National Media Decisions on Election Night (Part 2)

Featuring guest blogger, Harrison Hickman, President & Founder of Hickman Analytics, Inc.

Last week’s blog post listed the different “calls” viewers can expect from national news outlets on election night.  I also addressed a common myth in noting that exit polls are not used to make these calls unless the race is very one-sided.

The biggest current use of exit polls is explaining why one candidate won and another did not.  The polls include a variety of questions about policy issues, impressions of the candidates’ strengths and weaknesses, and other factors that might influence how a person voted. Even before a race is characterized as “too close to call” or showing one candidate “leading,” the exit poll results can be used to explain, for instance, that the economy was the most important issue, that voters believe one candidate is more honest than the other, and to compare voting with positions on policy questions.

The exit poll results are also used to show that voting is not as monolithic as voters might expect given the way elections are covered by the media.  Here’s an example of something that could be reported on election night: “20% of voters who support abortion rights are supporting the pro-life Republican candidate in this race, while 25% of those who want a tougher policy on immigration are supporting the Democrat.” What the news outlets do not report is that the reason this seems odd is that their own reporting is what led to the widespread belief that all pro-choice voters support Democrats and that voters who want tougher immigration policy support Republicans.

Statistical models.  If the decision desks at national media outlets do not rely on exit polls, how do they decide on where a race stands before the final votes are in? The current basis for the reporting comes from comparing partial early results with sophisticated statistical models based on past election results and current registration information. These models have become increasingly important as the number and sample sizes of exit polls have been reduced.

On election night, early partial results are compared with results to past elections in different areas (or in precincts) and among different groups of voters. The use of models for making these judgments helps avoid mistakes made because one area of a state or district is reporting results faster than others. The two main factors that these models are designed to measure are the support for the candidates of the major parties and the turnout in different areas.

Geography is a critical element of the analysis, and the precincts or counties included in the models are selected to be representative of specific areas. These areas might include places where Democrats or Republicans traditionally do very well, or areas where both parties have a history of winning depending on the year and the candidates. One additional key is to measure the proportion of vote that is coming from each area so that the model adjusts to current turnout rather than the historical norm.

Demographics are important as well. This year, for instance, expect the news media to spend a disproportionate amount of time reporting on how this Presidential race compares to past races among Black and Latino voters. The reason we know they will focus on this is because they have spent a lot of time in their election coverage talking about the special attention the Trump campaign has been paying to these groups and to Kamala Harris’s lower-than-expected support from them.

But you don’t need to be an experienced analyst to know that gender will be a critical factor.  The so-called "gender gap," the difference between the support a candidate gets from men and from women, is the highest it has ever been in pre-election polling. Part of this is the fact that the major party candidates are of different genders. But the issue landscape is playing a role as well, especially since the Supreme Court reversed its position on abortion rights since the last campaign.

Remember that both turnout and candidate support matter. In this regard, it will be interesting whether women or men vote in larger numbers, and a comparison of Trump’s margin among men with Harris’s margin among women.

Conclusion

I hope this helps you better understand what is going on behind the scenes when you are watching election results roll in on election night. In addition to watching the network coverage, I will be logged into the reporting at the New York Times  which has emerged as providing the best election night predictions in races up and down the ballot all over the country.