Why Nations Fail. Key Arguments and Empirical Support of the Theories

I. Introduction

“An estimated 766 million people, or 10.7 percent of the world’s population, lived in extreme poverty in 2013.” (World Bank 2017, p. 1) As if these numbers itself weren’t enough sign of the great inequality in incomes after centuries of prosperity, following the World Income Indicators, more than half of the people living under these circumstances originate from one region, Sub-Saharan-Africa. Maybe as long as growth has been observable, controversies about the causes and its inherent erratic distribution flourished. Over time, many hypotheses have been proposed, discussed and rejected. Two of the ones that managed to establish themselves are subject of this essay. More specifically, what their key arguments and empirical support are. One the one hand, the institutional theory of growth promoted most notably by Acemoglu and fellows (2012; 2005). On the other hand the geographic theory of growth, proposed by Sachs et al. (1998; 1999) .

Plan of the essay is as follows. Chapter II will describe the institutional theory of growth as described in Acemoglu and Robinson (2012). Chapter III assesses the key factors and their empirical support of the institutional and geographic growth hypotheses respectively. Followed by Chapter IV, which gives insight on surrounding literature. Chapter V discusses the main problems of each line of argument, concluding that the institutional model offers more consistency.

II. The Institutional Theory of Growth and its Contractants

When assessing a theory explaining economic growth based on institutional endowments, there is one fundamental point to be clarified at the outset, the understanding of institutions that prevails in the following context. Going along with the interpretation Acemoglu, Johnson and Robinson (2005) emphasized in their main work (to which I will resort often as AJR in the course of this essay) presenting this approach to growth theory, one can think of institutions as “[…] constraints that shape human interaction[…]” and thus “[…] structure incentives in human exchange, whether political, social, or economic” (North 1990, p. 3).

But in how far do these institutions matter? Resorting to Chapter 2 of Acemoglu and Robinson (2012), they are key in explaining the current erratic income distribution around the world, whereas Western Europe and its former colonies with high proportions of European settlers exhibit the highest per capita GDP in U.S. $ in 2008, while all other regions lack behind. Especially the countries from the Sub Saharan region with average pc. GDP below $2,000, only curtly outperformed by most of Latin America as well as Central and South East Asia (pc. GDP between $2,000 and $ 7,500). Of course, these regions are not uniform, but also show clear income rankings among them. (Compare Map 1)

What makes economic institutions fundamental in that matter can already be deduced from the North citation above. They give incentives for economic agents, most importantly to invest into physical capital, technology and productional structure. (Acemoglu et al. 2005) Thus, in line with the basic assumption of common growth models, increasing output through higher factor availability (particularly capital) and efficiency. Logically, some institutional settings bare better circumstances for investment and growth, whilst others impede it. The authors identify such good institutions “[..] as those that provide security of property rights and relatively equal access to economic resources to a broad cross-section of society”. (Acemoglu et al. 2005, p. 395)

This economic Institutions in turn are no ominous exogenous construct, but consequence to social processes, whereas people have preferences over certain institutional sets due to their distributional consequences. Which institutions now prevail within a nation is determined by distribution of influence, namely political power, across different interest groups. Political power stems from two sources. First, from the ability to impose interests against others by force, which is determined by resource endowment of a group, which again is defined by the economic institutions yielding the certain resource distribution. Second, by (generally more durable) legal rights given to certain groups through political institutions. Consequently, the party with biggest political power will implement political and economic institutions best fitting their interests. These are in general, assuming egoistic utility maximising behaviour, sets that secure them their place in power. Thus, economic institutions that allocate most resources of an economy towards them and political institutions providing long-term legal advantages. This is the core of the institutional hypothesis proposed by Acemoglu et.al. (2005, 2012).

One can see, that the institutional sets succeeding this process not necessary need to be optimal in terms of economic performance including growth, as long as they fit the interests of the group in charge. There is only little compromise between groups to increase general wealth and opportunity to expect due to the lack of a sovereign entity enforcing contracts between opposing parties. Thus, the possibility to infringe upon the agreements and still act after own interest remains. If it comes to change, it is often abrupt and with vast effects, characterized by big shifts in the ability to impose interests by force. (Acemoglu et al. 2005)

The depicted income differences in Map 1 can therefore be traced back to differing institutional sets among the nations, whereas some exhibit `good institutions` (especially the European democracies and their direct colonial heirs) whilst others do not. This became particularly important in the 18th century with beginning industrialization and rapid technological progress, from which, as AJR argue, the countries with `good institutions` benefitted the most, being able to adopt new technologies and utilize great investment opportunities. By this means the current income pyramid was shaped, portraying a harsh turn-around (especially among erstwhile European colonies) in relative incomes around the globe, away from the former prosperous regions in the tropics towards Europe and its descendants.

This observable process is basis for the fundamental empirical support, presented in Acemoglu et al. (2002), of the institutional hypothesis as well as main argument contradicting the Geographic hypothesis of growth. Mainly promoted by Sachs et. al (1998; 2001; 1999) on observations of Sub-Saharan Africa and Diamond (1997), the theory connects differing growth prospects to geographic and climatic circumstances (who can either be seen as constant over time or only potent for a certain period) of a region. As also depicted in Map 1, the poorest regions on earth all resort within the geographical tropics, thus the argumentation that especially the tropical conditions impede growth due to their negative influence on public health and labour productivity, agricultural productivity, transport costs and diffusion of technology. Given that geographic conditions at best change marginally, the growth prospects for certain regions should be, following the simplest Geographic thesis, unalterable, leading to durable growth rates and global income distributions. The observed turn-over in relative incomes sharply contradicts that (Acemoglu and Robinson 2012).

III. Institutions vs Geography

As mentioned in the previous chapter, the institutional hypothesis puts certain institutional sets at the centre of growth prospects. Figure 1 shows an undeniable positive relationship between income (in form of log GDP in PPP in 1995) and institutions, here measured in a variable displaying protection against expropriation of investment through a government between 1985- 1995. The data is obtained from World Bank (1999) for GDP and for the Expropriation risk directly from Political Risk Services in September 1999 by the authors. Of course, hardly any causality can be drawn from that, as influence might very well be reversed, not to mention omitted factors playing a role. To further support the argument, that indeed there might be a causality running between these two variables, AJR use a series of historical examples and natural experiments tracing institutional development and connecting it with growth prospects.

Most notable is the observed turn-around in relative incomes among former European colonies, as emphasized in Acemoglu et al. (2002). According to that, relatively prosperous nations before the start of European Colonial rule are nowadays among the relatively poorer ones, globally and within the sample of Ex-Colonies (i.e. comparing North and Latin America). That is due to the differing institutional sets implemented by colonial rulers, focusing on exploitation of wealthy regions, which additionally were less amenable for European settlement, mostly because of an unfavourable disease environment (Acemoglu et al. 2005). Further research into that direction already suggests a strong correlation between amenity of regions, proxied by settler mortality between 1700 and 1900, and present quality of institutions as well as GDP, making it a good instrument for present institutions [compare Acemoglu et al. (2001)].

To consolidate the turn-around, one must first find a viable measurement unit describing income levels at the time before European colonisation. AJR do so by using estimated urbanization rates and population densities in 1500 as proxies for economic development. As argued by many scholars, it needs relatively progressive levels of agricultural productivity, yielding outcomes beyond self-sufficiency levels, as well as sophisticated structures of trade and transport to sustain bigger urban centres and greater populations. The applied data for urbanization estimates is collected from Bairoch (1988) and Eggimann (1999), supplemented by findings from Chandler (1987). To achieve a uniform structure of the data for a big enough sample of countries (especially colonies) required the restructuring of Eggimanns data, who listed settlements from the scale of 20.000 inhabitants upward, to a Bairoch-equivalent, who set the threshold for urban centres at 5.000 inhabitants. AJR do so in various ways, mainly by running a regression of Bairoch on Eggimann data for all overlapping countries in 1900 and using the resulting coefficient and constant to convert the findings [Appendix 1 and 2 in Acemoglu et al. (2002)]. Population density specifications are obtained from MacEvedy and Jones (1978), dividing total population of a country by arable land for the specific date. One can see there already, that this construction of basic data is afflicted by the lack of consistent research and, due to the large share of estimations and transformation, subject to great volatility. This limits the validity of results based on this data.

In the next step, AJR regress the log of 1995 PPP adjusted GDP on the proxies for economic development in 1500 among the sample of 41 former colonies. Both for urbanization and log population density the regressions yield a negative and significant (on the 1% to 5% level) coefficient (i.e. -0.078 for the basic regression model of log GDP PPP pc on urbanization in 1500). Results remain broadly in a similar pattern, even when testing for certain regions and adding additional geographical (who are insignificant) and cultural control variables [compare Tables III-V in Acemoglu et al. (2002)]. This negative relationship supports the argument of a relative income shift. The more prosperous a region in 1500, the lower is the present expected GDP to be. Table VI gives additional regression results when accounting for the measurement error the development proxies (especially urbanization rates) are ridden with, using different specifications of urbanization and population density and instrumenting urbanization rates in 1500 with log population density. The general pattern of influence remains unchanged. Worth mentioning are the results from Panel C, where urbanization and population density are simultaneously applied in the regression. Urbanization loses its significance in this specification, which hints that population density includes the explanatory power of the volatile urbanization measurement. Additionally, the R-Squared increases to a moderate level of about 50% for the first time, strengthening the validity of the empirical conclusions.

The missing link is now to demonstrate that the income shift was facilitated by institutions. AJR do so in a simple matter. Suppose the simple regression model 𝑌= 𝛼∗𝑋+ 𝛽∗𝑍+ 𝜖,where Y is present income, X a measurement for institutions and Z either urbanization and or population density. If now X were to include all effects of Z, coefficient of the latter should be insignificant and close to zero when running this regression model. However, in practice, regressing present institutions on present GDP is set to be error ridden due to reverse causality, influence of omitted variables and imprecision. To bypass these problems, AJR find in the log settler mortality, developed in Acemoglu et al. (2001), a viable instrument for institutions. This allows them to conduct a 2SLS regression on the basic model. If the 2SLS is to yield significant results, 3 assumptions need to be fulfilled. According to the paper mentioned beforehand, Relevance condition is satisfied. Second, the Exclusion restriction. This is the case due to differing immunities between the native population and the settlers, who didn’t bring the same immunity to disease influences in the Colonial regions. (Acemoglu et al. 2002) Thus, the variable doesn’t reflect general mortality rates and implicitly population development of a region, influencing growth through this channel. Seemingly, AJR assume the Monotonicity assumption to hold. The results of the 2SLS depicted in Table 1 significantly support the hypothesis that the shift in relative incomes was driven by economic institutions, hence also that institutions are the main driving force behind economic growth. The coefficient for the predicted institutional measurements (in different forms) is positive and mostly significant at the 1% level, also exhibiting adequate explanatory power (i.e. even when using the earliest measurement for institutions after independence, the coefficient remains between 0.37 and 0.46 and significant, hinting that institutions implemented during colonial rule had long lasting effects on the growth prospects of a country). More importantly, the controls of urbanization rate and population density are insignificant throughout all specifications, provided that the calculated standard error is correctly accounting for the uncertainty in the adjusted X variable (AJR do not mention this directly). One factor decreasing the value of the outcomes is the relatively small sample size of countries (who are generally also mainly ex-colonies), which leads to higher bias in the estimations and complicates generalisations of the institutional influence on countries not belonging to the colonies (for which log settler mortality would also be an improper instrument). Yet, institutions indeed seem to strongly influence growth prospects of a country.

[...]