TITLE

States and County Level

As the Financial Times reports, low turnout among low-income voters killed Harris. In NY, IL, FL, TX, OH, and VA, she lost 3.0m compared to 2020. Swing states were less harsh on her (with some gains, ie in WI and GA), but including them, her losses are 3.1m.

Let’s look at all the states.

We can see that the states that saw a more drastic fall than the nation in Democratic presidential votes were CA, NY, IL, FL, NJ, MA, OR, MS, LA, AL, KY, HI, WV, and AK. That is to say, nearly everywhere in the country. Some of these, such as IL and NY, we know urban turnout fall among low-income black and Latino voters was a big factor. Also notably, around 65% of Harris’s net-lost votes were in five states, albeit big ones - California, New York, Illinois, Florida, and Texas. You might say "these aren’t swing states, so it doesn’t matter". But we are trying to pick out trends here.

Overall, the states that were within 5% Dem:GOP margin were NC, NV, GA, PA, MI, WI, NH, MN. This is largely the familiar swing state group. Typically, AZ is considered a swing state, but in this election, the margin was 5.5%. Interestingly, MN looks like a swing state here (NH also tends to swing, but it’s only worth 4 points, so people don’t pay much attention it seems).

Counties

If we break down by counties, we see some interesting patterns. However, there is a big flaw, in that counties, while much smaller than states, are still a large geographic area. So while the city precinct-level data gives a fairly precise idea about how different groups voted, it’s more fuzzy at this level (consider the flaws in the rural/suburban/urban classification, per Pew (higher resolution version of that map here), of counties - which I use here. Yet looking at the map, you probably can find counties familiar to you (if you’re in the US) that don’t seem to fit well with their classification). Or for example, that "median income", the poverty metrics here ("poverty" ~ households below the poverty line, "poverty150" ~ person below 150% of the poverty line), or the Gini coefficient don’t give a precise idea of the income distribution in a county. Further, these aren’t perfect indicators of a person’s class position. The 2019 Gini coefficient of the US overall was 0.4811, whereas per the Federal Reserve, the wealth inequality Gini coefficient for the US in 2019 was 0.852. The US is already highly unequal in terms of income (the bottom 50% representing, roughly, 18.9% of the cumulative income), but it’s far more stark in terms of wealth (the bottom 50% representing, roughly, 4.7% of cumulative wealth). In terms of class (corresponding, roughly, with capacity to invest), the wealth inequality would certainly be a better metric, and a far starker one. But the income inequality data is easier to access, so I’ve gone with that.

Trends based on race are a bit easier to detect, because we have actual composition values for each county (it would be helpful if we had composition values of income groups for states). Race is indicative not only of the particular issues relevant to a racial group, but also income. Nationally, across all racial-groups, 59.1% of households make <$100k, and 31.3% <$50k. For "white alone, not Hispanic" households, it’s 55% and 27.8%. For "black alone" households, its about 74.0% and 44.8%. For "Hispanic (Any Race)" households, 68% and 37%. For "Asian alone" households, 44.0% and 22.6%. For "American Indian and Alaska Native alone" households, its 72.9% and 41.6%. Altogether, to be white or Asian is to be disproportionately higher-income, and otherwise to be disproportionately lower-income, compared to the typical household nationally. Of course, every racial group is highly classed (45% of white households $100k+, black 26%, Latino 32%, Asian 56%, AIAN 27.1%; and, as a specific example, Charles County MD has a relatively low Gini coefficient for a US county (0.3835), a high median income ($116,882), and is nearly majority black (48.5%)), but the huge differences for some groups (especially black households) can also give some indication of income distributions, given the fuzzy picture the income indicators give.

A more granular approach could be undertaken by looking at township-level data (or the relevant subunit for a county in a state). But this would be far more tedious data to collect, so I haven’t done that here. All that said, can we use county data to discern income-based trends? I think so. At least a rough picture.

Just added Utah, New Mexico and Wisconsin, need to comment on still

If lower-income turnout fell and higher-income turnout increased, then we would expect that (1) turnout and median income to positively correlate (higher median income in a county ⇒ higher turnout in a county); (2) turnout and poverty would negatively correlate (more poverty ~ more low income in a county ⇒ lower turnout in a county); (3) turnout and gini coefficient - an index of income inequality in a group - should correlate...? A Lorenz curve falls below the equality line, as wealthier individuals distort the "rise" of the curve. This I’m not sure what should happen, theoretically, although it appears in all the above examples to be negatively correlated. That is, areas with high income inequality have less turnout.

For the most part, the three behave as expected. There are some exceptions. Florida sees the opposite sign correlations for median income and poverty, but the expected correlation for the gini coefficient - although each correlation has a small magnitude. In Michigan, the correlations are all small, poverty close to zero. Likewise for Georgia, with small correlations for median income and gini coefficient (but not for poverty).

This isn’t just a problem of overall turnout though. While Michigan and Georgia see increasing turnout, so do Pennsylvania and South Carolina. Florida overall saw falling turnout.

Issue: don’t precisely know income distribution within counties. So county median income by itself doesn’t necessarily tell the whole story. Poverty rates give a better idea of how many low-income people there are (not all are in poverty, but all impoverished are low-income), so is likely a better indicator. Yet if there are sufficient numbers of reach people, this will balance out. Does gini help here? The correlation for poverty and gini in Michigan is also small, at 0.1500. At the same time, the average county gini score is around 0.43 - typical of American counties - so the impoverished are generally accompanied by higher-income.

It seems reasonable to conclude that Michigan turnout was less affected by income, although precise data is needed, for the above confounding issues.

Another problem is that income isn’t quite the same as class. People don’t change their voting patterns simply as a function of income; instead, that income (and developing wealth) correspond with evolving relations with society (ie increasingly subordinated to debt, perhaps invested in some securities, perhaps buying or selling, with the issues of taxes that appear, etc). In terms of a standard metric, wealth is a better indicator of one’s social relations: an owner or not. For example, the income Gini coefficient in the US is 0.47, but the wealth Gini coefficient is around 0.86! While a Gini coefficient of 0.51 would say that about 50% of the population makes 18.9% of the cumulative income per year, whereas with gini wealth coeff = 0.83 has 4.7% of the total wealth - a percentage factor of nearly 4! That is, not only are income and wealth inequality important to distinguish on their own merit, but there is a huge difference in the inequality betwee the two. However, income inequality, even if far more conservative than wealth inequality, is the more accessible data.

Another interesting observation: In the midwest states and Florida, turnout was more favorable than the state in rural and suburban areas (except PA, where only rural was more favorable). In Arizona, South Carolina, Alabama, only suburban turnout was more favorable than the state. In Georgia, like PA, only rural turnout was more favorable. Urban turnout was the least favorable in all states, falling much faster (or improving much slower) than rural or suburban - except in Arizona, where urban turnout fell faster than suburban turnout, but slower than rural.

In all {state} × {region type} correlations - except for FL rural+suburban, FL urban, MI rural, MI suburban, and PA suburban - does the expected pattern of increasing county income → increasing county turnout hold. Possibly, these exceptional cases are exceptions to the rule, or possibly there are issues masked in the data (and likewise, for the apparent "positive" cases). Not all of these correlation values are even modestly significant, but that is at least the consistency in terms of sign.

In the majority of cases, a correlation with the predicted sign had moderately significant correlation (ie ≥ ~0.3). Only in (setting aside mixtures) FL rural and urban, and MI rural, is this not the case.

Regarding the expected-sign exceptions: for FL rural+suburban, the strongest correlation (with Gini coefficient) is consistent with the prediction. For FL urban, all correlations are inconsistent with the prediction, though it’s worth noting only five counties are classified as urban (this sample size smallness is why values aren’t reported except for OH and FL). At the same time, Floridian urban turnout collapse was enormous, which is interesting alongside these observations (ΔΔ -2.6% compared to FL rural). For MI rural, the strongest correlation (with income) is consistent with the prediction. For MI rural, the strongest correlation (with the Gini coefficient; which is also significant in magnitude), is consistent with the prediction. For PA suburban, the strongest correlation (with income) is inconsistent with the prediction, at -0.3580. A close second however is with Gini coefficient, at -0.3372.

Notably, if Miami-Dade county is removed from the Urban list for Florida, the anomalous income and two poverty correlations persist, albeit weaker (-0.3808, 0.2075, 0.3440 vs overall -0.6530, 0.6460, 0.7988), but the Gini coefficient correlation becomes highly significant in the expected direction (-0.9217 vs 0.7348). In terms of race groups, the trend for white, black, and Asian voters varies (a few away from expectations) (-0.2511, 0.6136, 0.3865 vs -0.5657, 0.0743, -0.7420), but Latino significantly negatively correlates with turnout (-0.4486 vs 0.7343) (AIAN barely changes, with -0.5345 (excluding Miami-Dade) vs -0.5749). Notably, Miami-Dade county (which is comfortably majority Latino, at 68.7%) saw a drop for Harris of 137509 votes, a rise for Trump of 72757 votes, and a loss of overall voters of 64825. It appears plausible Harris loss voters to Trump here, rather than simply losing them to no-vote. Further, around 50.1% of Miami city’s Latinos are Cuban, whereas 25.5% of Latinos in Florida overall are Cuban, and thus are largely over-represented here (based on exit poll analysis, explored in a related article - although considering the exit poll issues in general - all Latinos appear to have swung towards Trump this election, not just Latinos, but that has to be taken with a large grain of salt, given it’s an exit poll). In light of the income-based correlations, it’s plausible (though by no means firmly indicated) that the anomalies in Florida are, to an extent, a result of the specific political interests of Cubans in Miami-Dade.

From 2019, it appears that, for Miami-Dade, "White, not Hispanic" had a median income (poverty level) of $82,099 (9.5%), black $37,839 (24.9%), Hispanic/Latino $49,272 (16.8%), Asian $70,150 (15.0%), American Indian $48,828 (11.8%). For Florida altogether, it’s "White, not Hispanic" $61,682 (10.0%), black $41,702 (22.0%), Hispanic/Latino $49,266 (17.7%), Asian $72,205 (11.8%), American Indian $48,608 (16.6%).

Even these then, should be taken with a grain of salt.

Swing States

If you point out that Harris lost millions of votes across the country, some will retort that swing states actually saw, generally speaking, a rise in turnout. Here’s the NYT (my emphasis):

If you’ve been reading post-election coverage, you’ve probably seen one of the big takeaways from the returns so far: In counties across the country, Kamala Harris won many fewer votes than Joe Biden did four years ago.

...

As such, it’s tempting to conclude that Democrats simply didn’t turn out this year — and that Ms. Harris might have won if they had voted in the numbers they did four years ago.

This interpretation would be a mistake.

For one, the story doesn’t apply to the battlegrounds, where turnout was much higher. In all seven battleground states, Mr. Trump won more votes than Mr. Biden did four years ago.

More important, it is wrong to assume that the voters who stayed home would have backed Ms. Harris. Even if they had been dragged to the polls, it might not have meaningfully helped her.

How is that possible? It’s because the low turnout among traditionally Democratic-leaning groups — especially nonwhite voters — was a reflection of lower support for Ms. Harris: Millions of Democrats soured on their party and stayed home, reluctantly came back to Ms. Harris or even made the leap to Mr. Trump. And if those who stayed home had voted, it wouldn’t have been an enormous help to Ms. Harris, based on Times/Siena polling linked to validated records of who did or didn’t vote.

On the last remark - that soured-Dem voters might have made the election worse: maybe Harris shouldn’t have run a campaign that soured these voters!

But on to the main point.

The idea of "but swing states saw higher turnout" is that, unlike the rest of the country, turnout rose; therefore, where Harris lost votes, maybe (A) those were due to voters switching to the Republicans from the Democrats and/or (B) Dem turnout largely held out, sometimes eroding, sometimes even gaining (which is true, in terms of gross numbers). So then the picture isn’t quite that Harris "lost votes".

There’s clearly some issues in this argument. But the data bears it out quite clearly. In Michigan, Arizona*, Wisconsin, Georgia, Dems saw rising turnout in higher-income counties and falling turnout in counties with increasing poverty. In Pennsylvania, Dems saw falling turnout in counties with higher poverty and inequality, with no accompanying uptick in higher-income counties.

For the GOP, there was no significant trend with economic factors in Michigan, Georgia, Pennsylvania, and Wisconsin. In Arizona, they saw falling turnout with income and rising with poverty and inequality, though Arizona’s correlations should be taken with a grain of salt, as they have only 15 counties (thus fewer "data points").

What this means is that in 2024 "battleground states", voters by these economic factors didn’t change their opinion much on Trump, but generally lower-income voters soured on Harris, and higher-income voters found her more appealing.

In terms of race, the Dems saw rising turnout in whiter counties and falling turnout in blacker counties in Michigan and Georgia, and the same in Pennsylvania except with falling turnout also in more Latino and Asian counties. In Wisconsin, the negative trend was with more Asian counties, and positive with American Indian. In Arizona, turnout rose in whiter, blacker, and more Asian counties, and fell in American Indian counties (again, with the above caveat).

For the GOP, there were no significant trends in Michigan. In Georgia, turnout increased with whiter counties, and fell in blacker and more Asian counties. In Wisconsin, turnout increased with whiter counties, and fell with more Latino and Asian counties. In Pennsylvania, turnout fell with Asian and American Indian voters. In Arizona, they saw the opposite trend as Democrats in the state - falling with whiter, blacker, and more Asian counties, and rising with American Indian counties.

Overall, for Dems, we see a fall in support among non-whites, and a rise with whites. For the GOP, it’s more scattered, but whites appear to have generally voted for them more, and Asians fell off more frequently.

Overall then, it’s the same story we see nationwide. But state-by-state, the pattern is more scattered outside the swing states. In terms of income, the trend is similar in Texas and Ohio. In Florida, Dems saw falling turnout with increasing inequality, and the GOP saw no significant trends. Further outside the trend, in California, it appears the GOP made gains in more impoverished counties, while Democrat turnout fell off in higher income counties.

In terms of race, Democrats fell off with blacks and Latinos in California, but saw correlation with white and American Indian counties. The GOP saw nearly the opposite case. In Texas, there was no overall trend for the GOP, but the Dems saw rising turnout in white counties, and falling turnout in Latino counties. In Florida, its similar to Texas, except the GOP saw near-moderate negative correlation with more Asian counties, and the Dems also saw fall off in black and Asian counties. In Ohio, the Dems turnout was correlated with whiter counties, and negatively correlated with blacker counties, while the GOP saw near-moderate negative correlation with black and Asian counties.

To temper your interpretation of these correlations, let’s consider the change in actual vote share for the parties (change in {total vote for a party divided by the eligible voters}).

First, the swing states. The GOP saw gains of 2.2 pp and the Dems fell 0.7 pp in Michigan. In Georgia, the GOP gained 1.6 pp and the Dems 0.08 pp. In Pennsylvania, the GOP gained 1.7 pp, and the Dems lost 0.3 pp. In Wisconsin, the GOP gained 1.8 pp, and the Dems gained 0.7 pp. In Arizona, the GOP gained 0.9 pp, and the Dems lost 2.5 pp.

For the other states now. In Ohio, the GOP gained 0.3 pp, and Dems lost 1.5 pp. In Florida, the GOP gained 0.9 pp, and the Dems lost 4.8 pp. In Texas, the GOP gained 1.1 pp, and the Dems lost 2.9 pp. In California, the GOP gained 0.7 pp, and the Dems lost 7.5 pp.

Overall then, we see (1) a rise in actual Trump vote, with weaker rise in actual Dem vote (GA, WI), (2) a rise in actual Trump vote, with a weak fall in actual Dem vote (MI, PA), and (3) a rise in actual Trump vote, and a substantially larger fall-off in Dem vote (OH, FL, AZ, TX, CA). The first two patterns seem more common in the swing states. For (3) especially, interpretations of the correlations should be tempered with these varying turnout changes.

What all of the above indicates is that (A) Harris’s alienization of low-income and non-white voters was especially strong in the swing states, and (B) this cost her significant turnout, and likely multiple swing states (consider that Trump’s vote change largely didn’t correlate with economic factors, whereas Harris’s did).

The idea then that "the swing states saw increasing/stable turnout, so let’s ignore the national electoral problems" is an utter mistake. It’s the same basic problem, albeit compressed into less dramatic turnout changes.

Ohio

See here for analysis of Ohio. Particularly, long-term trends and how labor populist Sherrod Brown performed.

Georgia

In terms of racial groups, the GOP turnout was moderately correlated with whiter counties, and moderately negatively correlated with blacker counties (and near moderate negative correlation with more Asian counties). The Dems saw similar trends for white and black counties (albeit less pronounced, but still moderate). In terms of income factors, the GOP didn’t see significant trends. The Dems saw moderate correlation with higher income counties, and moderate negative correlation with counties with higher poverty and more inequality. Overall, the GOP saw a rise in actual turnout of 1.6 pp, and the Dems a rise in actual turnout of 0.1 pp.

Florida

Overall, in terms of race, the GOP didn’t see significant trends, except a near-moderate negative correlation with more Asian counties. By contrast, the Dems saw moderate correlation with whiter counties, and moderate negative correlation with black, Latino, and Asian counties. In terms of income, the GOP didn’t see any significant trends, and the Dems saw falling turnout with increasingly unequal counties. Overall, the GOP saw a rise in actual turnout of 0.9 pp, and the Dems saw a fall in actual turnout of 4.8 pp.

Texas

These results are interesting. In terms of race, there aren’t any state-wide changes for the GOP. But for the Dems, whiter counties saw increasing turnout, and more Latino counties saw decreasing turnout (weak correlation for other racial groups). In terms of income factors, the GOP didn’t see significant trends. The Dems, however, saw modest positive correlation with income, modest negative correlation with the poverty metrics. Overall, the GOP saw an increase in actual turnout by 1.1 pp, and the Dems, saw a fall in actual turnout by 2.9 pp.

California

In terms of racial groups, the GOP saw falling turnout with white voters, near-moderate gains with black voters, modest gains with Latino voters. By contrast, the Dems saw significant gains with whiter and American Indian counties (though the latter is a small group), while seeing moderate losses with black, Latino, and Asian voters. In terms of income, Republicans see moderate gains with counties with more poverty, and near-moderate gains in counties with higher income inequality. By contrast, it appears Dems saw moderate losses in counties with increasing income. Overall, the GOP saw a rise in actual turnout of 0.7 pp, and the Dems a fall of 7.5 pp.

Pennsylvania

In terms of racial groups, the GOP saw near-moderate falling turnout with Asian and AIAN voters. The Dems saw moderate rising turnout in whiter counties, moderate falling turnout in blacker, more Latino, and more Asian counties. In terms of income factors, the GOP saw no significant trends. The Dems saw near-moderate falling turnout in counties with higher poverty and inequality. Overall, GOP actual turnout rose by 1.7 pp, and the Dems actual turnout fell by 0.3 pp.

Wisconsin

In terms of racial groups, the GOP saw moderate increasing turnout with whiter counties, moderate falling turnout with more Latino and Asian counties. The Dems saw little significant trends, other than moderate increasing with more Asian counties and near-moderate falling turnout with more American Indian counties. In terms of income, the GOP saw no significant trends, while the Dems saw moderate rising turnout in higher income counties, moderate falling turnout in counties with increasing poverty. The GOP saw a rise in actual turnout of 1.8 pp, and the Dems a rise in actual turnout of 0.7 pp.

Arizona

These results should perhaps be taken with a grain of salt, as there are only a total of 15 counties in Arizona. Nevertheless... In terms of racial groups, the GOP saw moderate falling turnout with whiter, blacker, and more Asian counties, while seeing moderate rising turnout with American Indian counties. For the Dems, turnout strongly increased with whiter counties, near-moderately rose with blacker and more Asian counties, and moderately fell with American Indian counties. The GOP saw moderate falling turnout with higher income, and moderate increasing turnout with increasing poverty and inequality. It’s the inverse for the Dems. Overall, the GOP saw an increase in actual turnout by 0.9 pp, and the Dems saw a fall in actual turnout of 2.5 pp.

Kentucky

In terms of racial groups, the GOP saw not significant trends. The Dems saw moderate increases in whiter counties, and moderate decreases in blacker counties, and near-moderate decreases in more Asian counties. In terms of economic factors, the GOP saw no significant trends, and neither did the Dems. Overall, actual GOP turnout rose by 0.1 pp, and actual Dem turnout fell 2.1 pp.

Edit this:

What happens if we just look at income correlation w turnout changes though?

For example, So far I’ve only looked at Michigan and South Carolina (they have the advantage of having their 2024 county-by-county results in nice Wikipedia tables, easy to copy/paste into a spreadsheet; I may add more analysis of other states as the vote counts are completed in total elsewhere). I’ll look at these by type (rural vs urban) and by income (ofc, there are other factors worth looking at in future analysis). As we’ll see with Michigan, this isn’t a surefire check, since counties themselves may have income-differentiated behavior (certainly as we’ve seen, this is the case in cities). Even where an apparent correlation is found though, we should be cautious, as counties are still classed. However, if the class stratification within counties is less than the stratification across counties, then a county-by-county analysis will be useful to check the trends predicted in this article.

In Michigan, turnout overall rose - the values used here give 70.0% and 71.8%, or +1.8 pp (these may differ from elsewhere, as I used the 2018-2022 18+ percentage for 2016, 2020, and 2024 to estimate eligible voter population, as a quick way to estimate such for counties. Just to get a rough idea of voter trends). In all counties, turnout rose +2.2 ± 1.9 pp (<0 pp in 3.6% of counties; <1.76 pp in 32.1% of counties); in rural counties, it rose +2.1 ± 2.3 pp (<0 pp in 5.8% of counties; <1.76 pp in 32.7%); in urban counties, it rose 2.4 ± 1.0 pp (<0 pp in 0% of counties; <1.76 pp in 32.3% of counties). Adding the populations and votes of the rural counties altogether, 2020 → 2024 turnout rose by 2.62 pp; for urban, 2020 → 2024 rose by 1.61 pp.

If we correlate their change in turnout from 2020 to 2024, we see for Michigan an overall correlation coefficient of 0.10 (0 being no correlation, 1 being full correlation, -1 being anti-correlation). This isn’t that remarkable. If we break down by rural vs suburban+urban (the only "urban" county here being Wayne), we get a respective correlation of 0.13 and -0.05. The latter seems to imply that turnout falls with higher income, but both of these values aren’t really that significant.

It’s possible however, as initially suggested, that class stratification within counties is comparable to stratification across counties. (Certainly, the exit poll data in the swing state analysis section seems to indicate lower <$50k/$100k turnout, higher $50k+/$100k+ turnout given the sampling, but as I’ve repeated a lot, exit polls are prone to error, and we want verifications outside of them). For example, in the Detroit section above, it was reported Detroit had a lower turnout than expected, although it appears Wayne county (where Detroit is located, along with the Muslim locales discussed above (Hamtramck, Dearborn, and Dearborn Heights)) saw an overall increase in turnout of 0.25 pp. If this is any indication, Michigan county-by-county behavior may not be representative of the voter trends discussed in this article.

This could be addressed by looking at turnout in sub-county units (ie townships), although that sounds a bit tedious for the moment.

In South Carolina, the results are much more stark. County-by-county, the average turnout change was -1.6 ± 4.9 pp (falling in 68% of counties); for urban, -0.5 ± 2.8 pp (falling in 62% of counties); for rural, -2.5 ± 5.9 pp (falling in 76% of counties) (standard deviation, not error of mean, reported here). For all urban counties eligible populations and votes added together, turnout rose 0.27 pp from 2020 → 2024. For rural, turnout fell 0.45 pp. Overall, the correlation with increased turnout and median county household income was 0.60. For rural counties, it was 0.67, and for urban counties, it was 0.46. In other words, the more income, the more turnout here.

Looking at Florida, I noticed a few interesting patterns.

TO DO: include links to sources of data used here, etc

What if we look at black [very super] majority counties? Below, I’ve computed change in votes for counties that are 70%+ black, along with their median income and poverty rate (as per the Census Office).

While a more thorough county-by-county analysis of each state in question is needed for a better conclusion, it’s clear that (A) black very-super-majority counties are very poor and (B) that turnout fell here far more dramatically than the states they fall within (with only a couple exceptions, all in Louisiana, and these were barely exceptions; also all except Clayton County in Georgia fall in the <$50k household income bracket). This seems consistent with the argument of this article.

The above table seems to imply that $100k+ turnout may have actually decreased, just slower than the <$100k turnout. (Although I need to look at overall turnout for their respective states for full picture; and still, counties aren’t perfect indicators). However, it’s worth noting that in those places where ΔTotal is worse than the state’s trend, such counties have a disproportionately large Asian population - and as we saw here, there was probably decline in Asian turnout across income groups. So if they are over-represented, it might bring down $100k+ turnout more than otherwise. However, it isn’t always the case that Asian over-representation leads to worse trends than the state. That said, further analysis is required.

I’ve also included results for Charles County, MD, as this is a nearly majority black county with high income. Notice, contra the declining black turnout in southern very-super-majority black counties (which are all low income), and the low-income black turnout in cities discussed above, that we see rising turnout overall, and more votes for Harris (and a decline for Trump). This seems to highlight the classed behavior among black people. I’ve included Napa, Santa Cruz, and Ventura counties (CA) for similar reasons, as they are high-income with a large share of Latinos (but not near majority, ranging from 34.8% - 43.3%). They trend slightly better for turnout than California as a whole.