![]()
Part I
Alternative methodologies
![]()
1 Critical dilemmas in the methodology of economics facing the crisis
Alessio Moneta
1. Introduction
The aim of this chapter is to provide a personal assessment of the status of economics in the face of the current economic crisis, from the point of view of the methodology of economics. Many scholars have observed that the status of economics, and in particular that of macroeconomics, is far from being in a good shape. For example, Joseph Stiglitz has recently written that:
(Stiglitz, 2015)
Here I shall point out some dilemmas in the methodology of economics. A dilemma is a problematic situation in which there are offered two possible solutions, neither of which seems satisfactory. The methodology of economics is âthe study of how economics functions, how it could function, and how it should functionâ.1 In the study of the functioning of economics, it is very likely you would come across some impasses, which probably have always existed in the discipline but which have become particularly evident and critical after the economic crisis. I call these problematic situations âdilemmasâ; but this term should not be taken too literally: in most cases the duality of the (failed) solutions is only prima facie. However, this term may help in framing the different problems and their connections.
The first dilemma concerns the problem of causal inference, i.e. the problem of uncovering causal-effect relationships that can in turn be useful for predicting, explaining and intervening in the economy. The two âhornsâ of the dilemma are here the data-driven and the theory-driven approaches. The second problematic situation is the fact that economic phenomena are usually explained through (mathematically) formalised models. Here there are two possible approaches: the first aims at searching for pragmatic relevance. In other words, scholars who take this route want models to be able to give precise, simple and useful answers to economic-relevant questions, rather than to mirror the complex economic reality. This is typically done through abstraction and idealisation of the reality. The alternative route is to aim at developing explanatory models which are true and verified, but at the cost of facing the problem of replicating a complex and seemingly unstructured reality, neglecting pragmatic relevance. The third dilemma concerns the problem of falsification of economic-theoretical hypotheses: should we develop methods that attempt to severely falsify them or should we renounce this task given the fact that economic-theoretical hypotheses cannot be easily confronted with the data? The fourth dilemma concerns the alternative view of economics as a scientific and moral discipline.
Beside the note of caution about âdilemmasâ that may contain more than two âhornsâ, I should also add that the four dilemmas presented here may not exhaust all the difficulties one might encounter in studying how economics functions. But this simplification may be useful to illustrate the difficulties the methodology of economics should address. Moreover, in the rest of the chapter, for reasons of space, I will devote much more effort to explaining the first dilemma (see section 2) than the other three, which will only briefly be introduced in section 3 and shown how they are intertwined with the first.
2. The dilemma of causality
2.1 Theory-driven vs. data-driven approaches
Questions about cause-effect relationships are pervasive in economics, but they come in many different forms. We may ask for the cause of the great recession, where ârecessionâ is an event or an historically defined set of events. We may enquire about the effect of conventional and unconventional monetary policy on macroeconomic variables in a particular country, monetary area or even in general in the market economy. We may pose retrospective (e.g. what was the effect of austerity policy?) or prospective (e.g. what would be the effect of Italy exiting the Euro area?) causal questions. Another important issue is measurement. Sometimes it is crucial to attach numbers to causal influences. What is the size of the government spending multiplier (the ratio between change in national income to the change in government spending that causes it) in a particular country, in a particular period?2
We should note that questions about causality are important in any scientific discipline, but in economics they assume a particular relevance because economics is a political discipline: it attempts to provide reliable knowledge on which to ground policy. Policy is about intervening in the (social and physical) reality to influence or control some output. Causal knowledge is, among other things, knowledge of connections that permit intervention and control. Thus, it is no wonder that policy discussions in economics are also discussions about what causes what.
Questions about causality in economics concern events (or types of events) of quite a different nature, can be framed in many different manners, and are addressed in the literature using diverse methods. Nevertheless, they share a common problem, which I call here the âcausality dilemmaâ.3 Indeed, to settle these questions, one needs to substantiate some sorts of causal claims, and in economic research there seem to be two radically different traditional ways to justify cause-effect relationships. One way is to let (economic) theory guide us in our attempts to infer causal relationships from observations. The other way is not to let theory guide us and to rely instead on statistical methods jointly with rules of inference and restrictions which allow us to estimate causal relationships from data.
The first horn
If we let theory guide us, we build theoretical models on the basis of our theoretical knowledge and background knowledge in general. But in economics theoretical or background knowledge is often uncertain or contentious. At least there is no consensus on the assumptions upon which theoretical models are built. It is easy to build models with conflicting policy implications by (sometimes slightly) modifying the initial conditions.
Consider, for example, the question about monetary policy: which instruments are controlled by the monetary policymaker, and how is their modification transmitted to real economic activities? This is, undoubtedly, one of the most discussed issues in macroeconomics: the way it is framed and addressed is at the basis of the different schools of thought in economics. An answer is crucial for the solution to the current economic crisis, but the issue goes back to at least David Humeâs essay âOf Moneyâ (1985 [1752]), which, as pointed out by Hoover (2001), is framed in explicit causal terms. (Incidentally, David Hume is also the most prominent figure in the modern philosophical analysis of causality.) But let us consider here the way the issue of monetary policy was framed by mainstream economists in the 1950s and 1960s. The most used approach was, at that time, the so-called âCowles Commissionâ econometric approach. Essential to this approach was a clear distinction between exogenous and endogenous variables. The variables controlled by the monetary policymaker were assumed to be exogenous, while the variables representing final outcomes of the policy were taken as endogenous. As illustrated by Favero (2001, p. 103), the Cowles Commission approach starts by specifying a theoretical model, which is usually a large-scale macroeconometric model, i.e. consisting of a large number of variables and equations describing the economic system. The model is then identified by imposing a number of a priori restrictions that ascribe exogeneity status to a number of variables. It is also essential in this approach that the error terms entered into the behavioural equations of the model follow a definite probability distribution, so that the model can be analysed by standard statistical tools (Hoover, 2012). Further steps are estimation of the relevant parameters (through standard regression models) and simulation of the effects of policy interventions.
The consensus in mainstream economics around the Cowles Commission approach broke down in the mid-1970s for various reasons, one of which (perhaps the most relevant) was the lack of trust on the theoretical assumptions used in this approach. Particularly popular were the critiques of Lucas (1976) and Sims (1980). Lucas attacks the identification framework proposed by the Cowles Commission by pointing out...
