Critical dilemmas in the methodology of economics facing the crisis Alessio Moneta Institute of Economics Scuola Superiore Sant Anna, Pisa amoneta@sssup.it 28 November 2014 Seminar Series Critical economics in the times of crisis Scuola Superiore Sant Anna Dilemmas in methodology of economics 1/22
Introduction Introduction Causality The crisis of the economy and the crisis of economics The point of view of methodology Need of a critical assessment Dilemmas in methodology of economics 2/22
Outline Introduction Causality Some dilemmas in the methodology of economics What is a dilemma? The dilemma of causality: data vs. theory driven Other dilemmas : Explanation through models: abstraction vs. idealization Confirmation: falsifiability vs. calibration Economics as a moral science vs. natural science. Dilemmas in methodology of economics 3/22
Causality Introduction Causality Importance and variety of causality issues in economics. understanding the causes of recession / current economic crisis predicting the effect of a policy intervention what is the effect of monetary/fiscal policy? retrospective and prospective questions singular vs. general causes measurement: can we estimate the fiscal multiplier? Notwithstanding the variety of problems all these issues share a dilemma Dilemmas in methodology of economics 4/22
The dilemma of causal inference We face a dilemma in our attempts to uncover causal relationships: We let theory guide us We build theoretical models. But our theoretical/background knowledge is uncertain. 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. We do not let theory guide us We start from the data. We estimate empirical/econometric models and we try to find evidence for causality from them. (Cfr. VAR models, Instrumental Variables, natural experiments, etc.). We cannot guarantee that our conclusions are true of the sample (problem of induction). But suppose we get good evidence for that. Without an understanding of the underlying mechanisms, it is difficult to generalize to other settings / applications (problem of external validity). Dilemmas in methodology of economics 5/22
Deductivist vs. inductivist Disputes in the history of economic thought that partially reflect this dilemma: J.S. Mill (1837) and his critics. Methodenstreit (C. Menger vs. G. Schmoller), beginning XX cent. Measurement without Theory controversy (A. Burns, W.C. Mitchell, D.R. Vining vs. T. Koopmans), 1940s - 1950s. Cowles Commission approach vs. Granger causality VAR (Cooley vs. C. Sims), 1980s. Mostly Harmless Econometrics debate (J.D. Angrist. J.S. Pischke, G. Imbens vs. J. Heckman, A. Deaton), current. Dilemmas in methodology of economics 6/22
Solutions? Introduction Causality Integration of both sources of knowledge. Searching for robust empirical and theoretical knowledge making explicit the sensitivity problem. Dilemmas in methodology of economics 7/22
The dilemma of philosophers What is causation? The philosophy-of-science point of view: regularity/probabilistic account counterfactual accounts interventionist accounts mechanistic accounts Each of these accounts presents problems/counterexamples pluralistic view inferentialist /epistemic account (cfr. Reiss 2012, Williamson 2006): causality has several indicators Dilemmas in methodology of economics 8/22
The probabilistic indicator Scholars who choose the second horn of the dilemma emphasize the importance of the probabilistic indicator. let us the data speak for themselves (cfr. Sims 1972, 1980) Granger-causality Dilemmas in methodology of economics 9/22
The probabilistic indicator Sloppy issues: implicit causality and hidden background theory Cfr. the (in)famous paper by Reinhart and Rogoff (AER 2010): Our approach is decidedly empirical...our main finding is that across both advanced countries and emerging markets, high debts/gdp levels (90 percent and above) are associated with notably lower growth outcomes. Critique by Herndon, Ash and Pollin (CJE 2013): selective exclusion of available data, coding errors and inappropriate weighting of summary statistics lead to serious miscalculations that inaccurately represent the relationship between public debt and GDP growth among 20 advanced economies. Dilemmas in methodology of economics 10/22
Probabilistic/statistical dependence One important indicator of causation is probabilistic dependence. But correlation is not causation. More in general: statistical (i.e. probabilistic) dependence is not causation What is statistical dependence? intuitively, two random variables X and Y are statistical associated (i.e. dependent) if the the realization of X gives useful information about the likely realization of Y statistical dependence is a property of the distribution function f XY (x, y) = f X (x)f Y (y) Dilemmas in methodology of economics 11/22
Statistical dependence and causation There are different measures of statistical dependence, e.g.: correlation (Pearson correlation coefficient): ρ XY = corr(x, Y) = linear regression coefficient β = Granger causality f (X t+1 Ω t ) = f (X t+1 Ω t \Y t ), cov(x, Y) σ X σ Y = E[(X µ X)(Y µ Y )] σ X σ Y cov (X,Y) σ 2 X = ρ XY σ Y σ X etc. one difference between statistical dependence and causality: s.d. is symmetrical (but measures of s.d. can be asymmetrical) causality is asymmetrical Dilemmas in methodology of economics 12/22
Statistical dependence and causation In which sense is s.d. an indicator of causality? Principle of the common cause (cfr. Reichenbach 1956): if X and Y are statistical dependent either (i) X causes Y, (ii) Y causes X, (iii) or there is a common cause Z causing X and Y. But there is also the following possibility: X and Y are prima facie statistical dependent but they result in this way because of a not adequately specified statistical model. Thus, it is crucial to correctly specify a statistical model. How to do that? statistical testing integrating theoretical and background knowledge The particularity of economics as social science: statistical dependencies generated by social structures / interactions. Dilemmas in methodology of economics 13/22
Other indicators Introduction Causality counterfactual dependence interventions mechanism Dilemmas in methodology of economics 14/22
Intervention and causation The idea here is that X causes Y if the manipulation of X will result in the manipulation of Y. J.S. Mill (1837) and the impossibility of running controlled experiments in economics Dilemmas in methodology of economics 15/22
Ideal experiments in economics The impossibility of running controlled experiments is not seen as insurmountable Haavelmo (1944) nature can run experiments for us variety of independent sources of variation conformity to well-defined distributions similarity to randomized controlled trial Dilemmas in methodology of economics 16/22
Experiments as benchmark? In Mostly Harmless Econometrics (2009) J.D. Angrist and J.S. Pischke claim that the exploitation of natural experiments (random assignment of treatment independent of potential outcome) has induced a credibility revolution in empirical economics Much of the research we do... attempts to exploit... readily available sources of variation. We hope to find natural or quasi-experiments that mimic a randomized trial by changing the variable of interest while other factors are kept balanced. Can we always find a convincing natural experiment? Of course not. Nevertheless, we take the position that a notional randomized trial is our benchmark (Angrist and Pischke 2009: 21). But without an understanding of the underlying mechanism this reduces to knowledge of (conditional) statistical independence. Dilemmas in methodology of economics 17/22
Mechanisms Introduction Causality Mechanistic account of causation: if X causes Y, we expect to be a mechanism from X to Y. A mechanism is something which can be decomposed into parts which transmit a causal message, such that the transition from one part to another is governed by some understandable principles. There is this (pernicious!) idea in mainstream economics that understandable means that it has to be reduced to rationality of the agents, and that rationality means optimal allocation. Dilemmas in methodology of economics 18/22
Failure of macroeconomic models in the crisis This idea is at the heart of DSGE models, seen as responsible of the failure of academic economics in the face of the crisis by Colander et al. (2008, 2009), Buiters (2009). Importance of the other horn of the dilemma. Here econometrics has been based on calibration, not on estimation, let alone on testing. Dilemmas in methodology of economics 19/22
The connection to the other dilemmas The riddle of explanation through economic models: abstraction vs. idealization. Empirical validation: falsifiability vs. calibration. Economics as a moral science / natural science / practical science. Dilemmas in methodology of economics 20/22
Conclusions Introduction Causality Three pleas to reorient economics in the face of the crisis: Need of a critical view in two (quite different) meaning: Popperian sense: severe empirical testing (notwithstanding the difficulties of falsificationism). Marxian sense: debunking implicit assumptions, recovering important variables or structures that are left out in the abstraction of economic modelling. More pluralism, but refusal of the anything-goes view. Rethinking of the facts/values distinction. The failures of economics as objective science is also a failure of an idea of objectivity. Thus, what kind of science is economics? Dilemmas in methodology of economics 21/22
...the master-economist must possess a rare combination of gifts. He must be mathematician, historian, statesman, philosopher in some degree. He must understand symbols and speak in words. He must contemplate the particular in terms of the general, and touch abstract and concrete in the same flight of thought. He must study the present in the light of the past for the purposes of the future. No part of man s nature or his institutions must lie entirely outside his regard. He must be purposeful and disinterested in a simultaneous mood; as aloof and incorruptible as an artist, yet sometimes as near the earth as a politician. J. M. Keynes, Alfred Marshall, 1842-1924, The Economic Journal, (Sept. 1924). Dilemmas in methodology of economics 22/22