Rodrik, Subramanian, and Trebbi (2002) contribute to the discussion on the impact of geography, institutions and international trade on cross-national differences in income levels. They find that only institutions exert a significant direct impact on incomes. Meanwhile, international trade has no direct effect and geography has at best weak direct effects on incomes.
Examining its empirical framework, I find that the way the study measures geography is inadequate and that the framework therefore does not treat the three “deeper determinants” equally in the sense of giving the geography-hypothesis a smaller chance to prevail. Based on this finding, the underlying structure is adjusted and developed further to provide an improved basis for future analyses.
Table of Contents
List of abbreviations
1 Introduction
2 Instruments
2.1 AJR instrument
2.2 FR- instrument
3 Measures of geography
4 Empirical framework
5 Conclusion
Appendix
References
List of abbreviations
Abbildung in dieser Leseprobe nicht enthalten
1 Introduction
In 2002, when Rodrik, Subramanian, and Trebbi (2002) (hereafter, RST) was released, average income levels of the world´s richest and poorest countries according to the World Bank (2014) differed by a factor of more than 100.[1] Until today, these differences haven´t altered significantly (World Bank, 2014). Finding the reasons for these imbalances is one of the most important tasks in economics, since it can help policy makers and others in defining actions that support sustainable growth. In this context, the large crossnational differences in income levels have been explained by economists, historians, and other scientists in various ways. RST is an article that has joint the discussion over the impacts of geography, international trade, institutional quality, and other factors on cross-national differences in income levels. Former studies on this debate were carried out by Easterly and Levine (EL, 2003), Alcalá and Ciccone (AC, 2004), Dollar and Kraay (DK, 2003), Acemoglu, Johnson, and Robinson (AJR, 2001), Frankel and Romer (FR, 1999), Hall and Jones (HJ, 1999), and others.
Together with HJ (1999), AJR (2001), and EL (2003), RST is one of the key contributions to the line of literature that emphasizes the primacy of institutions over geography and other potential determinants of cross-national variation in income levels. On the other side, especially Jeffrey Sachs and coauthors have argued in a series of papers that measures of geography that focus on the disease ecology may have a direct impact on income levels in addition to the undisputed effects of institutional quality (Bloom and Sachs 1998, Gallup et al., 1999, Gallup and Sachs, 2001, Sachs, 2001, Sachs and Malaney, 2002, Sachs, 2003, Carstensen and Gundlach, 2006 (hereafter CG)).
RST focuses on the three “deeper determinants” of economic growth. In the authors´ opinion, geography, integration and institutions determine which countries will innovate and accumulate faster and therefore develop, and which will not. Admitting that economic growth is a complex topic with an enormous variety of factors playing a role, the goal of the paper is to estimate and compare the individual impacts of the three factors on cross-national income variation.
RST views geography as an exogenous variable influencing the other two “deeper determinants” as well as income levels. On the other side, institutions and integration in the analysis are endogenous, influenced by each other as well as by geography and income levels. To deal with the problems of reverse causality and omitted variables bias resulting out of the various interdependencies and the multiplicity of potential variables, the authors employ a two stage least squares estimation procedure utilizing two instruments that were used before in literature. First, following FR (1999), RST instruments for actual trade/GDP ratios by using trade/GDP shares constructed on the basis of a gravity equation for bilateral trade flows. Second, RST follows AJR (2001) in using mortality rates of colonial settlers as an instrument for institutional quality. RST finds that once institutions are controlled for, geography has at best weak direct effects on income while integration has no impact at all. Meanwhile, RST finds a strong and significant positive impact from institutions on income.
Regarding the interdependencies between the “deeper determinants”, the article notes that institutional quality exerts a positive and significant effect on integration. On the other side, integration and geography in return have an impact on institutional quality, which RST interprets as a sign for indirect effects of these factors on income.
This review aims at evaluating RST´s empirical framework as well as its choice of variables and instruments. Therefore, chapter two reviews the appropriateness of the AJR-instrument and the FR-instrument used by RST. Chapter three evaluates RST´s measures of the impact of geography on incomes. This includes a critical examination of RST´s measurement of malaria. I argue that RST´s measures of geography are inadequate and that the framework therefore does not treat the three “deeper determinants” equally in the sense of giving the geography-hypothesis a smaller chance to prevail. In chapter four, the RST´s framework is adjusted and developed further to provide a better basis for future analyses.
2 Instruments
2.1 AJR instrument
To address various problems arising from reverse causality and omitted variables bias, RST makes use of the so called AJR- instrument. Following AJR (2001), RST argues that settler mortality had an important effect on the type of institutions that were built in countries which were colonized by the Europeans. In regions where the settlers encountered relatively few health hazards, they built up strong institutions. In places where the disease environment was less favorable to European settlement, their interests were limited to extracting as much resources as quickly as possible. Furthermore, assuming that differences in institutions persisted over time and that settler mortality does not affect by itself current income levels, RST notes that it is appropriate to use settler mortality as an instrument for institutional quality today.
Taken into account a variety of arguments elaborated by Acemoglu, Johnson and Robinson in numerous studies,[2] this argumentation appears to be reasonable. Additionally, RST states that the settler mortality instrument, as having passed the American Economic Review test, is the best available instrument for detangling the net of interdependencies between the deep determinants. Taken into account the general lack of alternatives for the measurement of the exogenous source in the variation of institutions, it does not surprise that until today, the AJR settler mortality is used by over 20 published articles and many more working papers. Even though studies like Albouy (2012) remark eligible criticism regarding AJR´s choice of data,[3] the use of the AJR- instrument is appropriate, since it is widely accepted and therefore simplifies comparison between RST and related literature.
2.2 FR- instrument
Following FR (1999), RST instruments for actual trade/GDP ratios by using trade/GDP shares constructed on the basis of a gravity equation for bilateral trade flows. The procedure is as follows: First, bilateral trade flows are regressed on measures of country mass, distance between the trade partners, and other geographical variables.[4] In a second step, predicted trade shares are constructed for each country on the basis the estimated coefficients. This constructed trade share is then used as an instrument for actual trade shares.
As the literature on the gravity model of trade shows, geography is an important determinant of bilateral trade (see, for example, Hans Linnemann, 1966, Frankel et al., 1995, and Frankel, 1997). It is furthermore hard to think of other variables than geography ones as important sources of exogenous variation of international trade. Furthermore, by including multiple geography variables, FR (1999) provides an instrument that takes into account the complexity of the matter.
AC (2002) criticizes that this measure of openness (nominal trade divided by nominal GDP) is biased downwards. The authors argue that an increase in trade raises productivity, but mainly in the tradables sector. In their view, this raises the relative price of non-tradables, unless non-tradables are inferior in demand. According to AC (2002), this tends to depress the ratio of trade to nominal GDP. As a result, the increase in the openness ratio would be reduced. Therefore, AC (2002) prefers to use nominal trade divided by PPP GDP, called “real openness”. RST argues in detail why the use of “real openness” produces results inferior to theirs. RST notes that any increase in the productivity, whether driven by trade or not, raises non-tradables prices at home and the price level of an economy relative to others. Therefore, using PPP GDP instead of GDP overestimates measured openness. Appendix A in RST provides a detailed and useful illustration of this point. It shows that the openness measure preferred by AC (2002) yields more biased results than RST´s when productivity in the tradables sector is driven by non-trade factors.
[...]
[1] Average income levels refer to GDP per capita (current US$).
[2] Among these studies are AJR (2001) and Acemoglu, Johnson, and Robinson (2002).
[3] Albouy´s main points of criticism refer to the matching of mortality rates to neighbouring countries and the fact that mortality rates use by AJR never come from actual European settlers, even though some settler rates are available.
[4] Other variables included by FR (1999) are whether the respective countries are landlocked, and whether they share a border.