Exploring consumption-based asset pricing model with stochastic-trend forcing processes

Abstract

Using Canadian data, the consumption-based asset pricing model is studied, defined in terms of nondurable and durable goods consumption. A two-stage estimation procedure is used, which takes account of the presence of common stochastic trends in the forcing processes. This method yields more reasonable estimates of the preference parameters than the previous studies did, and the asset-pricing equation is not rejected by the data. Moreover, the preference specification adopted in this paper allows a number of useful economic information to be obtained. The additive separability assumption and the Cobb–Douglas functional form of the utility function are ruled complements in the sense of Edgeworth and Pareto.

Notes

 This two-step estimation strategy has been used by Wirjanto (1992, 1993), Cooley and Ogaki (1996), Amano and Wirjanto (1998), and Amano et al. (1998). Ogaki and Reinhart (1998) independently estimate a similar model using postwar US data.

 Although {C _t } and {D _t } are I(1) sequences, {D _t C _t } can obviously be I(1) if C _t and D _t are not cointegrated. The pre-tests (not reported in the paper but available from the author on request) confirm this to be the case.

 Another potential source of the downward bias of this measure is liquidity constraint faced by consumers (See Wirjanto, 1995).

 In a simulation study, Amano and Van Norden (1992) find that this joint testing procedure can reduce the frequency of incorrect conclusion regarding the data generation process of a time series substantially.

 See Elliott et al. (1996) and Stock (1991) for Monte Carlo evidence.

 The test statistic is calculated using the prewhitened quadratic spectral kernel estimator proposed by Andrews and Monahan (1992), and the leads-and-lags estimator of Stock and Watson (1993) with a fifth order. The KPSS critical value is taken from Shin (1994).

 The spectral density matrix for this test is estimated using an AR(6) spectral estimator. See Stock (1991). The critical value for this test is calculated from MacKinnon (1994).