The Effects of Mass Migration on Labour and Educational Outcomes. Evidence from Chile. Dante Contreras

Home , Chileans, Regions of Chile

The effects of mass migration on labour and educational Outcomes. Evidence from Chile. Dante Contreras. Department of Economics, Universidad de Chile

1 Introduction

Over the last decades, migratory phenomena have become a key topic in most countries. In this context, the Latin America and the Caribbean (LAC) region has played a major role. According to the 2017 UN International Migration Report, nearly 15% of the world’s migrants arrive from this region. Even though 70% of LAC migrants reside in the USA, it is relevant to note that 16% of people migrate within this region, constituting the second most relevant destination.

Regarding this intra-regional migration, Chile is among the countries that have received the largest number of foreigners. More specifically, since 2015, Chile has received a large contingent of immigrants from Venezuela and Haiti. According to the 2017 Census, considering the 2015-2017 period only, said groups grew by 254% and 273% respectively1, while National Statistics Institute (INE) estimates for late 2018 place them as the first and third largest migrant bodies in Chile. Undoubt- edly, Venezuelans constitute the most emblematic example of intra-regional migration of the last years. In this regard, a 2018 report by the International Organization for Migration highlights that although Colombia, the USA, and Spain have received the largest absolute quantity of Venezuelan migrants, Chile has experienced the world’s sharpest increase, with the number of people from that country displaying a 15-fold increase in the last 3 years. Chile’s relative economic and political stability may be a possible influence on this migratory process.

Therefore, Chile provides an interesting opportunity for empirically examining the consequences of rapid mass migration phenomena. Indeed, given Chile’s small population, any eﬀects of mass immigration phenomena should be statistically identiﬁable.

Although the empirical literature has examined the effects of migration on a variety of outcomes2, there is insufficient information about the effects of recent migratory shocks in countries of a similar level of development. Studying migration in Chile can shed light on said effects. The relative size of the country, combined with the mass arrival of immigrants, provides an opportunity for identifying any possible effects.

1Own calculation with 2017 Census data. 2Lozano and Steinberger (2010) record at least 145 studies on the topics of migration and labour markets published between 1990 and 2010. Okkerse (2008), Longhi et al. (2005), and Dustmann et al. (2016) also provide good literature summaries, while at the same time outlining the main challenges and empirical diﬀerences involved in identifying the eﬀects of migration. Likewise, other relevant studies of this literature include Borjas (1999, 2014), Card (2001, 2009), and Peri (2016), among others.

1 2 Recent immigrant inﬂows in Chile: Some statistics

In Latin America and the Caribbean, Chile is one of the countries that has received the most immigrants over the last years. Panel (a) of Figure 1 shows that the share of immigrants relative to the total population of Chile tripled between 2011 and 2017 (from 1.4% to 4.4%), with the total number of people whose mothers resided in another country when they were born increasing from 243,878 to 777,4073, according to National Socioeconomic Characterization Survey (CASEN) data. Likewise, panel (b) of this ﬁgure reveals a major increase in the number of children born in another country who have enrolled in the Chilean school system. The number of immigrant children has increased 3.5 times between 2015 and 2018, with students from other countries reaching 114,000 according to information issued by the Research Center of the Ministry of Education (Mineduc).

This trajectory is also characterized by the distribution of immigrants by country of origin. Panel (a) of Figure 2 shows the evolution of the number of migrants in Chile for the six main communities, from 2011 to 2017. The graph has three salient aspects. First, the stable upwards trend of the population of Peruvian origin. Second, the drop in the relative importance of the Argentine colony, which was the second largest in 2011 and fell to the sixth place in 2017. Third, the rapid increase in Venezuelan and Haitian immigrants between 2015 and 2017, with the former becoming the largest colony in Chile (24.2% of the total number of migrants), surpassing the Peruvian community (22.2% of the total number of migrants) for the ﬁrst time.

As for panel (b) of Figure 2 , it shows the progression in the number of migrant students in the Chilean school system for the same six colonies presented in panel (a), from 2015 to 2018. Again, the period is marked by the rapid increase in enrollees from Venezuela and Haiti since 2016, with students from these countries becoming the ﬁrst and second colony in the Chilean school system in 2018. Likewise, as in panel (a), the evolution of the number of Peruvian students is quite similar; however, from being the largest colony in the Chilean school system, it was surpassed by Venezuela and Haiti as previously noted.

3Migrants are deﬁned as people residing in a country diﬀerent from that in which they were born[*][people whose mothers resided in another country when they were born].

2 Figure 1: Evolution of the number and percentage of migrants (migrant students) in Chile between 2011-2017 (2015-2018)

(Number and percentages, total population) (Number and percentage, total enrollment) 900,000 200,000 (4.4%) 800,000 777,407 180,000

700,000 160,000 140,000 600,000 (3.19%) (2.7%) 120,000 114,326 500,000 465,319 100,000 (2.1%) (2.2%) 400,000 354,581 77,608 80,000 (1.7%) 300,000 (1.4%) 61,086 243,878 60,000 200,000 (0.9%) 40,000 30,625

100,000 20,000

0 0 2011 2013 2015 2017 2015 2016 2017 2018 Year Year

(a) Migrants (b) Migrant Students

Source of Panel (a): 2011, 2013, 2015, and 2017 CASEN surveys. Calculations are performed using regional expansion factors; Source of Panel (b): Sistema General de Estudiantes (SIGE), 2015-2018.

Given Chile’s particular geography, the regional distribution of migrants is another relevant consid- eration. This is illustrated in Table 1 for the year 2017. The Table reflects the strong concentration of migrants in the Metropolitan Region (71%), with the Tarapacá,Valpara´ıso,and Antofagasta Regions lagging far behind. The fourth column of the Table shows the share of migrants relative to the total population of each region. This indicator is significantly heterogeneous. The Tarapacá Region stands out in this regard, as migrants constitute 12.8% of its population. This region is followed by the Metropolitan Region, the Arica and Parinacota Region (7.7% in both), and the Antofagasta Region (6.1%).

A recent report issued by the National Institute of Statistics (INE) states that, as of late 2018, four regions have a share of migrants of over 10%: Tarapacá(16.9%), Antofagasta (13.6%), Arica and Parinacota (10.4%), and Metropolitan (10.2%). A comparison between the 2017 CASEN survey (generated toward the end of that year) and the INE data (from late 2018) reveals that the share of migrants has increased significantly, especially in the Antofagasta Region, which rose from the fourth to the second place in a single year. This illustrates the constant growth of the migratory influx in all areas of the country.

The last 3 columns of the same Table paint a similar picture regarding the share of migrant students across all the regions of Chile in 2018. However, a significant difference is that Antofagasta, not Tarapacá,has the largest concentration of foreign students (over 11%). This is consistent with the INE information mentioned in the above paragraph.

The major increase in migration from 2011 to date, as revealed thus far, can have relevant consequences for the Chilean labour market. Before examining these possible eﬀects, it is necessary to identify the sectors where migrants are ﬁnding jobs and determine the characteristics of these jobs.

3 As suggested by Altonji and Card (1991), if immigrants tend to work in the same sectors as a speciﬁc subgroup of natives, like those with a lower educational level, the eﬀects of migration on this subgroup should be stronger.

Figure 2: Evolution of the number of migrants (migrant students) in Chile, by country of origin between 2011-2017 (2015-2018)

(Number of individuals) (Number of students) 200,000 20,000 180,000 18,000 160,000 16,000 140,000 14,000 120,000 12,000 100,000 10,000 80,000 8,000 60,000 6,000 40,000 4,000 20,000 2,000 0 0 2011 2013 2015 2017 2015 2016 2017 2018 Year Year

Perú Arg Perú Arg Bolivia Col Bolivia Col Haití Ven Haití Ven

(a) Migrants (b) Migrant students

Source of Panel (a): 2011, 2013, 2015, and 2017 CASEN surveys. Calculations are performed using regional expansion factors; Source of Panel (b): Sistema General de Estudiantes (SIGE), 2015-2018.

To determine this, Table 2 shows the distribution of migrant workers along with the distribution of low-education natives, divided by economic activity (2-digit ciiu). These data show that migrants tend to occupy all sectors rather heterogeneously. The Table, which is sorted descendingly according to the distribution of migrants among sectors, shows that the wholesale and retail commerce sector employs most migrant workers (21.7%), followed by Hotels and Restaurants (14.2%), Real Estate (12.2%), and Private Homes with Domestic Service (10.1%). All other sectors are under 10%. The fourth column shows the distribution of Chileans with less than 12 years of education (Low skilled) in each economic sector. In general, Chileans with the lowest level of education tend to work in the sectors where migrants are most numerous. The distribution of low-schooling migrants in all sectors displays a similar pattern, as shown in the ﬁfth column. Speciﬁcally, the wholesale and retail commerce sector –the main employer of migrants– concentrates the largest share of low-skilled natives and migrants. Finally, the last two columns show the average number of years of education of natives and migrants, with the latter surpassing the former in most cases.

In summary, recent statistical reports indicate that Chile’s immigrant population has sharply increased and that it is heterogeneously distributed across the economic sectors, as migrants are concentrated in the areas that Chileans with the lowest educational levels occupy. Therefore, it is relevant to examine the possible consequences of this migratory phenomenon for local labour markets. In line with this situation, the number of students born in other countries and who enroll in the Chilean school system has also increased greatly; thus, it is important to study how this experience has inﬂuenced student performance in the educational system.

4 Table 1: Regional distribution of migrants (migrant students) in 2017 (2018)

Migrants Migrant Students Region N Distribution Proportion N Distribution Proportion Region XIII: Metropolitan 557,965 71.77% 0.077 69,919 61.16% 0.051 Region I: Tarapac´a 44,537 5.73% 0.128 8,289 7.25% 0.104 Region V: Valpara´ıso 39,691 5.11% 0.021 4,976 4.35% 0.013 Region II: Antogafasta 35,817 4.61% 0.061 13,154 11.51% 0.100 Region VIII: Biob´ıo 15,632 2.01% 0.009 2,290 2.00% 0.005 Region VI: O’Higgins 13,300 1.71% 0.014 2,625 2.30% 0.013 Region XV: Arica 12,451 1.60% 0.077 3,669 3.21% 0.070 Region IV: Coquimbo 10,726 1.38% 0.013 2,561 2.24% 0.015 Region VII: Maule 9,731 1.25% 0.009 2,008 1.76% 0.009 Region IX: Araucan´ıa 8,893 1.14% 0.008 852 0.75% 0.004 Region X: Los Lagos 8,774 1.13% 0.009 1,063 0.93% 0.005 Region XII: Magallanes 5,505 0.71% 0.036 629 0.55% 0.019 Region III: Atacama 5,163 0.66% 0.018 1,763 1.54% 0.026 Region XVI: Nuble˜ 3,649 0.47% 0.007 --- Region XIV: Los R´ıos 3,048 0.39% 0.008 305 0.27% 0.003 Region XI: Ays´en 2,525 0.32% 0.024 223 0.20% 0.009 777,407 100.00% 0.044 114,326 100% 0.032

Source of migrant data: 2011, 2013, 2015, and 2017 CASEN surveys. Calculations are performed using regional expansion factors; Source of migrant student data: Sistema General de Estudiantes (SIGE), 2015-2018. Data for the Nuble˜ Region (XVI) are unavailable.

Table 2: Distribution of migrants among economic sectors, and of low-skilled natives and migrants (2017)

% Low-Skilled Mean years of education Sector N % Migrants Natives Migrants Natives Migrants and retail commerce 110,830 21.69% 20.05% 19.87% 11.54 13.19 Hotels and restaurants 72,705 14.23% 4.18% 8.55% 11.83 13.29 Real estate and business activities 62,323 12.20% 3.45% 6.63% 13.93 15.21 Private homes with domestic service 51,815 10.14% 11.47% 15.49% 9.31 12.17 Manufacturing industries 47,490 9.29% 9.92% 14.17% 11.58 12.56 Construction 46,978 9.19% 13.49% 15.44% 10.79 12.08 Transport, storage, and communication 25,078 4.91% 6.67% 3.10% 11.95 14.28 Social and health care services 21,989 4.30% 1.36% 0.79% 14.82 16.11 Other community service activities 20,236 3.96% 2.59% 2.74% 12.78 13.79 Agriculture, livestock farming, hunting, and forestry 16,845 3.30% 18.61% 9.20% 8.78 10.82 Teaching 13,337 2.61% 1.69% 2.04% 15.10 15.27 Finance 6,303 1.23% 0.28% 0.20% 14.73 16.59 Public administration and defense 5,388 1.05% 2.35% 0.53% 13.72 15.50 No data 4,772 0.93% 0.59% 0.21% 12.65 14.70 Exploitation of mines and quarries 2,549 0.50% 0.96% 0.55% 12.98 12.63 Fishing 1,130 0.22% 1.73% 0.37% 9.86 11.61 Electricity, gas, and water supply 800 0.16% 0.60% 0.10% 12.61 13.66 Extraterritorial organizations and bodies 438 0.09% 0.00% 0.00% 14.66 17.16 511,006 100% 100% 100% 11.77 12.63

Source: CASEN 2017. Calculations are performed using regional expansion factors. Low-skilled deﬁned with less than 12 years of schooling (incomplete secondary education).

5 3 Proposed Methodology 3.1 Empirical identification of labour outcomes This research proposal is aimed at shedding light on the impacts of migration in Chile, specifically by identifying and analyzing the effects of recent migration on the salaries of lower-skilled natives. In order to achieve this goal, the proposed methodology is represented by the following equation:

Yirs,t = α + δHMrs,t + γLSi + βHMrs,t ∗ LSi,t + Xirs,tρ + λr + λs + εirs,t (1)

Where Yirs,t is the logarithm of the nominal hourly salary of native individual i who lives in region r and works in economic sector s, HM (High Migration) is a dummy that indicates whether the individual works in a sector s and a region r where the average share of migrants is higher than in the region-sector cell at a national level, i.e. a sector-region with a high percentage of employed migrants, above the average proportion. LS (Low Skills) is a dummy that indicates whether individual i has a level of accumulation of human capital below 12 years of education. The interest coefficient is parameter β which accompanies the interaction between HM and LS, and which is interpreted as the percentage difference between the salary of low-skilled natives (LS) working in a high-migration sector (HM) and the salary of those working in a low-migration sector, i.e. β is similar to an estimator of difference in differences: salary differences between low-skilled natives, depending on whether they belong to sectors with a high/low percentage of employed migrants4. For its part, Xirs,t is a vector of individual variables such as sex, years of education, potential experience, and its square, dummies that indicate whether i resides in an urban area and whether he/she is married. To this, we add regional and economic sector fixed effects, denoted by the terms λr and λs respectively. Lastly, α is the intercept and εirs,t is the error term. All variables are measured in period t.

It should be noted that the selection of the dummy HM depends on the variation in the share of migrants working in each region-sector cell, given the high heterogeneity in migrants’ chosen sector and region, as shown in Tables 1 and 2.

Given the endogeneity of variable HM, which is potentially biased by immigrants’ chosen region of residence, we propose using an instrumental variable, similar to that used in the literature5. This variable is the past share of migrants in each region-sector cell. The use of this instrument is consistent with the fact that migrants tend to be drawn to cities with a high prior concentration of migrants (Bartel (1989); Greenwood McDowell (1986)). Following the same idea, immigrants are likely to be clustered in economic sectors with a high prior concentration of migrants. Therefore, in the ﬁrst stage, HM is estimated as follows:

ˆ 0 HM rs,t = π0 + π1MRrs,t−1 + Xirs,tπ2 + λr + λs + νirs,t (2)

4From a conceptual point of view, the latter difference is analogous to a “treatment” in the difference in differences strategy. 5This literature includes studies by Altonji and Card (1991) and Card (2001). For instance, in the first study, the authors use the percentage of migrants in each U.S. city in 1970 as an instrument of the same percentage for 1980.

6 6 Where MRrs,t−1(Migration Rate) is the share of migrants employed in each region-sector cell in 0 the previous period t−1. Vector Xirs,t includes the same control variables as (1), but an interaction between MR and LS is added as well as LS by itself. In (2), π0 is intercept, λr and λs are the fixed effects by region and sector, respectively, and νirs,t is the error term. All variables are measured in t , except from instrument MR. From (2), the predicted value of HMˆ is included in (1) to estimate coefficient β without bias.

3.2 Empirical Identification of educational outcomes As in the above identification, the strategy used to evaluate the effect of migration on native students’ performance is represented by equation (3):

SIMCEis,t = α + βHMs,t + Xis,tρ + εis,t (3)

In this equation, the dependent variable is the Mathematics or Reading SIMCE score 7 in 4th grade, depending on the test administered, measured in standard deviations.8 The variable of interest is HM ,which is a dummy indicating whether native student i attends a school s with a percentage of foreign students higher than the national mean. A higher proportion of foreign students (measured at school level) can have negative or positive effects on native students’ performance. If schools have limited resources, if teachers received insufficient training to work with diverse students, or if government programs are unable to respond to foreign students’ needs, local students may experience negative effects. In contrast, if migrant students come from families with a higher educational level or if local students benefit from increased diversity, the latter should attain higher SIMCE scores. Vector Xis,t contains control covariables which have been extensively used in the education literature, such as parents’ educational level and income, fixed effects by region, sex and rural area dummies , and categorical variables of school type and school SES. Lastly, εis,t is the error term at the school cluster level. All variables are measured in period t.

As in the strategy used to evaluate the eﬀects of migration on native salaries in the local labour market, the potential endogeneity of variable HM in equation (3) is tackled using two-stage least squares (2SLS), where the ﬁrst stage is:

HMˆ s,t = π0 + π1MRs,t−1 + Xis,tπ2 + νis,t (4)

In equation (4), the instrument for variable HM is MRs,t−1 , which contains the percentage of migrant students in school s in period t − 1. This instrument is consistent with the model advanced by Bartel (1989), Greenwood, and McDowell (1986), among others, given that, if migrants are drawn to cities with a high prior concentration of foreigners, a similar phenomenon should be true

6To address the heterogeneity of the rate of migrants in each region-sector cell, we follow the same logic used to construct variable HM. 7The Education Quality Measurement System (SIMCE) is a set of standardized tests used in Chile to measure students’ command of topics of the school curriculum. These tests are annually administered to students of several grades. 8This grade was selected because the SIMCE test is administered to this level every year.

7 of students from other countries. Continuing with the description of equation (4), Xis,t contains the same control variables as (3) and νis,t is the error term at the school cluster level. All the variables in (4) are measured in t, except from the instrumental variable HM, as previously noted.

3.3 Data and hypotheses The data to be used to study the eﬀect of migration on natives’ salary (equations (1) and (2)), come from the National Socioeconomic Characterization Survey (CASEN), a cross-sectional public database executed biannually since 1985. It is a representative database of Chilean homes at the country level and contains a wide range of variables that identify the sociodemographic characteristics of the Chilean population.

To estimate equation (2) in the ﬁrst stage, the 2011 wave of the CASEN survey will be used (i.e. t − 1 = 2011), because immigration levels in Chile remained relatively stable until then and began rising more and more markedly after that year, as shown in ﬁgure 1.

Equation (1) will be estimated separately with the 2013, 2015, and 2017 CASEN waves (i.e. t = 2013, 2015 o 2017). In addition, equation (1) will be separately estimated in restricted samples of men and women in order to identify possible heterogeneous eﬀects by sex.

We expect the estimator of difference in differences of equation (1), β, to differ by sex and across the periods analyzed. In this regard, β might be higher in periods closer to 2017 due to the increases in migrant population in Chile from 2011 onwards. If migrants’ work is a strong substitute of the work performed by lower-skilled natives, which is where most migrants tend to be employed, β should be negative.

To study the eﬀect of the increased enrollment of students from other countries on educational outcomes (equations (3) and (4)), we will use the 4th grade SIMCE databases from 2015 to 2018. This database derives from a census-type measurement performed annually by the Education Qual- ity Agency at several levels of the primary and secondary system. It comprises sociodemographic variables such as parents’ educational level and the income level of students’ families, individual and school variables, and each student’s scores in the Mathematics and Reading SIMCE test. General Student Information System (SIGE) databases, generated by MINEDUC’s Research Center, reveal whether a student was born in Chile or abroad, making it possible to impute this information with SIMCE data.

With this source of information, equation (4) is estimated in the ﬁrst stage with 2015 data (i.e. t − 1 = 2015)9 to correct for the endogeneity of variable HM and estimate (3) with data from 2016 to 2018 (i.e. t = 2016, 2017 or 2018). In order to identify heterogeneous eﬀects, (4) will be estimated using samples restricted by student sex as well as by school type and SES.

The sign of coeﬃcient β of equation (3) will indicate how prepared the Chilean school system was to respond to the increase in foreign students, as reﬂected by native students’ performance. Thus, a negative β might be associated with poor preparation in schools with a high percentage of migrant

9The analysis starts in this period because information about students’ migrant status is only available from this year onwards.

8 students, with the opposite being true if the coeﬃcient were found to be positive. By estimating β of equation (4) across several years, it is also possible to determine whether schools have adapted to these migrant enrollment shocks: if β is negative and decreases in magnitude between 2016 and 2018, or if it shifts from negative to positive between these years, then schools with a large proportion of foreign students could be said to have adapted to these changes through contingency plans, for example. The opposite is true if β increases in magnitude during this interval. Consequences for academic outcomes can shed light on relevant implications and/or inform public policies aimed at improving the educational system’s adaptation to the new and culturally diverse makeup of school enrollment in Chile.

9 References

Altonji, Joseph G., and David Card. 1991. “The Eﬀects of Immigration on the Labor Market Outcomes of Less-skilled Natives.” In Immigration, Trade and the Labor Market, edited by John M. Abowd and Richard B. Freeman, pp. 201–234. University of Chicago Press

Bartel, A. P. (1989). Where do the new US immigrants live?. Journal of Labor Economics, 7(4), 371-391.

Borjas, G.J., 1999. “The Economic Analysis of Immigration.” In: Handbook of Labor Economics, Volume 3. Elsevier.

Borjas, G. J., 2014. Immigration Economics. Harvard University Press.

Card, D., 2001. “Immigrant Inﬂows, Native Outﬂows, and the Local Labor Market Impacts of Higher Immigration.” Journal of Labor Economics 19.

Card, D., 2009. “Immigration and Inequality.” American Economic Review 99: 1–21.

Dustmann, C., U. Sch¨onberg, and J. Stuhler, 2016. “The Impact of Immigration: Why Do Studies Reach Such Diﬀerent Results?” Journal of Economic Perspectives 30(4).

Greenwood, M. J., McDowell, J. M. (1986). The factor market consequences of US immigration. Journal of Economic Literature, 24(4), 1738-1772.

Longhi, S. , Nijkamp, P. and Poot, J. (2005), A Meta-Analytic Assessment of the Eﬀect of Immi- gration on Wages. Journal of Economic Surveys, 19: 451-477.

Naciones Unidas, OrganizaciónInternacional para las Migraciones, (2018). Tendencias Migratorias Nacionales en Américadel Sur: RepúblicaBolivariana de Venezuela. Febrero 2018.

Lozano, Fernando A. and Steinberger, Michael D., Empirical Methods in the Economics of Inter- national Immigration. IZA Discussion Paper No. 5328.

Okkerse, L. (2008). How to measure labour market eﬀects of immigration: A review. Journal of Economic Surveys, 22(1), 1-30

Peri, G., 2016. “Immigrants, Productivity, and Labor Markets.” Journal of Economic Perspectives 30.

United Nations, Department of Economic and Social Aﬀairs, Population Division (2017). Interna- tional Migration Report 2017: Highlights.