Evaluating how predictable errors in expected income affect consumption

Abstract

This article studies whether anomalies in consumption can be explained by a behavioural model in which agents make predictable errors in forecasting income. We use a micro-data set containing subjective expectations about future income. This article shows that the null hypothesis of rational expectations is rejected in favour of the behavioural model, since consumption responds to predictable forecast errors. On average, agents who we predict are too pessimistic increase consumption after the predictable positive income shock. On average, agents who are too optimistic reduce the consumption.

Keywords:

• behavioural economics
• subjective expectations
• rational expectations
• consumption and saving

JEL Classification::

• D11
• D12
• D84

Acknowledgements

The authors thank Luigi Guiso and Tullio Jappelli for useful comments, as well as seminar participants at the University of Urbino. We are responsible for any errors in the manuscript. The views expressed in this article are those of the authors and do not necessarily reflect those of the Ministry of Economy and Finance. The usual disclaimer applies.

Notes

¹Brown and Taylor (2006) rely on financial expectations and realizations to link past financial optimism and pessimism with current financial prediction accuracy and determine how it affects savings and consumption. They find that past financial optimism has a positive effect on current expectations formation whilst past financial pessimism has a negative effect. Financial optimism is inversely associated with saving and that current financial expectations serve to predict future consumption. From a consumer confidence survey dataset covering 10 European countries over 22 years, Bovi (2009) compare average values of the differences between prospective views versus retrospective views as well as personal views versus general views on economic stances. Findings suggest the presence of structural psychologically driven distortions in people's judgements and expectations formation.

²This conclusion should not be new to an economist. It was clearly an implication of the original work of Friedman on permanent income (Friedman, 1957), where agents were myopic and time horizon was shorter than the entire life. Oddly, this explanation have been forgotten by the literature that follows Hall (1978).

³The percent chance of y ≤ y _max is not asked and it is implicitly assumed to be 100.

⁴Further analyses on the subjective expectation data are available in the Data Appendix.

⁵A macroeconomic shock occurring in the observed years and potentially affecting the Dutch household consumption behaviour is the final constitution of the European Monetary Union (EMU) and the consequent adoption of a single currency. Jappelli and Pistaferri (2011) study whether financial integration and liberalization brought about by the introduction of the euro has affected the sensitivity of consumption with respect to income shocks in Italy. The authors do not find a significant effect on consumption smoothing.

⁶Moreover, because of the pooling of the data into a single cross section, we have calculated SEs corrected to take into account those observations that belong to the same respondent.

⁷The rreg procedure first performs an initial screening based on Cook's distance >1 to eliminate gross outliers prior to calculating starting values and then performs Huber iterations followed by biweight iterations with tuning constant of 7 (Li, 1985). A more detailed description of rreg and some Monte Carlo evaluations are provided by Hamilton (1991).

⁸All these tests were performed by means of the STATA 11's routines ‘ivreg2’ and ‘robivreg’.

⁹Moreover, a visual analysis of error versus fitted values seemed to give support to this view.

¹⁰These results are available from the authors upon request.

¹¹For instance, see Hall and Mishkin (1982), Runkle (1991), Garcia et al. (1997) and Jappelli et al. (1998). More recently, Johnson and Li (2010) distinguish between a household with low liquid assets (liquidity-constrained household) and a household without ready access to credit (borrowing-constrained household) and find that only the consumption growth of households that are both liquidity- and borrowing-constrained is excessively sensitive to lagged income.

¹²See also Jappelli and Pistaferri (2010) for an extended and updated review on empirical approaches and evidence on the sensitivity of consumption to predicted income changes, including works combining realizations and expectations of income or consumption in surveys, in which data on subjective expectations are available.

¹³We have also replicated estimations reported in the previous section including those observations declaring not to be able to put money aside and have found similar results.