A cross-country analysis of the roles of border openness, human capital and legal institutions in explaining economic development

Abstract

Globalisation, human capital, and institutions have been widely recognised in the literature to be causally important for economic development. Most of the available studies, however, treat measures of these determinants either separately or as substitutes. In this paper, we study the income effects of border openness to migration, education, and the rule of law (our proxies for globalisation, human capital, and institutions, respectively). Using cross-country data covering all regions of the world, and employing instrumental variables for all three factors, we establish that they each have a robust, positive, and strong association with economic development. We then consider whether there are any useful interrelations between the three factors in explaining income. On the interaction effects, the results show that the impact on income of: (i) migration can be materially affected by cultivating good institutions but this effect is not dependent on the education level; (ii) education is important irrespective of the levels of migration and institutions; and (iii) institutions is significantly improved by raising the level of education but is not influenced by migration level. Our paper makes a significant contribution as the first investigation into the effects of migration, education, and institutions jointly and as complements.

KEYWORDS:

• Globalisation
• Migration
• Human capital
• Education
• Legal institutions
• Interaction effects

JEL CLASSIFICATIONS:

• E61
• F22
• I25
• O43

Acknowlwedgements

I would like to thank two anonymous referees and Charles Van Marrewijk (Joint Editor), without implications, for helpful comments. I thank also my wife, who has brought much improvement to this paper through her literary eyes.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 In this paper, “economic development”, “economic prosperity”, and “economic progress” are used interchangeably to refer to increases in national income, which we represent by real gross domestic product (GDP) per capita—and is the outcome measure of interest. Whereas “economic performance” or “economic outcome” are used to capture the level of national income.

2 In our discussion here, the why and what questions capture, respectively, the notions of “fundamental” and “proximate” causes of economic development (North and Thomas 1973; Acemoglu, Gallego, and Robinson 2014).

3 See, for example, North and Thomas (1973), North (1981, 1990), Kremer (1993), Engerman and Sokoloff (1997, 2002), La Porta et al. (1997, 1998), Galor and Weil (2000), Acemoglu, Johnson, and Robinson (2001, 2002) Nunn (2008), Comin, Easterly, and Gong (2010), Putterman and Weil (2010), Jones (2013), Easterly and Levine (2016), and Oyèkọ lá (2021b).

4 See, for example, Banfield (1958), Kamarck (1976), Putnam (1993), Guiso, Sapienza, and Zingales (2006), Diamond (1997), Gallup, Sachs, and Mellinger (1999), Tabellini (2008, 2010), and Nunn and Puga (2012).

5 It is necessary to underline that the use of “policy” or “policies” in this paper is liberally construed, purely descriptive and utilized largely to capture the idea of the “right choices” implied in Sachs (2005, 1–2) or that of “human choices” mentioned by Rodrik, Subramanian, and Trebbi (2004, 133). In this paper, therefore, man-made decisions, factors, or choices should be read similarly to invoke variables largely within a policy maker’s sphere of control.

6 This implies that we explicitly control for cultural, geographical, historical, and regional characteristics in our model; see McCord and Sachs (2015) for a similar interpretation.

7 For convenience, henceforth we will use ‘migration’ for border openness to migration, ‘education’ for human capital, and ‘institutions’ for legal institutions. We review the literature that informs our choice of these three variables (i.e., determinants) in the next section and provide detailed definitions of variables in section 3.

8 But, as Sassen (1996, 59) puts it, ‘there is a growing consensus in the community of states to lift border controls for the flow of capital, information, and services and, more broadly, to further globalisation. But when it comes to immigrants … the national state claims all its old splendour in asserting its sovereign right to control its borders.’ In this sense, there has been a benefit for the developing countries whose citizens aspired to resettle and up-skilled themselves but ended up working in their mother nations (Clemens 2011).

9 Pritchett (2001) and Clemens (2007) cast doubt on the degree of human capital’s positive externalities in development.

10 See also Acemoglu and Robinson (2010) and Acemoglu, Gallego, and Robinson (2014) for similar elucidations.

11 In line with the above, it is easy to imagine how the frontier of research in comparative economic development became that of identifying fundamentals, as opposed to proximates, in the process of income determination.

12 See also Borjas (2015).

13 See section 3 for the description of the instruments and section 4 for details of the econometric framework.

14 https://1997-2001.state.gov/travels/2000/000129clinton_wef.html.

15 While the literature on the income effects of trade and migration mostly developed parallel to each other, their paces have been different. There exists a vast literature exploring the links between trade and cross-country variations in economic performance (e.g., Kormendi and Meguire 1985; Grossman and Helpman 1991; Dollar 1992; Sachs et al. 1995; Frankel and Romer 1999; Alcalá and Ciccone 2004; Gao 2004; Rodrik, Subramanian, and Trebbi 2004; Lucas 2009; Melitz and Redding 2014). The results have been mixed because of differences in the economic and institutional mechanisms of individual countries and due to various reasons already raised in the relevant literature (e.g., Rodriguez and Rodrik 2000; Rodrik, Subramanian, and Trebbi 2004; Ortega and Peri 2014). Our focus in this paper however is migration.

16 It is not, however, all a walk in the field of gold. Apart from the frictions apparent between the economic and globalisation forces—pushing for open borders—and legal and political forces—pushing for closed borders (Hollifield 2004), De Haas (2010) gives a review of both an optimist’s and a pessimist’s perceptions regarding the migration-development nexus. In a no-holds-barred review essay on immigration and globalization, which touches on the remarkable trillion-dollar bills that Clemens (2011) indicated might be lying on the sidewalk, Borjas (2015) stated that the reason ‘nobody ever bothers to pick them up … is easy to summarize: those bills are probably fake’ (p. 967).

17 In comparison, panel data may be required if one were to be interested in the income effects of trade (Feyrer 2019). Also, a growing body of work now examines the factors impacting on the size and structure of the flow of migrants (e.g., Beine, Docquier, and Oden-Defoort 2011; Grogger and Hanson 2011; Belot and Hatton 2012; Bertoli and Moraga 2013; Beine and Parsons 2015; Razin and Wahba 2015; Uprety 2017). This literature stresses the roles that cultural, geographic, and linguistic distances play in determining who leaves, where to, and whence from. Further, another strand of the literature investigating the role of migration focuses on the historical immigration patterns of countries in accounting for differences in their present-day socio-economic outcomes. In a recent study, for example, Easterly and Levine (2016) argued and presented evidence that European colonial migration pattern has a strong and robust positive impact on current national income per capita. Utilising the same migration data of Easterly and Levine, Oyèkọ lá (2021b) finds that European migration during colonisation also has a direct effect on contemporary health outcomes (e.g., life expectancy and fertility rate). A review of the early theories of international migration is given by Massey et al. (1993). A recently edited volume, International Handbook on Migration and Economic Development, by Robert Lucas (2014) is also very relevant here.

18 Just as in the migration literature noted above, there is also a thriving strand of human capital research focused on the colonial, cultural, geographic, and linguistic determinants of human capital accumulation in the distant past (e.g., North 1981; Engerman and Sokoloff 1997, 2002; Glaeser et al. 2004; Lindert 2004; Galor, Moav, and Vollrath 2009). As an example, Crayen and Baten (2010) use age-heaping method to construct global trends in numeracy over the 1820–1949 period for 165 countries; they find that numerical skills mattered for the patterns of long-term growth observed around the world. Beginning with Becker (1960), a sizeable part of the human capital literature also examines the relevance of fertility choice for human capital formation.

19 Acemoglu, Johnson, and Robinson (2001, 2002) show that settlement conditions in the colonies informed the settlement strategies of the European colonisers. Given favourable settlement conditions (e.g., low settler mortality rates and low indigenous population density), Europeans were more likely to favour mass permanent settlement and seek to recreate inclusive institutions (e.g., protection of private property and constraints on the executive powers) akin to their home countries. In comparison, unfavourable settlement conditions indicate that Europeans were more unlikely to favour permanent relocation; hence, they put in place extractive institutions with the primary task of mass looting of the resources in the colonies. Accordingly, these two colonial institutional legacies are argued as the fundamental reasons behind present-day differences in cross-country income per capita.

20 The primacy of institutions over human capital, or the other way round, in the development process remains one of the most debated areas of economic research, being driven by leading scholars. In presenting their finding that ‘human capital is a more basic source of growth than are institutions,’ for example, Glaeser et al. (2004) wrote: ‘It is important to note that, even if one agrees that mortality risk or indigenous population density shaped the European settlement decisions, it is far from clear that what the Europeans brought with them when they settled is limited government. It seems at least as plausible that what they brought with them is themselves, and therefore their know-how and human capital’ (p. 289). In contrast, when discussing the fundament role of institutions vs human capital in causing long-run economic development, Acemoglu, Gallego, and Robinson (2014, 879) argue that: ‘Europeans appear to have brought more human capital per person to their extractive colonies than their settler colonies … If the United States is more educated today than Peru or Mexico, this is not because original colonizers there had higher human capital. Rather, it is because the United States established institutions that supported mass schooling, whereas Peru and Mexico did not.’ Based on the existing literature, and the degree to which a developing country may be able to, or not, affect the third explanatory variable of interest (migration), our research hypothesis necessarily implies that interrelations amongst migration, education and institutions are essential. Hence, we look to data to give credibility to all three.

21 In the words of Beck (2012, 4), legal institutions are ‘rules that govern commercial relationships between different agents of the soci­ety, that is, firms, households, and government. In the broadest sense, legal institutions thus support market-based transactions by defining property rights and allowing for their transfer and protection. They allow for writing and enforcing contracts between agents that do not know each other, in a cost-effective manner … [and] … also provide public goods and govern externalities[.]’

22 See Beck (2012) and the references therein on the different levels of legal institutions. On applications, various dimensions of legal institutions (including its outcomes) have been used in the existing literature to tackle issues (including on economic development and income inequality) at the cross-country, within-country, and inter-industry levels (e.g., Acemoglu, Johnson, and Robinson 2005; Beck, Demirgüç-Kunt, and Levine 2007; Oyèkọ lá 2021a). Beck and Levine (2005) review the literature that examines how important legal institutions are for financial development.

23 See, for example, Bhagwati and Hamada (1974), Miyagiwa (1991), Burda and Wyplosz (1992), Stark, Helmenstein, and Prskawetz (1997), Mountford (1997), Vidal (1998), Beine, Docquier, and Rapoport (2001, 2008), Stark and Wang (2002), Chakraborty (2006), Docquier, Faye, and Pestieau (2008), Easterly and Nyarko (2009), Batista, Lacuesta, and Vicente (2012), and Docquier and Rapoport (2012).

24 Interested readers are encouraged to see Ortega and Peri (2014, Table 1, p. 236) for additional data sources. As our reviewer noted, it would have been useful to employ data at different points to check for time-consistency of estimates or use a panel data set to validate our main findings. We are, however, unable to implement this due to lack of data.

25 The note attached to each table of results describes the relevant geographical endowment being used.

26 See Ortega and Peri (2014, 233–237) for details. In all our regressions, we have used their predicted migration share that is based on the non-linear model specification.

27 See Acemoglu, Gallego, and Robinson (2014, 887) for details. They have also used primary school enrolment rates in 1900 relative to the population aged between 6 and 14 as an additional instrument. We find, however, that this variable lacks predictive power when the three explanatory variables are simultaneously treated as endogenous regressors.

28 This approach was also used by Freund and Bolaky (2008).

29 The reported p-values are for t-tests. We reached similar conclusions to the results presented in table 2 when we applied the Mann-Whitney non-parametric U test.

30 Although we have charted figure 1 for a uniform 118 countries, similar patterns were confirmed when we used all the available data points for each of migration, education, and institutions vis-à-vis real GDP per capita.

31 This is the specification that has dealt with Rodriguez and Rodrik’s (2000) critique.

32 Letting $J=\text{Migration},\text{Education},\text{Institutions}$, we use $\mathrm{\Delta }\text{ln}{Y}_i=\widehat{{\alpha }_n}\times \mathrm{\Delta }\text{J}$, with $n=1,2,3$. Hence, for $J=\text{Migration}$, $\mathrm{\Delta }\text{ln}{Y}_i={\hat{\alpha }}_1\times \mathrm{\Delta }\text{Migration}=7.18\times (0.086-0.008)=0.56$. Following the same steps, it is easy to confirm that $\mathrm{\Delta }\text{ln}{Y}_i=3.12$ for $J=\text{Education}$ and that $\mathrm{\Delta }\text{ln}{Y}_i=2.99$ for $J=\text{Institutions}$.

33 The values for “Americanising” Burundi are obtained as follows: (i) $\mathrm{\Delta }\text{ln}{Y}_i=7.18\times (0.086-0.008)=0.56$—as Burundi’s actual natural log of GDP per capita is 5.98, the level of income per capita based on migration level in the United States would be $\approx \mathrm{\$}691(=e^{6.54})$; (ii) $\mathrm{\Delta }\text{ln}{Y}_i=0.29\times (12-1.26)=3.12$—using now the education level in the United States would yield for Burundi the level of income per capita of $\approx \mathrm{\$}8917(=e^{9.10})$, and (iii) $\mathrm{\Delta }\text{ln}{Y}_i=1.03\times (1.57-(-1.332))=2.99$—based on quality of institutions in the United States, Burundi’s new level of income per capita would be $\approx \mathrm{\$}7832(=e^{8.97})$. Following the same steps, and observing that “Burundianising” the United States would result in the following corresponding log-differences: $-0.56$, $-3.12$, and $-2.99$, it is easy to confirm that the implied real GDP per capita for the United States (respectively) are: $\approx \mathrm{\$}22401(=e^{10.02})$, $\approx \mathrm{\$}1737(=e^{7.46})$, and $\approx \mathrm{\$}1978(=e^{7.59})$.

34 To our best knowledge, this paper is only one of two that have used more than two endogenous regressors; see also Ortega and Peri (2014).

35 In particular, we employ instruments that have been tried and acquitted through peer-review publications and document several first stage diagnostic tests on weak instruments below. See Dollar and Kraay (2003) for a discussion on the problem of weak instruments when multiple instruments are used.

36 In particular, the tabulated critical values for weak instruments in Stock and Yogo (2005) that are derived under the assumption of homoskedasticity for Cragg-Donald Wald F statistic is used for the Kleibergen-Paap rk F statistics that are based on the assumption of heteroskedasticity; this is now commonly adopted in the applied economic literature.

37 Besides, the values of the F statistics mean that we can also reject the hypothesis that the relative IV bias is greater than 5 per cent at the 5 per cent level of significance. Moreover, these F statistics of exclusion restriction exceed the rule of thumb threshold of 10 suggested by Staiger and Stock (1997)—this is particularly appropriate for models 1 and 2 as the rule of thumb was conceived in the context of models with one endogenous regressor, given that instruments can be weak in models with higher endogenous regressions because of collinearity. While the latter is probably true in our case, there is minimal evidence from these first stage statistics that our results were materially influenced.

38 Again, letting $J=\text{Migration},\text{Education},\text{Institutions}$, we use $\mathrm{\Delta }\text{ln}{Y}_i={\hat{\alpha }}_n\times \mathrm{\Delta }\text{J}$, with $n=1,2,3$. Hence, for $J=\text{Migration}$, $\mathrm{\Delta }\text{ln}{Y}_i={\hat{\alpha }}_1\times \mathrm{\Delta }\text{Migration}=8.73\times (0.086-0.008)=0.68$. Following the same steps, it is easy to confirm that $\mathrm{\Delta }\text{ln}{Y}_i=4.91$ for $J=\text{Education}$ and that $\mathrm{\Delta }\text{ln}{Y}_i=2.36$ for $J=\text{Institutions}$.

39 The values for “Americanising” Burundi are obtained as follows: (i) $\mathrm{\Delta }\text{ln}{Y}_i=8.73\times (0.086-0.008)=0.68$—as Burundi’s actual natural log of GDP per capita is 5.98, the level of income per capita based on United States migration level would be $\approx \mathrm{\$}780(=e^{6.66})$; (ii) $\mathrm{\Delta }\text{ln}{Y}_i=0.457\times (12-1.26)=4.91$—using now United States education level would yield for Burundi the level of income per capita of $\approx \mathrm{\$}53639(=e^{10.89})$, and (iii) $\mathrm{\Delta }\text{ln}{Y}_i=0.815\times (1.57-(-1.332))=2.36$—based on United States quality of institutions, Burundi’s new level of income per capita would be $\approx \mathrm{\$}4199(=e^{8.34})$. Following the same steps, and observing that “Burundianising” the United States would result in the following corresponding log-differences: $-0.68$, $-4.91$, and $-2.36$, it is easy to confirm that the implied GDP per capita for the United States (respectively) are: $\approx \mathrm{\$}19856(=e^{9.90})$, $\approx \mathrm{\$}289(=e^{5.67})$, and $\approx \mathrm{\$}3689(=e^{8.21})$.