(2) (10%) Suppose the true wage equation was $y_i = \beta_1 + \beta_2 x_{2,i} + \beta_3 x_{3,i} + \varepsilon_i$, where individual i's wage (y) is a linear function of education ($x_2$) and ability ($x_3$). $\varepsilon_i$ is white noise. Normally, "ability" is a variable that can hardly be qualified, so Peter estimates a false wage equation as $y_i = \alpha_1 + \alpha_2 x_{2,i} + v_i$, where $v_i$ is assumed to be white noise. Assuming that "ability" is positively correlated with wage and with education, then, is Peter's estimated $\hat{\alpha}_2$ unbiased, positively biased, or negatively biased? Why?