15We take the expectation of lagged consumption growth because in a small sample with only two shocks, the coefficient estimates from an ordinary least squares (OLS) regression would be severely biased by the tendency of OLS to want to fit the ex-post experience. If we had a large number of shocks in our example, this problem would be smaller, but we wanted to keep the example as simple and ‘toylike’ as possible to aid comprehension. And since our empirical work below will regress consumption on the expectation of lagged consumption, using E t-2ΔCCCt-1 increases the coherence between our exercises on simulated and empirical data.