This blog post examines the insights provided by the instrumental variable approach and colonial mortality analysis used to clarify the causal relationship between institutions and growth.
The 2024 Nobel Prize in Economics was awarded to Acemoglu and his two colleagues for demonstrating that institutions are the cause of economic growth. The attempt to locate the cause of growth in institutions has a long academic tradition. However, merely confirming a high correlation between institutional development and economic prosperity is insufficient to support the claim that institutions cause growth. This is because while good institutions can lead to growth, a reverse causality also exists: as economies grow, institutions improve. Ajemoelou et al. sought to resolve this issue by employing instrumental variables to clearly establish the causal relationship between institutions and growth.
Typically, to present evidence that X causes Y, the commonly used statistical method involves demonstrating that the statistical inference is sufficiently reliable to reject the hypothesis that the slope estimated from the sample is zero, under the assumption of a linear relationship between them. However, if X affects Y but Y also affects X, or if a third factor exists that correlates with both X and Y but was not considered, or if the observed values of X are imprecise, the estimated slope from the sample may not accurately reflect how changes in X affect changes in Y. In such situations, we must find an instrumental variable Z that is highly correlated with X but uncorrelated with any other factors. Using samples of Z and X, we then estimate the predicted value of X—X-hat—and examine causality based on this. In other words, one must estimate the slope of Y using X-hat—derived from an instrumental variable that is independent of Y except for its relationship with X—rather than relying on the sample value of X, which may be affected by reverse causality, the influence of a third factor, or measurement error. The inference that this slope is non-zero must be reliable.
Ajemoglu et al. focused on regions with post-modern European colonial experiences to find evidence supporting the claim that institutions cause growth. They observed that regions relatively wealthy before colonization tend to be poor today, while those poor at the time tend to be wealthy today. They concluded this ‘reversal of prosperity’ resulted from an institutional reversal. Specifically, Europeans established exploitative institutions to plunder minerals and crops in regions with relatively advanced civilizations. Conversely, in underdeveloped and sparsely populated areas, they adopted inclusive institutions to encourage large-scale settlement and attract European immigrants. This strategy, they argued, produced the prosperity reversal. Therefore, the positive slope observed in the linear relationship between a region’s level of institutional development and its per capita income does not necessarily support the claim that some regions stagnated or grew slowly due to the development of extractive institutions, while others grew rapidly due to the development of inclusive institutions. Ajemoglu et al. presented the high correlation between the predicted values of current institutional development levels (estimated using the mortality rates of early European settlers as an instrumental variable) and the observed values of current income levels as evidence of causation.
But is the mortality rate of early European colonists truly an appropriate instrumental variable? Criticisms of this approach fall into two main categories. First, it is difficult to see a strong correlation between this mortality rate and the current level of institutional development. Ajemoglu et al. counter that the colonial strategy adopted based on the mortality rate of early European colonists was reflected in the formation of past institutions. Despite many subsequent changes, these institutions maintained sufficient persistence to retain a significant correlation with the current level of institutional development. Second, there is criticism that this mortality rate correlates with various factors influencing per capita income levels. Ajemoglu et al. counter that, conversely, the historical mortality rate of Europeans shows no correlation with current income levels when the influence of institutions is excluded. For instance, regarding the point that this mortality rate is also related to climate or geographical environments affecting economic activity today, they explain that it is not problematic because that mortality rate was that of Europeans at the time, not that of indigenous peoples or today’s mortality rate.
In this way, Acemoğlu et al. sought to strengthen the argument that institutions are the cause of economic growth by demonstrating that early colonial mortality rates could be linked to income levels through the formation and persistence of institutions. Their analysis remains significant today for its theoretical and empirical value, as it meticulously demonstrates, based on statistical and historical data, that institutions are a key determinant of long-term economic performance.
