The Aggregate Impact of Explanatory Variables in Logit and Linear Probability Models