Further argument for the view that theories are useful but not "true." This may seem to contradict the realist view that theories are general descriptions of causal relationships. But I don't think that this is what is is implied. Rather, useful theories can be viewed as fitting the data because they reveal underlying structures that are not observed directly but only indirectly.
There is a often a tendency to transfer simple analogies too complicated and complex situations and events. Some causal relationship are observable, as it a hammer driving a nail, with the physical theory explaining it in terms of simple variables related in a function.
But most interesting issues are much more complicated and nuanced and may be complex, e.g., subject to emergence owing to synergy. There may a constellation of factors involved, and this may be difficult to order in a hierarchy. Some factors may be catalysts that are necessary for an operation but do not themselves enter into it. These may be presumptions that are hidden assumptions.
In addition, statistics is by definition "inexact" in that it deals with probabilities, unlike deterministic functions in which the variables are all known and measurable, and are expressible in terms of a simple function.
While physics is mostly tractable other than at the edges, life sciences are less so, and social sciences and psychology even less. Economics combines social science and psychology, especially macroeconomics and political economy. Economic sociology and economic anthropology take this into account, global economic history also demonstrates it.
This is coming to the fore now as some critics of MMT, the Green New Deal, and "socialism" demand to see data-based model that "prove" proposed solutions have worked in the past. Of course, the record is important, but the demand for "proof" requires a degree of stringency that is not applied in social science and psychology because it is unattainable. Nor is this standard applied to conventional economics either, its econometric approaching being based on formalism rather than being empirically based.
Another important point that Andrew Gelman makes is the futility of pitting theories against each other. That is a recipe for disagreement in that the party that determines the framing wins. Whose assumptions are going to set the criteria? Why?
And, no, I don’t think it’s in general a good idea to pit theories against each other in competing hypothesis tests. Instead I’d prefer to embed the two theories into a larger model that includes both of them.
This is a good suggestion but it is general. Often, the disagreement is over fundamental criteria that determine a frame of reference. This should be obvious in the different approaches to economic theory and economic practice., e.g., econometric and institutional, static and dynamic, simple and complex, natural and historical.
Obviously, a short post like this can only suggest matters that need deeper reflection, open inquiry and sincere debate aimed at solutions to pressing design problems. This is no long just "theoretical." Humanity has to get this right to survive, let alone prosper. We have seemingly dug ourselves into a hole based on policy that is has turned out to impractical in the extreme, such as socializing negative externalities that have led to environmental degradation and threaten ecological collapse if not addressed successfully in a timely fashion. So, let's get with it.
Statistical Modeling, Causal Inference, and Social Science
Our hypotheses are not just falsifiable; they’re actually false.
Andrew Gelman | Professor of Statistics and Political Science and Director of the Applied Statistics Center, Columbia University
