Julian Debenedetti

Empirical research report Effect of Immigration on GDP per Capita A cross-country analysis of the effect of immigration on GDP per capita Author: Julian Debenedetti and Fadekemi Lawal Examiner: Reza Mortazavi Course: NA3011 Higher education credits: 1.5 hp Högskolan Dalarna 791 88 Falun Sweden Tel- Table of contents 1. Introduction ..................................................................................................... 1 2. Literature review ............................................................................................ 2 3. Empirical analysis ........................................................................................... 4 3.1 Data Sources ..................................................................................................................... 4 3.2 Model Specification ......................................................................................................... 4 3.3 Results .............................................................................................................................. 6 3.4 Limitations of the research project ................................................................................... 8 4. Conclusion........................................................................................................ 8 References .......................................................................................................... 10 Appendix 1: White’s test for Heteroscedasticity............................................ 12 Appendix 2: Ramsey RESET Test .................................................................. 13 Appendix 3: Correlation Matrix ..................................................................... 14 Appendix 4: Regression Results ...................................................................... 15 Appendix 5: Multicolinearity Results ............................................................. 16 1. Introduction GDP per capita is one of the most important indicators of a country’s standard of living. As expats living abroad, our interest and the purpose of this research task is to analyse the effects of immigration on the GDP per capita in the receiving economies. In very recent years Europe has received the largest amount of immigrants since the Second World War. It is therefore an increasingly important and pressing issue for all countries, especially those of Western European who has seen the greatest influx of immigrants. The consequences of responding in a slow or inaccurate way could be devastating for the economy of the countries receiving immigrants, hence the countless studies being done on this particular topic to avoid unnecessary complications. There are many studies on this issue and a vast majority have concluded that the level of income per capita for host economies increases along with the arrival of immigrants (Alesina, Harnoss, and Rapoport, 2016; Ortega and Peri, 2014). Furthermore, other studies suggest that the GDP per capita will increase regardless of the education level or productivity of the immigrants, claiming that migration restrictions are in any case inefficient (Michael Clemens and Lant Pritchett, 2016). Thus the purpose of this note is to estimate the GDP per capita, with cross- sectional data of 138 countries from the World Bank Open Data of 2015. We have developed an econometric model using Ordinary Least Squares method (OLS) that contributes to the on-going literature on the impact of immigration of certain economies. Controlling key variables such as the literacy level, the trade openness, labour force and employers we found a positive relationship between immigration and GDP per capita. The goal of this paper is of course to identify tendencies that serve as a starting point for future research. We realize that since we are not considering panel data, there is undoubtedly a bias on our estimation due to the fact that we are not including the value of the variables of the former periods. Also, due to the lack of data, we are allowing endogeneity issues on more than one occasion, such as in immigration to per capita income, in further investigations this issue can be solved by introducing an instrumental variable. The remainder of the note is organized as follows; a literature review in the first section to establish a framework and also to analyse concepts and relations between the main variables. 1 The second section is comprised of the empirical analysis, the detailed methodology and the results of the estimations. The last section consists of conclusions and the final discussions based on the results found. 2. Literature review Immigration may be defined as the movement of people into countries other than the country of their nationality, either temporarily or permanently. In other words, it is the international migration of people from their countries of origins, to new countries in which they do not have citizenship. There are several factors which contributes to the decision of people to become immigrants. According to Borjas (2014), a person decides to become an immigrant when the gains from immigration outweigh the costs of immigration. Typical immigration cost includes the cost of transportation and relocation, while immigration benefits include education, employment opportunities, safety, better standards of living and economic opportunities among others. Hence, it is not difficult to understand why a large number of immigrants are drawn to high income countries. Host countries have to develop stringent selection criterion so that the ideal immigrant is given the opportunity of settlement. For instance, Sweden requires a minimum salary of Kr 13,000 before tax for the provision of a work permit to a non EU- citizen. The goal is to have a high level of assimilation of immigrants with the society. Standard economic theories suggest that the major mechanisms through which immigration affects the economy of host countries is the labour market. Immigration increases the supply of labour in a host country. This could have several plausible impacts. The increase in labour supply could cause competition with natives and drive down the average wages of natives in the short run. However, in the long run where capital can be varied, the reduced spending on labour due to the lower equilibrium price for labour, could increase the capital available to employers and stimulate investments. This increase in capital stock could encourage an increase in demand for labour. Hence the short term negative impact of immigration may be diluted or even cancelled out in the long run (Edo et al, 2018). Longhi et al (2005) found a negative but low impact of an increase in immigration on native born employment. The impact differed based on gender as the effect of immigration on female employment was a little larger than for male employment. 2 According to Edo et al (2018), immigration could lead to a change in the skill composition of the workforce in the host country. The impact of this change depends on the nature of the relationship between the two goods (immigrant labour and native labour). If the two goods are viewed as substitutes, this implies that the skill of native labour is similar to and can be replaced by immigrant labour. In such scenario, immigration will negatively affect the wages of native workers as the competition and excess availability of labour in the market would drive down equilibrium prices. On the other hand, if the two goods are complementary, then immigration will increase labour productivity of the native workers and have a positive effect on wage levels. Similarly, if an immigrant has the same level of education with a native but they differ in languages or other technical skill, this could lead to the employment of natives in more competitive paying jobs while the immigrants are pushed to more manual jobs. This is because immigration may cause natives to specialise in more communication intensive jobs. The resulting effect is that immigrants compete with natives of lower level of skill or education, thereby driving down the average wages of low skilled workers. Peri and Sparber (2009) demonstrated how the inflow of less educated migrants could cause occupational reallocation of native workers with similar education to tasks or jobs which are more communication intensive. Where the stock of immigrant labour is highly skilled however, this could increase labour productivity of the host country, as well as lead to increases in innovations that could positively affect the economy as whole. For this reason, host countries aim to attract highly skilled immigrants to their countries. (Edo et al, 2018). Hunt and Gauthier-Loiselle (2010), studied the impact of immigration on innovation in America. Using panel data of states in the USA, over a period of 1940 - 2000 and patent per capital as a measure for innovation, they found a positive relationship between skilled immigration and innovation. To mention a few other similar studies which have found positive relation between immigration and host countries. Ortega and Peri (2009), investigated the cause and effect of international migration on the economies of OECD countries. They employ the use of a pseudo gravity model on panel data over 14 OECD countries and 70 sending countries. They found that immigration increases employment without crowding out native labour. They concluded that immigration has a positive impact on the GDP of the host countries in the short run and it does not have negative impact on average income. Aleksynska et al (2014) investigated the relationship between immigration and the host countries income and 3 productivity levels. They employed the use of panel data from 20 OECD countries spanning a period of 1960 to 2005. They found that that immigrants of prime age have a positive impact on income per capita and productivity levels of host countries. Feridun (2005) investigated the causal relationship between immigration and GDP per capita, using a case study of Norway. He found that increase in immigration has a positive influence on the GDP per capita. Boubtane et al (2013) also found a positive relationship between immigration levels and GDP per capita. He employed the use of a vector autoregressive model on panel data on 22 OECD countries spanning over the period of-. He also found a negative between immigration and unemployment levels. From our review of literatures, we find that little studies have been done using contemporary data. Our research therefore adds to the literature available on the impact of immigration on the economies of the host countries in recent times. 3. Empirical analysis 3.1 Data Sources The source of all our data excluding literacy rate was obtained from the world development indicators dataset, which was compiled by the World Bank. The data for literacy rate was obtained from the Human development report compiled by the United Nations Development programme. To analyse the relationship between immigration stock and GDP per capita, we employed the use of cross sectional data spanning 138 countries for the year 2015. We eliminated countries based on the lack of availability of data for the year in focus. We also searched for outliers and eliminated the top two countries whose GDP per capita differed markedly from the rest of the sample in other avoid strong influences on the results. 3.2 Model Specification We employed the use of ordinary least squares regression to analyse the relationship between immigration and GDP per capita in model stated below: Iny = β 0 + β1 x1 + + β 2 x2 + β 3 x3 + β 4 x4 + β 5 x5 + β 6 x6 + ε (1) Where: Y represents the GDP per Capita adjusted for purchasing power parity and expressed in USD 4 X1 represents the immigrant stock as a percentage of the total population. Immigrant stock refers the number of foreign or foreign born residents in a host country. X2 represents the literacy rate as measured by the percentage of the population with at least some secondary school education. X3 represents trade openness. Trade openness refers to the sum of imports and exports of goods and service as a percentage of the GDP X4 represents Gross savings as a percentage of GDP (%) X5 represents the labour force participation rate. “Labor force participation rate is the proportion of the population ages 15 and older that is economically active: all people who supply labour for the production of goods and services during a specified period”. (World development indicators, 2018) X6 represents the employers as a percentage to employment. “Employers are those workers who, working on their own account or with one or a few partners, hold the type of jobs defined as a "self-employment jobs" i.e. jobs where the remuneration is directly dependent upon the profits derived from the goods and services produced), and, in this capacity, have engaged, on a continuous basis, one or more persons to work for them as employee(s)”. (World development indicators, 2018) We decided to transform the dependent variable to its logarithmic form so as to reduce the risk of heteroscedasticity. The exogenous variables were all measured in percentages, and further logarithmic transformations would have reduced the sizes of the variables. We decided to estimate a log-linear model, while testing for its functional form and heteroscedasticity with a Ramsey test and white test respectively. The main exogenous variable of interest is the immigrant stock, with the dependent variable being GDP per capita. We included literacy rate to control for the differing levels of education, as we expect that high levels of literacy is positively related to the GDP per capita. We also included the variable trade openness as a measure of the country’s level of financial integration. We believe that a country that is open to international trade may be more liberal towards immigration. Similarly, gross savings, labour force participation and ratio of employers to total employment are expected to be positively related to GDP per capita. 5 3.3 Results Table 1. Descriptive statistics of the variables. Variable N Mean Ln GDP per- Capita Immigrant stock- Literacy rate- Gross Savings- Labor force- participation rate Trade openness- Employers ratio- Std. Dev- Min- Max- - - - - - - - The table above shows a descriptive analysis of our data. We see that the mean immigrant stock is around 8% of the population. This is not surprising considering high levels of immigrant stock will be concentrated in a few developed countries given the attraction of immigrants to countries with good institutions and great opportunities. We noted that the country with the highest share of immigrant population in our sample was Kuwait. This high level of immigrants can be attributed to the large numbers of expatriates employed in the country since the development of their dominant oil industry. Literacy rates of our sample is slightly above average and the share of employers in the workforce is low as one would expect. We estimated the sample using ordinary least squares method of regression. We also performed white test for heteroscedasticity as this is of particular concern given the cross – sectional nature of our data. Although we could not reject the null hypothesis of homoscedasticity, we decided to err on the side of caution and estimate our regression using robust standard errors. We also performed a RESET test for the functional form and our results suggested we cannot reject the current specification (See appendix 2). We ran the regression using OLS and tested for heteroscedasticity given the nature of our sample. Our estimated model is therefore as follows: Inyˆ = 6.425 + 0.033x1 + 0.023x2 − 0.001x3 + 0.025x4 + 0.005x5 + 0.077 x6 + e Where e represents the estimated residuals of the model 6 (2) Table 2. Ordinary least squares regression results Variable Immigrant stock Literacy rate Gross Savings Labor force participation rate Trade openness Employers ratio Constant Observations = 138 Coefficient - - - - Std. Dev. - - - - t-value- -.000015 - - - - R2 = 0.7071 - p-value- Prob > F = 0.000 From the results above, we can see that all the variables have significant impacts on the dependent variable, except trade openness and labour force participation. The trade openness variable was calculated as the sum of import and export, as a percentage of the country’s GDP. While we expect a high volume of exports to be positively related to the country’s GDP per capita, we don’t expect high volumes of import to be positively related to the GDP per capita. Our sample may have contain a mix of export intensive countries and import intensive countries, thereby counteracting to produce an overall insignificant impact. While labour force participation may be insignificant because having a large labour supply may not be indicative of the productivity of labour and labour productivity is what actually contributes to a high GDP. Interpreting our results, we see that the stock of immigrants has a positive relationship with the GDP per capita. To be specific, a 1 percentage point (pp) increase in the share of immigration in a population, would on the average increase the GDP per capita by 3.4%, ceteris paribus. We included all other variables in the model as control variables because we believe they are factors which contributes to having a higher GDP per capita. For example we believe that a country with a high literacy rate will likely have a high GDP because literacy provides a lot of positive externalities in a society. Similarly, a country with a high percentage of gross savings would likely have high stock of capital that can be invested in productive activities, thereby raising the gross domestic product of a country. Hence, we say that the magnitudes of the sign of the coefficients in our model are aligned with our economic expectations. 7 The model explains 70% of the variation of our dependent variable, GDP per capita, about its mean. This suggests that the model has a high explanatory power, given the cross-sectional nature of our sample. We also checked for multicolinearity by obtaining the variance inflation factor (VIF) of the variables in our model. All the variables had VIFs of less than 5, we therefore concluded that the level of multicolinearity in our model was acceptable. (See results in appendix 5) 3.4 Limitations of the research project A major limitation of our research project was lack of availability of data. We would have preferred to estimate the relationship between immigration and GDP per capita, while controlling for the level of skill or education of the immigrant stock and country specific effects. This is because we expect that the positive benefits of immigration are largely related to highly skilled immigrant. Highly skilled immigrants can raise labour productivity and increase innovations thereby raising the gross domestic product and consequently the GDP per capita. We also recognise that countries have various country specific effects that have not been adequately captured by our model. This could be improved by employing a fixed effect model on panel data. Another limitation of our research is the risk of endogeneity in our model due to the possibility of reverse causality. This because it is possible that having a high GDP per capita contributes positively to some of our independence variable such as literacy rate and gross savings. Due to the time constraint, we shall not be able to explore other possibilities such as the use of instrumental variables and panel data. 4. Conclusion. The results of our investigation are aligned with previous research in the field, with the evidence proving (as expected) that immigration has a significant positive impact on the GDP per capita of a country. The aim is to contribute to the large amount of literature already existing on the topic of how GDP (per capita) is impacted by immigration through the use of cross sectional data of 138 countries for our estimations. Our findings show that both Trade Openness and Labour Force Participation are not statistically significant, however, we have decided to include them in the regression nonetheless because they are variables that we consider important when modelling GDP per capita, and wanted to remain true to what we believe is the adequate regression. This note has 8 reviewed pertinent related literature and highlighted the most important conclusions of different authors, so that our new model and our insight had a solid background to be built upon. A limitation of the analysis is it being restricted to data from 2015 rather than of using panel data. It would have given more accurate results with the insight being more reliable, that is why further analysis should include this to catch, for instance, the seasonal effects of variables and to expand our conclusions. One may therefore be sceptical in relying on the results. Also it is important to note that we did not choose the 138 participating countries, it was all the information that we managed to obtain. It is fair to assume that more advanced countries possess better quality information and therefore our sample excludes a greater proportion of countries with lower GDP per capita, thus there is a bias on the sample. It would also be interesting to analyse countries first by region, then by schooling and the age of immigrants, in order to make the analysis a little more specific and not so high level. It would allow the performance of benchmarking, highlighting the main differences and therefore identifying possible courses of action in order to promote the desirable effects and minimizing the undesired. 9 References Angrist J. and Kugler A., 2003. Protective or counter-protective? Labor market institutions and the effect of immigration on EU natives. Economic Journal, 113, 302-331. Aleksynska, M. and Tritah A., 2015. The Heterogeneity of Immigrants, Host Countries’ Income and Productivity: A Channel Accounting Approach. Economic Inquiry 53(1): 150-72. Borjas, G. J., 2014. Immigration Economics. Harvard University Press. Boubtane, E., D. Coulibaly, and C. Rault. (2013). “Immigration, Unemployment, and GDP in the Host Country: Bootstrap Panel Granger Causality Analysis in OECD Countries." Economic Modeling 33:261-269. Edo A., Ragot L., Rapoport. H, Sardoschau S. and Steinmayr A., 2018. The Effects of Immigration in Developed Countries: Insights from Recent Economic Research. CEPII Policy Brief 2018- 22 , 2018 , CEPII. Felbermayr, G. J., S. Hiller, and D. Sala. 2010. Does immigration boost per capita income? Economics Letters 107, no. 2:177–179 Feridun, M. (2005). “Investigating the Economic Impact of Immigration on the Host Country: The Case of Norway." Prague Economic Papers 4:350-359. Hunt, J. and M. Gauthier-Loiselle. 2010. “How Much Does Immigration Boost Innovation?” American Economic Journal: Macroeconomics 2 (2): 31–56. Jaumotte, Ms Florence, Ksenia Koloskova, and Ms Sweta Chaman Saxena (2016). Impact of Migration on Income Levels in Advanced Economies. International Monetary Fund Kirk, A. (2019). Mapped: Which country has the most immigrants?. [online] Telegraph.co.uk. Available at: https://www.telegraph.co.uk/news/worldnews/middleeast/-/Mapped-Which-countryhas-the-most-immigrants.html [Accessed 20 Jan. 2019]. Longhi, S., Nijkamp, P. and Poot, J. (2005). A Meta-Analytic Assessment of the Effect of Immigration on Wages. Journal of Economic Surveys 19(3): 451-477 Ortega, F., and G. Peri. 2009. “The Causes and Effects of International Migrations: Evidence from OECD countries-.” NBER Working Paper 14833, National Bureau of Economic Research, Cambridge, MA. Peri, G. and Sparber C., 2009. Task Specialisation, Immigration and Wages. American Economic Journal: Applied Economics 1 (3): 135–69. United Nations Development Programme. Human Development Reports. Available online: http://hdr.undp.org/en/data [Accessed 20 Jan. 2019]. 10 World Bank. World Development Indicators. Available online: http://data.worldbank.org/data-catalog/ world-development-indicators [Accessed 20 Jan. 2019]. 11 Appendix 1: White’s test for Heteroscedasticity . estat imtest, white White's test for Ho: homoskedasticity against Ha: unrestricted heteroskedasticity chi2(27) Prob > chi2 = = - Cameron & Trivedi's decomposition of IM-test Source chi2 df Heteroskedasticity Skewness Kurtosis - 27 6 1 - Total 34.36 34 0.4504 12 p Appendix 2: Ramsey RESET Test . estat ovtest Ramsey RESET test using powers of the fitted values of lnGDPPCI Ho: model has no omitted variables F(3, 128) = 2.15 Prob > F = 0.0970 13 Appendix 3: Correlation Matrix . corr lnGDPPCI imm Lit GS LFP TO Employers (obs=138) lnGDPPCI imm Lit GS LFP TO Employers lnGDPPCI imm Lit GS LFP - - - - -0.1271 - -0.3037 14 TO Employ~s 1.0000 -0.0155 1.0000 Appendix 4: Regression Results Linear regression Number of obs F(6, 131) Prob > F R-squared Root MSE lnGDPPCI Coef. imm Lit GS LFP TO Employers _cons - - - .005338 -- - Robust Std. Err. - - - .007278 .000896 - - t- - P>|t|- 15 = = = = = - .64965 [95% Conf. Interval] - - - -- -.001787 - - - - - - - Appendix 5: Multicolinearity Results . estat vif Variable VIF 1/VIF Lit imm TO Employers LFP GS - - Mean VIF 1.18 16