The Bansal and Yaron (2004) model of long run risks (LLR) in aggregate consumption and dividend growth and its extension that captures potential co-integration of the consumption and dividend levels, are tested on a cross-section of asset classes and rejected using annual data over the period 1930-2006 and using both annual and quarterly data over the post-war period. The reversal of earlier empirical conclusions is partly due to the increase in the power of the tests resulting from two observations under the null. First, the latent state variables and, therefore, the pricing kernel are known affine functions of observables such as the interest rate and the market-wide price-dividend ratio. Second, the parameters of the time-series processes of consumption and dividend growth, the LLR variable, and its conditional variance impose constraints on the parameters of the pricing kernel. The value of the persistence parameter of the LRR variable that best fits the data implies that its half-life is shorter than that of the business cycle.
