TODAY'S PAPER
36° Good Morning
36° Good Morning
OpinionCommentary

TV ads still win elections. And Democrats are buying a lot more of them.

Democratic presidential candidate former Vice President Joe Biden

Democratic presidential candidate former Vice President Joe Biden at The Queen theater on Wednesday in Wilmington, Del. The Wesleyan Media Project's data show that between Sept. 5 and Oct. 11, pro-Biden broadcast ads outnumbered pro-Trump ads by 5,301 in Harrisburg, Pa., and by 4,291 in Green Bay, Wis. Credit: AP/Andrew Harnik

The vast quantity of television advertising aired by candidates, parties and outside groups is proving to be one of the many ways that the 2020 election is remarkable. According to data gathered by the Wesleyan Media Project, more than 4.9 million ads have been aired on broadcast and national cable television in federal races this year — more than double the number run in 2012 or 2016.

Voters in competitive states may experience this as an onslaught. They — and we — may even wonder whether so many ads can possibly make a difference. Aren't viewers zoning out, changing the channel or even fast-forwarding as soon as yet another ad comes on? And anyway, given the decreasing number of swing voters in U.S. politics, is anyone left for all these ads to persuade?

Our new research finds that ads do matter — often quite a bit. The effects are not large in presidential races, although they may make a difference in a swing state in a tight year. But they are substantial in down-ballot races for the House and the Senate, and even more so in statewide races. Perhaps the most surprising finding is that the effects of each additional ad diminish only slightly, even at extremely high numbers. So what seems like excessive spending to voters just trying to watch a football game may well be worth it. TV ads, archaic though they seem in an online era, still pack an electoral punch.

Most people studying this issue have focused on presidential campaigns. But to investigate how much televised campaign advertising matters, we gathered data on more than 2,250 elections from 2000 through 2016, including not only the five presidential elections, but 276 Senate races, 176 gubernatorial elections, 1,655 House elections, and about 150 other statewide elections for offices such as treasurer and attorney general.

For each of these elections, we know how many ads were aired in each media market on each day of the campaign — ads aired by campaigns as well as by political parties or outside groups to support candidates. We matched this information to how many votes each candidate received in each county (it helps that most counties lie within a single media market). Then we estimated whether the candidate who aired more ads ended up with a higher share of the vote. Using various statistical models, we can mitigate the obvious chicken-and-egg problem, which is that candidates typically run ads in places where they expect to do well. (For example, our models allow us to take account of the unique features of counties — including demographics — as well as the trends that affect states across time, such as the shift of West Virginia from blue to near-solid red. We also isolated pairs of counties adjacent to each other but on different sides of a media market boundary. These counties are similar on many dimensions except for the level of advertising they experience.)

We found that in presidential elections, ads matter mainly at the margins. Imagine that a presidential candidate was able to run 1,000 more ads than the opponent in a given county in the last two months of the campaign. (This is far from an outlandish hypothetical, given how many ads campaigns and affiliated groups run.) We estimate that this candidate would end up with about 0.2 percentage points greater vote share. That two-tenths of a point won't be a game-changer unless the election is razor-close, as in Michigan in 2016, when Donald Trump won by 0.22 points.

What might this mean in 2020? Former vice president Joe Biden has had a substantial advertising lead that has surprised many observers (because his campaign had financial struggles early on). For example, the Wesleyan Media Project's data show that between Sept. 5 and Oct. 11, pro-Biden broadcast ads outnumbered pro-Trump ads by 5,301 in Harrisburg, Pa., and by 4,291 in Green Bay, Wis. If Biden's advantage persists through Election Day, television ads could net him perhaps half a point of vote share in these markets. Clearly, these are votes Biden wants, but they probably will constitute only a small fraction of his overall margin of victory, if current polling is correct. (Washington Post polling averages have Biden up by six points in Pennsylvania and eight points in Wisconsin.)

But in states such as Florida and Georgia, where the polls are neck and neck, even the small effects of ads could prove consequential. In Georgia, for instance, Trump had a substantial lead in advertising between Sept. 5 and Oct. 11 — anywhere from a 900- to 4,800-ad advantage, depending on the media market.

Ads outside the presidential general election are far more important. They matter, our research found, three times as much in Senate races as in the presidential race, four times as much in gubernatorial races, and 12 times as much in races for state treasurer. This makes intuitive sense: The candidates in those races are less familiar than national candidates such as Trump and Biden. Television ads are therefore more likely to provide new information.

Because they are running more ads than Republicans in many key Senate and House races this year, our findings bode well for Democrats this cycle. In the much-watched race in Arizona between the former astronaut Mark Kelly, a Democrat, and Sen. Martha McSally, a Republican, for instance, the Wesleyan project finds that in the first half of October alone, 10,421 ads have backed Kelly, vs. 8,091 for McSally. In Montana, ads supporting Democratic Gov. Steve Bullock outnumber those supporting his opponent, Republican Sen. Steve Danes 18,329 to 13,429.

However those races turn out, Democrats will have done better because of TV advertising. (There is also a lesson here for donors: Their money will go further in down-ballot races than in presidential general elections, which soak up so much attention and cash.)

Surprisingly, we did not identify a spillover effect from race to race: An imbalance of ads in one contest — even the presidency — did not appear to affect elections at other levels.

We also expected to find more evidence of diminishing returns to TV ads than we did, because it seems only natural that voters would get tired of ads. But in fact, we found only modest evidence that the 1,000th ad is less effective than the 10th. One reason is that the persuasive effect of advertising may — as other research suggests — be short-lived, so candidates have to stay on the air to secure any gains in vote share. The implication for candidates is: Spend more, if you can.

But the timing of that spending matters, we also found. It is the ads that are aired in the last two months of the campaign that are correlated with vote share; we did not find that ads aired earlier make much difference by the time voting occurs.

This calls into question the Trump campaign's strategy of spending significant resources over the summer. If the goal was to influence how people vote — as opposed to, say, attracting donations — Trump's advertising advantage during that period shouldn't matter much now. Biden's advantage in September and October is far more important.

This new research doesn't shed light on which types of ads may be the most persuasive, or which voters may be the most affected. But other research suggests that many types of ads — promotional or attack, made by the campaigns or produced by outside groups — can have small but significant effects on voters of all kinds: Democratic, Republican and independent alike.

Political observers are often writing the epitaph for television advertising in this age of streaming, cord-cutting and social media. But our research suggests that this old-school tactic still moves voters.

Sides is a professor of political science at Vanderbilt University. He is co-author, with Lynn Vavreck and Michael Tesler, of "Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America."

Vavreck is Marvin Hoffenberg chair of American politics and public policy at UCLA.

Warshaw is an associate professor in the department of political science at George Washington University. This piece was written for The Washington Post.

Columns