Fertility implications of family-based regularizations

Depósito de investigación de la Universidad de Sevilla https://idus.us.es/ “This is a pre-copyedited, author-produced version of an article accepted for publication in Journal of Economic Geography following peer review. The version of record Catalina Amuedo-Dorantes, Cristina Borra, Noelia Rivera- Garrido, Fertility implications of family-based regularizations, Journal of Economic Geography, Volume 23, Issue 2, March 2023, Pages 449–484, https://doi.org/10.1093/jeg/lbac023 is available online at: https://academic.oup.com/joeg/articlebstract/23/2/449/6652867?redirectedFrom=fulltext.” Article Title Fertility Implications of Family-Based Regularizations Authors and Affiliations Catalina Amuedo-Dorantes Economics, University of California – Merced, 5200 North Lake Rd., Merced, CA 95343 E-mail: [email protected]du Tel.: 1-619-300-6362 Cristina Borra Economics and Economic History, Universidad de Sevilla, Ramon y Cajal 1, 41018 Sevilla, Spain. E-mail: [email protected] Tel.: +34 954551699 Noelia Rivera-Garrido Economics, Universidad Loyola Andalucía, Avda. de las Universidades s/n, Dos Hermanas, 41704. Sevilla, Spain. E-mail: [email protected] Tel.: +34 955 641 600 Corresponding autor Noelia Rivera-Garrido Economics, Universidad Loyola Andalucía, Avda. de las Universidades s/n, Dos Hermanas, 41704 Sevilla, Spain. E-mail: [email protected] Tel.: +34 955 641 600 Disclosure Statements and Acknowledgements Catalina Amuedo-Dorantes has nothing to disclose. Cristina Borra acknowledges support by the Spanish Ministry of Science, Innovation, and Universities (RTI2018-098217-B-I00, MCI/AEI/FEDER, UE). Noelia Rivera-Garrido acknowledges support by the Spanish Ministry of Science, Innovation, and Universities (PGC2018-093506-B-I00), and the Excelencia-Junta (PY-18-FR-0007). We are grateful to Iñigo Iturbe-Ormaetxe, Pedro Albarrán, Lola Collado, Sara de la Rica, Ana Sanz, Javier Vázquez-Grenno and Marcos Vera, for their helpful comments and suggestions. We also thank conference participants at the XIII Labour Economics Meetings, at the Symposium of Economic Analysis, and at the University of Strathclyde as well as Lidia Farré, Martín Fernández-Sánchez, Joan Monras, Hillel Rapoport, and other seminar participants at the Economics of Migration Senior Seminar series for valuable comments and suggestions to earlier drafts. 1 Fertility Implications of Family-Based Regularizations August 6, 2021 Abstract We examine the fertility impact of a family-based regularization policy granting temporary legal status to unauthorized immigrants based on their offspring’s citizenship. The policy, enacted through a 2011 Royal Decree in Spain, allows for unauthorized parents of eligible nationalities to become temporary legal residents if they have a Spanish minor child. Using data from the Spanish Labor Force Survey (2007-2016), along with a quasi-experimental approach that exploits the change in legal residency eligibility requirements, we show that the policy has significantly increased the childbearing likelihood of eligible mothers, even though the overall increase in fertility nationwide remains trivial. Keywords: Immigrants, Legalization, Fertility, Spain. JEL Codes: J13, J15, K37. 2 1. Introduction This paper examines the fertility impact of a family-based regularization policy aimed at keeping families together by granting undocumented parents legal status based on the citizenship of their offspring. Understanding the fertility impacts of immigration policies is critical for various reasons. First, widening social and economic differences between the developed and developing world continue to stimulate unauthorized migration. 1 In the European Union, approximately 600,000 third country nationals were undocumented in 2017 – 44,000 of whom were residing in Spain. This figure may, nonetheless, understate the total number of unauthorized migrants, as “only persons who are apprehended or otherwise come to the attention of national immigration authorities are recorded in these statistics” (Eurostat, 2019). In the case of Spain, the population of unauthorized migrants averaged 300,000 in 2018 –a figure estimated by subtracting the number of residency permits from the overall number of immigrants recorded in local population registers. 2 Even though the size of the undocumented population has been reduced to half its size in 2009, immigration is still listed as one of the three top issues of concern by 15 percent of the Spanish population (Centro de Investigaciones Sociológicas, 2018). These expressions of concern, along with recent work documenting a clear association between the inflow of less educated immigrants –as it is often the case with undocumented immigrant flows– and a vote shift towards nationalist political parties in Europe (Moriconi et al., 2018; Edo et al., 2019), underscore the continued relevance of undocumented immigration. 1 For this study, we will be using the terms “unauthorized” and “undocumented” interchangeably. 2 Table A.1 in Appendix A shows the estimated number of undocumented immigrants from 2009 to 2018 using data from the population registers, which are deemed to reflect quite accurately the size of the undocumented population in Spain (see Bertoli et al. 2013). 3 Second, social and political reactions to the presence of undocumented immigrants are further reinforced by differences in fertility between immigrants and natives. Immigrant women are generally mothers at an earlier age than their native-born counterparts and have more children (OECD, 2015). In Europe, gross birth rates of the immigrant population averaged 16 newborns per 1000 non-citizens in 2017, compared to about 9 births per 1,000 inhabitants among the native population. In Spain, despite being on the decline, births to noncitizens more than doubled births to Spanish nationals in 2017. 3 While these differential fertility patterns are considered beneficial for the sustainability of the Social Security system in the context of an aging population (Adserà and Ferrer, 2015), as well as essential for the developed world to address population stagnation (Cave, Bubola and Sang-Hun, 2021), they can also be perceived as a threat to national identity (Davis and Deole, 2021; Russo, 2021). Therefore, understanding how immigrant fertility responds to immigration policy is crucial in informing immigration debates and policymaking. We examine the impact of a family-based regularization policy that grants legal status to undocumented parents from certain eligible countries if they have a minor with Spanish citizenship. The policy, enacted by Royal Decree in 2011, was aimed at promoting orderly migration flows and regularizing the immigration status of mixed-status households. Although the purpose of the policy was to protect Spanish children residing in mixed-status households, it may have unintentionally induced eligible migrants to have children. We use a quasiexperimental approach that exploits the change in immigration status available to migrants from eligible nationalities to gauge the policy’s fertility impact. Ideally, we would like to have register data on fertility, nationality, and immigration legal status of women of childbearing age. Unfortunately, such data are not available. Instead, we use data from the Spanish Labor Force Survey on a sample of women 16 to 45 years old over the 2007 through 2016 period. 3 See gross birth rates by European country and citizenship in Table A.2 in Appendix A. 4 Our treatment group is composed of women from eligible countries, which might include some legal migrants. As with other official datasets, information on immigration status is not available in the Spanish Labor Force Survey. However, the inclusion of legal migrants in our treatment group can only bias the estimated policy impact downwards. Our control group are women from non-eligible European countries, other than the EU-15, who are clearly legal and more comparable to migrant women in the treatment group, although we experiment with alternative control groups as part of our robustness checks. Using a difference-in-differences approach, we show that the 2011 Royal Decree has increased the fertility of migrant women from eligible countries by 11.5 percentage points or 48 percent, raising the overall fertility rate by 0.4 percent. The effect is not driven by preexiting differential trends in childbearing across migrants in treatment and control groups; plus, it proves robust to the use of alternative control groups and model specifications. Comparisons of: (a) the policy impact on women’s propensity to have one child born in Spain relative to their propensity to have a second child born in Spain, (b) the policy impact on migrant women’s propensity to have a child born in Spain before vs. after the enactment of the policy, as well as (c) the fertility impact of an earlier generalized amnesty program, all point to the impact primarily stemming from migrants’ yearning to acquire temporary legal status. Lastly, the response appears to have originated from women who were already in Spain when the Royal Decree was enacted, as opposed to new arrivals, suggesting the fertility increase was not the byproduct of a “call effect.” In fact, the size of nuclear households (excluding newborns) did not increase, appeasing concerns of ensuing immigration inflows as part of family reunification. Finally, we do not find evidence of migrant women having children any earlier, although the share of them finding a partner did rise along with childbearing. Relatedly, the increase in fertility stemmed from women with a partner, as opposed to single mothers, who might be more likely to be exposed to poverty. 5 To our knowledge, the study herein is innovative in its analysis of fertility implications of family-based regularizations that grant temporary legal status to undocumented parents based on the citizenship of their offspring. As such, it contributes to an immigration policy literature examining the impact of programs granting legal status on a variety of outcomes ranging from employment and wages (Kossoudji et al., 2002; Kaushal, 2006; Monras et al., 2017; Devillanova et al., 2018), to schooling (Amuedo-Dorantes and Antman, 2017; Felfe et al., 2020), remittances (Amuedo-Dorantes and Mazzolari, 2010), criminal behavior (Pinotti, 2017), and consumption decisions (Dustmann et al., 2017), to name a few. Our focus is on migrant fertility –a topic of concern given fertility busts and the reliance of developed economies on immigrants as a substitute for population stagnation (Cave, Bubola and Sang- Hun, 2021). The analysis also contributes to an extensive literature on fertility responses to various types of policies. Some are non-immigration policies that alter the cost of childbearing through tax incentives (Milligan, 2005; Baughman and Dickert-Conlin, 2003.), child subsidies (Cohen et al., 2013), parental leaves (Lalive and Zweimuller, 2009; Cygan-Rehm 2016), or cash transfers (Nandi and Laxminarayan, 2016). Others are immigration policies, as in the case of studies examining fertility responses to intensified immigration enforcement in the United States (Amuedo-Dorantes and Arenas-Arroyo, 2021), or the granting of birthright citizenship in Germany (Avitabile et al., 2013; Avitabile et al., 2014). Yet, to date, we are unaware of studies examining the fertility implications of regularization programs granting temporary legal status to undocumented parents based on the citizenship of their offspring. Understanding the consequences of this type of policy is relevant, not only for gaining a better understanding of how fertility responds to incentives, but also for identifying unintended effects of alike policy measures. Policies granting parents legal status based on the citizenship of their children are not abundant. Yet, different versions of such policies have 6 been proposed in key immigrant receiving countries, including the United States. The policies can be easily replicable given their underlying motivation is a concern common to many developed nations worldwide. In the case of Spain, the growing share of this type of mixedstatus households in recent years (Figure A.1 in Appendix A) underscores the relevance of policy aimed at protecting children in mixed-status households and keeping families together. However, as we just noted, this challenge is not unique to Spain. In 2014, President Barack Obama proposed the Deferred Action for Parents of Americans and Lawful Permanent Residents (DAPA) in the United States. Even though DAPA was never implemented and did not grant temporary legal status, it allowed certain parents of U.S. citizens and lawful permanent residents to apply for temporary protection from deportation and work permits renewable every three years. As the 2011 Royal Decree examined herein, DAPA’s intent was to protect children and keep families together –also a goal of President Joe Biden’s agenda. 4 Given the incidence of undocumented migration worldwide, understanding the consequences of policies aimed at keeping mixed-status households together is well warranted. The rest of the study is organized as follows. Section 2 reviews the institutional framework, with a detailed description of how children born in Spain to immigrant parents can acquire the Spanish citizenship and of the changes introduced by the 2011 Royal Decree. Section 3 discusses the conceptual framework used to derive some testable hypotheses. Section 4 presents the methodology, whereas Section 5 describes the data and some descriptive statistics. Section 6 discusses our key findings, along with several identification and robustness checks to assess the reliability of our findings. Section 7 considers welfare implications of the family-based regularization program, and Section 8 gauges the relevance of the two alternative mechanisms in explaining the observed fertility impacts. Section 9 concludes the study. 4 See section on “Protect Dreamers and their Families” at: https://joebiden.com/immigration/ 7 2. Institutional Framework 2.1 The Regulation of Spanish Citizenship The Spanish Civil Code establishes under which circumstances an individual can acquire the Spanish citizenship. In general, there is no birthright citizenship. Children born in Spain from immigrant parents are not necessarily Spanish as Spain grants nationality based on the principle of jus sanguinis. Regardless of the place of birth, a child is Spanish if at least one of the parents is Spanish. In some instances, however, the principle of jus soli (birthright citizenship) can be applied. According to the Civil Code art.17, children born in Spain of immigrant parents can acquire the Spanish citizenship if both parents are stateless or the parents’ countries of nationality do not grant nationality to the children. Argentina, Belarus, Brazil, Cabo Verde, Colombia, Costa Rica, Cuba, Guinea Bissau, India, Pakistan (Pakistani born outside of Pakistan), Panama, Paraguay, Peru, Uruguay, Senegal (Senegalese mother and father from another country of those included in this list), Palestine and Sahara (because of stateless parents), Venezuela (father or mother of Venezuela, but not both) and Morocco (Moroccan other and stateless father or Moroccan mother and father from some of the previous countries). 5 , 6 Finally, citizenship may be obtained after one year of continued legal residence (Civil Code art. 22) –a process that, however, only applies to children whose parents are legal residents. 5 Children born in Spain whose parents are from Portugal and Switzerland could also qualify for Spanish citizenship based on art. 17. However, both Portugal and Switzerland are part of the European Union and, as such, migrants from those nationalities would not be irregular immigrants in the first place. Also listed in art. 17 are parents from Principe or San Tome; however, we have no individuals from those nationalities in our sample. 6 See https://www.boe.es/buscar/doc.php?id=BOE-A-2007-7482 for further details. 14 𝛽3= 0 would imply that the policy had no significant impact on migrant fertility, either because they were uninformed, misinformed, or because income and substitution effects cancelled each other out. Equation (1) is estimated by OLS and standard errors are clustered at the country of nationality level. 5. Data and Descriptive Evidence To evaluate the impact of the 2011 Royal Decree on immigrant fertility, we use data from the 2nd quarters of the Spanish Labor Force Survey (Encuesta de Población Activa, EPA). 19 The EPA collects data from 65,000 families (about 160,000 individuals) each trimester, and provides demographic, economic and educational attainment data for each member of the household. Given its reliance on local population registers for sampling design (see footnote 2), the data on the foreign-born documented and undocumented population in the EPA can be reasonably accurate and up to date compared to other sources of data. 20 We focus on the period spanning from 2007 to 2016, which enables us to circumvent changes in legal status due to the entry of Bulgaria and Romania in the EU in 2007. Our sample 19 In line with prior studies, such as Amuedo-Dorantes and De la Rica (2010) or Farré and Rodriguez-Planas (2014) among others, we use the second trimesters of the EPA. This enables us to circumvent having repeated, yet unidentifiable, respondents in the survey, as well as seasonality concerns. Nevertheless, results prove robust to using all trimesters, as shown in Table A.6 in Appendix A. 20 We do not use the Spanish Vital Statistics because it lacks data on non-mothers and mothers’ year of arrival. Consequently, we are unable to model the impact of the policy on likely undocumented women’s childbearing decision. While one could try to exploit the data at the aggregate level, doing so would require data we currently lack on: (1) arrival year of mothers to identify those still likely undocumented, and (2) population figures by nationality and arrival year –information needed to identify those who legalized through prior amnesties and, therefore, might respond differently to the policy. Similarly, microdata from Municipal Population Registers not only lacks information on key demographic controls, such as education, but also on arrival year or years since migration. As such, we are unable to distinguish those who legalized through prior amnesties and respond differently to the policy. Likewise, administrative data from the Spanish Social Security, available for a sample of affiliated individuals, is not suitable for our purposes since it lacks critical information on: 1) illegal workers, 2) non-workers, and 3) migrants’ year of arrival. 15 is composed of immigrant women from countries in the treatment and control groups detailed in the prior section, who reside in Spain and are between 16 and 45 years old, with two exclusions: (1) students, who are most likely legal residents –a requirement to obtain a noncompulsory education in Spain; and (2) migrants who arrived before the prior 2005 regularization, as they would be legal. Figure 2 details how we group respondents in our study. Information on childbearing, nationality and year of arrival allows us to identify those individuals who would have been potentially affected by the 2011 Royal Decree. If they have a child born in Spain, are unauthorized and nationals from one of the eligible countries, they could apply for legal residence through the so-called family settlement criterion. In principle, only unauthorized immigrants would be interested in doing so. However, as pointed out earlier, our treatment group might include some legal migrants, resulting in a lower-bound estimate of the fertility impact of the reform. Table 1 reports summary statistics for women in our sample, distinguishing between those in the treatment versus control groups. 21 Most control variables display similar values across both groups prior to 2012. Note that, although the reform was announced in February 2011 and implemented in April 2011, its impact would not be observed until January 2012 because of the duration of most pregnancies. 22 Additionally, even though fertility differed across the two groups, it trended similarly and in a parallel fashion for both groups of migrant women, as shown in Figure 3. While migrant women in the treatment group display higher 21 See Appendix A, Table A.4, for variable definitions and A.5 for further information on the number of observations per country before and after the reform. As can be observed therein, the proportion of countries remains similar before and after the passage of the 2011 RD, hinting on the lack of significant sample compositional changes during the period under examination driven by inflows or outflows of specific migrant groups. 22 In any case, results prove robust to excluding 2011, as shown in Table A.7 in Appendix A. 16 fertility rates than their counterparts, both groups exhibited similar fertility trends prior to the reform. Nevertheless, their fertility rates diverged thereafter, soaring for women in the treatment group after the enactment of the 2011 Royal Decree, while they continued their prepolicy trend for women in the control group. As such, Figure 3 is suggestive of the appropriateness of our control group (with pre-policy fertility trending parallelly to that of migrants in the treatment group), and of the likely impact of the 2011 Royal Decree on migrant fertility. Finally, Table 2 reports preliminary difference-in-difference estimates of the impact of the 2011 Royal Decree on immigrant fertility. The share of migrant women in the treatment group with a child born in Spain rose by about 21 percentage points from before to after the 2011 Royal Decree –an increase statistically significant at the 1 percent level. Fertility also increased for migrant women in the control group, although by a much lower extent (11 percentage points). As such, the Royal Decree seemingly raised fertility by 10 percentage points. Nevertheless, this estimate fails to account for several characteristics potentially affecting fertility. In what follows, we address this shortcoming with a thorough analysis that controls for all such traits. 6. Did the 2011 Royal Decree Alter Migrant Fertility? 6.1 Main Findings Our main goal is to assess if and how the 2011 Royal Decree, which granted temporary legal status to undocumented migrants if they had a minor child with Spanish citizenship, might have altered immigrant fertility. To that end, Table 3 displays the results from estimating equation (1), where we look at the likelihood of having a child born in Spain. To explore the sensitivity of the coefficient capturing the impact of the Royal Decree on immigrant fertility based on the inclusion of specific regressors, we estimate several specifications that progressively add controls. Our baseline model in column (1) only includes year, province, 17 and nationality of origin fixed effects. In column (2), we further control for basic demographic traits –namely, age, indicators for whether the individual is married or has a college education, the number of children, job tenure and the duration of their migration spell in Spain. Finally, in column (3), we include a province-specific time trend, which we replace with province-year fixed-effects in column (4). Regardless of the model specification used, the 2011 Royal Decree appears to have significantly increased immigrant fertility. The effect only strengthens with the inclusion of additional controls, rising from an 8-percentage-point increase in the likelihood of having a child born in Spain in column (1), to an 11.5-percentage-point increase in column (4). 23 Given that migrant women from eligible nationalities had a 23.9 percent likelihood of having a Spanish child prior to the 2011 Royal Decree, the estimated impact ranges between a 34 percent and a 48 percent increase in the most complete model specification. Consequently, granting undocumented migrants temporary legal status based on having a minor child with Spanish citizenship seems to have significantly raised fertility. While the estimate might seem large, it has not meaningfully altered fertility nationwide because only 0.91 percent of women in Spain were from eligible countries prior to the reform. As such, the 2011 Royal Decree only raised the propensity of having a child born in Spain among women between 16 and 45 years of age by 0.4 percent. Furthermore, the estimated fertility impact is consistent with prior findings in the literature. According to Amuedo- Dorantes et al. (2012), legal immigrants earned 30 percent more than their unauthorized counterparts in Spain back in 2007 –about €200 more per month. Over a 10-year period, this would represent a €30,000 sum assuming a 2 percent interest rate. Hence, the model suggests 23 We also estimate standard errors using bootstrap with resampling weights because of the limited number of clusters and the variation in the number of observations per country of birth. Results remain statistically significant when using bootstrap standard errors with resampling weights (i.e. t=5.7731, p-value=0.0200). 18 that a windfall gain of €30,000 would raise fertility by 48 percent. González (2013) reports how the introduction of a universal, one-time, child benefit amounting to €2,500 raised fertility in Spain by 6 percent in 2007. Based on that estimate, a monetary incentive of €30,000 alone –other legalization benefits set aside– would raise fertility by 72 percent. 6.2 Identification Checks An important threat to identification stems from the possibility that the estimated fertility impacts might be capturing pre-existing differential trends in childbearing across eligible and non-eligible migrants. In other words, fertility for individuals in the control and treatment groups should trend similarly in the absence of the reform. While this assumption cannot be fully tested, we conduct several identification checks to partially address this concern. First, we re-estimate equation (1) including additional interaction terms of our treatment group indicator with dummies for up to four years prior to the enactment of the 2011 Royal Decree. In the absence of any pre-existing differential impacts, the coefficients for the four additional interaction terms should be non-statistically different from zero, whereas the estimated coefficient for the true policy indicator should remain statistically significant and of similar sign and magnitude to the one in Table 3. According to the estimates in the top panel of Table 4, none of the coefficients on the additional interaction terms are statistically different from zero, suggesting that the fertility patterns of migrants in our treatment and controls groups were not statistically different from each other prior to the enactment of the 2011 Royal Decree. However, fertility patterns did differ across the two groups after the enactment of the Royal Decree. The propensity of eligible migrants to have a child born in Spain rose by 12 percentage points (50 percent) –an amount comparable in magnitude to the estimated coefficient on the last column of Table 3. Second, we experiment with restricting our sample to the period prior to the enactment of the 2011 Royal Decree and creating a time trend, which we interact with our treatment group 19 indicator. If migrants in our treatment group already exhibited a differential childbearing pattern when compared to migrants in our control group prior to 2011, the coefficient on the interaction term should be statistically different from zero. According to the estimates displayed in the bottom panel of Table 4, there is no evidence of that being the case. Third, we conduct an event study, which enables us to assess if fertility of migrant women from eligible countries was trending differently from fertility among migrant women in the control group prior to the 2011 Royal Decree, while also allowing us to gauge the longlasting impact of the policy change on migrant fertility. To that end, we estimate the following model: (2) 𝑦𝑖𝑗𝑠𝑡 = 𝛽0+∑𝛽1,𝑡𝑌𝑒𝑎𝑟𝑡 2016 𝑡=2007 + 𝛽2𝐸𝑙𝑖𝑔𝑖𝑏𝑙𝑒𝑖𝑗 +∑𝛽3,𝑡(𝑌𝑒𝑎𝑟𝑡∗ 𝐸𝑙𝑖𝑔𝑖𝑏𝑙𝑒𝑖𝑗) 2016 𝑡=2007 + 𝑋𝑖𝑗𝑠𝑡 ′𝛾 + 𝜆𝑗+ 𝛿𝑠+ 𝛿𝑠𝑡 + 𝜀𝑖𝑗𝑠𝑡 Figure 4 displays the estimated coefficients (𝛽3𝑡), along with 95 percent confidence intervals. All estimates for the years preceding the 2011 Royal Decree are close to zero, strongly supporting the notion of no differential pre-trends in fertility between migrant women in the treatment vs. control groups. In addition, after 2011, we observe a clear break in the fertility trend of migrant women from eligible countries, with fertility rising and the increase remaining statistically different from zero from 2013 through 2015. Altogether, the results in Table 4 and Figure 4 support the assumption of parallel fertility trends among migrant women in the treatment and control groups prior to 2011. Another identification concern refers to the possibility that the captured impacts might be driven, not by the policy change but, rather by changes in the population of migrants in the sample due to either new migrant inflows or outflows. This is unlikely to be the case given the robustness of our findings to the use of alternative migrant samples restricted to those entering the country between 2005 and 2011 (to be discussed in Section 6.3), as well as the similar sample composition by country of origin before and after the enactment of the Royal Decree 20 (Appendix A, Table A.5). Nevertheless, we still assess if the enactment of the 2011 Royal Decree altered the composition of our sample, as captured by their age, years since arrival, marital status, or educational attainment. As shown in Table 5, there is no strong evidence of that being the case, as only the likelihood of having a university degree appears to be marginally different from zero post-2011. 6.3 Robustness Checks We also conduct a variety of robustness checks, all of which are included and discussed in detail in Appendix B. Specifically, we experiment with using alternative treatment and control groups, restricting the sample to new parents who might have distinct childbearing incentives, and using alternative fertility outcomes. We examine the impact of including legal migrants among our treatment group and check the extent to which our findings might be driven by migrants from a specific origin. Results prove robust to the various checks and do not seem driven by migrants from any country. 7. Welfare Implications With the fertility impact of the 2011 Royal Decree clearly established, we consider welfare implications of such a regularization effort. We start by, first, assessing if the estimated fertility impacts are driven by new arrivals, to gauge the potential role of a “call effect”. To address that question, we exclude migrants who arrived after the enactment of the 2011 Royal Decree. As can be seen in column (2) of Table 6, we continue to find evidence of a significant increase in the likelihood of having a child born in Spain of 13.1 percentage points (55 percent) following the adoption of the Royal Decree. Hence, the impact measured by our baseline (a 48 percent increase) is not driven by new arrivals. Next, we explore if the policy appears to have led to family reunification efforts. We can only explore this question at the nuclear household level. Specifically, in column (3), we model household size, excluding new births. As shown therein, household size did not 21 significantly change in response to the policy, assuaging concerns regarding subsequent immigration inflows as part of family reunification. Finally, we explore if the policy appears to have pushed migrant women into having children earlier by modeling their age at first birth. As shown in column (4), we find no evidence of that being the case. However, in column (5), we do find an increased propensity for migrant women to be in a relationship (married or unmarried), which might be a byproduct of settling down after adjusting their immigration status. Relatedly, we explore if fertility increases stemmed from single women. A long-standing child development literature has noted the benefits of being raised in a two-parent household, as opposed to by a single mother (Lundberg et al., 2016). According to the estimate in column (6), the Royal Decree significantly raised the childbearing likelihood of female migrants in a relationship. In conclusion, by granting undocumented migrants temporary legal status if they had a Spanish minor child, the 2011 Royal Decree effectively raised the childbearing propensity of migrants from qualifying nationalities. This impact does not seem to be driven by new arrivals, nor does it seem to have increased household size due to subsequent family reunification efforts. In addition, the policy does not appear to have pushed women into having children earlier, nor does it seem to be driven by single mothers, appeasing concerns of the policy raising the number of children living in poverty-prone singly headed households. 8. Mechanisms: Incentive to Legalize and Income Effects Thus far, we have found that the 2011 Royal Decree significantly rose migrant fertility. In this section, we aim to assess the role played by two potential mechanisms at play –namely, migrants’ incentive to qualify for temporary legal status under the new policy and/or potentially higher incomes from earnings or public benefits they may qualify for once legalized. With that aim in mind, we conduct three analyses. First, we examine how the policy change impacted the likelihood of having a second child born in Spain among mothers from 22 eligible nationalities. These mothers would have already been eligible for temporary legal status. Any further increases in their childbearing would not be reflective of their incentive to qualify for legalization. Column (2) in Table 7 shows that legalization had a positive, albeit far from being statistically different from zero, impact on these mothers’ propensity to have a second child. This finding is suggestive of the fertility increase captured in Table 3 being distinctly driven by something other than solely income effects, such as migrants’ incentive to qualify for temporary legal status. To further check on this interpretation, we next explore if the broad amnesty from 2005, which granted legal status irrespective of parenthood, had similar impacts on fertility to the ones found after the 2011 Royal Decree. If both policy changes (the one from 2005 and the one from 2011) had similar impacts on fertility, it would suggest that the fertility effects from the 2011 Royal Decree were not unique to its policy design. In other words, migrants did not raise their fertility to qualify for temporary legal status under the new policy. To perform this analysis, we select immigrant women from eligible nationalities under the 2011 Royal Decree who arrived in Spain after 2001 –when another broad amnesty took place. Because some EU migrants included in our control group might not have been legal (their countries had not yet joined the EU in 2001), we compare the likelihood of having a child in Spain of migrant women arriving before August 2004 (i.e. women who would have qualified for the amnesty) to the one of women arriving afterwards. Note that we can only observe these women in 2006, prior to yet another amnesty in 2007. 24 In other words, our treatment group is composed of women with eligible nationalities (based on the 2011 Royal Decree) who had been in Spain for 2-3 years, and their counterparts with less than 2 years. Our control group are migrant women with the same nationalities observed in 2004 and 2005. 24 The continuous regularization mechanism in place since 2006 would have impacted fertility from 2007 onwards. 23 According to the estimate in Column 3, Table 7, the impact is, once more, positive but not statistically different from zero at conventional levels, potentially due to the lack of power. 25 As such, income effects stemming from higher earnings following legalization, while undoubtedly important, cannot explain the observed increase in fertility that followed the 2011 Royal Decree. Finally, to further gauge the role of migrants’ desire to obtain temporary legal status relative to that of an income effect in the observed fertility increases, we conduct a triple differences analysis that compares changes in fertility of migrant women with a child born in Spain before 2011 to changes in fertility of migrant women without a child born in Spain prior to 2011 as follows: (3) 𝑦𝑖𝑗𝑠𝑡 = 𝛽0+ 𝛽1𝑃𝑜𝑠𝑡𝑡+ 𝛽2𝐸𝑙𝑖𝑔𝑖𝑏𝑙𝑒𝑖𝑗 + 𝛽3(𝑃𝑜𝑠𝑡𝑡∗ 𝐸𝑙𝑖𝑔𝑖𝑏𝑙𝑒𝑖𝑗)+ 𝛽4(𝑃𝑜𝑠𝑡𝑡∗ 𝑁𝑜𝐶ℎ𝑖𝑙𝑑𝑖𝑗)+ + 𝛽5𝑁𝑜𝐶ℎ𝑖𝑙𝑑𝑖𝑗 + 𝛽6(𝑃𝑜𝑠𝑡𝑡∗ 𝐸𝑙𝑖𝑔𝑖𝑏𝑙𝑒𝑖𝑗 ∗ 𝑁𝑜𝐶ℎ𝑖𝑙𝑑𝑖𝑗)+ 𝛽7(𝐸𝑙𝑖𝑔𝑖𝑏𝑙𝑒𝑖𝑗 ∗ 𝑁𝑜𝐶ℎ𝑖𝑙𝑑𝑖𝑗)+ + 𝑋𝑖𝑗𝑠𝑡 ′𝛾+ 𝜆𝑗+ 𝛿𝑠+ µ𝑡+ 𝛿𝑠𝑡 + 𝜀𝑖𝑗𝑠𝑡 where 𝑁𝑜𝐶ℎ𝑖𝑙𝑑𝑖𝑗 equals 1 if the ith migrant woman with nationality j did not have a Spanish minor in 2011. As shown in Table 8, the policy raised the childbearing propensity of migrant women from eligible nationalities who did not have a child born in Spain prior to 2011 by 10 percentage points –that is, by an amount similar to the impact in Table 3. However, it had no significant impact on the childbearing propensity of women who already had a child born in Spain prior to the reform. All in all, the evidence from columns (2) and (3) in Table 7 and the results in Table 8 point to migrants’ incentive to qualify for temporary legal status as a crucial factor in explaining the impact of the 2011 Royal Decree on migrant fertility. 25 Table A.8 in Appendix A conducts the analysis using migrant women from all eligible nationalities –that is, migrant women with non-EU nationalities (the latter are considered legal migrants). The use of a larger sample allows us to identify a marginally significant positive impact of the 2005 broad amnesty on fertility of 17 percent. 30 Hotz, V. J., Klerman, J. A., Willis, R. J. (1997) The economics of fertility in developed countries. Handbook of Population and Family Economics, 1: 275-347. Hotz, V. J., Miller, R. A. (1988) An Empirical Analysis of Life Cycle Fertility and Female Labor Supply. Econometrica: Journal of the Econometric Society, 56.1: 91–118. Kaushal, N. (2006) Amnesty Programs and the Labor Market Outcomes of Undocumented Workers. Journal of Human Resources, 41.3: 631–647. Kossoudji, S. A., Cobb‐Clark, D. A. (2002) Coming out of the Shadows: Learning about Legal Status and Wages from the Legalized Population. Journal of Labor Economics, 20.3: 598-628. Lalive, R., Zweimüller, J. (2009) How Does Parental Leave Affect Fertility and Return to Work? Evidence from Two Natural Experiments. The Quarterly Journal of Economics, 124.3: 1363-1402. Larramona, G., Sanso‐Navarro, M. (2016) Do Regularization Programs for Illegal Immigrants Have a Magnet Effect? Evidence from Spain. The Manchester School, 84.2: 296- 311. Lundberg, S., Pollak, R. A., Stearns, J. (2016) Family Inequality: Diverging Patterns in Marriage, Cohabitation, and Childbearing. Journal of Economic Perspectives, 30.2: 79–102. Milligan, K. (2005) Subsidizing the stork: New evidence on tax incentives and fertility. Review of Economics and Statistics, 87.3: 539-555. Monras, J., Vázquez-Grenno, J, Elias Moreno, F. (2017) Understanding the Effects of Legalizing Undocumented Immigrants. IZA Discussion Papers 10687. Moriconi, S., Peri, G., Turati, R. (2018) Skill of the Immigrants and Vote of the Natives: Immigration and Nationalism in European Elections 2007-2016. NBER Working Papers 25077. 31 Nandi, A., Laxminarayan, R. (2016) The unintended effects of cash transfers on fertility: evidence from the Safe Motherhood Scheme in India. Journal of Population Economics, 29.2: 457-491. OECD. (2015) Socio-demographic characteristics of immigrant populations. Paris: OECD. Pinotti, P. (2017) Clicking on heaven's door: The effect of immigrant legalization on crime. American Economic Review, 107.1: 138-68. Sabater, A., Domingo, A. (2012) A new immigration regularization policy: The settlement program in Spain. International Migration Review, 46.1: 191-220. Willis, R. (1973) A New Approach to the Economic Theory of Fertility Behavior, Journal of Political Economy, 81.2: 14–64. 32 Table 1: Descriptive Statistics Variables Treatment Group Control Group Pre-RD Post-RD Pre-RD Post-RD Panel A: Outcomes Child Born in Spain 0.24 0.46 0.15 0.29 (0.43) (0.50) (0.36) (0.45) Child Born in Spain <5 0.23 0.39 0.14 0.23 (0.42) (0.49) (0.35) (0.42) Child Born in Spain <3 0.21 0.28 0.12 0.16 (0.41) (0.45) (0.33) (0.37) Number of Children Born in Spain 0.27 0.68 0.16 0.36 (0.52) (0.87) (0.40) (0.62) Number of Children Born in Spain <5 0.25 0.48 0.15 0.26 (0.50) (0.67) (0.38) (0.51) Number of Children Born in Spain <3 0.22 0.29 0.13 0.17 (0.45) (0.50) (0.35) (0.40) Panel B: Controls Age 30.45 32.08 29.76 31.56 (7.14) (6.85) (7.70) (7.00) Job Tenure 7.46 12.60 8.97 16.22 (13.98) (24.92) (15.88) (25.88) Married 0.61 0.65 0.53 0.55 (0.49) (0.48) (0.50) (0.50) University Education 0.14 0.14 0.11 0.15 (0.35) (0.35) (0.31) (0.35) Number of Children 0.58 0.70 0.51 0.52 (0.98) (0.97) (0.83) (0.78) Years in Spain 2.08 5.49 2.18 5.83 (1.45) (2.42) (1.50) (2.41) Observations 1,893 2,936 1,076 1,378 Source: EPA 2007-2016 and authors’ own tabulations. Pre- and Post-RD stand for before and after the enactment of the 2011 Royal Decree. 33 Table 2: Difference-in-Differences in the Probability of Having a Child Born in Spain Treatment Group Control Group DD Pre Post DT Pre Post DC (DT-DC) Child Born in Spain 0.248 0.458 0.211*** 0.160 0.271 0.110*** 0.101** (0.432) (0.498) (0.034) (0.367) (0.444) (0.015) (0.037) N 1,893 2,936 4,829 1,076 1,378 2,454 7,283 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group (CG) is composed of women from European nationalities, excluding EU- 15 countries. The treatment group (TG) is composed of mothers from eligible nationalities under the reform. Standard deviations and robust standard errors clustered at the country of nationality level are displayed in parentheses. All regressions include a constant term. *** p<0.01, ** p<0.05, * p<0.1. Source: EPA 2007-2016. 34 Table 3: The Impact of the 2011 Royal Decree on Fertility Model Specification: (1) (2) (3) (4) Baseline With Demographic Traits With Demographic Traits + Province Time Trend With Demographic Traits + Province- Year FE Post * Eligible 0.081** 0.092*** 0.107*** 0.115*** (0.033) (0.025) (0.021) (0.019) Year FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls N Y Y Y Province-Year FE or Trend N N Province-Trend Y Dep. Var. Mean 0.239 Observations 7,283 7,283 7,283 7,283 R-squared 0.165 0.241 0.246 0.290 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group (CG) is composed of women from European nationalities, excluding EU- 15 countries. The treatment group (TG) is composed of mothers from eligible nationalities under the reform. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. All model specifications include a constant term. Specification (1) includes year and country of nationality fixed effects. Specification (2) adds age, marital status, university education, number of children, job tenure and years since arrived in Spain. Specification (3) adds province-year time trend. Finally, specification (4) includes province-year fixed effects. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. The dependent variable mean refers to eligible women with at least one child born in Spain before the RD 2011. *** p<0.01, ** p<0.05, * p<0.1. Source: EPA 2007-2016. 35 Table 4: Identification Check #1: Assessing the Assumption of Parallel Pre-trends Key Regressors Using Placebos for Up to 5 Years Earlier Post x Eligible 0.122*** (0.026) Placebo 2010 x Eligible -0.019 (0.024) Placebo 2009 x Eligible -0.010 (0.041) Placebo 2008 x Eligible 0.042 (0.064) Placebo 2007 x Eligible -0.004 (0.026) Observations 7,283 R-squared 0.290 Key Regressors Using a Trend Eligible x Trend 0.005 (0.010) Observations 2,969 R-squared 0.271 Year FE Y Province FE Y Nationality FE Y Other Controls Y Province-Year FE Y Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The treatment group (TG) is composed of mothers from eligible nationalities. The control group (CG) is composed of women from European nationalities, excluding EU-15 countries. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. The model specification includes a constant term; year, country of nationality and province-year fixed effects, age, gender, marital status, an indicator for university education, number of children, job tenure and years since arrived in Spain. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 36 Table 5: Identification Check #2 – Assessing Changes in Sample Composition Model Specification: (1) (2) (3) (4) Married University Education Age Years since Arrival Post * Eligible -0.003 -0.046* -0.228 -0.119 (0.024) (0.025) (0.391) (0.102) Year FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls Y Y Y Y Province-Year FE Y Y Y Y Observations 7,283 7,283 7,283 7,283 R-squared 0.280 0.205 0.306 0.609 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The treatment group (TG) is composed of mothers from eligible nationalities. The control group (CG) is composed of women from European nationalities, excluding EU-15 countries. All model specifications include a constant term; year, province, and nationality fixed effects; province-year fixed-effects, and the following demographic controls if they are not the ones being modelled, i.e. age, marital status, university education, number of children, job tenure and years since arrived in Spain. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 37 Table 6: Welfare Implications Model Specification: (1) (2) (3) (4) (5) (6) Baseline Excluding Post-2011 Arrivals Number of Household Members Age at Motherhood Partnership Single Mothers Post * Eligible 0.115*** 0.131*** -0.051 -0.522 0.111*** 0.042 (0.019) (0.023) (0.079) (0.571) (0.033) (0.031) Observations 7,283 6,796 7,283 2,329 7,283 1,799 R-squared 0.290 0.296 0.296 0.310 0.197 0.332 Dep. Var. Mean 0.239 0.239 3.437 27.618 0.320 0.061 Year FE Y Y Y Y Y Y Province FE Y Y Y Y Y Y Nationality FE Y Y Y Y Y Y Other Controls Y Y Y Y Y Y Province-Year FE or Trend Y Y Y Y Y Y Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. The treatment group is composed of mothers from eligible nationalities potentially affected by the reform. The control group is composed of women from European nationalities excluding EU-15 countries. Column (1) includes baseline estimates from Table 4. Column (2) excludes migrants who arrived after 2011. Column (3) models household size excluding children born in Spain to better capture family reunification. Column (4) models age at motherhood, whereas column (5) models the likelihood of having a partner. Column (6) goes back to modeling the likelihood of having a Spanish child among single women. All models include a constant term; year, country of nationality and province-year fixed effects, age, gender, marital status, an indicator for university education, number of children, job tenure and years since arrived in Spain. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 38 Table 7: Mechanisms #1 Model Specification: (1) (2) (3) Baseline Model Having a Second Child Having One Spanish Born Child after the 2005 Royal Decree Post * Eligible 0.115*** 0.070 0.082 (0.019) (0.074) (0.052) Year FE Y Y Y Province FE Y Y Y Nationality FE Y Y Y Other Controls Y Y Y Province-Year FE or Trend Y Y Y Dep. Var. Mean 0.239 0.347 0.272 Observations 7,283 1,881 938 R-squared 0.290 0.448 0.337 Notes: In columns (1) and (2), the sample are women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group (CG) is composed of women from European nationalities, excluding EU-15 countries. The treatment group (TG) is composed of mothers from eligible nationalities under the reform. Post equals 1 for observations from after 2011. In column (1), the dependent variable is a dummy equal to 1 if the woman has a child born in Spain. In column (2), which further restricts the sample to mothers, the dependent variable equals 1 if the mother has at least two children born in Spain. In column (3), the sample are women between 16 and 45 years of age from eligible nationalities potentially affected by the 2011 reform, who are not students and arrived in Spain after the 2001 regularization. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. The CG is composed of women who had been in Spain for less than two years, whereas the TG are women who had been to Spain two or three years. Post equals 1 for observations from after 2005. All models include a constant term, as well as the controls noted in the table. The dependent variable mean refers to eligible women with at least one child born in Spain before the 2011 Royal Decree in columns (1) and (2), and prior to 2005 in column (3). Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Source: EPA 2007-2016 in columns (1) and (2). EPA 2004-2006 in column (3). 39 Table 8: Mechanisms #2 Model Specification: (1) (2) (3) (4) Baseline With Demographic Traits With Demographic Traits + Province Time Trend With Demographic Traits + Province-Year FE Post * Eligible * No Child 0.094** 0.092** 0.093** 0.104** (0.043) (0.044) (0.044) (0.045) Post * Eligible -0.026*** -0.023** -0.014 -0.016 (0.008) (0.011) (0.011) (0.012) Post * No Child 0.087*** 0.083*** 0.086*** 0.082*** (0.011) (0.014) (0.011) (0.007) No Child -0.984*** -0.967*** -0.970*** -0.970*** (0.008) (0.011) (0.008) (0.007) Year FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls N Y Y Y Province-Year FE or Trend N N Province-Trend Y Dep. Var. Mean 0.239 Observations 7,283 7,283 7,283 7,283 R-squared 0.769 0.774 0.776 0.784 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group is composed of women from European nationalities, excluding EU-15 countries. Eligible is an indicator variable for mothers from eligible nationalities. No Child is an indicator variable for lacking a child born in Spain in 2011. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. All model specifications include a constant term. Specification (1) includes year and country of nationality fixed effects. Specification (2) adds age, marital status, university education, number of children, job tenure and years since arrived in Spain. Specification (3) adds province-year time trend. Finally, specification (4) includes province-year fixed effects. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. The dependent variable mean refers to eligible women with at least one child born in Spain before the RD 2011. *** p<0.01, ** p<0.05, * p<0.1. 46 Table A.3: Applications and Concessions of Family Settlement (2006-2014) Year Applications Concessions 2006 1,199 585 2007 2,823 1,917 2008 1,745 920 2009 1,183 767 2010 1,123 484 2011 12,754 5,728 2012 10,859 11,961 2013 8,518 7,367 2014 7,915 6,447 2006-2010 8,073 4,673 2011-2014 40,046 31,503 Total 48,119 36,176 Source: Cerezo (2016) and authors’ own tabulations. 47 Table A.4: Definitions of Key Variables Age Individual Age (in years) Job Tenure Work Length (in months) Married Dummy Variable 1- Be married 0- Otherwise University Dummy Variable 1- Have University Education 0- Otherwise Number of Children Number of children prior to the modeled birth. Years in Spain Number of Years of Residence in Spain Child Born in Spain Dummy Variable 1- Have a Child Born in Spain 0- Otherwise Child Born in Spain <5 Dummy Variable 1- Have a Child Born in Spain (< 5 years old) 0- Otherwise Number of Children born in Spain Number of Children Number of Children born in Spain <5 Number of Children (< 5 years old) 48 Table A.5: Sample Sizes by Country Pre- RD Post- RD N Share N Share Control Group Bulgaria 111 10.32% 133 9.65% Cyprus 0 0.00% 2 0.15% Slovenia 4 0.37% 2 0.15% Estonia 2 0.19% 1 0.07% Hungary 8 0.74% 14 1.02% Latvia 1 0.09% 2 0.15% Lithuania 3 0.28% 15 1.09% Poland 45 4.18% 99 7.18% Czech Republic 4 0.37% 12 0.87% Slovakia 4 0.37% 13 0.94% Romania 894 83.09% 1.085 78.73% Treatment Group Belarus 4 0.2% 14 0.5% Cabo Verde 0 0.0% 6 0.2% Morocco 579 30.4% 1.145 38.72% Senegal 26 1.4% 45 1.5% Western Sahara 6 0.3% 6 0.2% Guinea-Bissau 8 0.4% 1 0.0% Costa Rica 1 0.1% 3 0.1% Cuba 84 4.4% 136 4.6% Panama 2 0.1% 5 0.2% Argentina 134 7.0% 145 4.9% Brasil 190 10.0% 212 7.2% Colombia 315 16.6% 427 14.4% Paraguay 197 10.4% 279 9.4% Peru 200 10.5% 250 8.5% Uruguay 56 2.9% 52 1.8% Venezuela 82 4.3% 159 5.4% Palestine 0 0.0% 1 0.0% India 19 1.0% 70 2.4% Pakistan 0 0.0% 1 0.0% Source: EPA 2007-2016 and authors’ own tabulations. 49 Table A.6: The Impact of the 2011 Royal Decree on Fertility (Using All Trimesters) Model Specification: (1) (2) (3) (4) Baseline With Demographic Traits With Demographic Traits + Province Time Trend With Demographic Traits + Province- Year FE Post * Eligible 0.066** 0.074*** 0.079*** 0.092*** (0.032) (0.022) (0.019) (0.022) Trimester FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls N Y Y Y Province-Trimester FE or Trend N N Province-Trend Y Dep. Var. Mean 0.228 Observations 28,022 28,022 28,022 28,022 R-squared 0.167 0.249 0.253 0.294 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group (CG) is composed of women from European nationalities that do not include the countries in the UE-15. The treatment group (TG) is composed of mothers from eligible nationalities potentially affected by the reform. All model specifications include a constant term. Specification (1) includes trimester and country of nationality fixed effects. Specification (2) adds age, marital status, university education, number of children, job tenure and years since arrived in Spain. Specification (3) adds province-trimester time trend. Finally, specification (4) includes province-trimester fixed effects. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 50 Table A.7: The Impact of the 2011 Royal Decree on Fertility (Excluding 2011) Model Specification: (1) (2) (3) (4) Baseline With Demographic Traits With Demographic Traits + Province Time Trend With Demographic Traits + Province- Year FE Post * Eligible 0.074* 0.082*** 0.095*** 0.111*** (0.036) (0.025) (0.022) (0.021) Year FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls N Y Y Y Province-Year FE or Trend N N Province-Trend Y Dep. Var. Mean 0.211 Observations 6,492 6,492 6,492 6,492 R-squared 0.172 0.246 0.525 0.295 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group (CG) is composed of women from European nationalities that do not include the countries in the UE-15. The treatment group (TG) is composed of mothers from eligible nationalities potentially affected by the reform. All model specifications include a constant term. Specification (1) includes year and country of nationality fixed effects. Specification (2) adds age, marital status, university education, number of children, job tenure and years since arrived in Spain. Specification (3) adds province-year time trend. Finally, specification (4) includes province-year fixed effects. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 51 Table A.8: The Impact of the 2005 Normalization on Fertility Model Specification: (1) (2) (3) (4) Baseline With Demographic Traits With Demographic Traits + Province Time Trend With Demographic Traits + Province- Year FE Post * Eligible 0.045** 0.034 0.033 0.039* (0.020) (0.025) (0.023) (0.022) Year FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls N Y Y Y Province-Year FE or Trend N N Province-Trend Y Dep. Var. Mean 0.226 Observations 2,327 2,327 2,327 2,327 R-squared 0.142 0.203 0.229 0.245 Notes: The sample are immigrant women of non-European nationality between 16 and 45 years of age, who are not students and who arrived in Spain after the 2001 regularization. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. The CG is composed of women who had been in Spain for less than two years, whereas the TG are women who had been to Spain two or three years. Post equals 1 for observations from after 2005. All models include a constant term, as well as the controls noted in the table. The dependent variable mean refers to eligible women with at least one child born in Spain prior to 2005. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 52 Figure A.1 Share of Non-European Women with a Spanish Child Source: EPA 2006-2011 and authors’ own tabulations. 53 Appendix B B.1. Robustness Checks After addressing identification concerns, we conduct several robustness checks aimed at gauging the reliability of our findings to further specification tests. First, we explore if the found impacts are robust to the use of alternative control and treatment groups (See summary statistics for theses alternative treatment and control groups in Table B.1). This is done in columns (2) through (5) of Table B.2 (column (1) replicates our baseline result from column (4) in Table 3). Specifically, column (2) documents how the propensity to have a child born in Spain rises by 7 percentage points (almost 30 percent) when the control group includes all European migrants. In column (3), we use, instead, Spanish women in the control group. We continue to find that the 2011 Royal Decree raised the likelihood of having a child born in Spain by 8 percentage points (slightly over 30 percent). Lastly, in column (4), we experiment with restricting the entire sample to women from eligible nationalities. Our control group becomes women with a child born in Spain prior to the reform. Women without a child born in Spain before the enactment of the Royal Decree continue to be our treatment group. We conduct the analysis for the years after the enactment of the Royal Decree. To make the two groups as comparable as possible, we focus on women in prime childbearing ages (16 to 35 years of age) living with a partner since mothers are more likely to live with a partner than nonmothers. Again, we find that the 2011 Royal Decree raised the probability of having a child born in Spain by 5 percentage points. 26 Next, we experiment with changing the treatment group. Because some migrants, unaware of which nationalities were eligible to request temporary legal status after having a child born in Spain, might have responded to the Royal Decree, we include all foreign-born 26 We do not compute a percentage increase since the treatment group had no Spanish children prior to the reform. 54 parents in the treatment group, except for those from EU member countries. As shown in column (5), we continue to find that the Royal Decree raised these women’s childbearing propensity by 10 percentage points (48 percent). After assessing the robustness of our results to alterations to both treatment and control groups, we experiment with limiting our sample to new parents. Their childbearing incentive might differ and, in turn, their responsiveness to the policy change. This check also allows us to rule out the selection of women into motherhood as the mechanism driving the estimated policy impact. According to the coefficient in column (6) in Table B.2, these parents responded similarly to the ones included in our baseline specification, raising their propensity to have a child born in Spain by 9 percentage points. In columns (7) through (9) in Table B.2, we conduct additional robustness checks aimed at assessing the impact of potentially including legal migrants among our treatment group. As noted earlier, the EPA does not have information on the legal immigration status of migrants. If legal migrants are included in the treatment group, our baseline estimate might be potentially biased downwards. To assess the degree to which that might be the case, we first consider excluding migrants who could easily be legally in the country by fulfilling one of the following criteria: (a) being married to a Spanish citizen (b) having a father or mother originally born in Spain, or (c) being born in Spain and having lived in the country for 3 years before turning 18 years old. According to the estimate in column (7), migrants in this newly redefined treatment group are 12 percentage points (51 percent) more likely than migrants from European countries from outside the EU15 to have a child born in Spain. Next, in column (8), we experiment with excluding migrants with, at least, a 6-month tenure in the job. As noted earlier in Section 2, unauthorized migrants could gain legal status through the labor settlement criterion if they could prove they held a job for more than 6 55 months. As shown therein, the new estimated impact suggests that fertility rose by 12 percentage points (42 percent) among this restricted group of eligible migrants. We also consider restricting our treatment group to individuals in couples in which the other person is also a migrant from one of the countries eligible to apply for temporary legal status based on the 2011 Royal Decree. In this manner, we exclude women with partners who might have provided them with an alternative path to a legal immigration status. This sample should, therefore, have an even greater incentive to have a child to qualify for temporary legal status. As seen in column (9), these migrants became 15 percentage points (69 percent) more likely to respond by having a child in Spain when compared to alike migrants from European countries from outside the EU-15. To conclude, in column (10) of Table B.2, we use the same control and treatment groups included in our baseline specification, but restrict our attention to the 2009 through 2014 period. Since their entry in the EU in 2007, migrants from Bulgaria and Romania enjoyed legal status; however, they required a special permit to work until 2013. 27 If we focus on the period during which they would have lacked a work permit requirement, we continue to find evidence of a significant childbearing response to the 2011 Royal Decree, with the propensity to have a child born in Spain rising by 10 percentage points (37 percent). Finally, we explore if the fertility response to the policy change is driven by migrants from a particular nationality. To that end, Figure B.1 depicts how the estimated fertility impact of the 2011 Royal Decree varies if we remove migrant women from each eligible nationality one group at a time. It is evident from the figure that no single country of origin is responsible for the results. 27 We include 2014 to gauge the impact of fertility decisions made in 2013. We start the sample in 2009 to have a balanced panel of years before and after the reform. 62 Table B.5: The Impact of the 2011 Royal Decree on Fertility (Having a Child Current Year) Model Specification: (1) (2) (3) (4) Baseline With Demographic Traits With Demographic Traits + Province Time Trend With Demographic Traits + Province- Year FE Post * Eligible 0.009*** 0.009*** 0.016*** 0.019*** (0.003) (0.003) (0.003) (0.004) Year FE Y Y Y Y Province FE Y Y Y Y Nationality FE Y Y Y Y Other Controls N Y Y Y Province-Year FE or Trend N N Province-Trend Y Dep. Var. Mean 0.028 Observations 7,283 7,283 7,283 7,283 R-squared 0.022 0.029 0.037 0.099 Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The control group (CG) is composed of women from European nationalities that do not include the countries in the UE-15. The treatment group (TG) is composed of mothers from eligible nationalities potentially affected by the reform. All model specifications include a constant term. Specification (1) includes year and country of nationality fixed effects. Specification (2) adds age, marital status, university education, number of children, job tenure and years since arrived in Spain. Specification (3) adds province-year time trend. Finally, specification (4) includes province-year fixed effects. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 63 Table B.6: Impacts of the 2011 Royal Decree on Fertility Using Alternative Outcomes for Women without Children Between 1994-2011 Model Specification (1) (2) (3) (4) (5) (6) Baseline Child Born in Spain <5 Years Old Child Born in Spain <3 Years Old No. of Children Born in Spain No. of Children Born in Spain <5 Years Old No. of Children Born in Spain <3 Years Old Post * Eligible 0.086** 0.088** 0.076** 0.099** 0.100** 0.083** (0.033) (0.033) (0.031) (0.043) (0.043) (0.036) Observations 5,402 5,402 5,402 5,402 5,402 5,402 R-squared 0.217 0.220 0.209 0.217 0.220 0.207 Dep. Var. Mean -- -- -- -- -- -- Year FE Y Y Y Y Y Y Province FE Y Y Y Y Y Y Nationality FE Y Y Y Y Y Y Other Controls Y Y Y Y Y Y Province-Year FE or Trend Y Y Y Y Y Y Notes: The sample are women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. The basic control group (CG) is composed of women from European nationalities that do not include the countries in the UE-15. The basic treatment group (TG) is composed of mothers from eligible nationalities potentially affected by the reform. The model specification includes a constant term; year, country of nationality and province-year fixed effects, age, gender, marital status, an indicator for university education, number of children, job tenure and years since arrived in Spain. Standard errors are clustered at the country of nationality level. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 64 Figure B.1: Impacts of the 2011 Royal Decree on Fertility When We Remove One Nationality at a Time Notes: Sample: Women between 16 and 45 years of age, who are not students and who arrived in Spain after the 2005 normalization. Each point corresponds to the average effect of the 2011 Royal Decree on fertility without the country indicated in the horizontal axis. The horizontal solid line corresponds to the average effect, including all countries. The horizontal dotted lines are the corresponding confidence intervals. The dependent variable is a dummy equal to 1 if the woman has a child born in Spain. All coefficients are estimated controlling for age, marital status, university education, number of children, job tenure, years since arrived in Spain, year and country of nationality fixed effects and province-year fixed effects. Standard errors are clustered at the country of nationality level. The control group (CG) is composed of women from European nationalities, excluding EU-15 countries. The treatment group (TG) is composed of mothers from eligible nationalities under the reform.