousand in an immigrant s home country increases by one, there is almost a three percentage point increase in the probability that an immigrant gets a divorce. In other words, immigrants from Russia, the country with the highest CDR of 4.28, are about 10 percentage points more likely to be divorced than immigrants from Italy facing a CDR of 0.65, the lowest among the countries considered. 9

10 In column 2 controls for age, gender, and education are added to the specification. The inclusion of these variables has no effect on our parameter of interest the estimated effect of the origin country CDR. Some studies have found that older individuals are less likely to get divorced conditional on being married (Peters 1986). Our results however indicate that older individuals are more likely to be divorced presumably because they have had more time to have ended a marriage. Men are less likely to be divorced, most probably because they are more likely to get remarried. Consistent with the divorce literature, higher levels of education are associated with lower probabilities of divorce, and the college-educated have especially low divorce rates (Becker et al. 1977, Peters 1986). In the third column, metropolitan statistical area (MSA) fixed effects are added to the model and the estimated coefficient on the divorce cultural proxy decreases by almost 25%. This finding suggests that immigrants from high divorce rate countries are indeed more likely to reside in high divorce US cities. We also run a battery of robustness checks and results hold. Panel A in Table 3 show the results from running Equation (1) as a probit instead of as the linear probability model described in Equation (1). Panel B in Table 3 shows the results of running Equation (1) dropping Germany (the biggest country) from the sample. Panel C shows the results dropping Russia (the country with the highest CDR) and Italy (the country with the lowest CDR). Although we do not observe age at marriage directly, Panel D limits the sample to those individuals above the age 30 to see if the positive coefficient on the CDR is picking up the fact that countries with high divorce rates have high divorce rates in the US because they marry young. In all alternative specifications our main results hold. 4.1 Gender and Culture Table 4 presents results separately for men and women. The first and fourth columns show that in models with no controls, women are more heavily influenced by culture than men. In the full specification, shown in the second and fifth columns, both coefficients decrease, but women remain 10

11 about 70 percent more sensitive to home country divorce rates than men. It may thus be that culture, like income, has different effects on men and women. Higher incomes of husbands are associated with smaller likelihoods of divorce while, the earnings potentials of women, if anything, are associated with increases in the probability of divorce (Burgess et al. 2003, Jalovaara 2003). This differential effect of culture on divorce by gender could thus be interpreted in light of identity models (e.g., Akerlof and Kranton, 2000), whereby women loose identity for being divorced to a greater extent than do men given the gendered convention that women should stay by their men. An additional advantage of separating the sample by gender is that it enables us to include more controls. Because divorced women and women who will eventually divorce have higher labor force participation rates than married never-divorced women (Johnson and Skinner 1986), controlling for female wages in divorce equations may bias coefficients and so we do not include wages in our baseline specifications. However, since most men work regardless of marital status, we do not expect this variable to result in biased estimated coefficients in an all-male sample. As can be seen in column 3 of Table 4, a one percent increase in wages results in a five percentage point decrease in the probability that a man is divorced. When wages are included in the specification, the divorce culture coefficient decreases in magnitude slightly, but remain positive and statistically significant. 6 Similarly, because divorced fathers may not reside with their children, controlling for the presence of children in the household may result in endogeneity bias in our male sample (and there are very few observations to actually undertake this analysis). However, since mothers typically reside with their children regardless of marital status, it seems reasonable to control for this variable in our female sample. As can be seen in column 6 of Table 4, the presence of a child in the household decreases the probability that a woman is divorced by about ten percentage points. When 6 Sample sizes are smaller in specifications controlling for wage because they do not include individuals who are not employed. When we compare results from regressions with and without wage controls using just samples of employed workers, conclusions are the same: Although wages do impact divorce tendencies, home country divorce rates remain significant predictors of divorce probabilities even in specifications which control for wages. 11

12 this variable is added to the model, the coefficient on divorce culture decreases in magnitude but remains positive and statistically significant. Mindful of the potential bias which may result, we add wages to the model in Column 7. As expected, female wages are associated with higher divorce rates. Interestingly, the divorce culture coefficient decreases only slightly and remains statistically significant when this variable is added. We conclude, therefore, that divorce culture does not operate solely via changes in fertility or wages within individual families. 4.2 Other Ethnic-Specific Variables Interpretation of the cultural proxy coefficient may be problematic because there could be other determinants of divorce which vary by country of origin, but have little to do with culture. Here we explore the role played by household income, religiosity, marriage quality, and age at marriage, as possible sources of bias. This unobserved heterogeneity might be either embodied in the individual or in her ethnic network (see Fernández and Fogli, 2009). We lack information at the individual level, but we use country-average variables from different sources to capture these unobserved differences across ethnicities (See Data Appendix E for a description of the data sources, and Table D1 in Appendix D for the summary statistics of these variables). Immigrants from poor countries may have lower incomes in the US, even conditional on education. Given the negative relationship between household income and the probability of divorce (Becker et al. 1977), this may lead to a biased coefficient on our cultural proxy. Column 2 in Table 5, we add home country GDP (obtained from UN Statistics and measured in tens of thousands of current US dollars) to the model. There is no change in the estimated CDR coefficient. Another potential source of bias stems from the importance of religiosity in divorce decisions. A person s divorce tendencies may simply be related to personal religious beliefs, and religious beliefs vary systematically across ethnicities. 7 The US Census however does not contain a measure of religiosity. Instead, we add to the model a country of origin level variable obtained 7 The evidence of the effect of religiosity and divorce is mixed. Lehrer and Chiswick (2003) find that more religious individuals have a lower probability of divorce, however Trent and South (1989) fail to find any significant effect of religion on divorce probabilities. 12

13 from the World Values Survey measuring the percentage of the home country population which attends religious services weekly. 8 As shown in Column 3, the sign on religiosity is negative albeit not statistically significant. The CDR coefficient however remains positive and significant even after controlling for religiosity, although the magnitude of the coefficient decreases slightly in magnitude. Ultimately it is difficult to interpret the decrease in the magnitude because religiosity and culture are so intimately intertwined. More socially liberal countries have more lenient attitudes toward divorce and are also less religious (authors calculations from the WVS reveal indeed a negative correlation between church attendance and divorce rates of 0.42). In most contexts, the practice of religion is very social, making it very difficult to separate culture from religion even within a thought experiment. Nonetheless these results seem to suggest that although there is indeed a part of divorce culture that is associated to religiosity, the effect of culture on divorce probabilities extends beyond religious considerations. In column 4, we add country of origin fertility rates to the model since couples with children are less likely to divorce (Becker et al. 1977, Peters 1986 ), and fertility rates are also different across ethnicities (Fernandez and Fogli 2006). Just as is the case for religiosity, it is unclear whether fertility should be thought of as a mechanism through which divorce culture operates or a related variable which needs to be held constant when determining the effect of divorce culture on divorce rates. However, it is comforting that the coefficient on origin country divorce rates remains approximately the same when fertility rates are included as controls. If anything, the estimated divorce culture coefficient increases slightly suggesting a positive relationship between country of origin divorce rates and fertility rates. It is straightforward to interpret divorce culture as a set of social norms that make divorce easier conditional on marriage quality. Two couples that are equally as happy (or unhappy) may 8 We loose information for Norway and Switzerland since information on religiosity is not available for these countries in the WVS. We have re-run our baseline regressions without the observations from these countries and results hold. 13

14 have different divorce probabilities simply because one couple belongs to a culture which places social sanctions on divorcees and another which treats divorcees as equal members of social society. Thus, an alternative explanation for variation in divorce rates across ethnic groups is not related to differential divorce probabilities conditional on marriage quality, but variation in the probability to enter marriages of different qualities. It is certainly difficult, if not impossible for researchers to measure marriage quality. There are however certain marriage types that are more likely to end in divorce, and so we might infer something about their average quality (ref some lit on quality of marriage). For example, interethnic marriages are more likely to end in divorce (Kalmijn 1993). If members of certain ethnic groups are more likely to enter interethnic marriages, then these groups may have higher divorce rates even if sanctions against divorce were equal across groups. To examine this issue, we construct interethnic marriage rates for each ethnic group using Census data on married couples. 9 Results in column 5 suggest that ethnic groups with higher interethnic marriage rates are more likely to divorce. However, adding this variable to the specification does not significantly change the coefficient on divorce culture. Similarly, individuals who marry at a young age may lack maturity to choose optimal spouses, and this may lead to higher divorce rates. For example Becker (1977) suggests that one reason that young marriages are more likely to end in divorce is that they follow a shorter search process. In column 6, country of origin average age at marriage is added to the baseline specification. Again, this has no impact on our measure of divorce culture. Column 7 includes all these country of origin variables together in our regression Equation (1) and results do not changes. It might also be that individuals enter bad marriages to start because they know there are few social sanctions to exiting the marriage. To examine this issue more formally, we look at the relationship between country of origin divorce rates and the probability of never having gotten married. As can be seen in column 8, home country divorce rates have no statistically significant 9 The interethnic marriage rate is defined as the proportion of married men (spouse present) who arrived at the US at or under the age of 5 and that their spouses share a common birth place with him. 14

15 impact on the probability of never having been married. This result suggests that divorce culture operates through differential tendencies to divorce conditional on marriage quality as opposed to differential tendencies to enter marriages of low quality. 4.3 Other Cultural Proxies A country s crude divorce rate is not the only possible measure of home country attitudes toward divorce. Table 6 presents results for the main specification using different measures of divorce culture. To simplify comparisons, we reproduce the preferred specification from Table 2 in Column 1. In the second column, we substitute the origin country s crude divorce rate, defined as the number of divorces per 1000 inhabitants, with the country s total divorce rate, defined as the number of divorces per 1000 married inhabitants. The same country may have a relatively high total divorce rate but low crude divorce rate if it has a relatively small proportion of the population which is married. This measurement difficulty is likely to be exacerbated if, as the evidence seems to suggest, countries where divorce is more common also have higher cohabitation and lower marriage rates. Results shown in Column 2 of Table 6 indicate that total divorce rates also have a positive and significant effect on divorce probabilities. Although the coefficient on the TDR is half the size that of the CDR, the fact that it is economically and statistically significant suggests that it is actually divorce tendencies, as opposed to the marriage tendencies, that are driving our results. Concerned that our results may be driven by abnormally high divorce rates among late universal divorce law adopters, we also examine a more direct measure of attitudes about divorce: the percentage of the country s population believing that divorce is never justifiable. Information about divorce attitudes comes from the World Values Survey (WVS). Respondents are asked whether they think that divorce can always be justified, never be justified, or something in between. (1 Never justifiable,2,, 10 Always Justifiable). Data is available for four waves , , and We pool all the waves together since there is no wave in which all countries considered are available. We have also run further tests using wave 2 ( ), which 15

16 includes information for all countries considered except Greece, and wave 4 ( ), which includes information for all countries except Switzerland and Norway (see Appendix E for definition of variables and data sources). As seen in Column 3, this cultural proxy for divorce also has a significant impact on divorce rates of immigrants residing the US. A ten percentage point increase in the proportion of individuals believing that divorce is never justifiable results in a.2 point decrease in the probability that an immigrant is divorced. Because of the potential disconnect between what individuals say in a survey and their genuine attitudes, the crude divorce rate remains our preferred measure of culture, but again, it is comforting that the results are not sensitive to our choice of cultural proxy. 10 An alternative strategy used in the literature is to include dummy variables for the various countries of origin instead of controlling directly for the divorce rates in these countries (e.g., Carroll et al. 1994, Giuliano 2007, Antecol 2000). This approach does not impose any restrictions on the home country effects, whereas our estimation method from Equation (1) imposes the requirement that the cultural effect is linear in the home country divorce rate. Results in Table 7 suggest that the two approaches lead to similar conclusions. The reference country of origin is Ireland, which is the country with the second lowest CDR of 0.69, followed only by Italy with a CDR of Immigrants from countries with higher CDR than Ireland generally have a higher probability of divorce. We do not find statistically significant differences between the probability of divorce for Irish immigrants and those from Greece, Italy, and Poland, although it is not surprising given that CDR in these countries are also very low. More puzzling perhaps is the fact that immigrants from Denmark and Sweden, both high CDR countries, do not seem to have a significantly higher probability of being divorced than Irish immigrants. Also immigrants from The 10 There is generally a negative relationship between crude divorce rates and the percentage of the population believing that divorce is never justifiable, but Spain, for example, has particularly favorable attitudes toward divorce conditional on their very low divorce rates. This may explain the relatively high proportion of Spaniards divorcees in the US. 16

17 Netherlands seem to have a lower probability of divorce than Irish immigrants, despite The Netherlands having a higher CDR. 5 Cultural Transmission Up until this point, we have considered whether culture is an important determinant of divorce. In this section, we explore how divorce culture gets transmitted from person to person. Parents certainly instil in their children a set of values about family and divorce which gets passed down from generation to generation. In fact, this vertical transmission of culture is often cited as a possible explanation for why children of divorced parents are more likely to get divorced themselves (Amato 1996, Gruber 2004). However, culture can also be horizontally transmitted through the ethnic communities in which immigrants and their children typically live. Local communities can preserve culture either by providing role models for acceptable family arrangements or by punishing behavior which is different from the norm (see Fernandez and Fogli 2009). Several papers have found that communities whose members are more social integrated (as measured by church membership, urbanicity, and population change) have lower divorce rates (Glenn and Shelton 1985, Breault and Kposowa 1987). Since European divorce rates are lower than US divorce rates (see Table 1), if culture is transmitted within local communities, then we might expect that immigrants living in predominantly ethnic communities will be less likely to divorce than immigrants living amidst Americans. Moreover, the effect of ethnic concentration should be particularly strong for immigrants in ethnic groups with especially low divorce rates. To formalize ideas, consider the following equation: α1 Pjk + α 2Pjk * DR j + α3 D = X + δ + γ + e j k where D is an indicator variable for whether individual i of cultural origin j living in metropolitan statistical area (MSA) k is divorced. The proportion of individuals in the immigrant s metropolitan 17

18 area from the same country of origin is denoted, P jk, while δ j is a series of country of origin fixed effects and e is an error term. The other variables are defined as before. If culture is transmitted within communities, we may expect that an increase in the concentration of individuals from one s country of origin should result in a decrease in divorce rates so α 1 should be negative. The country of origin fixed effects will pick up any determinant of divorce which varies by country of origin. This certainly includes the country of origin crude divorce rate used in our original specification, but the fixed effects will also control for unobserved country of origin variables as well as personspecific determinants of divorce which happen to be correlated within ancestry. Our variable of interest is the interaction between ethnic concentration and the crude divorce rate in one s country of origin. As discussed above, an increase in the concentration of same-ethnicity immigrants should decrease divorce rates more for immigrants from countries with low divorce rates than for immigrants from high divorce countries. For example, since the crude divorce rate in Russia is higher than the crude divorce rate in Italy, an increase in the concentration of same-ethnicity immigrants should have a more negative effect on Italians than Russians. In fact, if divorce rates of the Russians that immigrate to the US, surely a non-random sample of Russians, are higher than the divorce rates of Americans, we might expect Russian immigrants surrounded by other Russians have higher divorce rates than Americans. In either case, we expect α 2 to be positive. Table 8 presents regression results of the coefficients of interest, but the full set of baseline control variables are included in the models. As can be seen in the first column, the ethnic concentration has a negative but insignificant effect on divorce rates. When the home country s crude divorce rates is added in the second column, the concentration coefficient remains insignificant although the CDR has the expected positive sign. In the third column, the interaction between origin country CDR and ethnic concentration is added to the model, and as predicted, the coefficient on the interaction is positive and statistically significant. Moreover, when the interaction is included, the coefficient on the concentration variable is negative and highly significant. 18

19 Because in this model, we are identifying the role of culture off of variation in the interaction between ethnic concentration and home country CDR, we can replace home country CDR with home country dummy variables. As can be seen in column 4, our coefficient of interest increases slightly and remains highly significant again confirming that an increase in the concentration of immigrants leads to a larger decrease in the probability of being divorced for immigrants from countries with relatively low crude divorce rates. For example, a ten percentage point increase in the concentration of co-ethnics leads to a 2.9 percentage point decrease in the probability of being divorced for Italians since the crude divorce rate in Italy is 0.65, but only a 0.4 percentage point decrease in the probability that a German is divorced since the German CDR is In fact, the same ten percentage point increase in the concentration of Russians results in a 2.3 percentage point increase for Russians since the Russian crude divorce rate is An attractive feature of this approach is that it can speak to some of the potential sources of bias in our main specification. By replacing the crude divorce rate with country of origin dummy variables, we are implicitly controlling for all of the individual characteristics that are correlated with home country divorce rates but that cannot be interpreted as culture. There may also be problems with this method of identification; immigrants that choose to reside amidst co-ethnics may have preferences and constraints which are similar to those in their ethnic groups. However, it is comforting that the different methods of identifying culture point to the same basic conclusion: The differences in divorce rates in Europe cannot be explained entirely by laws and institutions. Our evidence suggests that culture plays an important role. 6 Conclusion This paper aims at rigorously disentangling the effects of markets and institutions from the effects of culture in determining divorce decisions. Because immigrants that arrived in the US as very young children absorb home country culture from their parents and ethnic communities but are 19

20 exposed to US laws and institutions, we interpret the effect of home country divorce rates on their divorce probabilities as evidence of the role of culture. We find that origin country divorce rates have economically and statistically significant effects on their probabilities of being divorced. We view our results as strong evidence that cross-country variation in divorce laws, welfare policies, and economic conditions in Europe cannot entirely explain the observed variation in divorce rates. Using several techniques, we make a case for the importance of culture in divorce decisions, but acknowledge that our list of controls is rather limited. Determinants of divorce not considered in our analysis include age at marriage, marriage order, premarital childbearing (White 1990), unexpected economic shocks (Weiss and Willis 1997, Ermisch and Böheim 2001), and premarital cohabitation (Lillard et al 1995) among others. Our omission of these variables is partly due to data limitations, but it is unclear whether we would want to include a long list of controls even had the data been available. Attitudes about divorce may impact divorce rates through fertility, religion, and cohabitation decisions and so including these controls would limit the avenues through which culture is allowed to operate. All in all, we view our results as evidence in favor of the role of culture but believe that an examination of the mechanisms through which culture operates is an interesting question for future research. We also note that our identification strategies surely yield lower bounds on the effect of culture in Europe since we focus on immigrants arriving in the US at a young age. Given their decisions to leave Europe, the parents of these immigrants may not be representative of their home country populations. More importantly, childhood arrivers are surely affected by US norms and customs in addition to home country culture. There is a large literature studying the effects of divorce on the socioeconomic outcomes of women and children. Children with divorced parents are typically less well educated, have lower family incomes, marry earlier but separate more often, and have higher odds of adult suicide (see Gruber 2004). On the other hand, women residing in states where divorce is more accessible experience less domestic violence and lower suicide risks (Stevenson and Wolfers 2006). These 20

21 findings suggest that it is crucial to carefully consider all of the determinants of divorce in addition to how and why they matter. Our results can provide insights into the short term and long term effects of changes in divorce-related policies. Cross-country differences in culture may explain the findings in the literature that similar changes in divorce laws have very different effects on divorce outcomes (Poppel and de Beer 1993, Smith 1997, Allen 1998). Moreover, our findings may help explain why policies resulting in small short term increases in divorce rates may have large long term effects because of the way culture responds to changes in laws (e.g., Smith 2002). We leave the examination of these potential multiplier effects for future research. 21

22 References Akerlof G.A. and R. E. Kranton, Economics And Identity, The Quarterly Journal of Economics, 115(3), Allen, D.A., 1998, No-fault divorce in Canada: Its cause and effect, Journal of Economic Behavior and Organization, 37, Almond, D., Edlund, L. and K. Milligan, 2009, Son Preference and the Persistence of Culture: Evidence from Asian Immigrants to Canada, NBER Working Paper No Amato, P. R., 1996, Explaining the Intergenerational Transmission of Divorce, Journal of Marriage and the Family, v58, Antecol, H., An examination of cross-country differences in the gender gap in labor force participation rates, Labour Economics, 7(4), Becker, G.S., Landes; E.M. and M. Robert, 1977, An Economic Analysis of Marital Instability Journal of Political Economy, December 85, Bedard K. and O. Deschênes Sex Preferences, Marital Dissolution, and the Economic Status of Women, Journal of Human Resources, 40(2), Breault, K. D. and A.J. Kposowa, 1987, Explaining divorce in the United States: A study of 3,111 counties in 1980., Journal of Marriage and the Family, 49, Brügger, B., Lalive, R. and J. Zweimüller, 2009, Does Culture Affect Unemployment? Evidence from Rostigraben, DP IZA Burgess, S., Propper,C. and A. Aassve, 2003, The role of income in marriage and divorce transitions among young Americans, Journal of Population Economics. 16, Carroll, C., Rhee, B. and C. Rhee, 1994, Are there Cultural Effects on Saving? Some Cross- Sectional Evidence, Quarterly Journal of Economics, 109(3), Dickert-Conlin, S., 1999, Taxes and transfers: their effects on the decision to end a marriage, Journal of Public Economics, 73, Ermisch, J. and R. Böheim, 2001,. Partnership dissolution in the UK: the Role of Economic ircumstances, Oxford Bulletin of Economics and Statistics, 63(2), Fernández, R., and A. Fogli, 2005, Culture: An Empirical Investigation of Beliefs, Work, and Fertility, NBER Working Paper No Fernández; R., 2007, Women, Work, and Culture, Journal of the European Economic Association, 5(2-3),

23 Fernandez, R. and A. Fogli, 2006, Fertility: The Role of Culture and Family Experience, Journal of the European Economic Association,. 4(2-3), Fernandez, R. and A. Fogli, 2009, Culture: An Empirical Investigation of Beliefs, Work, and Fertility, American Economic Journal: Macroeconomics, American Economic Association, 1(1), Friedberg, L. 1998, Did Unilateral Divorce Raise Divorce Rates? Evidence from Panel Data., American Economic Review, 88(3): Giuliano, P Living Arrangements in Western Europe: Does Cultural Origin Matter?, Journal of the European Economic Association, 5(5), Glenn, N. D. and B.A. Shelton, Regional differences in divorce in the United States. Journal of Marriage and the Family, 47, González, L., and T. K. Viitanen, The Effect of Divorce Laws on Divorce Rates in Europe, European Economic Review 53: Gray, J. S., 1998, Divorce-Law Changes, Household Bargaining, and Married Women s Labor Supply, American Economic Review, 88: Gruber, J., 2004, Is Making Divorce Easier Bad for Children? The Long-Run Implications of Unilateral Divorce, Journal of Labor Economics, 22(4): Guiso, L., Sapienza, P. and L. Zingales, 2006, Does Culture Affect Economic Outcomes, Journal of Economic Perspectives 20(2), Jalovaara, M., 2003, The Joint Effects of Marriage Partners' Socioeconomic Positions on the Risk of Divorce, Demography, 40 (1), Jensen, P. and N. Smith, 1990, Unemployment and Marital Dissolution, Journal of Population Economics, 3, Johnson, W.R. and J. Skinner, 1986, Labor Supply and Marital Separation, American Economic Review 76, Kalmijn, M., 1993, Spouse selection among the children of European immigrants: a comparison of marriage cohorts in the 1960 Census, International Migration Review, 727, Kalmijn M., De Graaf, P.M. and A. Rigt Poortman, 2004, Interactions Between Cultural and Economic Determinants of Divorce in The Netherlands Journal of Marriage and Family, 66 (1), Kiernan, K., 2004, Redrawing the Boundaries of Marriage, Journal of Marriage and the Family, 66, Lehrer, E.L. y C.U. Chiswick 1993, Religion as a Determinant of Marital Stability, Demography, 30 (3),