(ts) Suppose now the reverse. You estimate the model in first differences, but, in fact, the model includes a lagged dependent variable. Calculate the bias of the regression coefficient of $\beta$. If $\theta < 1$ and $\gamma > 0$, what is the sign of the bias?