SO WHY BE ROBUST?
Why should we commend robustness? One justification is that we are looking for sound social scientific explanations of real-world outcomes. Levels of human welfare vary enormously across space and time. For a great deal of human history, severe poverty (McCloskey 2011) and violent conflict (Pinker 2012) were prevalent features of the human experience. This prompts the central question in political economy of how some regions and regimes underwent a change from a bad state of affairs to a comparatively remarkable degree of prosperity, or from what Adam Smith (1981, chap. 1) called a “rude state” to an “improved one”?
It is this broadly Hobbesian recognition that conflict and poverty are in some sense common or “natural” elements of the human experience that inspires part of the RPE stance. The RPE focus on worst-case scenarios is not necessarily a pessimistic undertaking as such but a mode of analysis for exploring possible causal explanations for transitions away from this natural condition. For most of history, and in many parts of the present world, this worst-case scenario is a realistic scenario. This means that we need an explanation for how people, similar in key respects to ourselves, who are ignorant of their environment, lacking scientific know-how, technology, physical resources, personal security, and assurance, are nevertheless in some circumstances able to improve their condition, typically over the course of many generations but sometimes much more rapidly. The best answer is robust institutions. By contrast, a theory that explains how people of goodwill and already in possession of essential knowledge for social cooperation is not as convincing an explanation of these real-world phenomena.
Cognizance of the possible depths of human experience also provides a normative justification for considering worst-case scenarios. It means critically evaluating potential changes in policy not just with a view to what they could achieve if they succeed, but also what the outcome would be if they failed. This precautionary principle is more attractive once we recognize that political institutions are the result of path-dependent, incremental evolutionary processes rather than the product of rational design alone. Policies that alter the incentive structures of actors, or deprive them of information that was previously known or even assumed, may mean that it is impossible to undo changes that turn out to produce poorer outcomes than anticipated. It is possible for “public capital” (Buchanan 2000, 163) to be destroyed and made unrecoverable.
Robustness also represents an attempt to integrate humility into scholarly research itself. It suggests we recognize that aspects of any model we are using to defend a causal narrative or justify a particular public policy could be wrong. While we cannot eliminate error from our analysis, we can approach problems in such a way that our answers remain relevant even if our model is substantially inaccurate or our measurements of its parameters mistaken in crucial respects. In this way, a robust research methodology has similar virtues to robust statistical analysis that sacrifices point precision for greater confidence in the general pattern of a result (Levy 2002). Robust results are those that remain trustworthy and valid even after accounting for the likely errors and biases that are generated by any inevitably imperfect research project. In this respect, robustness has parallels with the use of triangulation in research approaches, whereby findings are validated by using evidence from multiple perspectives and sources (Blau 2015).