Permanent Income, Liquidity, and Adjustments of Automobile Stocks: Evidence from Panel Data

Article excerpt


Bar-Ilan, Avner, and Alan S. Blinder, "The Life Cycle Permanent-Income Model and Consumer Durables," National Bureau of Economic Research Working Paper 2149, February 1987.

Bernanke, Ben, "Permanent Income, Liquidity, and Expenditure on Automobiles: Evidence from Panel Data," Quarterly Journal of Economics, XCIX (1984), 587-614.

Dagenais, Marcel, "Application of a Threshold Model to I. INTRODUCTION

In a recent article Bernanke |1984~ extends the Hall and Mishkin |1982~ approach of testing the permanent income hypothesis to the case of automobile expenditures. Applying the approach to a panel data set with 1,434 households, Bernanke finds no evidence refuting the permanent income hypothesis. An important specification in his model is that households adjust their stock to close a constant fraction of the gap between the level of desired and actual stock during each period. While this specification is convenient, it suffers from two drawbacks. First, as Bernanke himself pointed out, it is inconsistent with the data. Since the stock adjustment specification is inappropriate, one may question his conclusion regarding the permanent income hypothesis. Second, it does not help clarify the nature of the stock adjustment. Aside from being of interest in its own right, the nature of stock adjustment is also important for his conclusion. For example, if the stock adjustment process is a proxy for liquidity constraints, then the conclusion should be that the permanent income hypothesis fails. These considerations suggest that further research should focus on improving the stock adjustment specification.

This paper attempts to improve on Bernanke's specification of the stock adjustment by using a threshold adjustment model. Households are assumed not to adjust their stock when the gap between the actual and desired stock is less than some threshold level, but to adjust completely to close the gap otherwise. The model is estimated using the same data set as Bernanke, except that only two years of observations on automobile expenditures are included. The major findings are the following. First, desired stocks are not overly sensitive to transitory income. Second, the threshold levels vary across subsamples in a way that indicates liquidity constraints. Finally, upward adjustments are substantially quicker than downward adjustments, suggesting the presence of resale market imperfections. Overall, these findings suggest that the permanent income hypothesis fails, and that liquidity constraints and resale market imperfections constitute the sources of the failure.

The rest of the paper is organized as follows. Section II describes the data and documents evidence that motivates the model. The model is presented in Section III. Section IV describes the estimation procedures. Section V discusses the implications of the permanent income hypothesis and its alternatives. Section VI presents the results. The last section contains some concluding remarks.


Bernanke's data set is the Hendrick and Youman |1976~ four-year panel study of consumer behavior conducted between 1967 and 1970. This seems to be the only data set with the sufficiently detailed information necessary to construct reasonable estimates of car stocks and expenditures. The basic data series for this paper are the car expenditures and changes in income for 1967, 1968, and 1969. In addition, our estimation also requires data for car stocks at the beginning of 1967. Their construction is described in the Appendices. Table I presents the summary statistics of the data. Automobile expenditures are large, and automobiles are the most expensive item for many consumers. This feature of automobiles is highlighted in Table I. According to Table I, automobile expenditures constitute as much as 5 percent of a household's disposable income.(1)


Two features of the data are inconsistent with Bernanke's |1984~ stock adjustment specification. …