Eicher, Theo S., García-Peñalosa, Cecilia, and Kuenzel, David J., Journal of Macroeconomics, 2018, 57, 182-209.
Cecilia García Peñalosa is a senior research professor at the CNRS and a member of AMSE. She holds a PhD from Oxford University and has taught in Germany, Spain, Switzerland, and the UK. She works on long-term growth and various aspects of inequality, has been a member of the Conseil d’Analyse Economique and has just joined the European Economic Advisory Group.
What makes countries rich?
Neoclassical growth theories viewed differences in per capita output across countries as due to either differences in accumulated factors (notably physical and human capital) or differences in the level of total factor productivity (TFP). In a seminal article, Hall and Jones set out to measure the relative share of these two causes and found that factor accumulation accounted for only a third of differences, while two thirds were attributed to TFP discrepancies. Yet, if productivity is simply the capacity to combine factors, is it conceivable that machines and educated individuals are ‘combined so much better’ in some countries than in others?
The authors argued that factors are more productive in advanced countries due to better social infrastructure, which they defined as the “institutions and government policies that determine the economic environment within which individuals accumulate skills, and firms accumulate capital and produce output.” This concept, borrowed from political science, captures the idea that the right economic environment will reduce transaction costs, hence leading to higher output for a given level of technology and inputs.
From social infrastructure to political institutions
Economists then addressed economic institutions and proposed a social infrastructure index. The strong explanatory power of this index (accounting for 70% of cross-country GDP differences) gave birth to an entire literature seeking to untangle the determinants of social infrastructure itself. A number of authors have suggested that constitutionally-specified political constraints determine social infrastructure. The resulting empirical literature highlights two key constitutional features: whether a democracy is presidential or parliamentary, which provides different checks and balances on the executive, and whether the voting system is proportional or majoritarian, which affects the match between policy choices and the electorate’s preferences. Unfortunately, these measures are indirectly and subjectively constructed to proxy for a range of aspects of political institutions. Hence, even if the proxies are significant, it is difficult to identify exactly which institution exerts an influence.
The constitutional dataset
To circumvent this problem, we exploit the data in the Comparative Constitutions Project, compiled by political scientists to code all the world’s constitutions. The data covers every aspect of a constitution, which allows us to go beyond the broad categories so far examined. For example, parliamentary regimes are thought to be more accountable than presidential ones, resulting in less rent extraction. The new dataset allows us to ask which rules matter: that budgets are voted in parliament, or that a president is directly elected? Moreover, because it includes aspects of constitutions so far ignored, we can explore new angles such as the role of human rights, an aspect highlighted by Hayek and Sen as providing another layer of checks and balances on decision-makers.
To juxtapose the vast number of candidate regressors, we employ a statistical methodology, Bayesian Model Averaging, specifically designed to address model uncertainty and allowing us to test each individual political institution. Two key results emerge. First, the explanatory power of electoral systems and forms of government is dominated by specific constitutional rules, such as the freedom to form parties or the checks and balances on the executive. Second, we reveal an entirely novel set of social infrastructure determinants: constitutionally-guaranteed human rights. Absence of censorship, separation of church and state, and academic freedom all appear as key variables. These results point towards individual freedoms and responsibilities as core elements of high-quality social infrastructure.
The estimates we obtain allow us to create an “optimal constitution value” generated by an artificial country whose constitution contains all the variables that exert a positive effect and none of the variables that exert a negative one. We can then construct an index that measures a constitution’s distance from the optimal constitution, an excellent predictor of social infrastructure across countries. But does it apply to other aspects of economic life? The index provides a succinct measure of the quality of political institutions, so it could be used to address a number of empirical questions related to growth, to trade or to inequality across individuals, genders or ethnicities. Our approach focused on crosscountry regressions that take constitutional features as given and use them to explain social infrastructure. Reverse causality is, however, an important concern, and it would be desirable to examine endogeneity at the constitutional rule level. Addressing this issue by looking for suitable instruments is a promising avenue for future research.