Birthplace Diversity and Economic Prosperity

Transcription

1 Birthplace Diversity and Economic Prosperity Alberto Alesina, Johann Harnoss and Hillel Rapoport June 2013 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

2 Plan 1 Introduction 2 Skill complementarities and diversity 3 An index of birthplace diversity 4 Empirical Analysis 5 IdentiÖcation 6 Conclusion Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

3 Introduction Foreign-born individuals now represent on average ten percent of the workforce of the OECD, a twofold increase in twenty years, and a threefold increase for the highly educated and skilled This growing diversity in terms of birthplaces at the country level may have far reaching economic implications as economic theory suggests that higher diversity could lead to beneöcial skill complementarities in certain production processes but also to ine ciencies via higher transaction costs due to mistrust and lack of social cohesion. The empirical literature has so far focused on ethnic, linguistic, and genetic diversity. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

4 Introduction Ethnic and linguistic fractionalization: has been shown to have a negative e ect on economic growth (Easterly and Levine, 1997, Collier 2001, Alesina et al., 2003) in cross-country comparisons. Interestingly, however, these e ects tend to converge to zero or even turn mildly positive in richer countries (Alesina and La Ferrara, 2005). Genetic and cultural diversity: - Ashraf and Galor (2013a) Önd an inverted u-shaped relationship between genetic diversity and productivity, indicating the trade-o between beneöcial forces of diversity expanding the technology frontier and detrimental ones due to communication and coordination problems. - Ashraf and Galor (2011) Önd that cultural diversity (based on World Values Survey data) is positively correlated with contemporary development and suggest that cultural diversity facilitated the transition from agricultural to industrial societies. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

5 Introduction We re-address this question using a new measure and a new perspective: birthplace diversity. Ethnic and birthplace diversity are empirically (perhaps surprisingly) almost completely uncorrelated; same for genetic and birthplace diversity. Also conceptual di erences: First-generation immigrants, who grew up in di erent cultural contexts, are a more diverse group than second-generation immigrants, who grew up in the same cultural environment, went to the same schools, etc. Birthplace diversity measures thus have the potential to identify new economic e ects of diversity. However, the empirical evidence on birthplace diversity and development is scant and limited to the context of the United States. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

6 Introduction Ottaviano and Peri (2006) construct a measure of diversity for using migration data on US metropolitan areas; Önd positive e ects on the productivity of native workers measured by their wages. Peri (2012) Önds positive e ects of the diversity coming from immigration on the productivity of US states. Interpretation: unskilled migrants promote e cient task-specialization and adoption of unskilled-e cient technologies, more so when immigration is diverse. Ager and Br ckner (2011) study the link between immigration, diversity and growth in US counties in They Önd that fractionalization increases output while polarization decreases it. Finally, a paper by Ortega and Peri (2013) developed independently from this paper also analyzes the e ect of openness to and diversity of trade v. immigration, showing that the latter dominates the former. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

7 Introduction Three contributions: Data: we construct and discuss the properties of a new index of birthplace diversity for the workforce of 195 countries in 1990 and 2000, disaggregated by skill/education level, and computed both for the workforce as a whole and for its foreign-born component. We argue this is a fundamentally di erent dimension of diversity. Empirics: we investigate the relationship between birthplace diversity and economic development and Önd that birthplace diversity of immigrants is positively related to productivity, this e ect being stronger for skilled migrants and in richer countries. IdentiÖcation: we make progress toward solving endogeneity issues by specifying a gravity model to predict the diversity of immigration and conörm our initial Öndings in a range of 2SLS models. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

8 Plan 1 Introduction 2 Skill complementarities and diversity 3 An index of birthplace diversity 4 Empirical Analysis 5 IdentiÖcation 6 Conclusion Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

9 Skill complementarities and diversity The costs and beneöts of diversity: theory People born in di erent places have di erent productive skills because they have been exposed to di erent experiences and perspectives that help them interpret and solve problems di erently. These di erences can be complementary. Alesina et al. (2000) formalize this using a Dixit-Stiglitz production function where outputs increase in the variety of inputs (e.g., workers). Lazear (1999a,b) includes beneöts from diversity via production complementarities from relevant disjoint information sets and costs via barriers to communication that rise in the diversity of workers, suggesting there is an optimal degree of diversity. Also, diverse groups of immigrants have stronger incentives to assimilate. Hong and Page (2001) emphasize cognitive di erences between peopleís internal perspectives (interpretation of a complex problem) and heuristics (algorithms to solve these problems), showing diversity can compensate lower average skills in a group/team. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

10 Skill complementarities and diversity The costs and beneöts of diversity: micro/experimental evidence Management and organization literature: Önds diversity (in terms of gender, education, tenure, nationality) being often beneöcial for performance but also decreasing team cohesion and increasing coordination costs (Milliken and Martins, 1996, OíReilly et al., 1989, Hambrick et al., 1996). Recent experimental evidence with business school students Hoogendoorn and van Praag (2012). Micro-studies using Örm-level data: ñ Ethnic diversity of workers has negative e ect on plant performance in Kenya, as inter-ethnic rivalries lead to distortionary misallocation of resources (Hjort, 2012). ñ Birthplace diversity of workers has strong positive e ect on plant-level productivity in Germany (Trax, Brunow, Suedekum 2012), particularly strong in innovation-intensive manufacturing and high-tech sectors. Similar evidence for Austria (Boheim et al., 2012), Danemark (Parrotta et al., 2012), Holland (Ozgen et al., 2013). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

11 Skill complementarities and diversity The costs and beneöts of diversity: macro evidence Easterly and Levine (1997) show that ethnic fragmentation is associated with lower economic growth in Africa. Collier (1999, 2001) adds that this is less so in the presence of democratic institutions. Alesina and La Ferrara (2000, 2002) stress the role of trust, showing that individuals in racially diverse cities in the US participate less frequently in social activities, trust their neighbors less and have lower preferences for redistribution. This leads to lower provision of productive public goods (Alesina, Baqir and Easterly, 1999). Alesina, Michalopoulos and Papaioannou (2012) stress the inequality dimension of ethnic diversity (i.e., it is the interplay between ethnic fractionalization and ethnic inequality that leads to conáict). See also Desmet, Ortun and Wacziarg (2012) on the role of linguistic distance. Esteban, Mayoral and Ray (2011, 2012) Önd polarization and fractionalization to correlate positively with conáicts over public and private goods, respectively. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

12 Skill complementarities and diversity Measuring diversity A populationís diversity is commonly measured by fractionalization (Alesina et al. 2003, Fearon 2003) and polarization indices (Esteban and Ray 1994, Reynal-Querol 2002). HerÖndahl-type fractionalization indices (which attain their maximum when each individual belongs to a di erent group) have been shown to work better in models of economic growth while polarization indices (which attain maximum with two equally large groups) have been shown to be adequate predictors of civil conáicts. Ethnic fractionalization measures donít distinguish, for example, between a 1rst and 2nd-generation Italian in the US, or between Italians and Germans ("Caucasiansî); linguistic fractionalization more accurate in separating language groups but also fails to distinguish between 1rst and 2nd-generation. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

13 Skill complementarities and diversity Measuring diversity A dimension of diversity among people that remains largely understudied is the diversity caused by di erences in peopleís country of birth. If early pre-working age years are formative for oneís own values, perspectives and skills these di erences last a lifetime and may serve as variation to be exploited for economic analysis. Shaped by di erent education systems and social values, this type of diversity is more likely to result in production function complementarities than deep-seated di erences in skin color or language spoken at home. To explore this dimension of diversity, this paper introduces a new diversity index which is more likely to be closer to the correct one when we try to explain skill complementarity: diversity in peopleís birthplaces. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

14 Plan 1 Introduction 2 Skill complementarities and diversity 3 An index of birthplace diversity 4 Empirical Analysis 5 IdentiÖcation 6 Conclusion Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

15 An index of birthplace diversity: data Our computation of birthplace diversity indices relies on the Docquier, Ozden, Parsons and Artuc (2012) (henceforth DOPA) data set, which allows for characterising the size, origin-mix, and skill structure of a countryís foreign-born labor force. It provides bilateral data on immigration by country of birth, skill category (skilled = college educated v. unskilled) and gender, for 195 receiving countries in 1990 and Immigrants are deöned as foreign-born individuals aged 25 or more at census or survey date. Caveats: illegal immigration, children immigrants, and heterogeneity in skill levels (origin e ects). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

16 An index of birthplace diversity: decomposition We base our birthplace diversity measure on the HerÖndahl diversity index. This index measures the probability that two individuals drawn randomly have di erent countries of birth. Let s i refer to the share in the total population of individuals born in country i with i = 1,..., I. In particular, i = 1 for natives. The fractionalization index Div pop may be expressed as: Div pop = I Â i=1 s i (1 s i )=1 I Â i=1 (s i ) 2 (1) A given level of Div pop may come from a small-diverse or from a large-homogenous pool of immigrants. It is therefore useful to decompose Div pop to highlight these di erences. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

17 An index of birthplace diversity: decomposition We therefore decompose our diversity index as Div pop = Div between + Div within, where Div between is the diversity from immigration (as if all immigrants were from the same country) and Div within the additionnal diversity brought by immigrants : Div between = s 1 (1 s 1 )+(1 s 1 ) s 1 (2) Div within = I  i=2 [s i ((1 s i ) s 1 )] (3) However, this separation is not yet fully satisfying, since it does not separate clearly between size and variety e ects: Div within still depends on s 1 - the share of natives since  I i=2 s i =(1 s 1 ). We thus proceed to re-scale the Div within component so that it does not depend on s 1. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

18 An index of birthplace diversity: decomposition After transformations, we obtain: Div pop = Div between +(1 s 1 ) 2 Div Mig (4) where Div Mig = J h i  s j (1 s j ) j=2 (5) We can then rewrite (4) in terms of s F, the share of immigrants (deöned as foreign-born) and deöne s F =(1 s 1 ): Div pop = 2 s F (1 s F )+(s F ) 2 Div Mig (6) We have thus an expression of Div pop purely as a function of the size (or openness to) and diversity of immigration, s F and Div Mig. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

19 An index of birthplace diversity: descriptive analysis A Örst visual overview (see the maps in Figures 1) shows that ethnic fractionalization and Div pop di er considerably; this is also reáected in their low bilateral correlation of While birthplace diversity (Div pop ) is highly correlated (+0.98) with the share of immigration (s F ), these two and Div Mig di er considerably (overall diversity is dominated by the size over the variety of immigrants ñ but this is somewhat index speciöc). The pure variety of birthplaces is highest in many rich countries: Canada, Italy, Israel, Germany, Australia and the UK all have diversity of immigrants at about.9. The United States rank only 20 in a list of the most diverse immigrant countries (at.92) due to its relatively low diversity of unskilled workers (0.84). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

20 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

21 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

22 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

23 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

24 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

25 !!Birthplace!Diversity!and!Gene3c!Frac3onaliza3on!

26 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

27 Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

28 An index of birthplace diversity: descriptive analysis Correlation between ethnic fractionalization and birthplace diversity of immigrants very low (-.01 overall) and even negative at -.2 (in 2000) for skilled immigrants. Similarly, genetic diversity and birthplace diversity do not relate much (+.08 for diversity overall and -.08 for skilled diversity. Hence the "new dimension" claim. Correlation between s F and Div Mig surprisingly low, suggesting that the amount of immigration (size) and its composition (variety) are largely independent (holds irrespective of country size). But positive correlation when we look at Örst di erences: in the 1990s, the variety of immigrants rose when a country experienced an ináow of immigrants (true for skilled (+.22) but not for unskilled immigrants (bilateral correlation +.06). This is most likely due to the role of diaspora/immigrant networks. Skilled and unskilled diversity are highly correlated overall, with correlation coe cient of for the year 2000 (see the last panel of Figure 2). There are some interesting deviations (eg, the US). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

29 An index of birthplace diversity A horse race between birthplace diversity and fractionalization We run the following simple model to show simple correlations irrespective of confounding e ects: ln y kt = a + b 1 birthplace diversity + b 2 fractionalization + e (7) We expect the two coe cients b 1 and b 2 to have di erent signs and to be statistically di erent from each other (genetic diversity may have a positive linear and a negative quardratic term, following Ashraf and Galor, 2013a,b) We use Div pop in some speciöcations and replace it later by our size and variety components s F and Div mig, Alesina et al. (2003) for ethnic and linguistic fractionalization, and Ashraf and Galor (2013) for genetic diversity. Sample: 167 (GDP) and 134 (TFP) countries, cross section for the year 2000, estimation using OLS, heteroskedasticity-robust standard errors. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

30 An index of birthplace diversity A horse race between birthplace diversity and fractionalization We present results for GDP/capita and TFP/capita in Table 4a/b. The results conörm our main hypothesis: both measures of ethnic, linguistic and genetic fractionalization enter negatively (or U-shaped) whereas birthplace diversity (Div pop ) enters positively (at 1% signiöcance level) in all models. All fractionalization measures turn (signiöcantly) more negative once birthplace diversity is controlled for ñ s4uggesting that other established measures of fractionalization capture some positive e ects of birthplace diversity if used in isolation. Most importantly, our size and variety components both relate positively to development and remain highly signiöcant when entered jointly, a rming the relevance of both size and variety of immigration. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

31 !!Table!4b:!Birthplace!and!Gene3c!Diversity!!8!GDP! Dependent variable (log) (1) (2) (3) (4) (5) (6) GDP/capita GDP/capita GDP/capita GDP/capita GDP/capita GDP/capita Birthplace Diversity, Population 0.536*** 0.664*** (0.107) (0.0875) Birthplace Diversity, Immigrants 0.301*** 0.379*** (0.0924) (0.0743) Share of Immigration 0.419*** 0.545*** (0.129) (0.104) Genetic Diversity 10.54*** ** (2.887) (2.405) (2.424) Genetic Diversity (squared) *** ** (2.905) (2.423) (2.441) Ethnic Fractionalization *** *** *** (0.0936) (0.0825) (0.0718) Observations Adjusted R-squared All regressors standardized to mean of zero and standard deviation of one. Sample: Cross section for year All models include an intercept (not reported). Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

32 !!Table!4b:!Birthplace!and!Gene3c!Diversity!!8!TFP! (1) (2) (3) (4) (5) (6) Dependent variable (log) TFP/capita TFP/capita TFP/capita TFP/capita TFP/capita TFP/capita Birthplace Diversity, Population 0.309*** 0.370*** (0.0974) (0.0706) Birthplace Diversity, Immigrants 0.273*** 0.275*** (0.0696) (0.0599) Share of Immigration 0.211* 0.280*** (0.107) (0.0762) Genetic Diversity 8.119*** ** (2.087) (1.821) (1.844) Genetic Diversity (squared) *** ** (2.098) (1.838) (1.861) Ethnic Fractionalization *** *** *** (0.0648) (0.0635) (0.0545) Observations Adjusted R-squared All regressors standardized to mean of zero and standard deviation of one. Sample: Cross section for year All models include an intercept (not reported). Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

33 Plan 1 Introduction 2 Skill complementarities and diversity 3 An index of birthplace diversity 4 Empirical Analysis 5 IdentiÖcation 6 Conclusion Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

34 Empirical analysis: model and data In this section we test whether birthplace diversity is positively related to economic development, conditional on a range of confounding factors: ln y kt = a + b 1 diversity migrants skt +b 2 share immigration skt +b 3 origin e ects skt +b 4 years of schooling kt +b 5 market size controls kt +b 6 G kt + b 7 D k + b 8 F kt + b 9 Y kt + h t + e(8) where G kt is a vector of geographic characteristics, D k is a vector of fractionalization measures, F kt is a control for institutional development, Y kt is a vector of controls for trade openness and trade diversity, and h t is a time Öxed-e ect. We use indices s for skill levels (overall, skilled, unskilled), t for time (1990, 2000) and k for countries. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

35 Empirical analysis: model and data The results from our decomposition as well as our previous analysis point to the need to separate the share of foreigners, s F, and the diversity of immigrants, Div Mig, to isolate size and variety e ects. An alternative speciöcation with sub-samples for below/above median share of immigrants shows consistently higher positive estimates for diversity ñ see various robustness checks (e.g., Table 10) to ensure that our results are robust to the exclusion of small countries and of countries with very low immigration. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

36 Empirical analysis: model and data We control for a very wide range of potential confounding e ects: standard controls (e.g., education, population and area sizes, landlocked dummy) as well as additional control groups for trade structure, fractionalization, geography, and institutions. Trade: we use real trade openness from PWT 7.0 and also control for the structure of trade by constructing a measure of trade diversity (HerÖndahl index of exports and imports based on Feenstra et al., 2005). Fractionalization: both ethnic, linguistic and genetic fractionalization. Institutions: Polity-IV index of institutional quality Geography: absolute latitude, malaria intensity and share of population living within 100km of an ice-free coast. We test the robustness of our results to alternative geographical speciöcations, following Rodriguez and Rodrik (2001). We thus end up with a highly structured model (with our key variables and 15 covariates) and a short panel of 93 countries in 1990 and Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

37 Empirical analysis: results We run our model using an OLS estimator with standard errors clustered at the country level to account for serial correlation of standard errors and year Öxed e ects to account for year-speciöc shocks to all countries. We use a sample of 93 countries for which there are data for all variables, which amounts to 183 observations for the years 1990 and 2000 combined. All analyses show results separately for GDP and TFP per capita and also report results separately for overall, skilled and unskilled diversity. The OLS results are reported in Tables 5 to 12. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

38 Empirical analysis: results Table 5 presents initial results for our benchmark model and for the full sample (for GDP/capita and TFP/capita, respectively). We Önd that diversity of skilled migrants relates positively (and signiöcantly at the 5% level) to economic income, both for in terms of GDP and TFP/capita. For GDP, this estimate turns signiöcant once we control for geography and institutions, whereas the relationship for TFP is positive and signiöcant throughout the speciöcations. We interpret this as Örst evidence that variety in terms of skilled immigrants relates positively to economic development in a highly structured framework that accounts for many alternative channels of ináuence, such as pure skill e ects or better institutions. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

39 Empirical analysis: results We establish that the skill dimension of diversity matters: skilled diversity seems to exhibit more robust positive e ects than diversity of unskilled workers. The share of immigrants, the other key determinant of overall diversity, relates positively and signiöcantly to income for unskilled immigrants only (which drives the positive relationship for overall immigrants). If there were production function e ects of diversity, we should Önd that they are stronger in a subset of economies with more advanced production processes, which are closer to the technological frontier. We thus separate our sample into countries above and below the median GDP/TFP per capita in 1970 ñ see Table 6-7. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

40 !!Table!5:!Birthplace!Diversity,!Full!Sample!

41 Empirical analysis: results In Table 6, we report our results for rich countries only, and for poor countries only in Table 7 For rich countries we Önd the same results as in our overall sample, but the magnitudes on skilled diversity are a bit higher and signiöcance is also higher at the 1% level. On the contrary, for the poor countries sample we Önd no signiöcant relationship whatsoever between variety of immigrants and economic development. For robustness, we extend our split sample approach using patent data (Table 8): Subsample countries with above-mean patent intensity (2000): We replicate our rich country results at 1% signiöcance level. Patent intensity as dependent variable (2000): Diversity of skilled immigrants relates positively to patent intensity, while diversity of unskilled immigrants does not (signiöcantly). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

42 !!Table!6:!Birthplace!Diversity,!RICH!countries!

43 !!Table!7:!Birthplace!Diversity,!POOR!countries!

44 !!Table!8:!Birthplace!Diversity!and!Patent!Intensity!

45 Empirical analysis: robustness Robustness to geography controls: two alternative speciöcations suggested by Rodriguez and Rodrik (2001). Table 9 reports the results for share of tropics (in % of land mass area) as alternative control as well as a set of three regional dummies. Our main result on skilled diversity of immigrants is very robust: it holds at 10% in the GDP model and at 1% signiöcance in the TFP model. The magnitudes remain very stable. Interestingly, in both speciöcations, unskilled diversity tends to become positive and more signiöcant. This is particularly true for the speciöcation with regional dummies. We interpret this result as some (limited) evidence for the presence of productive e ects of diverse unskilled immigrants, e.g. via higher specialization of labor or better labor market integration of immigrant groups. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

46 !!Table!9:!Birthplace!Diversity!!geography!robustness!

47 Empirical analysis: robustness Table 10 shows the robustness of our results (for skilled diversity only) with Column 1 as benchmark. Columns (2) and (3) show our results remain valid when weighting observations by the size of immigration (Column 2) or when we drop the 10% of countries with lowest overall immigration (in % of population). Very similar results. Column 4 shows our results when dropping smaller countries (excluding countries below the 10th percentile of the size distribution, i.e. below 3.1 million population). Again very similar. Column 5 limits our sample to the 23 OECD countries in our overall sample. Column 6 shows results when dropping the smallest 25% of observations for skilled diversity to verify that our results are not driven by the skewed distribution of diversity and the corresponding outliers with very low diversity. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

48 Empirical analysis: robustness Table 10: Other robustness checks Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

49 Empirical analysis: robustness to borders pre 1989 We re-compute our diversity index with country borders pre avoids capturing "artiöcial" diversity from immigrants of countries that separated post Results: Our initial results for the full sample and rich subsample, for GDP and TFP remain fully robust at same magnitudes and signiöcance levels. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

50 !!Table!11:!Birthplace!Diversity!!Borders!pre!1989!

51 Empirical analysis: robustness to migration networks Finally, we control for migration networks in 1960 via a diversity index of migrants in 1960 (based on data from Ozden et al., 2011). Full sample: Results for diversity of skilled migrants in 1990 and 2000 lose signiöcance but remain positive (largely due to the high correlation between diversity in 1960 and 1990/2000, at about +.46) - suggests powerful long run e ects of variety of immigrants. Rich country subsample: Our measure of diversity in 1990/2000 remains fully robust and highly signiöcant at the 1% level. This indicates that lagged diversity - while an important long run e ect - does not drive our main results. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

52 Empirical analysis: robustness Interpretation: Correlation between diversity today and 1960 lower for skilled immigrants (likely due to migration network e ects). Descriptive analysis reveals that Div mig and s F have increased between 1960 and 2000 but changes do not correlate with changes in GDP per capita (hinting that reverse causality may not be such a concern) nor do they correlate one with the other (low interaction). Finally, countries that had a diverse workforce in 1960 are the ones having a large and diverse second and third generation of immigrants today: compensation e ects of birthplace and ethnic diversity could explain that diversity in 1960 does not a ect the level of productivity today or that it does not wash out the positive productivity e ects of current birthplace diversity. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

53 !!Table!12:!Birthplace!Diversity!!Migra3on!Networks!1960!

54 Plan 1 Introduction 2 Skill complementarities and diversity 3 An index of birthplace diversity 4 Empirical Analysis 5 IdentiÖcation 6 Conclusion Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

55 IdentiÖcation In this section we discuss and go part of the way towards addressing endogeneity concerns. A Örst concern is that richer countries could attract a more diverse set of immigrants, especially skilled ones. Reverse causality would seem a bigger concern for "size" than for "variety" (.1 v..02 correlations for changes between and income per capita growth). A second concern is the existence of omitted variables explaining both diversity and productivity We address these concerns by controlling for a very large range of factors and by specifying a gravity model to predict diversity using exogenous bilateral variables. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

56 IdentiÖcation: a gravity model We follow the gravity literature and specify a gravity model to generate predicted bilateral migration stocks. These stocks then serve as inputs to calculate predicted immigration sizes s F and a HerÖndahl index of predicted Div mig. However, we have to rely exclusively on bilateral determinants that do not directly a ect incomes, and so have to leave out income di erences or di erences in institutional quality (that could directly a ect bilateral immigration policy). We specify the following gravity model for migration, where y ikst is the emigration rate from origin country i to destination country k for immigrants of skill level s in year t: y ikst = a + b 1 DISTANCE ikt + b 2 BORDER ikt + b 3 LANGUAGE +b 4 COLONY + F i + c k + h t + e Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

57 IdentiÖcation: a gravity model Note that the model is (purposefully) mis-speciöed as we exclude bilateral determinants of migration that would a ect income through other channels. In this context, Head and Mayer (2013) Önd that OLS performs better than a Poisson estimator (more robust and consistent). We therefore use OLS, meaning that we do not include zero-cells, as our main speciöcation. However, the cost of neglecting them is minor given the way our index is computed (requires correctly predicting the main immigration stocks, not whether a given cell will be empty or slightly positive, as this will not a ect the aggregate index). Still, we also use PPML as robustness check. Multilateral resistance e ects are dealt with by including a set of origin-year Öxed e ects to account for any time-varying common origin shock to migration which ináuences migrantsí locations decisions (Bertoli and Fern ndez-huertas, 2013). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

58 IdentiÖcation: Örst-stage results Table 13 shows the results for our gravity model estimated with OLS and PPML. Generally, the models have a very high explanatory power of R2 >.4; all the coe cients on the migration determinants have the expected sign. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

59 !!Table!13:!Gravity!Model!Results!

60 IdentiÖcation: Örst-stage results We turn to assessing the Öt of the predicted values with actual diversity. The Örst panel of Figure 3 shows actual vs. predicted diversity of skilled immigrants based on the gravity model. Overall, the correlation between actual and predicted diversity is very high (+.6 for skilled and unskilled diversity). Our instrument should plausibly be lower (higher) than actual diversity in richer (poorer) countries if prosperity and diversity are positively correlated. The second panel of Figure 3 plots the ìprediction errorî as a function of GDP per capita. The di erence between predicted and actual diversity becomes negative as GDP/capita increases (slope -0.05, signiöcant at 1%), serving as illustrative evidence that our gravity model produces an instrument that takes out some diversity-increasing e ects in richer countries. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

61 !!Figure:!Actual!vs.!Predicted!Birthplace!Diversity!

62 !!Table!14a:!2SLS!results!8!GDP!

63 !!Table!14b:!2SLS!results!8!TFP!

64 IdentiÖcation: 2SLS results Tables 14a,b show our baseline models with s F and/or Div mig instrumented by our gravity-model based measures. We largely conörm our prior OLS Öndings: skilled diversity is now largely signiöcant at the 1% level for GDP/capita and TFP/capita (somewhat higher magnitudes suggesting prior attenuation bias). Interstingly, the coe cient on birthplace diversity is una ected by whether we instrument s F or not, meaning there is no bias on diversity arising from the endogeneity of the size of immigration. We also fully conörm the split-sample results for rich countries at 5-10% statistical signiöcance for both GDP and TFP per capita and Önd evidence for positive e ects of unskilled diversity in richer countries (Table 15). Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

65 !!Table!15a:!2SLS!results!!GDP!/!low!skilled!immigra3on!

66 !!Table!15b:!2SLS!results!!TFP/!low!skilled!immigra3on!

67 IdentiÖcation: 2SLS results We also pursue an alternative IV approach using the diversity of immigrants in 1960 (based on Ozden et al., 2011) as an additional instrument for diversity today. This approach is valid to the extent that diversity in 1960 is not correlated with unobserved factors that also determine diversity and income today (which is of course questionable) but allows us to test the exclusion restriction of the joint instruments directly. Table 16 shows that we replicate our OLS results using su ciently strong instruments (F> 10) at the 5% signiöcance level. This also holds for the split samples. We interpret these results as additional evidence for our prior OLS and IV model Öndings. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 68

68 !!Table!16a:!2SLS!results!!GDP!/!2!instruments!

69 !!Table!16b:!2SLS!results!!TFP/!2!instruments!

70 The channel: Skill complementarities E ect of birthplace diversity on productivity could be driven by origin-e ects or skill-complementarities Origin-e ects appear unlikely, since we control for immigrants level of income/tfp at origin in our full model Skill complementarities appear more likely, since diversity also increases the level of patent citations per capita in a given country Still, we analyze the channel of e ect more directly Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

71 The channel: Skill complementarities (Methodology) Similar to Desmet, Ortun and Wacziarg (2012) who calculate linguistic fractionalization measures at di erent levels of language tree and test for predictive value (significance and magnitude) on civil conflict and economic growth. We extend the Herfindahl-index formula for Div Mig by a distance variable d jk,followinggreenberg(1956): Div Mig,augmented = J i  hs j (1 s j ) d jk j=1 (9) We use genetic distance between immigrants j and natives k (Cavalli-Sforza et al. 1994), linguistic distance (Isphording Otten 2013) and income at origin (PWT 7.0) as inputs for d jk ( ) Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

72 The channel: Skill complementarities (Methodology) We define a standard logistic function to assign a d jk between 0 and 2 (centered on d jk = 1 for the theoretical case that all immigrant groups are equidistant to natives). d jk = e (q x ) (10) Where q is a parameter that ranges from -10 to +10 and x takes on standardized values (min=0, max=1) for genetic distance, linguistic distance and income (GDP at PPP) at origin. Use logistic function largely for convenience, since it includes Div Mig as special case at q=0 or when d jk = 1 (if all groups equidistant from natives at average distance) Varying q so that q > 0 allows us to give more weight to genetically (linguistically) more distant / richer immigrant groups (and vice-cersa) Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

73 The channel: Skill complementarities (Intuition) Intuition: Running our main model (Table 5) using Div Mig,augmented at di erent levels of q allows us to infer on the importance of genetics/linguistics/origin e ects If results for Div Mig,augmented remain significant at same or higher magnitudes as q! genetically distant groups matter for e ect of birthplace diversity on productivty. If significance drops - they don t Same for linguistic distance and GDP p.c. (PPP) at origin Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

74 !!Table!B1:!Gene,c!distance2augmented!diversity! FULL MODEL > Median GDP/capita in 1970 Birthplace Diversity (skilled) at different α-values p-value Birthplace Diversity (skilled) at different α-values p-value α-values Coefficient (std error) α-values Coefficient (std error) H0: Equality of coefficients α-values Coefficient (std error) α-values Coefficient (std error) H0: Equality of coefficients ** ** n/a *** *** n/a (0.0847) (0.0847) (0.106) (0.106) * ** *** *** (0.0806) (0.0862) (0.0933) (0.114) ** ** *** (0.0789) (0.0865) (0.0902) (0.114) ** *** (0.0787) (0.0867) (0.0905) (0.114) ** *** (0.0771) (0.0868) (0.0921) (0.113)

75 !!Table!B2:!Income!at!origin2augmented!diversity!(skilled)! FULL MODEL > Median GDP/capita in 1970 Birthplace Diversity (skilled) at different α-values p-value Birthplace Diversity (skilled) at different α-values p-value α-values Coefficient (std error) α-values Coefficient (std error) H0: Equality of coefficients α-values Coefficient (std error) α-values Coefficient (std error) H0: Equality of coefficients ** ** n/a *** *** n/a (0.0847) (0.0847) (0.106) (0.106) ** * *** *** (0.103) (0.0708) (0.118) (0.0891) ** *** *** (0.121) (0.0649) (0.123) (0.0801) ** *** *** (0.135) (0.0635) (0.128) (0.0783) ** *** *** (0.146) (0.0635) (0.133) (0.0786)

76 !!Table!B3:!Income!at!origin2augmented!diversity!(unskilled)! FULL MODEL > Median GDP/capita in 1970 Birthplace Diversity (unskilled) at different α-values p-value Birthplace Diversity (unskilled) at different α-values p-value α-values Coefficient (std error) α-values Coefficient (std error) H0: Equality of coefficients α-values Coefficient (std error) α-values Coefficient (std error) H0: Equality of coefficients n/a n/a (0.0531) (0.0531) (0.0629) (0.0629) * * (0.0673) (0.0467) (0.0752) (0.0596) * ** (0.0846) (0.0465) (0.0919) (0.0614) * ** (0.101) (0.0482) (0.105) (0.0641) * ** (0.114) (0.0497) (0.111) (0.0660)

77 The channel: Results Genetic distance: Results become insignificant as as q! +10, but remain stable as as q! -10. Hence, culturally closer immigrant groups matter (Spolaore Wacziarg 2009) Linguistic distance (correlated with genetic distance +0.16): Results remain significant as q! +10 or as q! +10, especially in rich subsample - linguistic distances do not matter Income at origin: Results remain significant as q! +10 or as q! +10 and statistically - origin-e ects not a driver of our rich country subsample results. Full sample results may show some (not yet statistically significant) origin-e ects for poor countries. Also, richer unskilled immigrants matter for e ect of unskilled diversity in rich countries (likely cause: heterogeneity of skills within category of unskilled immigrants) Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

78 Plan 1 Introduction 2 Skill complementarities and diversity 3 An index of birthplace diversity 4 Empirical Analysis 5 Identification 6 Conclusion Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

79 Conclusion We propose a new concept of "birthplace diversity" that captures the diversity in birthplaces in a country s work force. The index is decomposed into a size (share of foreign- born) and a variety (diversity among immigrants) component, available for 195 countries in 1990 and 2000, and for overall/skilled/unskilled components of the labor force. We show that birthplace diversity is, maybe surprisingly, largely uncorrelated with ethnic and linguistic fractionalization and that unlike fractionalization it appears to be positively related to a country s level of economic development. We then empirically investigate the relationship between birthplace diversity and measures of productivity (GDP or TFP per capita), controlling for a large range of factors such as the share of immigration, origin-e ects, education, institutions, trade openness and trade diversity as well as geography. We employ OLS and specify a gravity-based 2SLS model to address endogeneity. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

80 Conclusion We find a positive and robust correlation between the diversity of a country s population in terms of origin countries and productivity (in GDP and TFP per capita). This association is particularly strong for the diversity of immigrants, especially for skilled immigrants in richer countries. Expanding the variety of origin countries in a country s immigrant population by one standard deviation (e.g., from Iran to Ireland, or Ireland to USA) increases long run incomes by a factor of 1.2 to 1.5; or a 10 percentage points change increases long-run income by 6.7 percent. These results hold for OLS and 2SLS estimators in a dataset of 93 countries and are robust to a wide range of alternative explanations. We interpret these findings as suggestive of production function e ects of diversity arising from complementarities in skills, values or problem solving capabilities that emerge from the combination of workers with diverse origins in a joint production task, as suggested by micro and US-based studies. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

81 Conclusion Immigration policy debates have so far largely centered on immigration quotas or on targeting specific labor market skills. But few policies explicitly focus on the diversity of immigrants the US diversity lottery being a notable exception. Focusing exclusively on skill levels, but not on diversity of backgrounds, would therefore imply missing an important channel through which immigration contributes to growth in the receiving countries. Is the melting pot melting too much? Finally, a potential corollary of more diverse immigration is greater public support for more immigration conditional on the additional migrants being skilled and coming from a diverse set of origins. This creates the possibility of a virtuous circle between the quantity and quality of immigrants a more diverse world could also become a more open one. Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75

82 Next steps Research agenda explore the role of birthplace diversity in the formation of attitudes to immigration and in the formation of social values (redistribution, trust) explore the role of birthplace diversity (population, migrants, researchers) on innovation (patents) Alesina, Harnoss and Rapoport () Birthplace Diversity June / 75