Catálogo de publicaciones - libros

Economic policy making literally matters to all of us. Economics is supposed to provide the necessary knowledge which enables policy-makers to take effective economic policy decisions. This is why it is so important for all of us to understand how reliable this kind of knowledge is. What can we reasonably expect from economics and where are its limits? One particular scientific objective in the policy-context is to make correct predictions. And the following inquiry is about the limits of such macroeconomic predictions as well as about their sociopolitical consequences.

This chapter assesses the current performance of macroeconomic forecasts. It reviews the findings of empirical forecast evaluations and discusses them thoroughly while distinguishing absolute and relative performance, the performance of specific methods and that of different forecasting institutions. All things considered, the evaluation diagnoses a comprehensive failure of macroeconomic forecasts. The last section summarizes the main conclusions that can be drawn from the discussed studies.

This chapter assesses the performance of macroeconomic forecasts from a historical point of view. After briefly outlining the history of economic forecasting, it reviews and discusses empirical studies that consider whether economic forecast performance has been improved over time. In the light of these empirical studies, the answer to that question is negative. Likewise, there seems to be no predictive progress as to the different forecasting methods, either. The last section summarizes this chapter’s main observations.

This chapter evaluates the predictive performance of thermodynamics in distinguishing between experimental and non-experimental contexts. Two historic case studies, namely on the latent heat of fusion and the ideal gas law, warrant the conclusion that experimental thermodynamics is characterized by predictive progress as well as . The non-experimental case is explored by examining attempts to predict the climate. Climatology’s predictive performance is found to be poor.

The fact that the economy is a complex system explains some of the observations made in the previous part. Besides (i) yielding an argument from external effects which shows why macroeconomic forecasts generally fail, it gives rise to explanations of (ii) the more particular fact that directional forecasts do not perform any better than point forecasts, and (iii) the marginal role of macroeconomic theory for economic forecasting.

Data cannot possibly be compared with the ‘true data’ in order to assess its quality. Instead, its quality is determined by its precision as well as by the evidence in favor of its underlying assumptions. In general, forecasts are negatively affected by inferior data-quality. Both macroeconomic and climate-data suffer from typical problems of aggregation and their quality, in terms of precision as well as of evidence for underlying assumptions, is poor. Thus, a considerable part of macroeconomic and climate forecast failure can be explained.

This chapter starts by reconstructing an argument from Oskar Morgenstern against the possibility of economic forecasting which explains economic forecast failure by pinpointing forecast reflexiveness. Subsequently, two lines of attack against this argument are presented. One of them consists in the proposal by Grunberg & Modigliani to construct self-fulfilling forecasts. The other one is based on the rational expectations hypothesis. However, it is shown that both of these counterarguments either fail or give rise to new explanations of forecast failure. Thus, whether the critiques are appropriate or not, forecast failure can be explained with reference to expectations and reflexiveness.

Sensitive dependence on initial conditions can explain forecast failure. This chapter develops a “routes to chaos”-approach which shows that all empirically adequate models of the economy as well as the climate exhibit sensitive dependence on initial conditions. As a first step, that approach identifies certain minimal models, namely the coupled oscillator and the cellular automata, which sensitively depend on initial conditions. Next, they are shown to model some characteristic features of both the economy as well as the climate under their every-day description. As the minimal models sensitively depend on initial conditions, so does any adequate model of the economy and our climate system, too.

Simulation, according to this chapter’s main result, will never improve the predictive performance neither of economics nor of climatology by compensating the lack of experiments. It is argued that there are principally two approaches to simulation: a conservative and a constructive one, and that both approaches cannot successfully be applied to the sciences this inquiry deals with. Last but not least, the methodological framework is used to unmask the absurdity of the ensemble-prediction method in climatology.

Macroeconomic and climate forecasts are largely based on unrealistic assumptions. This fact as such, however, does not explain forecast failure as the counterexamples of Larry Laudan’s pessimistic meta-induction show. Once this is established, this chapter develops a more sophisticated explanation of forecast failure involving unrealistic assumptions by a critical discussion of the molecular kinetic deduction of the ideal gas law and its improvement. But, finally, this type of explanation cannot be applied in economics.