Immigration and Economic Growth: Do Origin and Destination Matter? Youngho Kang and Byung-Yeon Kim October 2012
Introduction • Theoretical predictions in early studies - Neo-classical growth model: immigrant inflow leads to a decrease in the long-run economic growth per capita, i. e. capital dilution effect. - Dolado, Goria, and Ichino (1994): immigration can increase the stock of human capital in the host country, and so facilitate the growth. - Lundborg and Segerstrom (2002): Since immigrants lower wages, firms spend more R&D, which promotes the economic growth. 2/31
Introduction • Empirical evidence in early studies - Felbermayr, Hiller, and Sala (2008): Using a cross-section of countries (n=160, t=2000), they find a non-negative effect of immigration on real per capita income. - Ortega and Peri (2009): the positive impacts of the inflow of immigrants on employment and investment. (U. S. ) - Dolado Goria, and Ichino (1994): the negative impacts of migration on growth in OECD countries decrease by more than half due to the high human capital of immigrants. - Orefice (2011): the overall effect of migration on the per capita GDP is negative despite some positive effects of highly educated migrants on per capita GDP. 3/31
Introduction • These conflicting findings suggest the need to employ a more focused approach. - All countries vs. OECD - A certain kind of immigration vs. general immigration • Immigrants are different from newborn babies. - We have to focus what they bring as well as how many they enter. - A different type of human capital : schooling year vs. knowledge on technology and institutions embodied in immigrants 4/31
Introduction • We assess the heterogenous effects of immigration on growth in the host country, depending on the characteristics of both the origin and host countries. • If any, the effect might be more clearly pronounced in immigration from developed to developing countries than from developing to developed ones. 5/31
Introduction • This study is related to the strand of literature that deals with the role of intangible assets in growth. - Spolaore and Wacziag (2010): the culture differences between individual and frontier countries may act as barrier to technology adoption. - Andersen and Dalgaard (2011): the temporary in and outflows of travelers is one of the channels that gives exposure to the country abroad. 6/31
Introduction • This paper aims to contribute to the literature in three ways. - We investigate the role of knowledge on better technology and institutions carried by immigrants from richer countries in facilitating growth in the host country. - We make our contribution by showing that the relationship between immigration and growth can be negative, neutral, and positive depending on the country of origin of immigrants and where they settle eventually. - Estimating the dynamic equation of immigration, which is more appealing because of the path dependency of immigration. 7/31
Theoretical Framework • We modify the Dolado, Goria, and Ichino (1994) by specifying a technology term in production function. - Labor is augmented by technology and institutions (A) with the growth rate g. where Y is aggregate output, A is technology, K is physical capital, H is human capital, and L is labor. - Labor force growth is given by where M is the immigrant stock, n is the growth rate of the labor force, m is the ratio of immigrant stock to the entire population, sh is the fraction of resources devoted to human capital accumulation, is the depreciation, is the relative average 8/31 human capital of immigrants compared with the average human capital per worker.
Theoretical Framework - The dynamics of physical capital are the same as those in the Solow model. - Using units of labor, the production function is given by where , , and . - The evolution of the economy is determined by 9/31
Theoretical Framework - The economy converges to a steady state defined by - We obtain the steady-state real GDP per effective worker. 10/31
Theoretical Framework • We use a term (A) in Andersen and Dalgaard (2011) - to understand a new role of immigrants who carry better knowledge about technology and market-supporting institutions - The evolution of technology (A) is characterized by - The parameter reflects the intensity of knowledge spillover from a world technology frontier (Aw). where and - Along a steady state, A is calculated (Assumption: ) - The rate of (local) technological progress is equal to the frontier one. 11/31
Theoretical Framework - To approximate around the steady state, the pace of convergence is obtained using the following equation. - The above equation implies that - We can obtain the equation for growth regression as follows 12/31
Theoretical Framework - To identify how immigrants affect economic growth, we differentiate the above equation with respect to m. If , immigration affects economic growth positively. it is not possible to identify and separately. - The three kinds of immigration flow should be distinguished. case 1. immigrant from all countries to all countries case 2. immigrant from developed countries to all countries case 3. immigrant from developed countries to developing ones. case 3 > case 2 > case 1 because of , , , and 13/31
Estimation Strategy: Model Specification • Based on theoretical prediction, we construct the following equation in panel data setting. where ln(y(i, t)) and ln(y(i, t-1)) are the current and lagged logarithms of real GDP per capita in country (i) at time (t), respectively; imm_index is the ratio of # of immigrants to # of total population ; Z contains other classic growth variables such as the average of investment/GDP ratio, the lagged value of second enrollment rate and the average of population growth rate. - We can check which type of immigration affects economic growth by combining the different indices of immigration with the varying sample of the host nations. Model 1: the effect of imm_index on economic growth of all ones. Model 2: the effect of imm. MIC_ratio on economic growth of all ones. Model 3: the effect of imm. MIC_ratio on economic growth of 14/31 developing ones.
Estimation Strategy: Data • We use the Global Bilateral Migration Database. - It comes from world bank spanning from 1960 to 2000, five subperiods of then years each. - These data are more reliable because of being collected by destination countries. • Penn World Table 7. 1. and the World Development Indicator are used. - Penn World Table 7. 1. : Real GDP per capita, the growth rate of population, the share of investment in real GDP (proxy for the saving rate), and the ratio of the sum of exports and imports to GDP. - WDI: the secondary enrollment rate (proxy for the rate of investment in human capital) 15/31
Estimation Strategy: Data List of 90 countries included in the sample The MICs (18) Australia, Austria, Belgium, Canada, Germany, Denmark, Finland, France, United Kingdom, Ireland, Italy, Netherlands, Norway, New Zealand, Sweden, United States, Switzerland, Luxembourg The other countries (72) United Arab Emirates, Argentina, Bahamas, Bolivia, Brazil, Chile, China, Congo, Colombia, Costa Rica, Cyprus, Dominican Republic, Algeria, Ecuador, Egypt, Guatemala, Honduras, Haiti, Indonesia, India, Iran, Israel, Jamaica, Jordan, Kenya, Korea, Kuwait, Lebanon, Sri Lanka, Morocco, Mexico, Malta, Malaysia, Nigeria, Nicaragua, Pakistan, Panama, Peru, Philippines, Paraguay, Saudi Arabia, Senegal, Singapore, El Salvador, Syrian, Thailand, Trinidad and Tobago, Tunisia, Turkey, Taiwan, Uganda, Uruguay, Venezuela, South Africa, Zambia, Zimbabwe, Spain, Greece, Ireland, Portugal, Japan, Hong Kong, Czech Republic, Note: MICs is defined as the major industrialized countries as joining OECD before Estonia, Croatia, Russia, Slovakia, Slovenia, 16/31 Poland, Hungary, Bulgaria, Romania 1970 and being ranked above 18 th place with respect to per capita real GDP, 1960
Panel(b): The rankings of top four origin countries with their associated host countries 1960 Italy ARG BRA VEN TUN France MAR DZA TUN ESP Belgium ESP ZAF BRA HUN U. K. ZAF HUN IRL ZWE 1970 Italy ARG BRA VEN URY U. K. ZAF IRL ZWE ZMB France ESP DZA MAR ARG U. S. MEX JPN GRC PHL 1980 USA MEX NGA GRC JPN U. K. ZAF IRL ZWE NGA Italy ARG BRA VEN URY France ESP DZA NGA MAR 1990 USA MEX JPN GRC ISR U. K. IRL ZAF ESP ZWE France ESP PRT DZA MAR Germany TUR ESP BRA ZAF 2000 USA MEX ISR JPN PHL U. K. IRL ZAF ESP ARE France ESP PRT ISR DZA Germany TUR ESP ISR PRT Notes: ARE=United Arab Emirates, ARG=Argentina, BRA=Brazil, DZA=Algeria, ESP=Spain, HUN=Hungary, GRC=Greece, IRL=Ireland, ISR=Israel, JPN=Japan, MAR=Morocco, MEX=Mexico, NGA=Nigeria, PHL=Philippines, PRT=Portugal, TUN=Tunisia, TUR=Turkey, URY=Uruguay, VEN=Venezuela, ZAF=South Africa, ZMB=Zambia, ZWE=Zimbabwe. 18/31
Econometric Issues • The endogeneity problem may arise. - Immigrants choose a country to live in mainly on the basis of differences in income per capita between the origin and destination. - Two findings in early studies suggest how to construct IV. (i) geographic variables are important in estimating bilateral migration flows (Mayda 2008; Berthelemy Beuran, Maurel 2009; Peri and Ortega 2009). External IV approach: the gravity equation (ii) migration is positively correlated with past settlement of immigrants (Beine, Docquier, and Ozden 2009; Colliner and Hoeffler, 2011). Internal IV approach: GMM estimator 21/31
Econometric Issues • External Instrumental Variable - Felbermayr, Hiller and Salo (2008), Mayda (2008), and Ortega and Peri (2009) use the IV based on the gravity equation of immigration. - We employ a Pseudo-Poisson ML estimator because of the zero values of the dependent variable (Santos, Silva, and Tenreyro, 2006). where ln(immi, j, t) is the logarithm of immigrant stock from all sample countries i to country j at time t, ln(imm. MIC i, j, t) is the logarithm of immigrant stock from countries i (the MICs), ln(pop i, t) is the logarithm of (origin) country i population, ln(pop j, t) is that of (destination) country j population, log(dist i, j ) is the logarithm of distance, com i, j is the dummy variable for common language, col i, j is the colonial history, and cont i, is the geographical contiguity. 22/31
Econometric Issues • Internal Instrumental Variable - Beine, Docquier and Ozden (2009): the existing diasporas in 30 OECD countries are the most important determinant of migration flows. - The system GMM estimator uses the lagged immigrant stock as the instrument variable. - This can address the endogeneity of lagged dependent variable in dynamic model. - Two criteria: the test for serial correlation and the Sargan test for overidentifying restrictions. 23/31
Estimation Results • Fixed-effects regressions:
- The magnitude of imm_ratio in Model 1 is smaller than that of imm. MIC_ratio in Model 2 and Model 3. • 2 SLS:
- The estimates from the 2 SLS estimator is smaller than ones in Table 3. • GMM:
- A 1% point increase in the ratio of immigrant stock raises the growth rate of per capita real GDP 8%. - Table 5 lies between the fixed-effect estimator and the 2 SLS estimator. • Excluding transition economies from our sample:
the effect of immigrants on per capita real GDP (fixed-effects) Model 1 Time dummy R-square Observations Number of id All 0. 684*** (0. 065) -0. 019 (0. 025) 0. 264*** (0. 065) 0. 0004 (0. 0006) 0. 002** (0. 0008) 0. 006 (0. 007) Sample of host countries Model 2 0. 679*** (0. 066) -0. 018 (0. 023) 0. 267*** (0. 062) 0. 0004 (0. 0006) 0. 002** (0. 0008) Model 3 Restricted to non- MICs 0. 676*** (0. 069) -0. 022 (0. 027) 0. 258*** (0. 070) 0. 0003 (0. 0006) 0. 002 (0. 0013) 0. 026* (0. 014) Yes 0. 784 300 90 0. 087** (0. 037) Yes 0. 753 234 72 Yes 0. 783 300 90 Notes: a: Clustered-adjusted standard errors are reported in brackets. Significant variables at the 10%, 5%, and 25/31 1% significance levels are marked with *, ** and ***, respectively.
Results from the 2 SLS: gravity IV Panel (a): Model 1 All 2 SLS 0. 688*** (0. 065) -0. 015 (0. 024) 0. 270*** (0. 062) 0. 0005 (0. 0006) 0. 002** (0. 0008) -0. 0000001 (0. 0000001) Sample of host countries Model 2 2 SLS 0. 688*** (0. 065) -0. 018 (0. 023) 0. 267*** (0. 063) 0. 0004 (0. 0006) 0. 002** (0. 0008) Model 3 Restricted to non- MICs 2 SLS 0. 705*** (0. 068) -0. 025 (0. 027) 0. 270*** (0. 067) 0. 0003 (0. 0006) 0. 0016 (0. 0013) 0. 0004 0. 053*** (0. 0006) (0. 013) Time dummy Yes Yes R-square 0. 783 0. 753 Observations 300 234 Number of id 90 90 72 a: Clustered-adjusted standard errors are reported. Significant variables at the 10%, 5%, and 1% Notes: 26/31 significance levels are marked with *, ** and ***, respectively.
Results from the GMM estimator Model 1 Time dummy AR(1)/AR(2) Hansen test Observations Number of id All 0. 889*** (0. 080) -0. 087* (0. 052) 0. 401*** (0. 130) -0. 0002 (0. 001) 0. 004* (0. 002) 0. 001 (0. 004) Sample of host countries Model 2 0. 844*** (0. 063) -0. 102*** (0. 035) 0. 516*** (0. 113) 0. 0005 (0. 0007) 0. 0047*** (0. 0015) Model 3 Restricted to non- MICs 0. 870*** (0. 073) -0. 039 (0. 054) 0. 460*** (0. 151) -0. 0002 (0. 0009) 0. 0041* (0. 0023) 0. 003 (0. 013) Yes 0. 001/0. 312 0. 791 300 90 0. 079* (0. 047) Yes 0. 002/0. 362 0. 710 234 72 Yes 0. 000/0. 199 0. 281 300 90 Notes: : a: Robust standard errors are reported in brackets. Significant variables at the 10%, 5%, and 1% significance levels are marked with *, ** and ***, respectively. b: This system-GMM uses lags up to t-4 as instrument to avoid overfitting biases. 28/31 Both imm_ratioi, t and imm. MIC_ratioi, t are treated as endogenous variables.
Results in the sample excluding transition countries Sample of host countries Model 3 Restricted to non- MICs Within 0. 676*** (0. 070) -0. 020 (0. 027) 0. 256*** (0. 071) 0. 0004 (0. 0007) 0. 0019 (0. 0013) 0. 082** (0. 037) Yes 0. 759 Model 3 Restricted to non- MICs 2 SLS 0. 703*** (0. 069) -0. 023 (0. 027) 0. 269*** (0. 068) 0. 00045 (0. 0068) 0. 0018 (0. 0013) 0. 051*** (0. 013) Yes 0. 759 Model 3 Restricted to non- MICs SYS GMM 0. 874*** (0. 076) -0. 025 (0. 062) 0. 473*** (0. 129) -0. 0005 (0. 00096) 0. 004 (0. 0026) 0. 130** (0. 059) Yes Time dummy R-square AR(1)/AR(2) 0. 001/0. 344 Hansen test 0. 766 Observations 218 218 Number of id 62 62 62 Notes: Significant variables at the 10%, 5%, and 1% significance levels are marked with *, ** and ***, 29/31 respectively.
Conclusion • This paper finds that the origin and the destination countries in immigration matter. - Immigrants from the MICs positively affect the economic growth, whereas general immigration neither promotes nor deters growth. - The effect of immigration from the MICs on growth in the host countries is significantly larger in estimations that involve only less developed countries. - The effect of immigration on economic growth in the host countries depends on the gap in the quality of technology and institutions. - Immigrants from the MICs to less developed countries are more likely to be carriers of better knowledge and institutions. 30/31
Conclusion • Our findings shed some light on the strategy for economic growth in less-developing countries. - The literature on development emphasized the importance of FDI, trade, business travel for growth (Caves, 1974; Coe and Helpman, 1999; Hovhannisyan and Keller, 2012). - Our findings suggest the attraction of foreigners from advanced countries. • Future research topics - why do people migrate from rich to poor countries? - what do policies encourage such movement? 31/31