|
Article Excerpt
THE HALLMARK OF the recent development and growth literature is the quest to identify institutions that explain significant portions of the observed differences in living standards across countries. These differences are staggering, about 35- or 94-fold (depending on the data set) (see Hall and Jones 1999, Caselli 2005). While the definitions of "institutions" may vary in previous studies, results are consistent and strong: institutions are consistently shown to explain economically and statistically significant differences in per capita incomes across countries. (1) The set of countries under observation is often dictated by data availability, but generally the literature examines either the global sample or developing countries. The "consensus institutions" that have been associated with economic performance relate to measures of government risk of expropriation, rule of law, bureaucratic quality, corruption, government repudiation of contracts, civil liberties, and openness to trade.
While it is certainly interesting to determine the institutions that are lacking in developing countries, there exists no comprehensive literature that analyzes the institutions that determine the economic fortunes in developed nations. One might expect, for example, that the above-cited consensus institutions vary only marginally across Organisation for Economic Co-operation and Development (OECD) countries--too little, perhaps, to provide insights into how these nations achieve and maintain their status at the development frontier. We examine parameter heterogeneity as it relates to the influence of institutions on output in OECD and non-OECD subsamples. (2) In the process, we analyze whether a set of institutions exists that contains explanatory power and economic influence across subsamples.
While the growth literature provides ample guidance as to which institutions are commonly lacking in developing countries, there exists only rudimentary understanding to what degree these institutions actually matter in advanced countries. Research that focuses on relevant institutions in OECD usually abstracts from the rest of the world. As a result, such studies focus on completely different sets of institutions, such as labor market institutions (e.g., Boeri, Nicoletti, and Scarpetta 2000, Nickell, Nunziata, and Ochel 2005), traditional factor markets such as human and physical capital (e.g., Bassanini, Hemmings, and Scarpetta 2001) or product regulations (e.g., Nicoletti and Scarpetta 2003). Our focus is different, however. We seek to establish whether those institutions that have been shown to hold strong explanatory power in global regression analyses also exhibit explanatory power in the OECD subsample. The goal is to establish one set of institutions that matters in both advanced and developing countries.
Since we seek to investigate whether a set of institutions exists that matters across all countries, our point of departure is the Hall and Jones (1999) methodology. Their approach includes instruments that control for endogeneity in the global model. While their institutional quality measure is clearly correlated with per capita output in the global sample (see Figure 1), the simple ordinary least squares (OLS) regression lines for the two subsamples seem to indicate a differential impact in OECD countries. OECD countries do seem to have a noticeably lower slope than non-OECD countries. Conclusive statements to that effect require, however, rigorous empirical analysis.
[FIGURE 1 OMITTED]
To address issues of parameter heterogeneity, we employ two approaches. First we split the sample, and second we employ the interaction methodology of Brock and Durlauf (2001), who used the approach to examine whether institutional indices can be expected to exhibit parameter homogeneity across "complex heterogeneous objects such as countries." Specifically, they cite the case of the United States and Russia, where civil liberties data can hardly be seen to have a similar impact on economic performance. Previous evidence for parameter heterogeneity was offered by Brock and Durlauf (2001) and Masanjala and Papageorgiou (2003, 2004) for the case of Africa. Eicher, Roehn, and Papgeorgiou (2007) document substantial parameter heterogeneity for the OECD but do not examine institutions explicitly. Another key difference here is that we explicitly address endogeneity and also examine heterogeneity in the instruments.
Our results are not limited to a simple assessment of the impact of established institutions on output in global and OECD samples. As we examine parameter heterogeneity, we find that the established instruments are invalid when we split the sample into OECD and non-OECD countries. This forces us to consider a new set of appropriate instruments in order to successfully control for endogeneity in all subsamples before we can examine the economic impact of institutions on output. Our new set of instruments is based on the hierarchy of institutions hypotheses. The hypothesis is laid out in detail in Acemoglu, Johnson, and Robinson (2005) and similar approaches have provided empirical validation for such a hierarchy (see Persson 2004, 2005, Eicher and Schreiber 2005). The basic argument is that the constitutional institutions/political rules set the stage for the economic institutions. We thus divide institutions into two dimensions: constitutional/political institutions that serve as instruments and economic institutions that are thought to exert direct influence on output.
Our robustness analysis confirms that the explanatory power of the established economic institutions is highly significant for OECD and non-OECD countries, but the effect is about two-thirds smaller in OECD than in non-OECD countries. We also highlight that the estimates obtained in the previous literature for the global sample are a weighted average of the impact of institutions on economic performance in advanced and developing countries. However, evidence for parameter heterogeneity is strong and the instruments established in the literature are weak and often overidentified. Our new set of political instruments performs strongly across subsamples. Most importantly, the instruments are robust to overidentification and weak instrument tests. The instruments are also robust to a number of alternative specifications and data sets. The coefficient estimates for the political instruments are highly significant in all subsamples indicating the important impact of such institutions on the fabric of economic institutions in both advanced and developing nations. In OECD countries, their impact is generally shown to be smaller than non-OECD countries.
1. PARAMETER HETEROGENEITY IN INSTRUMENTS AND INSTITUTIONS
1.1 Established Instruments and Institutions We approach parameter heterogeneity sequentially, first examining the possibility of heterogeneity in the instruments and subsequently focusing on economic institutions. This progression is necessary to establish valid instruments across samples (global, OECD, and non-OECD). In the absence of valid instruments, the impact of economic institutions on output is contaminated by endogeneity bias. Parameter heterogeneity would bias the global regressions even further. Among the established instruments for economic institutions, immediate candidates for parameter heterogeneity relate to the notion of Western European historical influence. Hall and Jones (1999) and Engerman and Sokoloff (1997) provide extensive motivation and historical analysis, respectively, that these instruments relate specifically to European influences. The thought is that the colonizers brought with them the basic prerequisites to establish economic institutions that are conducive to economic development.
For OECD countries, instruments relating to European origins can only be justified if they exhibit sufficient identifying variation and sufficient similar explanatory power. More importantly, however, the validity of the instruments in the OECD context is directly related to how well they can be motivated. Since most OECD countries were the source of the influence that the instruments are supposed to measure, the motivation is called into question. Specifically, measuring the positive influence of a country's own historical experience upon itself makes for weak instruments. A similar line of reasoning raises questions about Latitude (distance from the equator) as an appropriate instrument for OECD countries. Certainly the preference of European settlers to emigrate to similar Latitudes can be seen as a strong motivation for Latitude's influence on economic institutions in developing nations. However, one might argue that, by definition, Latitude holds little power in OECD countries--most of whom were the very source of the settlers. The last instrument Hall and Jones (1999) employ is the Frankel and Romer (1996, 1999) Implied Trade Share for a country. While deviations from the implied trade share can signal weak, or protectionist institutions, it is well known that trade barriers have been uniformly low across OECD countries.
We commence with simple diagnostics to ascertain the validity of instruments and economic institutions across OECD and non-OECD subsamples. Certainly, components of Hall and Jones' (1999) "social infrastructure" index (Rule of Law, Bureaucratic Quality, Corruption, Risk of Expropriation, Government Repudiation of Contracts, and Openness) are key traits for development, but one might be skeptical that these economic institutions explain income variation for developed nations. To establish a benchmark, we follow the methodology introduced by Hall and Jones (1999), who explore the effects of institutions on output by examining the structural model
log Y/L =[alpha] + [beta] I + [epsilon], (1)
where Y/L denotes income per worker and I is the measure of economic institutional quality.
Recognizing that economic institutions are potentially endogenous to income, perhaps being determined by a vector of exogenous factors, X, the regression identifying institutions is
I = [gamma] + [delta] log Y/L + X [theta] + [eta]. (2)
Hall and Jones (1999) provide instruments to address endogeneity. We adopt their instrumental variable estimation strategy, without wanting to imply that other variables are irrelevant.
Valid instruments fulfill two criteria: they are (i) uncorrelated with the error term...
|
