What to Expect on Election Night: Where to Look—and What to Look For

November 2, 2020
Michael Latner
Senior Voting Rights Fellow

On the eve of the 2020 General Election, the Center for Science and Democracy has put together several pieces of information to help voters keep track of the election after vote counts are released tomorrow. Throughout this year, we have documented the importance of election outcomes on health outcomes and questions of environmental justice, and now we have come to that moment, where over 150 million of us will collaborate in deciding our shared fate.

Previously we identified numerous counties in pivotal states, along with several comparison cases, in order to keep track of election results before and after Election Day. We have considered the consequences of a surge in vote-by-mail (VBM) ballots requested by voters during the COVID-19 pandemic, as well as the potential for structural inequities, including racial disparities in ballot rejection rates, to taint election results. In this final pre-election peak, we consider a revised Cost of Voting Index provided by the original authors, examine VBM return rates across counties, and provide a rough measure of overall turnout expectations, so that voters can assess the degree to which the election is going as expected, given the data we have.

The first update comes by way of the Election Law Journal, where Scot Schraufnagel, Michael J. Pomante II and Quan Li have recently updated their Cost of Voting Index (COVI) for US states as a measure of the restrictiveness of election laws that make it easier or more difficult for people to vote. The index is included at the end of this post. While UCS has closely followed legislative and litigation-based changes to election laws since 2018 to update the 2016 measure, this update includes hard-to-find variables such as polling hours and details about enforcement of voter registration laws. The Index reflects changes to VBM access this year, with states like California, Illinois, Maryland and Vermont moving up the list by expanding voting access, and Texas and Georgia moving down to the bottom, below even Mississippi and Alabama in overall cost of voting. The Index is a good measure of where people face barriers to voting (registration, access to early voting, ID requirements, voting options, polling availability) before and on Election Day.

Professor Michael McDonald and staff at the U.S. Elections Project are also providing an invaluable service to election observers with daily updates of early voting statistics, including many county-level data. We also have additional data on VBM and election returns collected from county officials by UCS staff. Compiled VBM returns to date (October 28) are compared to 2016 VBM to illustrate which counties are experiencing the largest surge of mail ballots.

The Surge in Vote-by-Mail

Figure 1 orders counties by COVI and 2016 VBM rates. Most counties are experiencing a substantial increase in VBM ballots returned. The only exceptions are the three universal-vote-by-mail (UVBM) counties in Colorado, Orange and San Luis Obispo counties in California, and Maricopa and Pima counties in Arizona, all of which relied extensively on VBM in 2016 (California has moved to UVBM for 2020).

The counties with larger shifts to VBM to date are those in states that eased restrictions on voting by mail or made it easier to do through web portals or ballot applications: North Carolina, Georgia, Wisconsin and Pennsylvania. (The Wisconsin numbers may not be as accurate of a comparison due to the way that officials report ballots cast in the EAVS database.) Even Texas counties have seen an increase in mail ballots received, where the state seems to have done everything it could to reduce turnout. One of the most substantial increases anywhere in the country is seen in Philadelphia county, where local election officials will not start processing its nearly 300,000 VBM ballots until Election Day, compared to 13,306 in 2016.

Election rules vary by state in ways that will impact the counting of these votes. For example, even with these substantial surges in VBM, both North Carolina and Georgia are processing mail ballots early, while Michigan, Pennsylvania and Wisconsin will not begin processing ballots until Election Day or the day before. As a result, voters in these states, and around the country, will need to be patient while local election officials process ballots cast before or on Election Day. Indeed, many states will be and always have counted ballots after Election Day. Whether or not the results on Election Night are clear enough that one of the presidential candidates concedes, the election is not final until all eligible ballots have been counted and states certify their results, which in some states takes several weeks.

Figure 1: Comparing 2016 Total Mail Ballots with 2020 Ballots Received (Log Scale)

Turnout Matters

Another way to visualize the increase in share of mail ballots in 2020 is to estimate voter turnout and the share of a county vote that is likely to come from mail ballots. Figure 2 shows more clearly that barriers to voting by mail do limit its use. A simple, if rough, range of turnout can be estimated using national forecasting expectations of a 10% increase in overall voter turnout, adjusted for population growth. This projection is weighted using the COVI to add or reduce turnout by one point for every standard deviation on the scale (i.e., a score of .33 would reduce turnout by a third of one point). Taking 2016 turnout as the low end of the range, an “average” turnout estimate is calculated, which is a conservative estimate. Indeed, fast growing Harris County, Texas, has already passed 2016 votes cast and could easily hit the high estimate of 61% turnout on Election Day.

Figure 2: Actual 2016 VBM rates and estimated 2020 VBM rates (estimate assumes Colorado rates stay the same).

To estimate mail vote share of overall turnout, we need to anticipate how many more mail ballots will arrive. We estimate a 2020 return rate using data from the UVBM counties in Colorado, which is basically running regular elections.  Assuming that the UVBM counties will have similar return rates to 2016, we calculate the difference between what they have and what they still need to receive to achieve those return rates, which on average is about two-thirds of remaining outstanding ballots. Most of the estimated return rates are lower than actual rates observed in 2016.

One final calculation: Subtracting the estimated returned mail ballots from overall turnout estimates provides an estimate of 2020 mail vote rate (and alternatively, an estimate of in-person voting). Even low VBM counties in Texas could double their share of mail ballots.  That’s also the case for Los Angeles County, historically one of the lowest VBM counties in California. This figure paints a more accurate picture of the magnitude of increased VBM in the Midwest and Rustbelt counties. Sixteen of these counties are expected to go from less than 10% VBM in 2016 to 30-50%VBM in 2020. Both Figure 2 and Figure 3 provide some guidance about what to expect in the county-level returns, and where final vote tallies may take some time.

Beware Rejection Rates

While mail ballot rejection rate data will not always be available until after the election, this is another set of important numbers to watch. Specifically, the possibility of rejection rates at or higher than 2016 levels in pivotal states could very well determine the outcome in a close race. Consider the fact that the presidential race in Pennsylvania was decided by just 44,292 votes. If Philadelphia County rejects the same percentage (3.46%) of ballots that it did in 2016, with the expected number of mail ballots cast, that is over 11,000 ballots in one county.

Moreover, the distribution of high rejection rates is not allocated equally, as previous analyses have shown. Arraying these counties by the percentage of the county that reports being non-Hispanic White (highest percentages to the right) shows that, if 2016 rejection rates are sustained, that will result in tens of thousands of rejected ballots in predominately African-American and Latinx communities. On the one hand, given the number of people voting by mail for the first time, we might expect even higher rejection rates. On the other hand, greater scrutiny, and at least in the case of North Carolina, greater effort to correct voter ballot errors, could result in lower rejection rates. This is one area where local media need to keep close track of data as it becomes available, as unusual rejection rates could also signal administrative errors or corruption of the integrity of results. By watching these counties, listening to local voices, and reporting on the election as it unfolds, we can all help ensure the integrity of the election.

Figure 3: Potential number of rejected ballots (log scale) based on 2016 rejection rates and 2020 ballots received, across counties with increasing % non-Hispanic white population (left to right).

Finally, turnout range estimates also provide useful comparisons for watching the votes come in. With the 2016 vote as a baseline, county-level returns provide a good overall clue as to whether turnout will reach expected high levels. For example, if Dane County, Wisconsin hits the high mark of 91%, the capital city of Madison could see record turnout despite the inability to hold normal on-campus GOTV efforts at the University of Wisconsin. Even so, it is striking that a “record year” for turnout would still leave so many eligible voters behind. Voter participation in Bexar County, Texas, may only be half of what it is in Wisconsin, for example. If everyone actually had an equal opportunity to vote, and were as actively welcomed into political action, these disparities would be reduced.

Figure 4: Turnout ranges and weighted estimates based on 2016 turnout base, population growth and Cost of Voting Index.

Turnout differences between counties also provide important information about inequalities in turnout and statewide outcomes. Low turnout counties tend to be more economically depressed, and the more populous counties tend to be more demographically diverse. For example, if the 13-point turnout difference between Philadelphia (63%) and Fayette (50%) counties changes dramatically, that has important implications, as Fayette is more economically depressed, older, whiter, and supported president Trump by 32 points in 2016.

Keeping an eye on how the election unfolds locally can provide crucial insight into political and behavioral shifts that are happening nationally. At the end of this post, we provide information on each county for comparative purposes so that you can follow changes in the 2020 outcome, compared to 2016, as results are released. There are several possible scenarios for how Election Night might play out. The more informed you are, the better prepared you will be.

And even if you have already voted, there is still more you can do to help ensure that every voice is heard and every ballot is counted, that would be great. Go to act.ucsusa.org/election-watch.

Appendix: Cost of Voting Index and 2016 Election Results

Source: Scot Schraufnagel, Michael J. Pomante II, and Quan Li. Election Law Journal: Rules, Politics, and Policy, ahead of print

2016 Turnout Data and High Turnout Estimates

State County Turnout 2016 Totalvote16 Clinton16 Trump16 High Turnout
AZ Maricopa 58 1567834 702907 747361 2123704
AZ Pima 59 421640 224661 167428 533280
CA Los Angeles 56 3434308 2464364 769743 4298890
CA Orange 60 1197521 609961 507148 1508912
CA San Luis Obispo 63 135009 67107 56164 171014
CO Arapahoe 72 303046 159885 117053 383832
CO Denver 72 331852 244551 62690 453410
CO El Paso 67 318967 108010 179228 415946
FL Broward 68 831951 553320 260951 1046582
FL Hillsborough 65 597660 307896 266870 775196
FL Miami-Dade 63 980204 624146 333999 1232796
FL Orange 64 546275 329894 195216 736291
FL Palm Beach 68 662332 374673 272402 834855
GA Cobb 68 330819 160121 152912 428846
GA DeKalb 64 314757 251370 51468 394624
GA Fulton 62 431391 297051 117783 554067
GA Gwinett 63 328331 166153 146989 421180
MI Genesee 63 196296 102751 84175 228403
MI Kent 69 308184 138683 148180 380640
MI Macomb 66 419312 176317 224665 510018
MI Oakland 74 664614 343070 289203 785407
MI Wayne 61 777838 519444 228993 897304
NC Durham 68 157022 121250 28350 170147
NC Mecklenburg 67 475,650 294562 155518 580942
NC Orange 71 82,818 59923 18557 82562
NC Wake 75 531,256 302736 196082 676871
OH Cuyahoga 65 617,350 398,271 184,211 712185
OH Franklin 68 593,435 351,198 199,331 746645
OH Hamilton 70 417,456 215,719 173,665 487572
OH Lucas 61 200,973 110,833 75,698 230382
PA Allegheny 68 650114 367617 259480 742728
PA Fayette 50 53767 17946 34590 62275
PA Montgomery 72 434687 256082 162731 510822
PA Philadelphia 63 707631 584025 108748 839815
TX Bexar 48 589645 319550 240333 792454
TX Dallas 52 758973 461080 262945 967037
TX Fort Bend 63 262066 134686 117291 371045
TX Harris 52 1312112 707914 545955 1745915
TX Montgomery 60 204632 45835 150314 39601
TX Tarrant 54 668514 288392 345921 897779
WI Brown 69 129,011 53,382 67,210 154057
WI Dane 81 309,354 217,697 71,275 378918
WI Kenosha 62 76,304 35,799 36,037 92925
WI Milwaukee, Washington, Waukesha 70 755,403 388,898 320,352 874344