If the model is $Y_i = \beta_0 + \beta_1 X_i + \epsilon_i$ and the true relationship is $Y_i = \beta_0 + \beta_1 X_i + \beta_2 X_i^2 + \epsilon_i$, what happens to the OLS estimator $\hat{\beta}_1$?