ABSTRACT Most of the evidence on consumption‐based asset pricing is based on seasonally adjusted consumption data. The consumption‐based models have not worked well for explaining asset returns, but with seasonally adjusted data there are reasons to expect spurious rejections of the models. This paper examines asset pricing models using not seasonally adjusted aggregate consumption data. We find evidence against models with time‐separable preferences, even when the models incorporate seasonality and allow seasonal heteroskedasticity. A model that uses not seasonally adjusted consumption data and nonseparable preferences with seasonal effects works better according to several criteria. The parameter estimates imply a form of seasonal habit persistence in aggregate consumption expenditures.

This paper studies the ability of long-run risk models to explain out-of-sample asset returns during 1931–2009. The long-run risk models perform relatively well on the momentum effect. A cointegrated version of the model outperforms the classical, stationary version. Both the long-run and the short-run consumption shocks in the models are empirically important for the models' performance. The models' average pricing errors are especially small in the decades from the 1950s to the 1990s. When we restrict the risk premiums to identify structural parameters, this results in larger average pricing errors but often smaller error variances. The mean squared errors are not substantially better than those of the classical CAPM, except for Momentum.

This paper studies the estimation of asset pricing model regressions with conditional alphas and betas, focusing on the joint effects of data snooping and spurious regression. We find that the regressions are reasonably well specified for conditional betas, even in settings where simple predictive regressions are severely biased. However, there are biases in estimates of the conditional alphas. When time-varying alphas are suppressed and only time-varying betas are considered, the betas become biased. Previous studies overstate the significance of time-varying alphas.

This article shows how to evaluate the performance of managed portfolios using stochastic discount factors (SDFs) from continuous-time term structure models. These models imply empirical factors that include time averages of the underlying state variables. The approach addresses a performance measurement bias, described by Goetzmann, Ingersoll, and Ivkovic (2000) and Ferson and Khang (2002), arising because fund managers may trade within the return measurement interval or hold positions in replicable options. The empirical factors contribute explanatory power in factor model regressions and reduce model pricing errors. We illustrate the approach on US government bond funds during 1986–2000.

We develop methods for testing factor models when the weights in portfolios of factors and test assets can vary with lagged information. We derive and evaluate consistent standard errors and finite sample bias adjustments for unconditional maximum squared Sharpe ratios and their differences. Bias adjustment using a second-order approximation performs well. We derive optimal zero-beta rates for models with dynamically trading portfolios. Factor models’ Sharpe ratios are larger but standard test asset portfolios’ maximum Sharpe ratios are larger still when there is dynamic trading. As a result, most of the popular factor models are rejected.

ABSTRACT Even though stock returns are not highly autocorrelated, there is a spurious regression bias in predictive regressions for stock returns related to the classic studies of Yule (1926) and Granger and Newbold (1974) . Data mining for predictor variables interacts with spurious regression bias. The two effects reinforce each other, because more highly persistent series are more likely to be found significant in the search for predictor variables. Our simulations suggest that many of the regressions in the literature, based on individual predictor variables, may be spurious.