Pandemic, BLM, and the 2020 Presidential Election

Over the last twelve weeks, I have looked at the various swing states in the presidential election, breaking them down by subsections of Democratic vote share (or Republican vote share). I examined what factors drove the differing outcomes in each of these states, where Joe Biden and Donald Trump picked up ground from the 2016 election (or lost it), and what it could possibly mean for each of their respective parties moving forward.

This article will summarize general trends seen in these swing states and the country as a whole. I should note ahead of time that this will not be the last 2020 election article I write. There are numerous questions I want to explore that deserve their own articles. But this article will explore two things.

First, it will serve as a general national overview of the 2020 election and the swing state series. While I will make additional articles examining the election results, this one will be where I place my thoughts on the race as a whole while applying the subsections I laid out to the entire country. From this, we can draw meaningful comparisons between the results in the swing states discussed in the twelve-part series with the United States at large.

But aside from capping off the swing state series, I will answer a few specific research questions regarding the 2020 election. Namely, how did the COVID-19 pandemic and accompanying economic recession affect the 2020 presidential election, if at all? What effects, if any, did the state reopening plans have on the election results? And how did the level of Black Lives Matter protest activity affect the election results, if at all? These were questions I explored somewhat in the swing state series, but they will take center stage in this article as I think they can help us explain how these “once-in-a-lifetime” catalysts influenced the election results (or perhaps didn’t).

First, let’s take a broad scope of the national subsections.

Solidly Democratic Counties

Based on the results of the 2016 presidential election, there are 274 solidly Democratic counties in the United States. These were counties where Hillary Clinton received at least 55 percent of the vote. In this case, she received 67.1 percent of the vote in this subsection, slightly up from the 67 percent President Obama received in 2012 and a net gain of over 1.6 million votes. While they are generally scattered across the country, there are notable concentrations in the northeast, California, and the South. This is consistent with conventional wisdom, given that the northeast and California primarily consists of safe Democratic states, while the South has vast pockets of rural, predominantly black counties that are solidly Democratic.

In 2020, this subsection accounted for 33.5 percent of all ballots cast nationwide. Joe Biden won 272 of these counties, only losing Frio and Zapata, both rural, predominantly Hispanic counties in southern Texas. In total, Biden received 68.4 percent of the vote in this subsection, translating to a net gain of about 5.2 million votes from Clinton’s total. On average, Biden added about 1.6 percentage points onto Clinton’s vote share in each county, although he did underperform Clinton in 90 counties. Meanwhile, 15 counties moved outside the solidly Democratic subsection following the 2020 election, most of which are in the South.

As of Election Day, there have been 3.3 million COVID cases in this subsection, setting the incidence rate at 2.9 percent (slightly above the national average). Still, there is substantial variation in the subsection, ranging from an incidence rate of just 0.09 percent in Kauai County, Hawaii to as high as 14.1 percent in Buffalo County, South Dakota (and the eighth highest in the country). On the other hand, there have been 107,778 deaths in this subsection as of Election Day, setting the fatality rate at 3.3 percent. And for unemployment, this subsection followed the rest of the country, going from 4.4 percent in March to 14.8 percent in April, a 10.4-point increase. Similar to the incidence rate, the unemployment change varies widely in this subsection, from a 2.9-point decrease in East Carroll Parish, Louisiana to a 32.4-point increase in Maui County, Hawaii (the highest increase in the entire country). While it has recovered somewhat, the unemployment rate was still at 9.5 percent as of September. Finally, there have been 6,650 BLM protests in this subsection between Memorial Day and Election Day.

Swing Counties

Based on the results of the 2016 presidential election, there are 278 swing counties in the United States. These were counties where Hillary Clinton received between 45 percent and 55 percent of the vote. In this case, she won 209 counties and received 50.5 percent of the vote in this subsection. While this is technically a net gain of 241,446 votes, this is down from the 243 counties President Obama won in 2012 and the 52.6 percent of the vote that he received. As shown on the map, there is a more even distribution of swing counties throughout the country, with a particularly notable presence in the Midwest (also known as the “Blue Wall”).

Now in 2020, this subsection accounted for 23.1 percent of all ballots cast and Biden’s performance was interesting. On one hand, he only won 228 counties, which is less than what Obama won in 2012. But on the other hand, Biden came away with 54.4 percent of the vote, an improvement over President Obama and a net gain of 4.1 million votes over Hillary Clinton. Following the 2020 election, 99 counties have moved out of the swing classification, with 88 counties moving into the solidly Democratic column and 11 moving into the solidly Republican column. Regarding the counties that moved into the solidly Democratic column, they counties are fairly spread out throughout the country and 47 were classified as solidly Democratic following 2012, indicating that their status as swing counties for this election was largely a byproduct of Clinton’s underperformance in 2016. Meanwhile, of the counties that moved into the solidly Republican column, almost all of them are in the South and 5 of them were classified as solidly Democratic following 2012 (and none of them were classified as solidly Republican), indicating that these areas have shifted dramatically since Donald Trump became the face of the GOP. But for the most part, the overarching story with the swing counties is that Biden consolidated support in counties that were previously Democratic while ceding some ground to counties that were particularly receptive to Trump’s rise in the national political landscape.

As of Election Day, there have been over 1.9 million COVID cases in this subsection, setting the incidence rate at 2.7 percent (slightly below the national average). There is also substantial variation in this subsection, but not as drastic as that of the solidly Democratic counties, with the lowest being 0.1 percent in Orange County, Vermont and the highest being 10.7 percent in Big Horn County, Montana. On the other hand, there have been 47,974 deaths in this subsection as of Election Day, setting the fatality rate at 2.4 percent. And for unemployment, this subsection followed the rest of the country, going from 4.7 percent in March to 15.3 percent in April, a 10.6-point increase. Similar to the incidence rate, the unemployment change varies widely in this subsection, from a 3-point decrease in Tensas Parish, Louisiana to a 28.1-point increase in Atlantic County, New Jersey (the fourth increase in the country). While it has recovered somewhat, the unemployment rate was still at 7.9 percent as of September. Finally, there have been 3,612 BLM protests in this subsection between Memorial Day and Election Day.

Obama-Trump Counties

Based on the results of the 2016 election, there are 222 Obama-Trump counties in the United States. These are counties that voted for President Obama in 2012, only to flip to Donald Trump in 2016. In 2012, President Obama won this subsection with 52.5 percent of the vote. In 2016, however, Clinton lost all of these counties and only received 43.4 percent of the vote, translating to a net loss of 638,504 votes. On average, Clinton lost about 12.1 percent in each county. While these counties are scattered throughout the country, there are noticeable pockets in the Midwest as well as rural stretches of the Northeast, South, and West. Furthermore, 52 of these counties overlap with the swing classification.

Now, in 2020, this subsection accounted for only 5.7 percent of all ballots cast. Even so, Biden only won back 32 of these counties and came away with just 45.7 percent of the vote. On average, Biden only increased Clinton’s vote share by 1.6 percentage points while President Trump increased his own 2016 vote share by about 1.9 percentage points. Furthermore, 22 of the counties Biden flipped overlapped with the swing classification, indicating that Biden found the most success in counties that were already competitive, versus those that Trump swung heavily into this favor. Generally speaking, the counties that Biden won were relatively large. 17 of the counties he won are in the top 35 for population in the subsection, all of which have a population of over 100,000. But despite minor improvements, Trump largely held his ground in this subsection, retaining much of his gains from 2016.

As of Election Day, there have been over 420,587 COVID cases in this subsection, setting the incidence rate at 2.4 percent (slightly below the national average). There is also substantial variation in this subsection, but not as drastic as that of the solidly Democratic counties, with the lowest being 0.1 percent in Aroostook County, Maine and the highest being 9.5 percent in Roosevelt County, Montana. On the other hand, there have been 11,962 deaths in this subsection as of Election Day, setting the fatality rate at 2.8 percent. And for unemployment, this subsection followed the rest of the country, going from 4.6 percent in March to 15.8 percent in April, a 11.2-point increase. Similar to the incidence rate, the unemployment change varies widely in this subsection, from a 1.1-point decrease in Hidalgo County, New Mexico to a 25.8-point increase in Manistee County, Michigan (the fourth increase in the country). While it has recovered somewhat, the unemployment rate was still at 6.7 percent as of September. Finally, there have been 838 BLM protests in this subsection between Memorial Day and Election Day.

Now that we’ve established these broad subsections, let’s dive a bit deeper into the data.

BLM Protest Activity

Throughout my swing state series, I have discussed possible factors that can contribute to the level of Black Lives Matter protest activity at the county level. While I mentioned factors such as whether the county is predominantly urban or rural as well as its voting behavior, a central component of my discussion centered around the prevalence of three demographic groups: black residents, college-educated residents, and young adult residents. I identified these groups as the most likely to organize or participate in BLM protests; however until now, I have not tested this assertion using data.

Several findings emerge from these regressions. One is that the share of black residents is negatively associated with BLM protest activity across all the models. While there is a slightly positive correlation (r=0.103) between these variables, this negative relationship holds nonetheless. One reason for this could be that while blacks live in urban counties, there are also many blacks (particularly in the South) that live in mostly rural counties, which naturally see less protest activity. Another reason could be that black people complete college at a lower rate than white people, which not only contributes to disparities in socioeconomic outcomes, but also contributes to disparities in civic engagement, such as voting and participating in protests. Even so, the models indicate that despite these factors, counties with larger black populations still experience more protest activity than counties with larger white populations. To that end, race plays somewhat of a factor.

Across all five models, the most powerful indicator of protest activity is educational attainment, particularly the share of college graduates. This makes sense, given that college-educated people are more likely to be exposed to social issues that motivate protest activity, such as racial disparities in police brutality, through the courses they take, the people with which they interact, and the prevalence of student organizations dedicated to civic engagement and political activism. Even if one doesn’t participate in protests while in college, their awareness of social issues and high level of civic engagement (relative to the rest of the population) carries with them beyond campuses, which can lead to activism after graduation. Now it should be noted that the share of the population without a high school diploma is also positively associated with protest activity, but across all the models, the coefficient is smaller than that of the share of college graduates.

Remarkably, the share of young adults is not statistically significant across any of the regressions and the magnitude of its coefficient varies widely depending on the control variables. A key reason for this, however, is that there is not substantial variation in the share of young adults across counties. While there are some pockets here or there, such as counties anchored around large universities, these are not substantial enough to formulate a statistically significant relationship.

Aside from these factors, there are a few other variables worth mentioning. First, both the county’s Clinton and Biden vote shares are positively associated with the level of protest activity. This makes sense, given that Black Lives Matter is predominantly a movement centered within the Democratic Party, meaning that Democratic strongholds are reasonable contenders for protest activity. And on a similar note, protest activity is higher in counties considered metropolitan, which tend to lean Democratic.

To this end, we can establish a clearer picture on the factors that drive BLM protest activity. Namely, the biggest drivers of protest activity are educational attainment, Democratic leaning, and urbanicity.

Swing State Overview

My swing state series revealed some interesting trends within each individual state for the 2020 election. Some of these trends reflect idiosyncrasies within these states, while others are more representative of the country as a whole. One of the most enduring trends found in the series was a widening education gap, where all else being equal, counties with more college graduates consolidated support behind Biden while counties with lower educational moved in the opposite direction. This isn’t too surprising, given that it was also a key factor in the 2016 election. This subject deserves further discussion in a separate article, but for the sake of this article, it’s important to understand that it played an important factor in this election as well.

Aside from the education gap, there are other trends observed to differing degrees. For example, some states saw a widening generation gap, with younger voters shifting farther into the Democratic column while older voters tended to retain their support with President Trump; however, this gap isn’t as prevalent or as two-sided in all states. There was also evidence of a growing urban-rural divide, where Biden consolidated support in major cities and the suburbs, while President Trump maintained support in more rural, “working class” counties which he carried in 2016. And race/ethnicity played an interesting role as well, with some states seeing Trump making gains in predominantly Hispanic counties (such as Texas and Florida). But in most cases, these drastic shifts were largely concentrated to specific states rather than being a broader development.

Now, with regard to the main variables in question, the results were largely mixed in the swing states. The main unifying trend is that the number of BLM protests had a minimal effect on both the county-level Biden vote share and the Democratic vote share change. Aside from that, the effects of the pandemic on the election results in the swing states varied quite a bit.

The tables below provide a general summary of the relationships found for different variables in the regression results for states reviewed in the swing state series. Arizona and Nevada were excluded from these tables, given that no regressions were run for those states. As a rule of thumb, “positive” means there was at least one regression model that produced a statistically significant positive coefficient for that variable for the state in question (same thing for “negative” in the opposite direction), while “none” means that there was either no statistically significant coefficient for that variable in any of the regression models or there were statistically significant coefficients in conflicting directions (e.g. one positive and one negative).

There was no evidence found in the swing states that the COVID incidence rate had a positive effect on Biden’s vote share at the county-level. In fact, for most swing states, the opposite was true, where Trump’s vote share tended to be higher in the pandemic hot spots. This is consistent with the results of my pandemic study, where I found that while the hot spots for COVID incidence were in the major cities during the early months, they shifted towards more rural counties as Election Day drew closer, which tended to support Trump at higher levels. And by Election Day, the political debate on the pandemic had turned into an ultimatum between containing the virus at the expense of the economy or allowing businesses to remain open with minimal restrictions, even if cases and deaths go up (as if these goals were mutually exclusive).

Furthermore, the months separating the start of the pandemic and Election Day gave voters and political leaders enough time to form new narratives that reinforce pre-existing political beliefs, rather than allowing the emergency to be a catalyst for reassessment. For those on the right, the pandemic was an unavoidable disaster (or for some, a “Chinese plot”) that President Trump handled as well as any reasonable leader could have, while the left would be committed to crippling the economy and limiting freedom through unnecessary mandates. Meanwhile for those on the left, Trump’s handling of the pandemic was yet another example of his incompetent leadership, through his failure to prepare the country before the virus hit, his downplaying of its threat to “avoid panic”, and his efforts to politicize essential safety protocols such as mask wearing. In both cases, the pandemic went from being a “once in a lifetime” opportunity to re-evaluate our values and priorities into ammunition for each side to further entrench themselves into their efforts to “fight the good fight”.

But while the medical effects of the pandemic seem to favor Trump, there’s mixed evidence that the economic effects favor Biden in the swing states. In four swing states, the unemployment rate increase from March to April is positively associated with Biden’s vote share. Of these four, Biden won two (Georgia and Wisconsin) while Trump won two (Florida and Texas). On some level, this reflects how unemployment tended to spike more in cities or in areas that are dependent on heavily-hit sectors (such as manufacturing and tourism), a grouping that has more Democratic-friendly territory than the counties with the highest COVID incidence rate as of Election Day. In addition, though, this economic shock reflects the dual nature of the crisis to those living in these counties, as well as the government’s failure to prevent it.

With that said, though, none of the swing states where Biden’s vote share has a positive relationship with the unemployment increase have a positive relationship with the September unemployment rate. Two of them have a negative relationship with the September unemployment rate (Biden carried neither of them) while two have no relationship in either direction (Biden carried both). Meanwhile, three swing states overall have a positive relationship between Biden’s vote share and the September unemployment rate. Interestingly, all three of these states are in the Midwest; however, Biden only carried one of them (Minnesota) while the two losses weren’t exactly nail-biters.

Below is the table for the relationships on the change in Democratic vote share.

This table solidifies how little the pandemic seemed to have on the swing states, with many of them having no statistically significant impact for any of the variables. Only two states have a statistically significant relationship with the COVID incidence, but both of these are negative, indicating that counties with a higher incidence rate shifted more towards Trump relative to 2016. This, coupled with a statistically significant negative relationship with the Biden vote share in those two states indicates a compounding effect that hurt Biden. Meanwhile, the other states do not have a statistically significant relationship, meaning we cannot definitively conclude that the incidence rate changed much of the electorate relative to four years ago.

Economically, four states have a positive relationship between the change in Democratic vote share and the unemployment rate increase from March to April, three of which also enjoy a positive relationship with Biden’s vote share (Florida, Texas, and Wisconsin). But on the other hand, five states have a negative relationship with the September unemployment rate. This presents a more mixed view of how the economic factors impacted Biden. On one hand, it seems that Biden benefitted relative to Clinton in counties that were hit hardest in the early months of the pandemic, which would appear to run counter to the idea that voters more concerned about the economy than the virus would gravitate towards Trump. But on the other hand, Trump improved from his 2016 performance in counties with a higher September unemployment rate, which is more consistent with that statement. In this case, it seems that while Biden benefitted somewhat from anti-incumbent sentiment that recessions tend to foster, it was Trump whose economic message resonated more in counties that either suffered a sluggish recovery from the early months of the pandemic or those that are naturally economically depressed due to poor fundamentals (i.e. already had a high unemployment rate even before the pandemic).

Overall, the evidence indicates that Biden performed poorly in counties with higher COVID incidence rates (both in absolute terms and relative to Hillary Clinton), meaning that the medical impact of the pandemic did not benefit him. On the other hand, there is some evidence that Biden benefitted more in counties that experienced large increases in unemployment during the early months of the pandemic; however, this benefit didn’t carry over as much in counties where the unemployment rate remained high in September.

But more generally, there are many swing states where one or more of these variables did not significantly change the Democratic vote share relative to 2016 at the county-level. This finding undercuts the idea that the pandemic’s impact compelled voters to switch their support. If this were the case, we would see Biden make serious inroads in solidly Republican (or even Obama-Trump) counties that were battered by the pandemic. Conversely, President Trump’s may have persuaded more voters in solidly Democratic and swing counties that his leadership during the crisis was more effective than anything Hillary Clinton or Joe Biden could muster, winning him votes in these regions. But instead, the extent to which there is statistical significance with Biden’s vote share in the absence of significance for changes from 2016 reveals where the locations of the pandemic’s hardest hit regions rather than a causal relationship. It indicates that the biggest unemployment increases occurred in the counties that were already friendly to Democrats and that the biggest COVID spikes occurred in the counties that were already friendly to Republicans. Any changes in vote share from 2016 were not brought on by these effects.

With all this established, let’s run some regression models to see how this plays out on the national scale.

Biden Vote Share

First, there’s the Biden vote share. Here, we notice some similar findings to that seen in the swing states. The overall COVID incidence rate (B= -0.43; -0.28) is negatively associated with the Biden vote share, indicating that Biden performed worse in the counties with higher incidence rates. But what’s interesting here is the significance of the early COVID incidence rate (the percentage of the county’s population that contracted the virus between March and May), which was not significant in any of the swing state regressions. But here, we see that the early COVID incidence rate (B=1.73; 1.01) is positively associated with the Biden vote share. This is consistent with the findings from my pandemic overview, which found that the pandemic disproportionately hit major cities and solidly Democratic regions during the early months, before spreading more broadly across the rest of the country in later months. This is further backed up by t-test results, which failed to find a significant difference in the average overall COVID incidence rate between counties that Biden won and counties that Trump won, but found that the average early COVID incidence rate was more than twice as high in the counties that Biden won (u=0.55 percent) as that of the counties that Trump won (u=0.27 percent).

Next, the unemployment rate is positively associated with Biden’s vote share in all four models, while the coefficient’s magnitude remains fairly consistent even when controlling for other variables. Again, this finding isn’t too surprising, since it was observed in multiple swing states. It also holds up on a t-test, where the unemployment increase was significantly higher in counties that Biden won (u=9.4 percent) versus those that Trump won (u=7.3 percent). One reason for this is that the sectors that incurred the biggest losses in the early months of the pandemic (tourism, hospitality, and entertainment) are disproportionately concentrated in major cities, which tend to lean Democrat. While manufacturing, a sector that’s spread across both Democratic and Republican-leaning regions, was also severely hit, overall many of the job losses were concentrated in larger cities.

On the other hand, the September unemployment rate is also associated with Biden’s vote share. This is a rather bold finding, especially when the evidence for this relationship was more mixed in the individual swing states. Indeed, t-test results indicate that the September unemployment rate was actually higher in the Biden counties (u=7.7 percent) than in the Trump counties (u=5.5 percent). And on the national level, the September unemployment rate (r=0.40) is more strongly correlated with the Biden vote share than the March-April unemployment increase was (r=0.27).

Outside these variables, there are several additional observations. One is that each of the race and ethnic variables are positively associated with the Biden vote share, indicating that Biden performed fairly well in counties with larger black, Native American, and Hispanic populations. This isn’t surprising, given that each of these groups have historically leaned Democratic in presidential elections. Furthermore, the education gap holds up at the national level across all four models, with more highly educated counties (measured by the share of college graduates) lean more towards Biden while the less highly educated counties (measured by the share of residents that have not completed high school) lean more towards President Trump. There is also a small, but statistically significant generation gap, where younger counties lean more towards Biden while older counties lean towards Trump. And socioeconomically, there is some evidence that Biden performed worse in counties with a higher median household income, although this relationship is not significant in all models. Finally, there is fairly consistent evidence that Biden performed worse in counties with a higher poverty rate. While this is partially an indication of educational attainment, it is nonetheless telling that these poorer counties were more likely to go Trump.

Now one additional element I want to introduce to this model is the openness of the state reopening plans, as analyzed by MultiState. For those who have not read my pandemic overview, MultiState evaluated each state’s economic reopening plan based on eleven factors and calculated an overall “openness” score ranging from 0 (least open) to 100 (most open). For this article, I will be using the coding and openness scores from October 6, about one month before the election. For the purposes of this regression, I will be including the state’s overall openness score and its plan’s level of local/regional preemption. Below are the results.

For the variables already discussed, their relationships remain largely the same, although the September unemployment rate’s negative relationship is now more tenuous to other control variables. To that end, this largely solidifies those relationships as they can hold up with the inclusion of other variables.

Meanwhile, the state’s openness score is negative associated with Biden’s vote share, meaning that Biden generally performs worse in counties within states that adopt more open economic plans. This makes sense as the states that adopt more open plans are those that lean Republican. Indeed, nine of the ten states with the highest openness scores are have Republican governors (the exception being Kansas, with Democrat Laura Kelly), all of them have Republican-controlled legislatures, and all of them were ultimately won by Trump in the election. Many of these Republicans have taken President Trump’s lead with favoring less restrictions while downplaying the public health risks. Along similar grounds, the local preemption score is also negatively associated with Biden’s vote share.

Now, let’s apply these same specifications to the change in Democratic vote share.

Change in Democratic Vote Share

Outside these variables, there are several additional observations. One is that the race variables mostly line up with the relationships seen in the Biden vote share, although the magnitude of those coefficients are small. With that said, the relationship with the county Hispanic population is unclear; the coefficient is either positive or negative, depending on the control variables. Even so, the magnitude is small across each model, indicating that the relationship (regardless of direction) is weak. Meanwhile, there’s evidence that the education gap widened from 2016, with highly educated counties shifting more in favor of Biden and less highly educated counties shifting more in favor of Trump.

Now we move into the changes from 2016, which should shed additional light into how the pandemic altered the electoral landscape. First, there’s the COVID incidence rate, which is negatively associated (B= -0.15; -0.12), meaning that, all else being equal, Biden not only fared poorly in counties with a higher incidence rate, but performed worse than Clinton did in 2016. Furthermore, the early COVID incidence rate is not statistically significant, meaning that while Biden performed well in counties that had higher incidence rates during the early months, he didn’t necessarily do better than Clinton in these areas. For these places, it seems less like the virus outbreak compelled their voters to side with Biden and more that these places sided with Biden due to being traditionally Democratic strongholds. So overall, the evidence presented indicates that the medical side of the pandemic did not benefit Biden, either in his overall vote share or relative to Clinton’s 2016 performance.

Next, the increase in unemployment from March to April is positively associated with the change in Democratic vote share, which lines up with their correlation (r=0.12). This offers a bit more evidence that Biden improved in these areas relative to Clinton; however, the magnitude of the coefficient in each of the models is fairly small compared to those seen for the COVID incidence rate. On the other hand, the September unemployment rate is negatively associated with the change, which differs from its relationship with the Biden vote share. When these relationships are merged, it indicates that while Biden generally performed well in counties with higher September unemployment rates, he lost ground from Hillary Clinton. While we know that the mean September unemployment rate was higher in Biden counties, this offers some evidence that Trump’s emphasis on the economic costs of keeping businesses closed peeled away some voters in these areas as the economic costs endured are already high and long-lasting.

Finally, let’s add in the effects of the state reopening plans.

For the variables already discussed, their relationships remain largely the same. To that end, this largely solidifies those relationships as they can hold up with the inclusion of other variables.

Conclusion

2020 saw a confluence of new developments that many describe as a point of reckoning: a “once in a lifetime” global pandemic, “historic” spikes in unemployment, and dramatic clashes in the pursuit of rectifying racial injustice. And this doesn’t include events directly concerning the Trump administration, such as facing impeachment and rushing through a third Supreme Court nomination just weeks before the election. In many ways, this was a jam-packed year in the lead up to this presidential election. And as many have reflected on the possibilities for reform in light of these catalysts, one might also reasonably conclude that these developments would shake the electoral landscape in 2020.

There is certainly an argument that change has been made, particularly how down-ballot Republicans performed above expectations and how Biden won big in many suburban counties. And of course, this analysis has indicated that the medical and economic effects of the pandemic made some shifts on the electorate. But by and large, I think the most interesting aspect of the 2020 election is how little changed compared to 2016. On average, the Democratic vote share in each county only changed by 1.7 percent between 2016 and 2020. As mentioned earlier, Biden was only able to win back 32 of the 222 Obama-Trump counties. And overall, while there is some statistically significant relationship between the pandemic variables and the change in the Democratic vote share, the coefficients themselves are rather small. For the sheer magnitude of the events that transpired in the months leading up to this election, this stability in the electoral map speaks to the level of party polarization in the United States.

In the conclusion of my pandemic overview, I noted how in the past, great crises such as World War II and 9/11, can generate a sense of national unity, at least temporarily. And with regards to the pandemic, I found how such unity was ultimately short-lived, giving way to new messaging that reinforced pre-existing political beliefs. To that end, the months following the initial outbreak suggest that the ongoing polarization in the United States is rigid and self-sustaining, meaning that “once in a lifetime” shocks to the system are not enough to transform this process in a meaningful, long-lasting way.

After studying the 2020 presidential election results in greater detail, I believe I have provided additional evidence that supports this claim. For one, I did not find any evidence in this article or in my swing state series that the Black Lives Matter protests had any significant effect on voting behavior. A key reason for this is that protest activity at the county level is associated with key demographics (particularly college graduates and racial minorities) and urbanicity, both of which are also correlated with Democratic voting behavior. Rather than prompting change in voting behavior, these protests were merely an expression of existing political leanings (particularly Democratic ones).

And as for the pandemic, the regression models for Biden vote share were strong indications of where its effects were most greatly felt while offering limited evidence that those effects significantly shifted voting behavior relative to 2016. None of this is to say that we can’t learn anything about how the pandemic affected the election. Indeed, we can see how the differing narratives concerning the nature of pandemic and the appropriate response plays out across the electorate. Particularly, we see Trump’s economic-centric argument resonate with counties with high September unemployment, places that have suffered a more sluggish recovery or suffer from chronically poor economic circumstances. For these places, the fear of additional recession from public health restrictions was enough for these voters to stay the course with Trump. We also see how by Election Day, some of the biggest viral outbreaks in the country had occurred not in the major cities, but in rural, predominantly Republican areas with fairly relaxed protocols. For these places, such outbreaks reflect neither a failure for Trump’s presidency nor a reason to pursue a different path. Finally, we see the major cities and their suburbs, many of which contain a mixture of racial and ethnic minorities, college graduates, and voters that had already been opposed to (or deeply uncomfortable) with Trump’s brash, chaotic, “unpresidential” approach and were willing to vote for Biden as a return to normalcy. For those places, the pandemic was a culmination of the various shortcomings they have already seen in Trump; crisis or no crisis, they were ready for a change.

Overall, there is a lot that can be said about how this past election (and the year surrounding it) was a “game changer”. Much can be said about how it generated extraordinarily high voter turnout, how Biden drove up support in the suburbs, how it demonstrated the possibilities of mail and early voting, and how it oversaw the election of multiple female, nonwhite, Hispanic, and LGBT candidates. While all these developments are worthy of discussion, I believe that it’s just as important to examine how 2020 demonstrated the persistence of other issues, particularly party polarization, and how much the map resembled that of 2016.

So that will do it for this article. I plan to have more content out soon. If you enjoyed this, please like and follow the Book Aisle. Also share this article on Facebook, LinkedIn, and other social media platforms.