The role of non-contributory pensions on internal mobility in Spain

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript of an article published by Elsevier in Labour Economics, Vol. 70, on June 2021, available at: https://doi.org/10.1016/j.labeco.2021.101980 Copyright 2021 Elsevier. En idUS Licencia Creative Commons CC BY-NC-ND The Role of Non-Contributory Pensions on Internal Mobility in Spain Catalina Amuedo-Dorantes, University of California, Merced Cristina Borra, Universidad de Sevilla* Abstract Spain is one of the Southern European nations characterized by very low inter-provincial mobility rates despite large inter-regional unemployment rate differences. We examine whether non-contributory pensions –an increasingly important component of many welfare systems, have curtailed the internal migration of younger cohorts over the past three decades. Using changes in the number of beneficiaries brought about by a policy reform in 1991, we show that old-age non-contributory pensions are associated with reduced internal migration of 25-to-35 years old men and women with less human capital. The effects do not appear to be driven by intergenerational reciprocity or quid pro quo exchanges, and do not extend to other types of pensions, hinting on systematic employment barriers faced by younger, less educated, or unexperienced men and women from lower income households qualifying for these pensions. The ability to rely on the support from older family members residing nearby might provide a lifeline but, simultaneously, curtail their propensity to out-migrate. JEL Classifications: I38, J61, R23. Keywords: Internal migration, non-contributory pensions, inter-vivo support, Spain. (*) Corresponding author. Department of Economics and Economic History, Universidad de Sevilla, Ramon y Cajal 1, 41018 Seville, Spain. Email: [email protected]s 1 I. Introduction Researchers have, for quite some time, recognized the role of the welfare state on migration. The literature on this topic has primarily focused on two main areas. One set of studies has examined the role played by unemployment insurance (Antolín and Bover, 1997; Hassler et al., 2005; Tatsiramos, 2009; Jofre-Monseny 2014). These analyses generally conclude that unemployment benefits need to be large to deter migration. Another group of studies has focused, instead, on the role of family benefits as welfare magnets, primarily in the United States (Gelbach, 2004; McKinnish, 2007). For the most part, this second literature concludes that welfare migration is limited and typically larger for families living close to state borders. Economists have also studied the impact of old-age pensions on other family member’s social and economic outcomes, primarily in developing countries. Studies in that literature have focused on the impact of old-age pensions on children’s schooling (Case and Deaton, 1998; Yanez-Pagans, 2008), children’s health (Duflo, 2003), living arrangements (Hamoudi and Thomas, 2014), consumption (Angelucci and Di Giorgi 2009), and labor market participation (Juarez and Pfutze, 2015; Galiani et al., 2016; Borella-Mas et al., 2016; Bratti et al., 2018). However, although strong family ties are recognized as a migration deterrent (Spilimbergo and Ubeda, 2004; Alesina et al., 2015), the literature has yet to explore the role of old-age noncontributory pensions received by extended family members on the migration of working-age individuals. In this paper, we look at how old-age non-contributory pensions received by extended family members might influence the internal migration of younger, working-age individuals in Spain. Spanish non-contributory pensions accrue to individuals who have not worked sufficiently to qualify for contributory pensions and who have no other sources of income (OECD, 2019). For those eligible, the program raises household income –an increase that may 2 be shared with other family members. This typically parental inter-vivo support is more likely to benefit the offspring residing nearby for various reasons. First, as emphasized by Fu (2019), close relatives are more likely to become future caregivers than those residing far away. As such, parents might be more inclined to assist their nearby offspring. Second, in-kind transfers (such as the provision of room and board, childcare or home cooked meals), which often have no close substitute, can only be provided if the two parties involved live nearby (Laferrere and Wolf, 2006). Spain shares with other Mediterranean welfare states, such as Greece and Italy, a generous public pension system and a strong reliance on the family unit as the ultimate safety net (Ferrera, 1996; Borsch-Supan, 2013). Despite having a structurally high unemployment rate and large spatial unemployment differences, 1 internal migration remains rather low in Spain (Ahn et al., 1999; Antolín and Bover, 1997). Internal migration is particularly low relative to other European countries, as well as relative to other developed and less developed economies. Spain is the only European nation with a 5-year internal migration intensity that falls below 10 percent (Bell et al., 2015). 2 This was not always the case. During the 1960s and 1970s, many Spanish workers emigrated from economically struggling regions like Andalucía or Extremadura to the prosperous regions of Madrid or Cataluña (Bover and Velilla, 2005). The expansion of the welfare state has been previously proposed as a potential explanation for the low level of internal migration in Spain (Antolín and Bover, 1997; Bover and Velilla, 2005). However, the role played by the welfare state in explaining migration has not been directly examined. Internal migration may no longer be indispensable if there exists a reliable public 1 For instance, the unemployment rate in the Southern province of Cadiz was 37.2 percent in 2015, relative to 12.5 percent in the Northern province of Navarra. 2 Migration intensities are computed using regression techniques that account for distinct geographic units and population densities. Using this comparable measure, Bell et al. (2015) report intensities over 40 percent in the United States, over 30 percent in France, and between 10 and 20 percent in Greece and Portugal. Note that mobility rates can differ depending on the period and sample object of study. For instance, De la Roca (2017) reports annual mobility rates between 2.6 and 3.9 percent when focusing on 18-52-year-old Spanish workers in urban areas of 40,000 plus inhabitants during the 1980-2013 period. 3 safety net. In Europe, where public safety nets characteristic of a welfare state are widespread, adverse shocks have translated into higher unemployment rates and lower labor force participation rates. On the contrary, in countries with a less extensive welfare state, as is the case of the Unites States, adverse shocks lead to higher migration rates instead (Blanchard and Katz, 1992; Jimeno and Bentolila, 1998; Jauer et al. 2019; Mitze, 2019). We follow previous literature and consider internal migration as the relocation to a new labor market in response to geographic differences in expected utility at destination and at the origin, focusing on the role that monetary and in-kind support from nearby relatives may play in predicting such moves. Working-age individuals may take into consideration the support from relatives –particularly, the non-transferable kind, such as room and board or childcare– when deciding whether to out-migrate. Ideally, to gauge the impact of non-contributory pensions on migration, we would like to have longitudinal data on individuals’ residential location, receipt of parental support, as well as parental residence and old-age pension receipt. This would allow us to directly gauge the impact of old-age pensions received by parents on children’s migration. Unfortunately, such dataset does not exist. However, it is reasonable to assume that the likelihood of having a family member receiving a non-contributory pension is positively correlated to the share of non-contributory pensioners in their location. Then, we can rely on data from the Spanish Labor Force Survey, the Residential Variations Statistics, and the Social Security Statistics for the 1988 through 2015 period, to examine the impact of noncontributory pensions –as captured by the policy reform that introduced new non-contributory pensions in 1991 (pensions that gradually replaced the existing social assistance pensions over the 1990s, 2000s, and early 2010s)– on the migration of working-age individuals. In an attempt to isolate the effect of the policy change, we exploit cross-province differences in the take-up rate of the newer pensions over the 1988 through 2015 period, while accounting for a wide range of individual demographic and human capital traits, as well as temporal and provincial 4 fixed-effects. In addition, we include several province-specific time varying traits potentially correlated with non-contributory pension receipt, such as differences between origin and potential destination provinces (weighted by historical internal migration flows) in per capita Gross Domestic Product (GDP), employment rates, housing prices, the share of the population receiving unemployment benefits, and the share 65 years of age or older. We find that old-age non-contributory pensions are associated with reduced migration among young, less skilled, 25-to-35 years-old men and women. Our results prove robust to the use of alternative samples that exclude the foreign-born, provinces providing alternative sources of income support, or the years during which a Dependency Law was in place. Additionally, they do not predate the regulation of non-contributory pensions in 1991, nor can the differential take-up rate of the pensions across provinces be predicted by differences in migration patterns across provinces prior to the 1991 reform. We also consider two potential mechanisms at play -namely, intergenerational reciprocity and limited employment opportunities (Chan and Ermisch 2015). We fail to find evidence of intergenerational reciprocity or quid pro quo exchanges, as disability pensions – more likely to be linked to the care of their recipients, do not generate the same response. Additionally, the observed impact of non-contributory pensions on the lack of regional mobility among younger, less skilled cohorts is not driven by those with young children who might benefit from the childcare and childrearing assistance provided by older family members. Rather, limited employment opportunities might be the primary reason for the role of noncontributory pensions on the limited out-migration of young men and women with less human capital for various reasons. First, the effect is concentrated on younger individuals with either less education or no work experience, for whom we know job openings are limited. Second, only the safety net provided by non-contributory pensions seems to matter, hinting on the relevance of family, possibly in-kind support more likely to dominate among low-income 5 families (Moreno-Minguez 2017), who also might be in a worse position to assist their offspring in their labor market search. Youth emancipation, educational programs and work internships targeted at this younger cohort of men and women might prove helpful in promoting their migration and, in the process, address labor market imbalances. This study is innovative in its effort to assess the link between public assistance in the form of old-age non-contributory pensions and migration of working-age individuals. Unlike prior work examining the role of unemployment insurance or welfare benefits, our focus is on old-age non-contributory pensions received by the elderly. Assessing their impact on the migration of younger cohorts is especially important in a country like Spain where, despite the expansion of the welfare state during the past decades (Guillen and Leon, 2011), families still provide the ultimate safety net (Bentolila and Ichino, 2008). The ability to rely on the assistance provided by extended family members living nearby may have contributed to the relatively low out-migration of working-age individuals facing employment challenges. Our intent is to gauge if that has been the case and, when applicable, to what extent. In addition, we wish to understand the types of public assistance associated to distinct migration responses, as well as the affected demographic groups, to gain insights into the mechanisms at play. If only certain types of pensions matter, it might be indicative of systematic migration challenges faced by the offspring of specific population subgroups qualifying for such pensions, i.e. low-income households. Likewise, learning about the demographic traits of those impacted allow us to further understand some likely determinants, such as limited employment opportunities for those younger, less educated, and unexperienced. Finally, to date, most of the research on the impact of non-contributory pensions has focused on Latin American and African countries (e.g. Duflo, 2003; Juarez and Pfutze, 2015). Analyzing the Spanish case allows to extend our understanding of the impact of old-age noncontributory pensions on the social and economic outcomes of family members in the context 6 of a developed economy –an analysis that may yield a different set of policy recommendations. In this regard, the findings contribute to the literature on emancipation decisions of youth in Southern European countries by informing about how, through diminished internal migration, the welfare state might be associated to ongoing emancipation delays (Aparicio-Fenoll and Oppedisano, 2015; Ahn and Sanchez-Marcos, 2017). II. Spanish Non-Contributory Pensions As other Mediterranean countries, Spain is characterized by a relatively generous public pension system and a strong reliance on family as the main safety net (Ferrera, 1996; Borsch- Supan, 2013). Both features are closely inter-twined. Elderly care is perceived as a family responsibility, with the government playing a secondary role. This implies that children and grandchildren might benefit from non-contributory pensions of extended family members. Our calculations using data from the Spanish Survey of Household Finances over the 2002 through 2011 period reveal that approximately 8 percent of family units with a household head aged 65 or above assists relatives in other households on a regular basis. This financial help increased during the recession years, from about 4 percent in 2002 to about 13 percent in 2011. Remarkably, even families below the poverty line (that is, those with labor incomes below 60 percent of median family income) helped relatives. In 2011, six percent of families with annual gross incomes below 9,000€ gave over 110€/month to relatives and friends. Inkind services are also very common. According to our calculations using data from the 2009 Spanish Time Use Survey, approximately 11 percent of women between 65 and 75 years of age engaged in the childcare of grandchildren during the previous day an average of 12 hours/week, regardless of whether they resided in the same household. This proportion reaches 14 percent when considering women with, at most, a primary education. Old age support in Spain accounts for almost 45 percent of total public social expenditures, representing 11 percent of GDP –almost double the share of public spending on 7 health (OECD, 2019). The Spanish social support system is structured around two types of assistance. The first type consists of contributory benefits, which aid those who have paid into the system during unemployment spells or during retirement. The other type of public assistance consists of non-contributory pensions, which protect individuals in need who are not entitled to contributory pensions. In 2017, expenditures on old-age non-contributory pensions amounted to 1.8 percent of total public expenditure on old-age support –approximately 0.2 percent of GDP. In Spain, two types of old-age non-contributory pensions coexisted during our period of analysis: (1) what we refer to in this paper as “assistance pensions” –first established in 1981, and (2) what we label “non-contributory pensions”, which were regulated in 1991. Assistance pensions were granted to individuals aged 66 and over who had no subsistence means according to the Decree 2620/1981. Non-contributory pensions were introduced in 1991 by the Decree 357/1991 to replace assistance pensions. Non-contributory pensions require being at least 65 years old and having lived in Spain for a minimum of 10 years. Compared to the 1981 assistance pensions, the 1991 non-contributory pensions were more generous (with almost double the benefit) and had a lower income eligibility threshold, covering a larger population. Indeed, social pension’s coverage rose from 0.56 percent of the population in 1988 under the old assistance pensions, to 0.72 percent of the population a decade later under the newer noncontributory pensions. Throughout the 1990s and the first half of the 2000s, non-contributory pensions gradually replaced assistance pensions, which became virtually extinct by 2007. We use this policy reform and the progressive adoption of non-contributory pensions as a quasi-natural experiment. Provinces substituted one scheme for the other at very different rates, as shown in Figure 1. Whereas some provinces (such as Huesca in the North) adopted the new noncontributory pensions rapidly, other provinces (such as Huelva or Cadiz in the South) did so at 14 constant euros averaged 24,000. Finally, the shares of the population covered by noncontributory disability pensions were 0.2 percent in 1981 and 0.5 percent in 1991, respectively, whereas the share receiving old-age contributory pensions was 13.1 percent. To conclude, Panel C displays the summary statistics for the provincial level differences created using the data from Panel B, as shown in equation (2). V. The Role of Non-Contributory Pensions on Internal Migration A) Main Findings Our main goal is to learn how the expansion of the welfare state, as captured by old-age non-contributory pensioners, could have contributed to the out-migration patterns exhibited by young working-age individuals in Spain –a developed economy with strong family ties characteristic of other Mediterranean countries. Table 2 displays the estimated coefficients for increases in the share of pensioners in origin vs. destination provinces for the two types of oldage non-contributory pensions Spain has ever offered –namely, the 1981 assistance pensions and the 1991 non-contributory pensions regulated to replace the former. As we would expect, given the relatively small share of the population receiving the 1981 assistance pensions, the effect of increases in the relative share of such pensioners in origin vs. destinations provinces on the migration decisions of 26-to-45-year-olds is smaller than the impact of the more recent and prevalent old-age non-contributory pensions. Still, a onestandard-deviation increase in the share of social assistance pensioners in origin vs. potential destination provinces (similar to the increase in the share of the population with an assistance pension in Seville from 0.7 percent in 1989 to 0.8 percent in 1991) is associated to a 0.04 percentage point (8 percent) reduction in the migration of working-age individuals. This effect almost doubles when we consider old-age non-contributory pensions. In that case, a onestandard-deviation increase in the share of non-contributory pensioner rates in origin vs. destination provinces (similar to the increase in the share of the population with a contributory 15 pension in Málaga from 0.5 percent in 1993 to 1 percent in 2015) is associated to a 0.06 percentage point (12 percent) drop in the migration of working-age individuals (Column 3). Columns 4 through 9 display the results from conducting the analysis separately by gender. Focusing on the most complete model specifications, we find that the impact is somewhat larger and more precisely estimated for women (Colum 9) than for men (Column 6). A one-standard-deviation increase in the share of social assistance pensioners in origin vs. potential destination provinces lowers the out-migration propensity of working-age women by 10 percent. Similarly, a similar increase in the share of non-contributory pensioner rates in origin vs. destination provinces lowers that working-age group’s mobility by 17 percent. In contrast, among men, changes in the relative share of social assistance pensioners have no impact on their propensity to move, although the same one standard deviation increase in the share of noncontributory pensioners at origin, relative to destination, lowers their out-migration propensity by 12 percent. B) Identification Checks The estimates in Table 2 suggest that non-contributory pensions might primarily curtail the migration of working-age men and women. In what follows, we first provide some evidence supporting our identification strategy. Subsequently, we assess the robustness of our findings by addressing several potentially confounding factors that could be driving the observed impacts. Our identification strategy is based on the staggered take-up of non-contributory pensions across Spanish provinces, which we presume exogenous to the migration of workingage individuals prior to the regulation of the new pensions. In what follows, we test this assumption by collapsing the data prior to the enactment of the non-contributory pension reform at the province level. Then, we estimate the following model (Amuedo-Dorantes et al. 2018): (3) 𝐼𝑌𝑒𝑎𝑟 𝑗= 𝛼 + 𝑋𝑗 1988 𝛽 + 𝑍𝑗 1988 𝛿 + 𝛾𝑀𝑗 1988 + 𝜀𝑗 16 where IYearj is the first year in which the share of pensioners in province j receiving the new non-contributory pension exceeded the share receiving assistance pensions. The various key regressors in equation (2) are collapsed at the province level and measured in 1988 –that is, prior to the 1991 reform. The vector 𝑋𝑗 1988 is the vector of demographic characteristics, whereas 𝑍𝑗 1988 includes the provincial macroeconomic controls. Our key regressor is 𝑀𝑗 1988 , which captures the share of the working-age population migrating from province j in 1988. According to the results in Table 3A, migration rates prior to the enactment of the 1991 reform do not help predict the timing of when non-contributory pensions came to replace most social assistance pensions, regardless of the specification used. In other words, the prevalence of the policy being examined, while non-random, does not appear to be explained by prior migration patterns. We also conduct an additional identification check attempting to gauge if our estimates are somehow biased by differential migration patterns that preceded the prevalence of noncontributory pensions, relative to assistance pensions, in the province. To that end, we look at migration rates up to four years prior. Collapsing the data at the province level, we estimate the following model: (4) 𝑀𝑗𝑡 = 𝛼 + 𝛾𝐷𝑗0 +∑𝜑𝑗𝑦𝐷𝑗𝑦 𝑦=−4 𝑦=−1 + 𝑋𝑗𝑡 𝛽 + 𝑍𝑗𝑡 𝛿 + 𝜂𝑗+ µ𝑡+ 𝜀𝑗𝑡 where 𝑀𝑗𝑡 is the share of the working-age population migrating out of province j between t and t+1; 𝐷𝑗0 is a dummy taking value of 1 from the first year in which the share of pensioners in province j receiving the new non-contributory pension exceed the share receiving assistance pensions onwards. 𝐷𝑗𝑦 is a vector of dummies taking the value 1 for each of the y years prior to the share of non-contributory pensioners in province j exceeding the share of pensioners receiving the old social assistance. The vector 𝑋𝑗𝑡 refers to the vector of demographic characteristics in equation (2) in year t. Similarly, the vector 𝑍𝑗𝑡 includes provincial macroeconomic controls in year t at the province level. If our identification strategy is valid, 17 the coefficients for the 𝐷𝑗𝑦 dummies should be non-significantly different from zero –that is, differences in migration patterns should not predate the main implementation of the noncontributory pensions. Table 3B shows the results from this exercise. None of the estimated coefficients corresponding to the years prior to the majoritarian adoption of non-contributory pensions in the province are statistically different from zero. In other words, the observed out-migration patterns did not predate the primary adoption of non-contributory pensions. Furthermore, the estimated coefficient on the dummy indicative of the year when the share of non-contributory pensioners exceeded the share of pensioners receiving the old social assistance continues to be negative and statistically different from zero at conventional levels. Specifically, the primary implementation of the new non-contributory pensions in the province is associated to a 0.2 percentage point or 40 percent drop in mobility –an effect significantly larger than the one in Table 2 since we are now capturing a discrete change in the share of pensioners receiving the new vs. old pensions, as opposed to a one standard deviation change in the former share. Recapping, we find no evidence of differential migration patterns across provinces predating the main adoption of non-contributory pensions. 14 C) Robustness Checks In addition to concerns regarding reverse causality or differential migration trends prior to the granting of old-age non-contributory pensions, one might be concerned about other confounding factors. In this section, we consider the sensitivity of our findings to accounting 14 We also conduct a placebo check to ensure the mobility impacts being captured are not spuriously correlated to the posterior implementation of non-contributory pensions in the province. To that end, we restrict our sample to the period prior to policy reform (1988-1990), and create placebo lead variables that capture differences in the take-up of non-contributory pensions four years later –that is, when the pensions were first regulated. If differences in mobility patterns across provinces are spuriously related to the differential implementation of the new non-contributory pensions later on across provinces, the coefficients on the placebo lead variables should be statistically different from zero. Table A.2 in the appendix shows that is not the case. In other words, the mobility impacts do not appear to be spuriously correlated to differences in the posterior adoption of non-contributory pensions across provinces. 18 for housing costs, per capita income, and employment cycles; or addressing differences in population composition or in other income transfer programs across provinces. Finally, we address concerns regarding the exclusion of short-term regional migration from our data. (1) Housing Costs, Per Capita Income Difference and Employment Cycles The housing boom and the subsequent rise in housing prices experienced by more developed provinces such as Madrid, Cataluña and the Basque Country may have disincentivized out-migration from lower-income provinces. Coastal provinces may have also benefited from larger employment opportunities in the construction sector, possibly deterring the out-migration of working-age individuals. To address these concerns, we conduct a couple of robustness checks. First, in Panel A of Table 4, we include housing prices and per capita income as additional controls to capture differentials in the cost of living and income levels across provinces –both crucial in explaining residential choices. Unfortunately, such data are only available for a more limited period. Therefore, our sample size drops by roughly 30 percent. While the observed impact of social assistance on the migration decisions of women is no longer statistically significant at conventional levels, the effect of old-age non-contributory pensions significantly strengthens. Specifically, a one-standard-deviation increase in the noncontributory pensioners’ rate differential between origin and destination provinces appears to curtail the migration of working-age individuals by 0.11 percentage points or 22 percent (relative to 12 percent in Table 2), relatively stronger for women (24 percent relative to 17 percent in Table 2). Next, in Panel B of Table 4, we account for changes in the share of individuals employed in the construction sector to address that sector’s boom during part of the period under study. Results remain virtually unchanged from those in Table 2. A one-standard-deviation increase in the share of non-contributory pensioners in origin vs. destination provinces lowers the outmigration of working-age women by 0.06 or 12 percent. 19 (2) Differences in Population Composition An additional concern refers to the role played by differences in population composition across provinces. Non-contributory pensioners are low income residents over 65 years of age. The impact of changes in the share of non-contributory pensioners in the province could be driven by the spurious correlation between migration and the prevalence of a low-skilled oldage population. Thus, in Panel A of Table 5, we control for the share of the 65-year-old-and- above population with a primary or less than a primary education in the province –a group more likely to qualify for these pensions. Our results remain virtually unchanged. We next turn to another population composition concern given the conversion of Spain into a migrant receiving nation between the 1980s and 2000s –that is, the need to disentangle the migration of native-born vs. foreign-born individuals. Other things equal, we would expect our effects to stem, primarily, from the behavior of natives whose parents are more likely to qualify for the pensions being examined. Migrant status is only available from 1992 onwards in the EPA. Hence, our sample size is cut by roughly 15 percent when we run the analysis separately by nativity. The estimates, shown in Panel B of Table 5, confirm how higher rates of old-age non-contributory pensioners at the origin are linked to significantly lower migration propensities among both native men and women. Specifically, a one-standard-deviation increase in the non-contributory pensioners’ rate differential between origin and destination provinces is associated with reductions in the moving propensity of working-age native men and women between 0.08 and of 0.06 (or between 16 and 12 percent). While the effect of oldage non-contributory pensioners on the moving propensity of working-age women is not statistically different from the one in Table 2, it is interesting to note how the impact of the smaller share of social assistance pensioners is no longer present. Most importantly, as we would expect, we find no significant impacts among the foreign-born population in Panel C. 20 Finally, also related to the potential role of changes in the population composition of provinces in explaining our results, is the share of up to 25 year-olds who may still be completing their studies and living with their parents, as is traditional in Mediterranean countries (e.g. Aparicio-Fenoll and Oppedisiano (2015) on emancipation decisions in Spain). Panel D in Table 5 shows no significant correlation between old-age social assistance or noncontributory pensions and the migration of men and women under 25 years of age. (3) Accounting for Other Overlapping Income Supports and Labor Reforms Third, we address the possibility that the impact of old-age non-contributory pensions might be spuriously correlated to other overlapping transfers/payments/labor reforms. For example, on January 1, 2007, the so-called Dependency Law came into effect in Spain. The law intended to assist those in need of support due to illness, disability, or old age by allowing their caretakers to sign a formal labor contract and get compensated for their services. It is possible that the restricted mobility attributed to old-age non-contributory pensions might have been partially due to such transfers. To assess if that was the case, we repeat the analysis restricting our attention to the period prior to the enactment of the Dependency Law –that is, to the 1988-2007 period (see Panel A in Table 6). The same one-standard-deviation increase in the difference between the share of non-contributory pensioners (social assistance pensioners) in origin vs. destination provinces lowers individuals’ propensity to move by 0.06 or 12 percent (0.003 or 6 percent) –estimates similar to those uncovered in Table 2. In a similar vein, we experiment with excluding provinces that provide minimum income support policies to more than one percent of all households to make sure the estimated policy impacts are not driven by those other financial supports. Once again, according to the estimates in Panel B in Table 6, we continue to find that a one-standard-deviation increase in the share of non-contributory pensioners in origin relative to destination provinces lowers men and women’s propensity to move by roughly 0.08 or 14 percent. And, as in Table 2, we continue 21 to find evidence of a similar impact in the case of social assistance pensions. As previously found, the impacts appear stronger among women. Lastly during the studied period there were several labor market reforms in 1994, 1997, and 2001 that modified the regulation of employment protection with the aim of containing the increase in temporary employment (Dolado et al. 2002). To address the potential differential impact of these labor market reforms on migration across provinces, we experiment with adding an additional control to capture changes in the share of temporary workers. As can be seen in Panel C of Table 6, the results are virtually identical to those in Table 2. Specifically, a onestandard-deviation increase in the share of social assistance pensioner rates (non-contributory pensioner rates) in origin vs. destination provinces is associated to a 0.04 (0.06) percentage point drop in the migration of working-age individuals. (4) Accounting for Short-Term Regional Migration A valid concern with the data being used refers to its exclusion of short-term regional migration lasting less than one year. 15 In this subsection, we use population level data from EVR on all interprovincial residential moves, irrespective of their duration, among prime-age working-age individuals (aged 25 to 44). We compute migration rates by gender and origindestination pairs dividing the number of moves by the population in the origin province. Subsequently, we estimate a version of equation (2) using aggregate level data. 16 Table 7 displays the results from such an exercise. As shown therein, a one-standard- deviation increase in the share of social assistance pensioners at the origin (when compared to destination) provinces is associated to a 3 percent reduction in the migration of working-age individuals. Similarly, a one-standard-deviation increase in the share of non-contributory 15 Our emphasis continues to be on internal migration, which is conceptually different from work commutes that do not involve a change of residence. 16 We estimate the following equation: 𝑀𝑗𝑘𝑡 = 𝛼 + (𝑍 𝑗𝑡– ∑𝜔𝑘𝑍𝑘𝑡𝑘)𝛿 + 𝜂𝑗+ 𝜑𝑘+ µ𝑡+ 𝜀𝑗𝑘𝑡, where 𝑀𝑗𝑘𝑡 stands for the proportion of the population aged 25-44 in province j migrating to province k in year t. 22 pensioner rates in origin vs. destination provinces is associated to a 10 percent drop in the internal migration of working-age individuals. These results are in line with our findings from Table 2. In sum, non-contributory pensions seem to inhibit the migration of working-age men and women. This result does not predate the regulation of non-contributory pensions in 1991, nor can the differential take-up rate of the pensions across provinces be predicted by differences in migration patterns across provinces before the reform. In addition, our finding appears robust to accounting for differential housing prices and per capita income levels, growth of the construction sector, or the share of less-skilled 65-year-olds more likely to receive these pensions, across provinces. Results also prove robust to restricting our attention to the period prior to the enactment of the Dependency Law granting caregiving support, to eliminating regions offering other supplemental benefits, to controlling for concurrent labor market reforms, and to including both short- and long-term regional migration. VI. Heterogeneous Impacts by Age and Educational Attainment To better understand why non-contributory pensions primarily curtail regional migration among young working-age men and women, we further investigate heterogeneous impacts by age (often a proxy for work experience) and educational attainment. Our aim is to discern the degree to which the observed impacts might be related to differences in human capital possibly linked to limited employment opportunities. The estimates in Panels A and B of Table 8 address this inquiry. We split the sample in two age groups –namely, 26 to 35-year-olds and 36 to 45-year-olds. It is evident that the presence of a greater share of social assistance and old-age non-contributory pensioners at origin vs. destination provinces is associated to significantly lower out-migration patterns among men and women 26 to 35 years old –a group more likely to have finished any university studies and to be entering the labor market. In particular, a one-standard-deviation increase in 23 the share of social assistance pensioners in origin vs. destination provinces curtails the outmigration of this group by 0.06 percentage points or 8 percent among women, and a similar increase in the share of old-age non-contributory pensioners in origin vs. destination provinces lowers the out-migration of these young men and women by 15 percent. Conversely, we observe no significant impacts on the migration of 36-to-45-year-olds –a group more likely to have a more established labor market history. Because migration is typically greater among the highly skilled, we also explore if the expansion of the welfare state through old-age social assistance and non-contributory pensions had a differential impact on the mobility of working-age individuals based on their educational attainment. To that end, in Panels C and D in Table 8, we display the estimated impacts of the two types of non-contributory pensions on individuals with a primary education or less vs. those with, at least, a secondary education. We find that non-contributory pension programs are associated to lower migration among all women, regardless of their educational attainment. However, in the case of men, that is only the case among those mid- to high-skilled. This might explain why, when lumped together with their low-skilled counterparts in Table 2, noncontributory pension programs do not appear to have significantly lowered male migration. Specifically, based on the findings in Panels C and D of Table 8, a one-standard-deviation increase in the share of old-age non-contributory pensioners in origin vs. potential destination provinces is associated to a 14 percent and a 10 percent reduction in the migration of workingage men and women with more than a primary education; with the effect being particularly large among less-educated women, for whom migration drops by approximately 17 percent. In sum, social assistance and old-age non-contributory pensions appear to be associated to reduced regional migration among young, especially low-educated women, but also among young men and women with more than a primary education. 30 Borrella-Mas, Miguel Ángel, Mariano Bosch-Mossi and Marcello Sartarelli. (2016). “Non- Contributory Pensions Number-Gender Effects on Poverty and Household Decisions,” IVIE Working Papers. Serie AD 2016-02. Börsch-Supan, Axel H. (2013). “Entitlement Reforms in Europe: Policy Mixes in the Current Pension Reform Process” in Fiscal Policy after the Financial Crisis edited by Alberto Alesina and Francesco Giavazzi, Chicago: University of Chicago Press, 405-435. Bover, Olympia and Pilar Velilla. (2005). “Migration in Spain: historical background and current trends” in European Migration: What Do We Know? Edited by Klaus Zimmermann, CEPR and Oxford University Press, Oxford, pp. 389–414. Case, Anne and Angus Deaton. (1998). “Large cash transfers to the elderly in South Africa.” Economic Journal 108: 1330–1361. Chan, Tak and John Ermisch, (2015). “Proximity of Couples to Parents: Influences of Gender, Labor Market, and Family.” Demography 52(2), 379-399. Compton, Janice and Robert A. Pollak, 2015. “Proximity and Coresidence of Adult Children and their Parents in the United States : Description and Correlates,” Annals of Economics and Statistics, GENES, vol. 117-118: 91-114. De la Roca, Jorge. (2011). “Selection in initial and return migration: Evidence from moves across Spanish cities.” Journal of Urban Economics, 100(1): 33-53. Devillanova, Carlo and Walter García-Fontes. (2004). “Migration across Spanish provinces: evidence from the social security records (1978-1992).” Investigaciones Economicas, 28(3): 461-487. Dolado, Juan J., Carlos Garcia-Serrano, and Juan F. Jimeno (2002). “Drawing Lessons from The Boom of Temporary Jobs in Spain.” Economic Journal 112, F270-F295. Duflo, Esther. (2003). “Grandmothers and Granddaughters: Old-Age Pensions and Intrahousehold Allocation in South Africa.” World Bank Economic Review, 17 (1): 1-25. Enchautegui, María E. (1997). “Welfare payments and other economic determinants of female migration.” Journal of Labor Economics 15, 529-554. Ferrera, Maurizio. (1996). “The ‘Southern Model’ of Welfare in Social Europe.” Journal of European Social Policy 6 (1): 17-37. Fu, Chien-Hao. (2019). “Living arrangement and caregiving expectation: the effect of residential proximity on inter vivos transfer.” Journal of Population Economics, 32(1), 247-275. Galiani, Sebastian, Paul Gertler and Rosangela Bando. (2016). “Non-Contributory Pensions.” Labour Economics 38: 47-58. Gelbach, Jonah. (2004). “Migration, the lifecycle and state benefits: how low is the bottom?” Journal of Political Economy 112: 1091–1130. 31 Guillén, Ana M. and Margarita León. (2011). “Introduction” in The Spanish Welfare State in European Context. Surrey and Burlington: Ashgate, 1-16. Hamoudi, Amar and Duncan Thomas. (2014). “Endogenous Co-residence and Program Incidence: South Africa’s Old Age Pension.” Journal of Development Economics 109: 30–37. Hassler John, José V. Rodríguez Mora, Kjetil Storesletten and Fabrizio Zilibotti. (2005). “A positive theory of geographical mobility and social insurance.” International Economics Review, 46(1): 263–303. Izquierdo, Mario, Jimeno, Juan F. and Lacuesta, Aitor. (2016). “Spain: from massive immigration to vast emigration?” IZA Journal of Migration 5, 10. https://doi.org/10.1186/s40176-016-0058-y Jauer, Julia, Thomas Liebig, John P. Martin, and Patrick A. Puhani. (2019). “Migration as an adjustment mechanism in the crisis? A comparison of Europe and the United States 2006– 2016.” Journal of Population Economics, 32: 1-22. Jimeno Juan Francisco and Samuel Bentolila. (1998). “Regional unemployment persistence (Spain, 1976–1994).” Labour Economics, 5 (1): 25–51. Jofre-Monseny, Jordi, (2014). “The effects of unemployment protection on migration in lagging regions.” Journal of Urban Economics 83(C): 73-86. Juarez, Laura and Tobias Pfutze. (2015). “The Effects of a Noncontributory Pension Program on Labor Force Participation: The Case of 70 y Más in Mexico.” Economic Development and Cultural Change 63(4): 685-713. Kennan, John and James R. Walker. (2011). “The Effect of Expected Income on Individual Migration Decision.” Econometrica, 79(1): 211–51. McKinnish, Terra. (2007). “Welfare-Induced Migration at State Borders: New Evidence from Micro-Data.” Journal of Public Economics, 91 (3-4):437-450. Laferrere, Anne, and Francois-Charles Wolff. (2006). “Microeconomic models of family transfers,” Handbook on the Economics of Giving, Reciprocity and Altruism, in: S. Kolm and Jean Mercier Ythier (ed.), Handbook of the Economics of Giving, Altruism and Reciprocity, edition 1, volume 1, chapter 13, pages 889-969, Elsevier. Mitze, Timo. (2019). “The migration response to local labour market shocks: Evidence from EU regions during the global economic crisis.” Oxford Bulletin of Economics and Statistics, 81: 271-298. Ministerio de Fomento. 2017. Valor Tasado de Vivienda Libre. Available at: http://www.fomento.gob.es/BE2/?nivel=2&orden=32000000. Last accessed on December 12, 2017. Ministry of Employment and Social Security. 2017. Anuario de Estadísticas Laborales. 1988- 2015. On-line version since 2001 at http://www.mitramiss.gob.es/ es/estadisticas/contenidos/anuario.htm. Last accessed on December 12, 2017. 32 Molloy, Raven, Christopher L. Smith, and Abigail Wozniak. (2011). “Internal Migration in the US: Updated Facts and Recent Trends.” Journal of Economic Perspectives 25 (3): 173–96. Moreno-Mínguez, Almudena. (2017). Understanding the Impact of Economic Crisis on Inequality, Household Structure, and Family Support in Spain from a Comparative Perspective, Journal of Poverty, 21:5, 454-481 National Statistics Institute. (2017a). Encuesta de Población Activa, EPA, microdata. 1995- 2015. National Statistics Institute. (2017b). Estadística de Variaciones Residenciales, EVR, microdata. 1995-2015. National Statistics Institute. (2017c). Cifras de Población. http://www.ine.es /jaxiT3/Tabla.htm?t=10272. Last accessed on December 12, 2017. National Statistics Institute. (2017d). Encuesta de Población Activa. http://www.ine.es /jaxiT3/Tabla.htm?t=4247. Last accessed on December 5, 2017. National Statistics Institute. (2017e). Contabilidad Regional de España. Available at: http://www.ine.es/dyngs/INEbase/es/operacion.htm?c=Estadistica_C&cid=125473616762 8&menu=enlaces&idp=1254735576581#. Last accessed on December 5, 2017. National Statistics Institute. (2017f). Consumer Price Index, Base 2011. Available at: http://www.ine.es/jaxi/menu.do?type=pcaxis&path=%2Ft25/p138&file=inebase&L=1. Last accessed on December 5, 2017. OECD (2019). OECD Social Expenditure database, (www.oecd.org/social/expenditure.htm). Sergio Puente and Ana Regil. (2020). “Tendencias laborales intergeneracionales en España en las últimas décadas,” Boletín Económico, Banco de España, issue 2/2020, pages 1-18. Rabe, Birgitta and Mark P. Taylor. (2012). “Differences in Opportunities? Wage, Employment and House‐Price Effects on Migration.” Oxford Bulletin of Economics and Statistics, 74: 831-855. Saks, Raven E., and Abigail Wozniak. (2011). “Labor Reallocation over the Business Cycle: New Evidence from Internal Migration.” Journal of Labor Economics, 29(4): 697-739. Sevilla-Sanz, Almudena, Jose Ignacio Gimenez-Nadal and Cristina Fernandez. (2010). Gender Roles and the Division of Unpaid Work in Spanish Households,” Feminist Economics, 16(4), 137-184. Spilimbergo, Antonio and Luis Ubeda. (2004). “Family attachment and the decision to move by race.” Journal of Urban Economics, 55(3): 478-497. Tatsiramos, Konstantinos. (2009). “Geographic labour mobility and unemployment insurance in Europe.” Journal of Population Economics, 22(2): 267-283. 33 Table 1: Sample Descriptive Statistics Variable Names Definition Mean S.D. Min Max Panel A: Data from the Spanish Labor Force Survey (Encuesta de Población Activa, EPA) Moved 1207489 =1 if respondent emigrated from province 0.005 0.07 0 1 Age 1207489 age in years 35.293 5.68 26 45 Male 1207489 =1 if male 0.505 0.5 0 1 Married 1207489 =1 if married 0.642 0.48 0 1 Immigrant 1207489 =1 if born abroad 0.121 0.33 0 1 Basic Emancipation 1207489 =1 if eligible for emancipation benefits 0.112 0.31 0 1 Less than Primary 1207489 =1 if s/he has less than primary education 0.035 0.18 0 1 Primary 1207489 =1 if s/he has a primary education 0.188 0.39 0 1 Secondary 1207489 =1 if s/he has a secondary education 0.492 0.5 0 1 Tertiary 1207489 =1 if s/he has a college education 0.285 0.45 0 1 Receiving Unemployment Insurance 1207489 =1 if s/he is receiving unemployment benefits 0.055 0.23 0 1 Panel B: Province Traits Social assistance pensioners rate 1308 Proportion of province’s population receiving old-age pensions per 100 inhabitants 0.153 0.23 0 1.075 Non-contributory pensioners rate 1377 Proportion of province’s population receiving old-age pensions per 100 inhabitants 0.629 0.39 0 2.114 Employment rate 1377 Proportion of 16-64-year-old population working 0.444 0.06 0.296 0.621 Share of population over 65 years 1377 Proportion of province’s population over 64 years old over total population per 100 inhabitants 17.8 4.35 8.04 30.4 Unemployment insurance rate 1377 Proportion of province’s population receiving unemployment benefits per 100 inhabitants 2.183 1.47 0.19 12.4 Housing prices 1050 Price of square meter in constant 2011 prices (hundred €) 13.125 5.22 5.61 33.0 GDP per capita 1150 Provincial Gross Domestic Product per capita in constant 2011 prices (thousand €) 24.237 5.8 11.57 43.5 Social assistance disability rate 1308 Proportion of province’s population receiving social assistance disability pensions per 100 inhabitants 0.221 0.3 0 1.864 Non-contributory disability rate 1377 Proportion of province’s population receiving non-contributory disability pensions per 100 inhabitants 0.493 0.32 0 2.271 Contributory pensioners rate 1377 Proportion of province’s population receiving old-age pensions per 100 inhabitants 13.056 4.22 2.04 26.51 Year 1377 Survey year 2001.9 7.97 1988 2015 Province 1377 Province of origin 26 14.72 1 51 Panel C: Differentials in Province of Origin and Destination Traits Differential in social assistance pensioners’ rate 1207489 Social assistance pensioners’ rate at province of origin minus weighted rate at potential destinations, standardized -0.131 0.77 -2.857 4.626 Diff. in non-contributory pensioners’ rate 1207489 Non-contributory pensioners’ rate at province of origin minus weighted rate at potential destinations, standardized -0.166 1.02 -1.956 5.786 Diff. in employment rates 1207489 Employment rate at province of origin minus weighted rate at potential destinations, standardized 0.242 1.01 -2.501 3.09 Diff. in the share of population over 65 1207489 Share of population over 65 at province of origin minus weighted index at potential destinations, standardized -0.280 0.88 -2.311 3.292 Diff. in unemployment insurance rates 1207489 Unemployment beneficiaries’ rate at province of origin minus weighted rate at potential destinations, standardized 0.290 1.03 -2.689 4.739 Diff. in housing prices 886093 Housing prices at province of origin minus weighted prices at potential destinations, standardized 0.402 1.14 -2.319 3.135 Diff. in GDP per capita 1020361 GDP per capita at province of origin minus weighted GDP per capita at potential destinations, standardized 0.334 1.11 -1.954 2.624 Diff. in the social assistance disability rate 1205302 Social assistance disabled rate at province of origin minus weighted rate at potential destinations, standardized -0.163 0.84 -2.222 4.816 Diff. in the non-contributory disability rate 1206928 Non-contributory disabled rate at province of origin minus weighted rate at potential destinations, standardized -0.176 1.07 -2.241 6.814 Diff. in the contributory pensioners rate 1164429 Contributory pensioners rate at province of origin minus weighted rate at potential destinations, standardized -0.161 0.91 -3.802 3.958 Source: EPA (1988-2015), EVR (1988-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1988-2015), Valor Tasado de Vivienda Libre (1995-2015), Contabilidad Regional de España (1993-2015), Consumer Price Index (1993-2015). 34 Table 2: Internal Mobility and Differences in Take-up of Non-contributory Pensions Sample: All Males Females (1) (2) (3) (4) (5) (6) (7) (8) (9) Model Specification: Specif. 1 Specif. 2 Specif. 3 Specif. 1 Specif. 2 Specif. 3 Specif. 1 Specif. 2 Specif. 3 Social Assistance Pensioners’ Rate Differential -0.0002 -0.0004** -0.0004** -0.0001 -0.0003 -0.0003 -0.0003** -0.0005*** -0.0004*** (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0005** -0.0005** -0.0006** -0.0004 -0.0003 -0.0006** -0.0007** -0.0006** -0.0007** (0.0002) (0.0002) (0.0003) (0.0003) (0.0002) (0.0003) (0.0003) (0.0003) (0.0003) Demographic and Human Capital Controls No Yes Yes No Yes Yes No Yes Yes Macroeconomic Controls No No Yes No No Yes No No Yes Province of Origin Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 1,207,489 1,207,489 1,207,489 595,408 595,408 595,408 612,081 612,081 612,081 R-squared 0.0013 0.0035 0.0035 0.0014 0.0035 0.0035 0.0012 0.0037 0.0037 Dependent Variable Mean (S.D. in parentheses) 0.005 (0.07) 0.005 (0.07) 0.004 (0.07) Notes: All specifications include a constant term, province of origin and year fixed effects, demographic and human capital controls (i.e. age, male sex, marital status, an immigrant dummy, educational attainment dummies using tertiary education as reference, a dummy indicating emancipation benefit eligibility, and a dummy for receiving unemployment benefits), and macroeconomic controls (standardized differentials between the province of origin employment rate, share of population over 64 years old, and share of the population receiving unemployment benefits and the weighted average of the corresponding rates at potential destinations). Observations are weighted using the individual weights in the EPA. Standard errors are clustered at the province level. Standard errors are shown in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1988-2015), EVR (1988-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1988-2015). 35 Table 3A: Identification Check #1: Predicting the Take-up of Non-contributory Pensions Dependent Variable: First Year Non-contributory Pensioners Exceed Assistance Pensioners Model Specification: (1) (2) (3) Population Share Moving in 1988 34.05 85.77 106.06 (77.62) (96.76) (89.48) Demographic and Human Capital Controls N Y Y Macroeconomic Controls N N Y Observations 51 51 51 R-squared 0.003 0.265 0.416 Notes: The table presents the results from estimating Equation (3). All variables measured at the provincial level in 1988. Demographic and human capital controls are age, male sex, marital status, educational attainment dummies using tertiary education as reference, and a dummy for receiving unemployment benefits; macroeconomic controls are province of origin employment rate, share of population over 64 years old, share of the population receiving unemployment benefits, and share of the population receiving contributory pensions. Observations are weighted using the individual weights in the EPA. Robust standard errors in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1988-2015), EVR (1988-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1988-2015). 36 Table 3B: Identification Check #2: Checking for Pre-Existing Differences in Female Mobility Patterns Model Specification: (1) (2) (3) Dummy for when (share of non-contributory pensioners)>(share assistance pensioners) -0.0026** -0.0025** -0.0025** (0.001) (0.001) (0.001) One Year Prior to (share of non-contributory pensioners)>(share assistance pensioners) 0.0030* 0.0030* 0.0029* (0.002) (0.002) (0.002) Two Years Prior to (share of non-contributory pensioners)>(share assistance pensioners) -0.0008 -0.0008 -0.0009 (0.001) (0.001) (0.001) Three Year Prior to (share of non-contributory pensioners)>(share assistance pensioners) 0.0012 0.0012 0.0012 (0.001) (0.001) (0.001) Four Years Prior to (share of non-contributory pensioners)>(share assistance pensioners) -0.0008 -0.0006 -0.0006 (0.001) (0.001) (0.001) Demographic and Human Capital Controls N Y Y Macroeconomic Controls N N Y Observations 1,427 1,427 1,427 R-squared 0.343 0.348 0.351 Notes: All specifications include a constant term, province of origin and year fixed effects. Demographic and human capital controls are age, male sex, marital status, educational attainment dummies using tertiary education as reference, and a dummy for receiving unemployment benefits; macroeconomic controls are province of origin employment rate, share of population over 64 years old, and share of the population receiving unemployment benefits. Observations are weighted using the individual weights in the EPA. Robust standard errors in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1995-2015), EVR (1995-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1995-2015). 37 Table 4: Robustness Checks #1 – Additional Controls for Housing Costs, Per Capita Income Difference, Employment Cycles Sample: All Males Females Panel A: Accounting for Confounding Impacts of Housing Prices and Per Capita Income Differences (1996-2015) Social Assistance Pensioners’ Rate Differential -0.0009* -0.0009 -0.0008 (0.0004) (0.0006) (0.0005) Non-contributory Pensioners’ Rate Differential -0.0011** -0.0011** -0.0012** (0.0005) (0.0005) (0.0006) Observations 886,093 436,018 450,075 R-squared 0.0038 0.0038 0.0041 Dependent Variable Mean (S.D. in parentheses) 0.005 (0.07) 0.005 (0.07) 0.005 (0.07) Panel B: Accounting for Changes in Construction Employment (1988-2015) Social Assistance Pensioners’ Rate Differential -0.0004** -0.0003 -0.0004*** (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0006** -0.0006** -0.0006** (0.0003) (0.0003) (0.0003) Observations 1,207,489 595,408 612,081 R-squared 0.0035 0.0035 0.0037 Dependent Variable Mean (S.D. in parentheses) 0.005 (0.07) 0.005 (0.07) 0.004 (0.07) Notes: All specifications include a constant term, province of origin and year fixed effects, demographic and human capital controls (i.e. age, male sex, marital status, an immigrant dummy, educational attainment dummies using tertiary education as reference, a dummy indicating emancipation benefit eligibility, and a dummy for receiving unemployment benefits), and macroeconomic controls (standardized differentials between the province of origin employment rate, share of population over 64 years old, and share of the population receiving unemployment benefits and the weighted average of the corresponding rates at potential destinations). The specification in Panel A additionally controls for standardized differentials in construction employment rates. The specification in Panel B controls instead for standardized differentials in housing prices and per capita income. Observations are weighted using the individual weights in the EPA. Standard errors are clustered at the province level. Standard errors are shown in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1995-2015), EVR (1995-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1995-2015), Valor Tasado de Vivienda Libre (1995- 2015), Contabilidad Regional de España (1995-2015), Consumer Price Index (1995-2015). 38 Table 5: Robustness Checks #2 – Accounting for Differences in Population Composition Sample: All Males Females Panel A: Accounting for Changes in the Age and Potential Income Composition of the Population (1989-2015) Social Assistance Pensioners’ Rate Differential -0.0005** -0.0005* -0.0005*** (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0007** -0.0007** -0.0007** (0.0003) (0.0003) (0.0003) Observations 1,164,429 574,087 590,342 R-squared 0.0035 0.0036 0.0037 Dependent Variable Mean (S.D. in parentheses) 0.006 (0.07) 0.006 (0.07) 0.005 (0.07) Panel B: Born in Spain (1992-2015) Social Assistance Pensioners’ Rate Differential -0.0002 -0.0002 -0.0003 (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0007** -0.0008** -0.0006** (0.0003) (0.0004) (0.0003) Observations 947,855 469,092 478,763 R-squared 0.0033 0.0030 0.0038 Dependent Variable Mean (S.D. in parentheses) 0.005 (0.07) 0.005 (0.07) 0.005 (0.07) Panel C: Born Abroad (1992-2015) Social Assistance Pensioners’ Rate Differential -0.0007 -0.0008 -0.0006 (0.0013) (0.0017) (0.0014) Non-contributory Pensioners’ Rate Differential 0.0008 0.0017 -0.0001 (0.0022) (0.0024) (0.0023) Observations 76,982 36,037 40,945 R-squared 0.0059 0.0078 0.0058 Dependent Variable Mean (S.D. in parentheses) 0.010 (0.10) 0.011 (0.11) 0.009 (0.10) Panel D: Individuals Younger than 25 (1988-2015) Social Assistance Pensioners’ Rate Differential -0.0002 0.0001 -0.0006 (0.0003) (0.0004) (0.0004) Non-contributory Pensioners’ Rate Differential 0.0000 0.0006 -0.0006 (0.0004) (0.0005) (0.0005) Observations 519,065 266,354 252,711 R-squared 0.0044 0.0042 0.0055 Dependent Variable Mean (S.D. in parentheses) 0.006 (0.07) 0.005 (0.07) 0.006 (0.08) Notes: All specifications include a constant term, province of origin and year fixed effects, demographic and human capital controls (i.e. age, male sex, marital status, an immigrant dummy, educational attainment dummies using tertiary education as reference, a dummy indicating emancipation benefit eligibility, and a dummy for receiving unemployment benefits), and macroeconomic controls (standardized differentials between the province of origin employment rate, share of population over 64 years old, share of the population receiving unemployment benefits, and international migration rates and the weighted average of the corresponding rates at potential destinations). The specification in Panel A additionally controls for standardized differentials in the share of population over 64 years old who are primary or less than primary educated. Observations are weighted using the individual weights in the EPA. Standard errors are clustered at the province level. Standard errors are shown in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1988-2015), EVR (1988-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1988-2015). 39 Table 6: Robustness Checks #3 – Addressing Confounding Impacts from the Dependency Law, Regional Income Support Programs, and Labor Market Reforms Sample: All Males Females Panel A: Sample Period Prior to the Dependency Law (1988-2007) Social Assistance Pensioners’ Rate Differential -0.0003* -0.0003 -0.0004* (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0006** -0.0007* -0.0006** (0.0003) (0.0004) (0.0003) Observations 916,318 452,950 463,368 R-squared 0.0041 0.0044 0.0040 Dependent Variable Mean (S.D. in parentheses) 0.0055 (0.074) 0.0055 (0.074) 0.0054 (0.073) Panel B: Excluding Regions with High Income Support Policies (Asturias, Basque Country, Navarra, and Cantabria) Social Assistance Pensioners’ Rate Differential -0.0004** -0.0004* -0.0005*** (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0008*** -0.0008** -0.0009*** (0.0003) (0.0003) (0.0003) Observations 1,103,058 544,248 558,810 R-squared 0.0035 0.0036 0.0037 Dependent Variable Mean (S.D. in parentheses) 0.0058 (0.076) 0.0058 (0.076) 0.0057 (0.076) Panel C: Accounting for Changes in the Share of Temporary Workers Social Assistance Pensioners’ Rate Differential -0.0004** -0.0003 -0.0004*** (0.0002) (0.0002) (0.0002) Non-contributory Pensioners’ Rate Differential -0.0006** -0.0006* -0.0006** (0.0003) (0.0003) (0.0003) Observations 1,207,489 595,408 612,081 R-squared 0.0035 0.0035 0.0037 Dependent Variable Mean (S.D. in parentheses) 0.005 (0.07) 0.005 (0.07) 0.004 (0.07) Notes: All specifications include a constant term, province of origin and year fixed effects, demographic and human capital controls (i.e. age, male sex, marital status, an immigrant dummy, educational attainment dummies using tertiary education as reference, a dummy indicating emancipation benefit eligibility, and a dummy for receiving unemployment benefits), and macroeconomic controls (standardized differentials between the province of origin employment rate, share of population over 64 years old, share of the population receiving unemployment benefits, and international migration rates and the weighted average of the corresponding rates at potential destinations). Observations are weighted using the individual weights in the EPA. Standard errors are clustered at the province level. Standard errors are shown in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1988-2015), EVR (1988-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1988-2015). 46 Table A.2: Placebo Check for Spurious Correlations Sample Males and Females Males Females Period: 1988-1990 – Are non-contributory pension differences spuriously correlated to mobility patterns in years prior? Social assistance pensioners rate differential t+4 -0.0007 -0.0005 -0.0007 (0.0007) (0.0009) (0.0007) Non-contributory pensioners rate differential t+4 -0.0002 0.0002 -0.0006 (0.0018) (0.0021) (0.0018) Observations 130,299 64,335 65,964 R-squared 0.0030 0.0036 0.0032 Mean dependent variable (S.D. in parentheses) 0.005 (0.07) 0.005 (0.07) 0.004 (0.06) Notes: All specifications include a constant term, province of origin and year fixed effects, demographic and human capital controls (i.e. age, male sex, marital status, an immigrant dummy, educational attainment dummies using tertiary education as reference, a dummy indicating emancipation benefit eligibility, and a dummy for receiving unemployment benefits), and macroeconomic controls (standardized differentials between the province of origin employment rate, share of population over 64 years old, and share of the population receiving unemployment benefits and the weighted average of the corresponding rates at potential destinations). Observations are weighted using the individual weights in the EPA. Standard errors are clustered at the province level. Standard errors are shown in parentheses. *significant at 10% **significant at 5%; ***significant at 1%. Source: EPA (1995-2015), EVR (1995-2015), Anuario de Estadísticas Laborales (1988-2015), Cifras de Población (1995-2015).