Here we take a look at the four of the most predictive factors based on our Random Forest Analysis
Our analysis found that there were four primary, and 39 secondary, factors that influnced the outcome of our Random Forest machine model. They are presented one by one below for consideration.
Reported Ancestry as 'American'
Number of Individuals in District who reported their ancestry as ‘American’ vs. District, with color coding showing which party won that district in 2018 - Sources: US Census Bureau & FEC
Our analysis of the random forest procedure results found that this self-reported identifier was extremely closely aligned with districts that voted Republican in the 2018 Congressional Election. This may be a result of nationalistic tendencies among this group which closely track with the messaging of that party.
Number of People Who Commute Using Public Transportaion
Number of Individuals in District who reported that they commute to work using public transportation vs. District, with color coding showing which party won that district in 2018 - Sources: US Census Bureau & FEC
Our random forest analysis found that this factor, the method of commuting was a large predictor in determining voting outcome in the 2018 Congressional Election, in this case it strongly predicted Democratic victories. We suggest that this is essentially a proxy for urban areas, which historically are strongly diverse and lean Democratic.
Employed in Mining, Quarrying, and Oil/Gas Extration
Number of Employees in District employed in Mining, Quarrying, and Oil/Gas Extraction vs. District, with color coding showing which party won that district in 2018 - Sources: US Census Bureau & FEC
Our random forest analysis also identified this particular employment sector as important in determining the outcome of the 2018 Congressional Election. We speculate that the Republican party did well with voters in these districts partly because of this employment sectors representation in the constituency, however there are some obvious outliers that do not fit this trend.
Number of Houses valued less than $50,000
Number of Households in District valued less than $50,00 vs. District, with color coding showing which party won that district in 2018 - Sources: US Census Bureau & FEC
This was the final strongly correlated outcome of our random forest analysis procedure. It, like mining and ancestry strongly correlated with Republican victories.