STANFORD GRADUATE SCHOOL OF BUSINESS — Intuition tells us advertising increases sales of products, but researchers have had a hard time proving it. A new study from Stanford Graduate School of Business focused on the 2000 U.S. presidential election to prove that point. It found that TV ads were influential enough to tip the 2000 election in favor of George Bush over Al Gore.
“What you need for measuring advertising effects is a simple, unambiguous context, and the presidential election is one such,” says Wesley Hartmann, associate professor of marketing and one of the authors of the paper. First, he says, all consumer choice happens on the same day - election day - and television advertising is restricted mainly to the two months prior to the election.
“This allows us to pinpoint the effects of one advertising campaign on one decision,” explains Hartmann. An election outcome is much more clear-cut than, for instance, a TV ad campaign by Coca Cola where consumers might have been influenced by a particular ad or the cumulative effect of decades worth of advertising. And since Coke sales would be measured over a long period, measuring one-to-one causality would be difficult.
Looking at data from 75 of the top media markets in the United States, Hartmann and coauthor Brett Gordon, an associate professor of marketing at Columbia University, compared how ad prices, ad volume, and voting behavior changed between the 2000 (Bush-Gore) and the 2004 (Bush-Kerry) presidential elections.
Finding that advertising did have statistically significant positive effects, they decided to measure the implications by asking what would have happened if there had been no TV advertising for either candidate? Holding all other factors fixed and shutting off the advertising, their model predicted the 2000 presidential election between Bush and Gore would have played out differently.
“In 2000, without TV advertising, Bush would have won Oregon, which would have given him 7 more electoral votes,” Hartmann explains. “But Gore would have won Florida and New Hampshire, gaining 29 more votes that Bush otherwise would have had. In our analysis, without advertising, Gore would have won a total of 289 electoral college votes to Bush’s 249, and he would have taken the election. Because of the influence of advertising in Florida and New Hampshire, however, Bush won 271 to Gore’s 267.”
In 2004, in contrast, the absence of advertising would have given Bush even more electoral votes (14) than he actually garnered, particularly in New Hampshire and Wisconsin, simply solidifying his win over John Kerry.
The election scenario also allowed the researchers to determine that advertising did indeed cause changes in voting behavior - something other researchers have tried but not been successful in establishing. “Typically, advertisers spend more where sales potential is greatest,” Hartmann explains. “This could make advertising appear to be effective, even when it is not - sales could be a function of the nature of the market itself, not the ads. Other researchers have tried to document the causality by focusing on whether strong competition for ads, as reflected by high ad prices, might lead to less advertising and, hence, fewer sales. Yet prices are not a good indicator for this competition because local sales potential alone could affect ad prices.”
The election scenario is unique because the researchers could establish competition for advertising using ad prices during a time period when there were no “sales” - no elections and no voting - when any potential to swing voters could not possibly have affected ad prices. ”You can use ad prices for the year before the election to estimate what the ad prices would have been during the election without all the political influences on them – influences like a candidate buying a lot of advertising in a particular market,” Hartmann says.
Say a market like San Francisco had ad prices going up or down between the elections of 2000 and 2004. “Looking to ad prices in 1999 versus 2003 gives you a safe measure of the changes of ad prices in the actual market,” Hartmann says. “If we looked in the election year, in contrast, it might be that there was a tougher battle in one of the states that led to a change in ad prices there,” Hartmann says. “But if we go with 1999 and 2003 prices, we have a change that is not being driven by the election.”
Differentiating the base-level market price changes between 1999 and 2003, from how ad spending actually changed between the 2000 and 2004 elections, allowed the researchers to determine which markets were more or less difficult to advertise in during the 2004 elections. If increases in the difficulty of advertising resulted in less advertising and consequently fewer votes, then causality could be established.
They discovered that where advertising prices went down the year before the election, ad spending went up –– and votes went up. “Through this analysis, we established the direction of causality - that ad prices influenced number of ads, and that number of ads influenced number of votes,” Hartmann says. In some markets, one candidate might have been more sensitive to ad prices than the other, creating changes in relative ad purchases and consequently relative votes received.
Hartmann says the paper suggests that for researchers and businesses to determine whether and how much advertising affects sales, they should look at discrete situations and scenarios, such as the Superbowl. “It’s a huge one-time change in TV ads which should create more visible changes in sales,” says Hartmann, who is currently involved in a study of that phenomenon.
The research also indicates that focusing on advertising prices can help determine the causal link between ads and purchasing behavior. Finally, examining shifts among all these variables within discrete markets (for example, the San Francisco region) between two time periods can be important in understanding the influence of advertising.
- Marguerite Rigoglioso
Also on Stanford Knowledgebase: