Do institutions really matter?
Adam Prezeworski
Government and Opposition, Vol. 39, No. 4 (AUTUMN 2004), pp. 527-540 (14 pages)
https://www.jstor.org/stable/44483087
This post is based on a discussion in course on ‘Advanced Political Economy’ by Dr. Anar Ahmadov
Link to the paper HERE
Summary and Interesting Takeaways
Mainstream Understanding of the Impact of Institutions on Economic Development
Institutions are the ‘rules of the game’, made up of rules, behavior patterns, and beliefs. Since the 1990s when Douglas North described the New Institutional Theory as we know it, NIT and the impact of institutions on economic development have been widely researched. Adam Przeworski in this paper critically examines some of the most fundamental assumptions of the literature on NIT.
Przeworski acknowledges the impact of institutions on economic development. His question, however, aims at the known ‘chicken and egg’ problem. How do we distinguish the effects of institutions from those of the conditions that give rise to them in the first place (p. 527)? Institutions are epiphenomenal, meaning they are the secondary effects or byproducts arising from underlying social, economic, or political structures rather than as primary drivers of change or actions (p. 528). Institutions are themselves shaped by the conditions they're meant to be influencing, which calls for a need for methods such as instrumental regression analysis to avoid bias in institutional research. However, in mainstream academia, these problems are being overlooked, leading to potential bias.
Furthermore, Przeworski underscores two main epistemological problems with institutional research:
Exogenous selection on observables
Example: Poor countries tend to be dictatorships; rich ones, democracies. Comparing their growth rates unfairly links dictatorship to slower growth because poverty itself, not the government type, is the primary factor.
Endogenous selection on observables
Example: Democracies failing during crises leave a sample biased toward surviving dictatorships. Any observed superior post-crisis growth in dictatorships is thus skewed; it doesn't prove they are inherently better at handling crises.
Five biases connected to institutional research:
Baseline Difference: Comparing countries with different starting conditions (e.g., wealth) before implementing a regime change makes it hard to isolate the regime's true effect.
Effect of Treatment on the Treated: The choice of regime itself might influence outcomes. Countries that choose a specific regime type might already possess characteristics that affect the outcome. For example, countries that choose democracy may already have characteristics conducive to success.
Post-Treatment Effect: Changing one thing (e.g., the regime) rarely happens in isolation. Other changes occur simultaneously, confounding the analysis of the single change's effect.
Distance Effect: Comparing cases with very different initial conditions leads to weaker causal inferences as the observable differences are too large.
Aggregate Effect: The actions of one country affect the environment for others. Therefore, observing outcomes for one country does not produce a valid estimation if other countries' behavior alters the outcome.
Przeworski proposes the following to avoid these biases in institutional research:
Because we cannot perfectly isolate the effect of the institution from the effects of all other factors, the best we can do is to try various statistical approaches, each making different assumptions about how those factors interact. If these different approaches yield similar results, then we can have more confidence in our conclusions. If the results diverge greatly depending on the assumptions, it means the effect of the institution is heavily intertwined with other things, and our understanding of causality is weak (p. 540).
