Long-run expectations in a learning-to-forecast experiment

ABSTRACT

We conduct a Learning to Forecast Experiment using a novel setting in which we elicit subjects’ short- and long-run expectations on the future price of an asset. We find that: (i) the rational expectations equilibrium is not a meaningful description for the whole time spectrum of subjects’ expectations; (ii) they are, instead, better described by an anchor-and-adjustment learning scheme; (iii) subjects exhibit a higher degree of heterogeneity in their long-run expectations vis-à-vis short-run expectations.

KEYWORDS:

• Long-run expectations
• learning-to-forecast experiments
• rational expectations
• coordination

JEL CLASSIFICATION:

• D84
• G12
• C92

Acknowledgement

The authors thank an anonymous referee for the useful comments. The authors are grateful for funding this research from the European Union Seventh Framework Programme (FP7-2007-2013) under the grant agreement no. 619255, the Universitat Jaume I under the project P11B2015-63 and the Spanish Ministry of Economics and Competition under the project ECO2015-68469-R.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplementary data can be accessed here.

Notes

¹ The choice of 20 periods is based on a trade-off between having a sufficiently long time series of observations and avoiding a too demanding task for the subjects.

² We used a pay-off mechanism similar to Haruvy et al. (2007). We calibrated the parameters of the pay-off functions such that , in order to give to the subjects similar incentives to provide accurate predictions in the short- as well as in the long-run.

³ In Figure 2, we show one group as an example, since the qualitative features of the other groups are very similar (material upon request).

⁴ Note that, in the moving average , we exclude to avoid collinearity problems in the regression.

⁵ A simple rule where expectations depend just on previous one-lag short-run predictions.

Additional information

Funding

The authors are grateful for funding this research from the European Union Seventh Framework Programme (FP7-2007-2013) under the grant agreement no. [619255], the Universitat Jaume I under the project [P11B2015-63] and the Spanish Ministry of Economics and Competition under the project [ECO2015-68469-R].