Examining the relationship between county demogrpahic factors and voting outcomes
Summary of our data
## 'data.frame': 3112 obs. of 82 variables:
## $ X.1 : int 30 31 32 33 34 35 36 37 38 39 ...
## $ X : int 29 30 31 32 33 34 35 36 37 38 ...
## $ combined_fips : int 1001 1003 1005 1007 1009 1011 1013 1015 1017 1019 ...
## $ votes_dem_2016 : int 5908 18409 4848 1874 2150 3530 3716 13197 5763 1524 ...
## $ votes_gop_2016 : int 18110 72780 5431 6733 22808 1139 4891 32803 7803 8809 ...
## $ total_votes_2016 : int 24661 94090 10390 8748 25384 4701 8685 47376 13778 10503 ...
## $ Clinton : num 0.2396 0.1957 0.4666 0.2142 0.0847 ...
## $ Trump : num 0.734 0.774 0.523 0.77 0.899 ...
## $ diff_2016 : int 12202 54371 583 4859 20658 2391 1175 19606 2040 7285 ...
## $ per_point_diff_2016: num -0.4948 -0.5779 -0.0561 -0.5554 -0.8138 ...
## $ state_abbr : chr "AL" "AL" "AL" "AL" ...
## $ county_name : chr "Autauga County" "Baldwin County" "Barbour County" "Bibb County" ...
## $ FIPS : int 1001 1003 1005 1007 1009 1011 1013 1015 1017 1019 ...
## $ total_votes_2012 : int 23909 84988 11459 8391 23980 5318 9483 46240 14562 9761 ...
## $ votes_dem_2012 : int 6354 18329 5873 2200 2961 4058 4367 15500 6853 2126 ...
## $ votes_gop_2012 : int 17366 65772 5539 6131 20741 1250 5081 30272 7596 7494 ...
## $ county_fips : int 1001 1003 1005 1007 1009 1011 1013 1015 1017 1019 ...
## $ state_fips : int 1 1 1 1 1 1 1 1 1 1 ...
## $ Obama : num 0.266 0.216 0.513 0.262 0.123 ...
## $ Romney : num 0.726 0.774 0.483 0.731 0.865 ...
## $ diff_2012 : int 11012 47443 334 3931 17780 2808 714 14772 743 5368 ...
## $ per_point_diff_2012: num -0.4606 -0.5582 0.0291 -0.4685 -0.7415 ...
## $ fips : int 1001 1003 1005 1007 1009 1011 1013 1015 1017 1019 ...
## $ area_name : chr "Autauga County" "Baldwin County" "Barbour County" "Bibb County" ...
## $ state_abbreviation : chr "AL" "AL" "AL" "AL" ...
## $ population2014 : int 55395 200111 26887 22506 57719 10764 20296 115916 34076 26037 ...
## $ population2010 : int 54571 182265 27457 22919 57322 10915 20946 118586 34170 25986 ...
## $ population_change : num 1.5 9.8 -2.1 -1.8 0.7 -1.4 -3.1 -2.3 -0.3 0.2 ...
## $ POP010210 : int 54571 182265 27457 22915 57322 10914 20947 118572 34215 25989 ...
## $ AGE135214 : num 6 5.6 5.7 5.3 6.1 6.3 6.1 5.7 5.9 4.8 ...
## $ AGE295214 : num 25.2 22.2 21.2 21 23.6 21.4 23.6 22.2 21.4 20.4 ...
## $ age65plus : num 13.8 18.7 16.5 14.8 17 14.9 18 16 18.3 20.9 ...
## $ SEX255214 : num 51.4 51.2 46.6 45.9 50.5 45.3 53.6 51.8 52.3 50.2 ...
## $ White : num 0.779 0.871 0.502 0.763 0.96 0.269 0.539 0.758 0.583 0.93 ...
## $ Black : num 0.187 0.096 0.476 0.221 0.018 0.701 0.44 0.211 0.395 0.046 ...
## $ RHI325214 : num 0.5 0.7 0.6 0.4 0.6 0.8 0.4 0.5 0.3 0.5 ...
## $ RHI425214 : num 1.1 0.9 0.5 0.2 0.3 0.3 0.9 0.9 0.8 0.3 ...
## $ RHI525214 : num 0.1 0.1 0.2 0.1 0.1 0.7 0 0.1 0.1 0 ...
## $ RHI625214 : num 1.8 1.6 0.9 0.9 1.2 1.1 0.8 1.7 1.1 1.6 ...
## $ Hispanic : num 0.027 0.046 0.045 0.021 0.087 0.075 0.012 0.035 0.02 0.015 ...
## $ RHI825214 : num 75.6 83 46.6 74.5 87.8 22.1 53.1 72.9 56.8 91.6 ...
## $ POP715213 : num 85 82.1 84.8 86.6 88.7 84.7 94.6 83.6 85.8 90.6 ...
## $ POP645213 : num 1.6 3.6 2.9 1.2 4.3 5.4 0.8 2.4 1.1 0.7 ...
## $ NonEnglish : num 3.5 5.5 5 2.1 7.3 5.2 1.7 4.5 1.3 1.1 ...
## $ Edu_highschool : num 85.6 89.1 73.7 77.5 77 67.8 76.3 78.6 75.1 78.3 ...
## $ Edu_batchelors : num 20.9 27.7 13.4 12.1 12.1 12.5 14 16.1 11.8 12.8 ...
## $ VET605213 : int 5922 19346 2120 1327 4540 636 1497 11385 2691 2174 ...
## $ LFE305213 : num 26.2 25.9 24.6 27.6 33.9 26.9 24 22.5 24.6 26.9 ...
## $ HSG010214 : int 22751 107374 11799 8978 23826 4461 9916 53289 16894 16241 ...
## $ HSG445213 : num 76.8 72.6 67.7 79 81 74.3 70.3 68.7 67.9 76.1 ...
## $ HSG096213 : num 8.3 24.4 10.6 7.3 4.5 8.7 13.3 13.8 11.1 4.6 ...
## $ HSG495213 : int 136200 168600 89200 90500 117100 70600 74700 100600 81200 99400 ...
## $ HSD410213 : int 20071 73283 9200 7091 21108 3741 8235 45196 13722 11656 ...
## $ HSD310213 : num 2.71 2.52 2.66 3.03 2.7 2.73 2.47 2.54 2.46 2.2 ...
## $ Income : int 24571 26766 16829 17427 20730 18628 17403 20828 19291 22030 ...
## $ INC110213 : int 53682 50221 32911 36447 44145 32033 29918 39962 32402 34907 ...
## $ Poverty : num 12.1 13.9 26.7 18.1 15.8 21.6 28.4 21.9 24.1 21.2 ...
## $ BZA010213 : int 817 4871 464 275 660 112 393 2311 515 379 ...
## $ BZA110213 : int 10120 54988 6611 3145 6798 0 5711 34871 6431 3864 ...
## $ BZA115213 : num 2.1 3.7 -5.6 7.5 3.4 0 2.7 0.6 -0.2 5.5 ...
## $ NES010213 : int 2947 16508 1546 1126 3563 470 1095 6352 2354 1560 ...
## $ SBO001207 : int 4067 19035 1667 1385 4458 417 1769 8713 1981 2180 ...
## $ SBO315207 : num 15.2 2.7 0 14.9 0 0 0 7.2 0 0 ...
## $ SBO115207 : num 0 0.4 0 0 0 0 0 0 0 0 ...
## $ SBO215207 : num 1.3 1 0 0 0 0 3.3 1.6 0 0 ...
## $ SBO515207 : num 0 0 0 0 0 0 0 0 0 0 ...
## $ SBO415207 : num 0.7 1.3 0 0 0 0 0 0.5 0 0 ...
## $ SBO015207 : num 31.7 27.3 27 0 23.2 38.8 0 24.7 29.3 14.5 ...
## $ MAN450207 : int 0 1410273 0 0 341544 0 399132 2679991 667283 307439 ...
## $ WTN220207 : int 0 0 0 0 0 0 56712 0 0 62293 ...
## $ RTN130207 : int 598175 2966489 188337 124707 319700 43810 229277 1542981 264650 186321 ...
## $ RTN131207 : int 12003 17166 6334 5804 5622 3995 11326 13678 7620 7613 ...
## $ AFN120207 : int 88157 436955 0 10757 20941 3670 28427 186533 23237 13948 ...
## $ BPS030214 : int 131 1384 8 19 3 1 2 114 8 2 ...
## $ LND110210 : num 594 1590 885 623 645 ...
## $ Density : num 91.8 114.6 31 36.8 88.9 ...
## $ Clinton_Obama : num -0.0262 -0.02 -0.0459 -0.048 -0.0388 ...
## $ Trump_Romney : num 0.008021 -0.000383 0.039339 0.038998 0.03359 ...
## $ Trump_Prediction : num 0.621 0.587 0.518 0.692 0.79 ...
## $ Clinton_Prediction : num 0.34 0.36 0.475 0.286 0.177 ...
## $ Trump_Deviation : num -0.1135 -0.18677 -0.00488 -0.07743 -0.10887 ...
## $ Clinton_Deviation : num 0.10092 0.16385 0.00809 0.07181 0.09279 ...Cleaning our data
## 'data.frame': 3112 obs. of 18 variables:
## $ DemPercentRepPercent: num -0.4948 -0.5779 -0.0561 -0.5554 -0.8138 ...
## $ State : chr "AL" "AL" "AL" "AL" ...
## $ County : chr "Autauga County" "Baldwin County" "Barbour County" "Bibb County" ...
## $ PopulationChange : num 1.5 9.8 -2.1 -1.8 0.7 -1.4 -3.1 -2.3 -0.3 0.2 ...
## $ PercentAgeUnder5 : num 6 5.6 5.7 5.3 6.1 6.3 6.1 5.7 5.9 4.8 ...
## $ PercentAgeUnder18 : num 25.2 22.2 21.2 21 23.6 21.4 23.6 22.2 21.4 20.4 ...
## $ PercentAgeOver65 : num 13.8 18.7 16.5 14.8 17 14.9 18 16 18.3 20.9 ...
## $ MedianHouseIncome : int 53682 50221 32911 36447 44145 32033 29918 39962 32402 34907 ...
## $ NumberFirms : int 4067 19035 1667 1385 4458 417 1769 8713 1981 2180 ...
## $ PercentBlackFirms : num 15.2 2.7 0 14.9 0 0 0 7.2 0 0 ...
## $ PercentNAFirms : num 0 0.4 0 0 0 0 0 0 0 0 ...
## $ PercentAsianFirms : num 1.3 1 0 0 0 0 3.3 1.6 0 0 ...
## $ PercentHawaiianFirms: num 0 0 0 0 0 0 0 0 0 0 ...
## $ PercentHispanicFirms: num 0.7 1.3 0 0 0 0 0 0.5 0 0 ...
## $ PercentWomenFirms : num 31.7 27.3 27 0 23.2 38.8 0 24.7 29.3 14.5 ...
## $ Density : num 91.8 114.6 31 36.8 88.9 ...
## $ PercentHSGrad : num 85.6 89.1 73.7 77.5 77 67.8 76.3 78.6 75.1 78.3 ...
## $ PercentCollegeGrad : num 20.9 27.7 13.4 12.1 12.1 12.5 14 16.1 11.8 12.8 ...Percent Over 65
Percent of Women Firms
Percent of Median House Income
Percent of College Graduate
Hypothesis Test
Null Hypothesis = None of our independent variables contributes significantly to our model
Alternative Hypothesis = At least one of our independent variables contributes significantly to our model
We are going to run a multi-linear regression…
##
## Call:
## lm(formula = DemPercentRepPercent ~ PopulationChange + PercentAgeUnder5 +
## PercentAgeUnder18 + PercentAgeOver65 + MedianHouseIncome +
## PercentBlackFirms + PercentNAFirms + PercentAsianFirms +
## PercentHispanicFirms + PercentWomenFirms + Density + PercentHSGrad +
## PercentCollegeGrad, data = relevant_1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.76022 -0.13809 -0.01807 0.12032 1.11070
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.878e-01 7.229e-02 3.980 7.04e-05 ***
## PopulationChange -1.229e-02 1.092e-03 -11.258 < 2e-16 ***
## PercentAgeUnder5 3.294e-02 6.662e-03 4.945 8.03e-07 ***
## PercentAgeUnder18 -2.616e-02 2.420e-03 -10.808 < 2e-16 ***
## PercentAgeOver65 -1.372e-02 1.233e-03 -11.133 < 2e-16 ***
## MedianHouseIncome -2.826e-06 5.269e-07 -5.363 8.80e-08 ***
## PercentBlackFirms 1.354e-02 5.754e-04 23.527 < 2e-16 ***
## PercentNAFirms 1.518e-02 1.328e-03 11.433 < 2e-16 ***
## PercentAsianFirms 1.411e-02 1.737e-03 8.124 6.43e-16 ***
## PercentHispanicFirms 1.086e-02 6.517e-04 16.662 < 2e-16 ***
## PercentWomenFirms 2.284e-03 3.156e-04 7.236 5.80e-13 ***
## Density 3.426e-06 2.307e-06 1.485 0.138
## PercentHSGrad -3.215e-03 7.970e-04 -4.034 5.62e-05 ***
## PercentCollegeGrad 1.652e-02 7.242e-04 22.807 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.204 on 3098 degrees of freedom
## Multiple R-squared: 0.5654, Adjusted R-squared: 0.5636
## F-statistic: 310 on 13 and 3098 DF, p-value: < 2.2e-16This is our more condensed model
##
## Call:
## lm(formula = DemPercentRepPercent ~ PercentAgeUnder18 + PercentAgeOver65 +
## PercentBlackFirms + PercentNAFirms + PercentHispanicFirms +
## PercentCollegeGrad, data = relevant_1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.67772 -0.14518 -0.02404 0.12196 1.21274
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.2102684 0.0543131 3.871 0.00011 ***
## PercentAgeUnder18 -0.0241880 0.0015322 -15.787 < 2e-16 ***
## PercentAgeOver65 -0.0162217 0.0011926 -13.602 < 2e-16 ***
## PercentBlackFirms 0.0168902 0.0005766 29.291 < 2e-16 ***
## PercentNAFirms 0.0170474 0.0013857 12.302 < 2e-16 ***
## PercentHispanicFirms 0.0130834 0.0006403 20.432 < 2e-16 ***
## PercentCollegeGrad 0.0117063 0.0004757 24.611 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.2163 on 3105 degrees of freedom
## Multiple R-squared: 0.5104, Adjusted R-squared: 0.5095
## F-statistic: 539.5 on 6 and 3105 DF, p-value: < 2.2e-16Extra Findings
Let’s set the sizes of the training and validation data sets.
Let’s partition the data
This looks like a good model.
model2 <- lm(DemPercentRepPercent ~ PopulationChange + PercentAgeUnder5 + PercentAgeUnder18 + PercentAgeOver65 + MedianHouseIncome + PercentBlackFirms + PercentNAFirms + PercentAsianFirms + PercentHispanicFirms + PercentWomenFirms + Density + PercentHSGrad + PercentCollegeGrad, data=trainSet)
summary(model2)##
## Call:
## lm(formula = DemPercentRepPercent ~ PopulationChange + PercentAgeUnder5 +
## PercentAgeUnder18 + PercentAgeOver65 + MedianHouseIncome +
## PercentBlackFirms + PercentNAFirms + PercentAsianFirms +
## PercentHispanicFirms + PercentWomenFirms + Density + PercentHSGrad +
## PercentCollegeGrad, data = trainSet)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.76595 -0.13878 -0.01657 0.12156 1.06828
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.861e-01 8.654e-02 3.306 0.000962 ***
## PopulationChange -1.200e-02 1.281e-03 -9.367 < 2e-16 ***
## PercentAgeUnder5 3.349e-02 7.848e-03 4.267 2.06e-05 ***
## PercentAgeUnder18 -2.514e-02 2.877e-03 -8.736 < 2e-16 ***
## PercentAgeOver65 -1.341e-02 1.491e-03 -8.995 < 2e-16 ***
## MedianHouseIncome -2.941e-06 6.327e-07 -4.648 3.56e-06 ***
## PercentBlackFirms 1.323e-02 6.983e-04 18.953 < 2e-16 ***
## PercentNAFirms 1.365e-02 1.650e-03 8.274 2.24e-16 ***
## PercentAsianFirms 1.236e-02 2.099e-03 5.891 4.45e-09 ***
## PercentHispanicFirms 1.078e-02 7.811e-04 13.802 < 2e-16 ***
## PercentWomenFirms 2.398e-03 3.759e-04 6.379 2.17e-10 ***
## Density 9.196e-06 3.790e-06 2.426 0.015338 *
## PercentHSGrad -3.612e-03 9.556e-04 -3.780 0.000161 ***
## PercentCollegeGrad 1.700e-02 8.561e-04 19.858 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.2035 on 2164 degrees of freedom
## Multiple R-squared: 0.5682, Adjusted R-squared: 0.5656
## F-statistic: 219 on 13 and 2164 DF, p-value: < 2.2e-16Let’s make predictions
Train RMSE
## [1] 0.2028Test RMSE
## [1] 0.2063212Model for a hypothetical county we would expect to vote more Democrat
predict(model2, newdata = data.frame(
PopulationChange = -8,
PercentAgeUnder5 = 8,
PercentAgeUnder18 = 15,
PercentAgeOver65 = 10,
MedianHouseIncome = 35000,
PercentBlackFirms = 15,
PercentNAFirms = 3,
PercentAsianFirms = 10,
PercentHispanicFirms = 15,
PercentWomenFirms = 40,
Density = 4000,
PercentHSGrad = 80,
PercentCollegeGrad = 35
))## 1
## 0.9994977Model for a hypothetical county we would expect to vote more Republican
predict(model2, newdata = data.frame(
PopulationChange = 8,
PercentAgeUnder5 = 2,
PercentAgeUnder18 = 25,
PercentAgeOver65 = 30,
MedianHouseIncome = 100000,
PercentBlackFirms = 0,
PercentNAFirms = 0,
PercentAsianFirms = 0,
PercentHispanicFirms = 0,
PercentWomenFirms = 0,
Density = 1,
PercentHSGrad = 50,
PercentCollegeGrad = 15
))## 1
## -0.9932248
Examining the relationship between county demogrpahic factors and voting outcomes