Uporaba vedenjske ekonomije v makroekonomskih modelih lahko vodi k boljši napovedni moči

Vedenjska ekonomija, za katero sta Nobelovo nagrado dobila Daniel Kahneman (2002) in letos Richard Thaler, vse bolj prodira tudi v makroekonomijo. Sicer še vedno kraljujejo diskreditirani DSGE modeli, v katerih se kriza lahko zgodi le, če pride šok od zunaj, vendar je kar nekaj makroekonomistov naredilo velik korak naprej k večji realnosti teh neokeynesianskih makro modelov (ki sicer s Keynesom nimajo nobene zveze) – predvsem Roger Farmer z vključitvijo “živalskih nagonov” in Xavier Gabaix  vključitvijo “omejene racionalnosti”.

De Grauwe in Yi sta v zadnjih dveh letih testirala tak behavioristični makro model na primeru prenašanja makro šokov med državami in na primeru monetarne politike v okolju, kjer strukturne reforme povečujejo fleksibilnost gospodarstva (če podjetja prilagodijo cene, zaposleni pa plače navzdol). V slednjem primeru seveda ugotovita, da so v bolj fleksibilnem gospodarstvu šoki manjši in ima monetarna politika manj dela. In to je ključna prednost te nove generacije modelov – da je mogoče modelirati šoke, ki nastanejo znotraj gospodarstva.

(Jaz imam sicer določene težave s temi teoretskimi predpostavkami, saj se v realnem svetu plače in in cene v času krize vedno pokažejo kot rigidne navzdol, medtem ko strukturne reforme na trgu dela v času krize le še poglobijo krizo. Kar pomeni, da bo preteklo še nekaj vode do večje realističnosti makro modelov. Toda nekje je treba začeti).

V zadnji kolumni De Grauwe & Yi na to temo se splača prebrati predvsem uvodni del, kritiko novokeynesianskih modelov. Spodaj je en izsek iz tega.

This is much less the case in mainstream macroeconomics, however. Homo economicus continues to reign supreme in dynamic stochastic general equilibrium (DSGE) models. In these models, individual agents maximise an infinite horizon utility function using rational forecasts based on all available information including the information embedded in the model. Nothing really can go wrong in models populated by supreme agents peacefully optimising and endowed with great cognitive abilities that allow them to understand the complexities of the world. Only exogenous disturbances can get these agents off the rail, forcing them to re-optimise. These models then lead to the view that business cycle fluctuations occur as a result of exogenous events (shocks) that force individuals to reconsider their optimal plans. Nothing in the model creates endogenous business cycle movements. Booms and busts are all the result of exogenous disturbances (Smets and Wouters 2007, Gali 2008).

We need to do better – and that is what we have been trying to do in a series of publications (De Grauwe 2012, De Grauwe and Corrado 2015, De Grauwe and Ji 2016, 2017a). We are, of course, not alone in exploring different tracks of macroeconomic modelling. There is a growing number of researchers developing ‘agent-based’ models and ‘behavioural’ macroeconomic models (Alfarano et al. 2005, Tesfatsion and Judd 2006, Colander et al. 2008, Farmer 2006, Farmer and Foley 2009, Gatti et al. 2011, Gabaix 2014, Westerhoff and Franke 2012, Hommes 2016, Hommes and Lustenhouwer 2016, Muellbauer 2017; see also the recent criticism of Blanchard 2017 and the chapters in Gürkaynak and Tille 2017).

There are many ways in which one can depart from mainstream macroeconomic models. We have chosen to do so by assuming that agents experience cognitive limitations preventing them from having rational expectations. Instead, these agents use simple forecasting rules (heuristics) and evaluate the forecasting performances of these rules ex post. This evaluation leads them to switch to the rules that perform best. It can be argued that in a world of great complexity that nobody fully understands, such a process of adaptive learning might be the rational way of deal handling this complexity (Simon 1957, Gigerenzer and Selten 2002, Ackerlof and Shiller2009).

This adaptive learning assumption introduced in an otherwise standard New Keynesian macroeconomic model produces endogenous waves of optimism and pessimism (animal spirits) that drive the business cycle in a self-fulfilling way. This also leads to a two-way causality. That is, optimism (pessimism) leads to an increase (decline) in output, and the increase (decline) in output in term intensifies optimism (pessimism) (De Grauwe 2012, De Grauwe and Ji 2017a).

An important feature of this dynamics of animal spirits is that the movements of the output gap are characterised by periods of tranquility alternating in an unpredictable way with periods of intense movements reflecting booms and busts. Technically, this means that the distribution of the output gap and output growth is non-Gaussian and exhibits fat tails.

There is now a significant body of empirical evidence showing that the output gaps (and also the growth of output) in OECD countries do not exhibit a Gaussian distribution, but are characterised by excessive kurtosis and fat tails. Fagiolo et al. (2008) and Fagiolo et al. (2009) carried out important econometric analysis documenting the non-normality of the distribution of output gaps and growth rates of GDP. Thus, our behavioural model predicts that in the real world the output gap does not follow a normal distribution, but is characterised by excess kurtosis and fat tails. This feature of the higher moments of the output gap is generated endogenously in the model. It is not the result of imposing such a feature on the stochastic shocks hitting the economy.

The contrast with standard DSGE-models is significant. These models find it difficult to explain the fat tails in the distribution of the output gap. They have to rely on large exogenous shocks as explanations of the boom and bust features of business cycles. Such an explanation is not satisfactory, as it shifts the burden of explaining the business cycle to outside forces. One good example is the recent effort to integrate the financial sector in DSGE models to explain the business cycle. However, the results of these models depend on the assumption that the shocks are serially-correlated. As a result, one can argue that standard DSGE models do not have an endogenous business cycle theory.

Vir: De Grauwe & Yi, VoxEU

 

%d bloggers like this: