Stop the Ritual: Why Cross-Country Research Should Let Go of Fake Causality

In cross-country research (such as institutional economics, some comparative political science, and many other fields), instrumental-variable (IV) analysis has become something close to a ritual. Reviewers often demand it. Authors feel compelled to provide it. And everyone pretends that it solves a problem that, in many cases, simply cannot be solved.

I have fiddled in institutional economics for about two decades. One thing I have learned is that institutions are almost never exogenous. They are historically embedded and culturally shaped.

Sorry Rodrik, “Institutions Rule” is trivially true, because we define institutions that way. The interesting observation is that institutions evolve.

Treating institutions as if they were randomly assigned treatments is not heroic, it is highly misleading. Yet we continue to rely on a small set of “classic” instruments: settler mortality, legal origin, shift–share designs, neighbor averages, and historical quirks pressed into service as sources of supposedly exogenous variation.

These strategies create the appearance of causal identification. They encourage a simplistic story, that institutions cause outcomes, while obscuring the feedback loops, co-evolution, and historical contingency that actually define institutional change.

The problems with many of these instruments are well known. Albouy’s critique of Acemoglu, Johnson, and Robinson (2001) made clear that even the most celebrated IV strategies can rest on fragile data. We should all remember the opening three sentences from the abstract of Albouy (2012):

Acemoglu, Johnson, and Robinson’s (2001) seminal article argues property-rights institutions powerfully affect national income, using estimated mortality rates of early European settlers to instrument capital expropriation risk. However, 36 of the 64 countries in the sample are assigned mortality rates from other countries, often based on mistaken or conflicting evidence. Also, incomparable mortality rates from populations of laborers, bishops, and soldiers—often on campaign—are combined in a manner that favors the hypothesis.

The deeper issue is not that IV is “bad” econometrics. On the contrary, many IV-designs are really clever and help answer important questions.

But at the country level, the identifying assumptions required for credible IV-estimation are almost never satisfied. When institutions shape income, culture, trust, conflict, and politics — and are themselves shaped by all of these — there is no clean lever to pull. Pretending otherwise does not advance knowledge.

This does not mean cross-country associations are meaningless. On the contrary, they can be deeply informative. Patterns across countries help us understand institutional complementarities, trade-offs, cultural fit, and long-run correlations that theory alone cannot generate. Description, comparison, and careful interpretation are not sins. They are often the right tools for addressing the question at hand, particularly in understanding how institutions evolve.

As reviewers, we should stop insisting on IV approaches “because causality.” When no credible instrument exists, demanding one only encourages bad practice: weak instruments, implausible exclusion restrictions, and overconfident causal language attached to fragile results.

Cross-country research has limits. Acknowledging those limits is not a failure of rigor. It is intellectual honesty.