This article extends the work of Hansen and Jagannathan by showing how to d
ecompose approximation errors in stochastic discount factor models by frequ
ency. This decomposition is applied to a number of consumption-based discou
nt factor models in order to investigate how well they fit at low frequenci
es. There is some evidence of improved fit at low frequencies, but only in
models with high degrees of risk aversion. In models with low degrees of ri
sk aversion, approximation errors at low frequencies are just as severe as
those at high frequencies.