As Democrat Doug Jones looks to shock the world by winning the special election for U.S. Senate in deep-red Alabama, let’s take a look at the places that may help win him the race today.
If Jones wants to win, he’ll have to win over a decent share of voters who have voted for Republican candidates in presidential elections. Alabama voted for President Trump by a massive 28-point margin 2016, and voted for Romney by a similarly large 22-point margin 2012. However, Republican Roy Moore only won the state by a 4-point margin against Democrat Robert Vance when he ran for Alabama Supreme Court Chief Justice in 2012. This is especially notable given Alabama’s inelasticity in elections due to racial polarization — there just aren’t that many persuadable voters in the state.
Let’s take a look at the counties in Alabama where Moore most underperformed Romney in 2012. Using election data from the Alabama Secretary of State website, I created a spreadsheet with county margins of victory for Moore in 2012, Romney in 2012, and Trump in 2016.
We can analyze a few trends in this data.
First, there was a strong inverse relationship (r²=.495) between percent black and Romney’s overperformance relative to Moore. The more African-American a county was, the less Romney outperformed Moore. This makes sense intuitively— Moore had fewer voters to lose in heavily black areas where Romney was already doing pretty badly.
Second, there was a positive relationship between Romney’s overperformance relative to Trump and Romney’s overperformance relative to Moore when controlling for race. On the surface, there appears to be no relationship (r²=.006) between Romney-Trump and Romney-Moore. However, when we subdivide the county data into segments based on percent black, a clearer picture emerges. Among >50% African-American counties, there was a strong positive relationship (r²=.512) between Romney-Trump and Romney-Moore. Among 40–50% African-American counties, there was also a strong positive relationship (r²=.499). The r² values for 30–40%, 20–30%, 10–20%, and 0–10% African-American counties were .492, .396, .323, and .110, respectively. Put another way, Moore tended to lose Romney voters in the same areas where Trump lost Romney voters.
Third, there was a very strong positive relationship (r²=.548) between median household income and Romney’s overperformance relative to Moore. Richer counties were more likely to have Romney-Vance crossover vote than poorer counties. Much of this is because of race (white counties tend to be richer than black counties in Alabama), but there’s still a decent relationship between income and Moore underperformance when controlling for race. The r² values for 50–100%, 40–50%, 30–40%, 20–30%, 10–20%, and 0–10% African American counties were .651, .237, .734, .269, .326, and .308.
On election night, Jones has the most potential to make inroads in areas that have either (1) a large share of white voters (2) a large share of Romney/Clinton voters or (3) a large share of affluent voters. We can expect Jones to massively overperform Obama/Clinton in counties like Shelby County, where Moore underperformed Romney’s margin by 29 points (compared to 19 points overall). Shelby County is 84.4% white, swung 6 points towards Clinton in 2016, and at $70,879 has the highest median household income in the state. Jones probably has to overperform Trump’s margin in Shelby County by about 25–30 points to win statewide.
At a more granular level, we can expect certain high-income white precincts to shift strongly to the left with Moore in the race. One proxy to identify these areas is to look at the DecisionDeskHQ National Precinct Map and examine the dark blue precincts on the ‘Estimated Margin Swing, 2012–16’ map. Many of these precincts are in the southern Birmingham suburbs in Jefferson County. Places like Vestavia Hills (median income $84,854), Mountain Brook (median income $126,534), and Homewood (median income $61,626) may see some of the most dramatic shifts to Jones in the state.
On the contrary, Jones may only overperform Trump’s margin by 0–10 points in Black Belt counties like Macon, Greene, Dallas, Lowndes, and Perry and still be on track to win statewide. Since these counties already vote so overwhelmingly Democratic, there’s less for Jones to gain here.
One interesting tidbit is that Etowah County (Moore’s home county and home of the infamous Gadsden Mall) was the county where Moore most underperformed Romney compared to what we’d expect given its income. Given Etowah County’s median household income of $42,145, we’d expect Moore to have underperformed Romney by 16 points. However, Moore actually underperformed Romney here by 29 points. This suggests that Moore actually suffered from a home field disadvantage. Perhaps voters near his home base grew weary of the negative attention generated by his controversies. We’ll have to see if this same dynamic plays out in 2017.
Note: This analysis is mine only and does not reflect the views of Google or any other company / organization.