The test of a good model is not whether it corresponds to the true underlying structure of the world, but whether it usefully captures some of the regularities in the concrete phenomena we observe. There are lots of different regularities, more or less bounded in time, space and other dimensions, so we are going to need lots of different models, depending on the questions we are asking and the setting we are asking them in. Thus the need for the "art of choosing".The Slack Wire
A Quick Point on Models
JW Mason | Assistant Professor of Economics, John Jay College, City University of New York
Jumping off on this, three levels of modeling — micro, meso, and macro — pertain in science in general. They are particularly significant in social sciences, where an overarching macro model is lacking in precision.
The natural sciences are the most integrated and well-developed. For example, physics has a pretty strong macro model that unites the various branches of physics at the meso level, and the subsets of these branches are also well-articulated in terns of micro models that are capable of being tested experimentally. Now physicists are attempting to complete the process with a grand unified theory that "explains" the basic processes of change in terms of a single set of equations that pulls it all together into a "theory of everything."
The life sciences are quite well-developed, too, and while there is no grand unifying theory similar to physics in sight to explain life as there is in physics to explain energy in its transformations, biologists have a pretty good ideal of how life develops at the macro level in terms of evolutionary theory, at the meso level in terms of the branches of life science, and also at the micro level in terms of biophysics and biochemistry. So there is hope that eventually, even the life sciences will be folded into the natural sciences in terms of a theory of everything.
The social sciences are much less developed at the macro level owing to the nature of the subject matter. While some speculate that eventually the social science will also be folded into a theory of everything based on a better understanding of the brain and nervous system, few think that this going to happen anytime soon, if ever.
Nevertheless, humans have a strong natural tendency to theorize based on a desire to understand the human condition in terms of basic principles that are causal. Humans have attempted to understand not only their world but also themselves. The result of this has been philosophy.
Aristotle held that an account is not a sufficient explanation. An explanation sufficient to ground understanding (episteme) in truth must causal. Causes are different from reasons. The history of the Western intellectual tradition has been a search for actual causes instead of a merely providing a reasonable account. The idea is that there may be many accounts that are reasonable but a causal account is based on connection with the real.
David Hume threw a spanner into the works with the assertion of his fork between logic, which is empty of content, and sense perception, which is based on sense data. Humans don't have direct insight into a world separate from mind and therefore, they don't have knowledge of "real" causes and can never have causal knowledge. In this view, the Aristotelian project of a search for real causes was still-born. This view led to the development of the logical positivism of the Vienna Circle and British empiricism.
This debate is still lively and unresolved, although many Anglo-American thinkers lean toward Hume's critique. This is at the basis of Milton Friedman's pragmatic claim about modeling that the premises don't matter as long as the output is consistently correct, based on material implication — an argument is true as long as the conclusion is true, regard of whether the premises are true. There is no causality, only correlation.
Critical realists like Roy Bhaskar disagree. Bhaskar's approach is being integrated into economics by some heterodox economists.
Heterodox economists like Tony Lawson, Lars Pålsson Syll, Frederic Lee or Geoffrey Hodgson are trying to work the ideas of critical realism into economics, especially the dynamic idea of macro-micro interaction.
According to critical realist economists, the central aim of economic theory is to provide explanations in terms of hidden generative structures. This position combines transcendental realism with a critique of mainstream economics. It argues that mainstream economics (i) relies excessively on deductivist methodology, (ii) embraces an uncritical enthusiasm for formalism, and (iii) believes in strong conditional predictions in economics despite repeated failures,
The world that mainstream economists study is the empirical world. But this world is "out of phase" (Lawson) with the underlying ontology of economic regularities. The mainstream view is thus a limited reality because empirical realists presume that the objects of inquiry are solely "empirical regularities"—that is, objects and events at the level of the experienced.
The critical realist views the domain of real causal mechanisms as the appropriate object of economic science, whereas the positivist view is that the reality is exhausted in empirical, i.e. experienced reality. Tony Lawson argues that economics ought to embrace a "social ontology" to include the underlying causes of economic phenomena.Conventional economists would likely respond by pointing out microfoundations as a methodological requirement that grounds economic models in reality through rational choice.
There are deep divisions among economists, especially between conventional and heterodox economists, over methodology and that includes assumptions regarding model construction.
Heterodox economists would say that there is much more involved that choosing the right model. Conventional economists make this point, too, but they limit the choices to conventional methodological assumptions.
Heterodox economists counter with the objection that these assumptions are themselves the problem. Choosing the right model involves more than choosing a model. It involves having correct assumptions about methodology. Without correct assumptions about methodology, a comprehensive modeling process that integrates the micro, meso, and macro levels is impossible to achieve, and causal explanation is elusive. As simple example of this is observing stock-flow consistency, which purely mathematical approaches often ignore and proceed to violate.