A Revision of the Regression Model
Given the strong relationship between proximity to Mexico and high population density with underestimated values of foreign born citizens and not citizens alike using Model 1, a revision is necessary. Accounting for proximity to Mexico represents a great challenge, but accounting for population density is manageable. The perpov variable was dropped because of its low significance in Model 1. The revised Model utilized is:
Model 2: %foreign born = α + β %doc + β %highs + β pcinc + β emp + β farm + β manuf + β popdens + β rec(For explanation of variables see Appendix A)
The regression coefficients are:
If we analyze the coefficients of the statistically significant variables, we see again that the coefficient magnitudes are different for citizens and for not citizens. The farming dummy variable is especially different. The difference between the amount of variation that Model 2 explains for citizens (43%) and not citizens (36%) is much larger. Adding population density increased the effectiveness of the model to predict foreign born citizens percentage much more than foreign born not citizens percentage.