New experimental research that asked undergraduate students to predict future inflation confirms the difficulties people have in making fully rational forecasts. The study by Klaus Adam, published in the April 2007 Economic Journal, shows that in what seems to be a realistic situation where people use relatively simple and unsophisticated ways to predict inflation, output and inflation will be more volatile and monetary policy will have stronger but more delayed effects on the economy.
Can limitations in the way private agents formulate expectations about the future be a source of business cycles? For example, what happens if agents do not have a fully articulated forecasting model but instead use simple models to forecast the future? Can such simple forecasts give rise to economic fluctuations in a way such that the agents’ simple models appear to be ratified?
This study analyses laboratory experiments with human subjects, the aim of which is to discern the nature of subjects’ forecasting ability and to draw implications for the business cycle behaviour of output and inflation.
In the experiments, groups of undergraduate students (from Germany and Italy) are asked to predict future inflation, given an observable history of output (gross domestic product) and inflation.
Students’ predictions serve as an input to an underlying economic model that is used to determine a new data point for output and inflation. The new data is announced to students and added to the sequence of historical observations; thereafter the process repeats itself.
In this setting, students’ forecasts determine outcomes, and at the same time, historical outcomes suggest how best to forecast. An ‘equilibrium’ is a situation in which forecasts generate outcomes that make those forecasts appear optimal.
The main objective of the research is to analyse whether students are able to coordinate on a situation with fully rational inflation forecasts, that is, forecasts that are fully confirmed by outcomes, or whether an equilibrium is better characterised by agents that employ simple and less sophisticated ways to predict inflation that are confirmed only in a more restricted sense.
The experimental evidence shows that forecasters, for a considerable period of time, coordinate on a situation in which their forecasts about future inflation rates are only ‘constrained optimal’.
This constrained optimality has important implications for the implied cyclical behaviour of output and inflation. In particular, compared to a situation with ‘fully optimal’ inflation forecasts, output and inflation display increased volatility and also much more persistent deviations from their average values.
In addition, monetary policy has stronger and more delayed effects on the economy than in a setting with fully rational predictions. Finally, the model’s predictions with
constrained optimal forecasts mimic empirical evidence more closely than those implied by fully rational forecasts.
Eventually, coordination on the equilibrium with constrained optimal forecasts breaks down in intriguing ways, as agents learn to improve some of their forecasts, but convergence to better forecasts takes considerable time.
The research thus shows that equilibria with less-than-fully-rational expectations may be an important ingredient for understanding the cyclical behaviour of economic time series, for example, the persistence in the behaviour of output and inflation, or the effects of monetary policy, and that convergence to rational expectations might take more time than generally assumed.
ENDS
Notes for editors: ‘Experimental Evidence on the Persistence of Output and Inflation’ by Klaus Adam is published in the April 2007 issue of the Economic Journal.
Klaus Adam is at the European Central Bank and the Goethe University of Frankfurt.
For further information: contact Romesh Vaitilingam on 07768-661095 (email: romesh@compuserve.com); or Klaus Adam via email: kladam@wiwi.uni-frankfurt.de