Exit Polls in Presidential Elections

Are voters influenced on Election Night?

Dan Rather decided the 2000 presidential election.  Not the voters, not the states, not even the Supreme Court.  Well, not quite.  But CBS’s early call giving Florida to Gore – based on data from exit polls – may well have influenced enough voters to swing the election.  Exit polling has become a staple of every modern election, yet the visibility – and the stakes – are raised exponentially during US Presidential elections.  It has been argued for decades what effect, if any, exit poll information has had on voter behavior.  Yet the advent of 24 hour news channels, combined with the hurdles of time zones, has created an environment where the debate has shifted from whether there is any effect, to how great that effect is. (Sudman, 333)  

With the media’s perpetual quest to determine the outcome before their competition, national outlets will often put out voter data, or even declare an outright winner, before all polling precincts are closed.  Since the continental United States stretches over four time zones, this can lead to a slew problems.  Voters tend to analyze the costs versus the benefits of voting, focusing on varying factors depending on the individual.  For many, however, one main factor is the “perceived closeness” of an election; essentially, the likelihood that their vote might make the difference. So, the theory goes, if the media or an interest group publishes data showing one candidate way ahead, potential voters may lose their desire to vote, as they perceive the election to be decided. (Ibid., 332)

Election 1980: Reagan’s Surprise Landslide

Exit polls can meaningfully affect voters in only two situations.  First, as was observed in 1980, it can occur when there is a discrepancy between the predicted results before Election Day and the early returns from the media.  When the closeness of an election is “news” to voters, it is more likely to affect their decision to vote. (Creskin & Wielen, 8)   The Presidential race that year was the prototypical horserace and was expected to be as decided very narrowly.

Yet President Reagan’s route was evident early in the day, and the media reported this widely during the day.  Though officially, the media did not begin making predictions until after 8:15PM, it was clear by as early as 6:00PM that Reagan would win the election.  That year polls in several states, including New York and Rhode Island, did not close until 9:00pm, consequently the effect of the early declarations was not confined to the West Coast, but rather voters across the nation were potentially dissuaded from voting. (Jackson, 616)

John Jackson’s comprehensive paper outlines this effect during the 1980 election and concludes that early polling data and the media did, indeed, have a significant effect on voters.  Essentially, Jackson isolated voters who were eligible (and likely) to vote, but had not yet done so prior to hearing about the media’s election coverage, early projections and/or President Carter’s concession speech.  He then “compared the differential turnout rate between those who had heard [any of these], and those who had not heard.” (Ibid., 619)

Jackson’s study demonstrated that the average Western voter was approximately 12 percent less likely to vote after hearing some form of election coverage based on exit poll projections.  This result was present even among voters with a very high probability of voting prior to Election Day.  Interestingly, Easterners were more likely to vote going into Election Day, and that group saw a six percent smaller drop in likelihood of voting after hearing early projections. (Ibid., 627)

Election 2000: Razor Thin Margins

The other situation when exit polls play a major roll is when the winner is officially declared before all precincts have closed, as occurred during the 2000 Presidential election.  Specifically, several national media outlets officially declared the race was over, even though there were 10 counties in Florida’s panhandle which had not closed all of its precincts. (Lott, 349)  The impact of the legal mayhem that year was profound and far reaching.  In fact, prior to the mid-term elections in 2002, a Florida court issued an injunction prohibiting news stations from stating inaccurately that a polling location was closed. (Ibid., 350)

While the actual effect on voter turnout is debated, the instances of preemptive declarations were well documented and frequent.  Between 8:00pm and 9:00pm Eastern time, several major television networks “incorrectly announced many times… that the polls were closed in the entire state.  For example, CBS national news made 18 direct statements in a one hour period that the polls had closed and another 15 implying the Florida vote was over.” (Ibid., 349)  These declarations went further to state that Florida had gone for Gore; yet CBS made this contention 12 minutes before the polls closed.

Republican were quick to assert that voters may have been discouraged from these statements as they may have felt that their side had already lost.  Among others, John Lott has shown that there was a considerable drop-off in Republican turnout in the Florida panhandle during the 2000 election.  Comparing the turnout in the 10 counties of the panhandle to not only the rest of the country, but in particular to the rest of Florida (which was on Eastern Time and would not have been effected by the early calls for Gore), Lott demonstrates that though Republican turnout had been trending down, the drop in 2000 was sudden and significant.  As much as 31% lower than previous elections, turnout was certainly affected in a major way.

Lott also took into account the varied nature of each country.  For any effect on turnout to be real and caused by the media, the effect must have been observed broadly in all of the counties of the panhandle. 51 of 60 precincts evaluated showed a statistically significant reduction in turnout, and each time the rate was “remarkably consistent.” (Ibid., 355)  Lott concludes that Bush lost as many as 10,000 votes as a result of fewer voters deciding to vote after hearing the media state that Gore had already won Florida.

Furthermore, Lott cites figures from the USA today and Miami Herald which claim Bush’s margin of victory would have been between 885 and 1,665 votes if no media outlet had called the election prematurely. (Ibid., 359) However, CBS was quick to attack Lott and his work as uncredible, inaccurate and politically motivated.  Using vague criticisms including the questionable accuracy of registration records and Lott’s failure to assume other factors were at work, CBS defends its own agenda by claiming the 10,000 votes Lott says were lost for Bush is a gross over exaggeration at best, and partisan hackery at worst. (CBS, 78)

But other studies have contradicted these claims. (Ibid., 350)  Considering the election was eventually decided by a mere few hundred votes in a recount, any reduction in vote total for Bush (or Gore) would have been significant.

CBS Autopsies Itself

CBS news was attacked roundly after the election and the news agency immediately launched an investigation into what went wrong that night.  One of the primary recommendations of the resulting report includes that CBS refrain from “calling” the election in states with multiple time zones.  Furthermore, it asserts that CBS should begin using a new designation, “leaning”, for races in which one candidate is clearly ahead, yet it still would be premature to make an official call. (CBS, 4)  However, the CBS report made it clear that it did not believe early election calls had any effect on voters.  Instead, the report claims that all the hysteria surrounding early calls stems from voter perception; voters think that there is an effect, and any evidence to the contrary is “politically motivated.” (Ibid., 82)

Calling mainly for more transparency in their policies, CBS cites several notable gauges of voter sentiment.  “As early as 1964,” the report claims, “Americans believed, by about a three-to-two margin, that television’s early predictions should be held off till the polls are closed.” (Ibid., 81)  Furthermore, according to an LA Times poll conducted in 1984, a full “61 percent of registered voters had an unfavorable view of exit-poll predictions…[whereas] just 24 percent were in favor.” (Ibid.)  Generally speaking, the CBS report absolves the network and questions the credibility of research which concludes early calls have any effect on turnout.

Effect on State and Local Elections

In addition to affecting the outcome of the Presidential elections, early calls become especially meaningful when discussing state and local elections. Early calls may greatly influence Congressional races, particularly in the West.  If a Republican man in California hears that his preferred presidential candidate is en route to a blowout (or getting blown out), then he may decide not to vote since the president has effectively already been decided.  However, if the man lives in a very tight Congressional district, then the Republican (assuming the man votes the Party ticket) in that race will suffer.  When enough people follow suit, the election results can be swayed or altered based on the decision of a minority of voters.

Of course, this effect can be observed across the nation in state legislatures, governors’ mansions, and even sheriff, county commissioner and other obscure local offices. In 1972, for example, Nixon was given the victory by NBC at 8:30pm and by CBS at 8:50pm, both Eastern Time.  Previous analysis suggests that the net impact of these early calls was a 2.7% decrease in turnout overall. (Crespin and Wielen, 4)  While seemingly small, this number becomes much more significant in highly contested Congressional races.  In fact, Reps. James Corman of California and Al Ullman of Oregon may have been victims of these calls, as they both lost reelection bids by a small margin.  Each in part blamed the media’s early calls and Carter’s concession speech at 10:00pm ET (7:00 PT).

Then, in 1980, President Carter’s decision to concede publicly at 10:00pm (he had conceded privately to Gov. Reagan by phone at 9:01pm), enraged Congressional Democrats in the West.

“What in God’s name is wrong with you people?” Speaker of the House Thomas P. “Tip” O’Neill fumed by phone from Boston to Carter’s congressional liaison, Frank Moore. When Moore told O’Neill that Carter just wanted to “get it over with,” damn the western Democrats, the speaker exploded with rage, yelling, “You guys came in like a bunch of jerks, and I see you’re going out the same way.” Representative Tom Foley of Washington State put it more succinctly: “It was vintage Carter at its dead worst.”  (Douglas Brinkly, NY Times Book Review, The Unfinished Presidency)

Bandwagon Effect

Bandwagon EffectOf course, each voter is unique and will therefore make decisions based on a unique set of beliefs, values and perspectives.  Some voters will still trudge to the polls through rain or long lines even after knowing the outcome already, because they have a desire to be on the winning side.  Known as the bandwagon effect, exit polls will not change their decision to vote.  Conversely, the underdog effect occurs when a voter’s preferred candidate is way behind, yet the voter still votes in order to reduce the margin of victory for the opponent.

However, according to political scientist Seymour Sudman, “neither bandwagon nor underdog effects have been observed in national elections.” (Sudman, 333) This was echoed by Creskin and Wielen in 2002. (7)  Epstein and Strom contend that voters make the ultimate decision to vote based on such a complicated assortment of factors that the media declaring an early victory is unlikely to affect voters. (Creskin and Wielen, 4)

However, exit polls have also been suggested to have the opposite effect as is typically assumed; that is, they increase voter turnout when the election is turning out to be closer than originally forecasted.  As perhaps was seen in 1948 and 1968, voter turnout was spurred higher when exit polls indicated a highly competitive finish, rather than the anticipated route. (Jackson, 632)  This theory is not repeated frequently, and data to support such a claim is lacking.

Looking towards the future, some have expressed concern that voters are learning to wait until later in the day on Election Day to vote in order to determine if it will be worth it to them, i.e. if the election is close.  In such a scenario, turnout would drop early in the day; and in a surprise one-sided contest, overall turnout would diminish markedly. (Ibid., 633)  A possible solution to this and other potential problems might be solved by any number of proposals. The most common solution put forth involves mandating that polls close at the same time across the nation, neutralizing the time zone effect. Nevertheless, the CBS report argued vehemently against forcing the media to wait until all polls are closed and instituting a nationwide time to close the polls. (CBS, 81) In its estimation, a common close-time would allow the dissemination of misinformation for hours late in the day which would only encourage confusion, not accuracy.

Exit Polls Hurt Everyone Equally

Additionally, there is no evidence to suggest that a voter of one particular party affiliation is more likely to vote early in the day or later in the day.  Therefore, a fairly representative proportion of voters will have already voted by the time election coverage begins on television.  For instance, if candidate A is beating candidate B by 10 points by 5:00pm (before election coverage and exit polls are available), it should be safe to assume that the same percentage of voters in candidate A’s party have voted as in candidate B’s.  By this point, a majority of the total votes have been cast. It seems logical, therefore, to extend the results as of 5:00pm as the likely final outcome, meaning the person already ahead is going to win.

So while early projections will certainly diminish overall turnout, it is less clear that this drop will actually change the final election results.  Just like in 6th grade algebra, if you subtract X from both sides of the equation, both sides are still equal.  Candidate B’s supporters may assume the election is lost and decide not to vote. This appears to be the rational decision, as – assuming that exit polls and early projections are accurate (which is highly questionable) – their candidate will not win regardless of their decision to vote.

Since their candidate will win or lose anyway, a voter’s decision not to vote after hearing early projections may seem moot, yet this is not the case.  It has been demonstrated repeatedly that exit polls do affect turnout, usually depressing it.  Downs’ celebrated equation for predicting turnout (R=PB-C) may help in discussing the impact of exit polls.  The ‘P’ in the equation refers to the degree to which voters perceive their vote will make a difference.  While in virtually every situation this number is effectively zero because a solitary vote virtually never determines an election, the important element is the perception of the voters.  Once a call has been made based on exit polls, many voters will lose the misguided perception that their vote will matter in determining the outcome.

Riker and Ordeshook’s (1968) extend Downs’ formula to include “civic duty”, represented by D, changing the equation to R=PB+D-C. (Creskin and Wielen, 7)  According to John Hanson:

“People are likely to vote despite a small P term if they believe that they are making an important civic contribution… However, when a projection is made, the voter is essentially alienated from making such a contribution, and therefore is less likely to participate.” (Creskin and Wielen 2002:8)

Further complicating the effect of exit polls and early calls is the bias of the media in general and particular networks specifically. Joseph E. Uscinski, of the University of Arizona, recently analyzed the 2000 election night coverage of the three major networks, NBC, CBS and ABC.  While Uscinski concludes that it is the competitiveness of each state which is the primary determinant in when the networks called the state, he also found significant partisan bias in when each state was called.  Uscinski offers suggestions of the cause of the bias (which he defines as calling states earlier for one candidate over the other, yet he does not conclusively identify a sole cause.

On Election Night, the media must “balance the uncertainty of making an accurate call with the competition of making the first call”. (Uscinski 2007:55)  In an independent report commission by CNN, the network admitted that the exit polls it relied on “overstated the Gore vote in 22 states and overstated the Bush vote in 9 states.” (Konner, Risser, and Wattenberg, 48)  To defend its late calls of several Bush states even when the margin was greater than five points, the report merely claims that those states “notoriously slow in counting their votes,” and in the case of West Virginia, CNN wanted to make sure that the state was “bucking the historical [Democratic] trend and voting Republican.” (Ibid., 49)

Even more damaging, the accuracy of exit polls was horrendous in 2000.  Exit polls inaccurately predicted the winner by as much as 16 points in Alabama (Bush won by 15 points, the exit polls had Gore up 1 point), an astounding figure. Five states in particular – Alabama, Arizona, Colorado, Georgia and North Carolina – were the subjects of grossly inaccurate data.  While all five went for Bush, by the time the polls closed in each state exit polls showed Gore winning – by nearly four points – in four of the states. (Ibid., 74)

There is room for little doubt that early calls on Election Night, and the exit polls they stem from, have real and measurable effects on turnout.  The controversy has its roots in the 1960’s, when network television was just entering the Election Night coverage game. Though the media had by and large been quite successful in calling states at an appropriate time over the decades, the debacle of 2000 demonstrated the failures of their system, and the potential harm of premature declarations.  In the official CNN report published after the election, “television news organizations staged a collective drag race on the crowded highway of democracy, recklessly endangering the electoral process, the political life of the country, and their own credibility.” (Ibid., 78)

It may be unfair to blame Dan Rather for the mess following Election Night 2000, especially since he relied on the same data as the other networks.  All were equally culpable.  Yet that election exposed the flawed system of exit poll data being used to declare Presidential winners.  Consistently, researchers report a very real decline in turnout following early calls by the media.  “Simply put, when a voter hears news that the presidential race is effectively over, the benefits received from the act of voting decrease and thus, the probability of voting declines.” (Creskin and Wielenl, 8)

In 2000, “Television interfered with the electoral process and the election result”. (Konner, Risser, and Wattenberg 2001:36)  Whether turnout drops by two or 15 percent, the ramifications can run deep: skewed state and local elections, voter apathy and potentially, incomplete Presidential election results.  Polls can have substantive value leading up to an election, and even following an election if limited to constructive (non-partisan) analysis, but Election Day polling is problematic.

“It is a flawed system that fails to take into full account many dynamic factors—absentee balloting, early voting, demographic change in key precincts, a declining response rate to polling generally, the quality of questionnaires, vote undercount, mistaken balloting, computer error, human error and more.” (Ibid., 56)

Polls provide little more than a momentary picture of a small piece of the electorate and any conclusions gleaned from such data are tenuous estimations at best. “Polls are statistical calculations, not factual realities.” (Ibid., 56)  For now, it may be better to disavow those calculations and return to the drama and intensity of learning who the next president is over coffee and cereal.

+Cliff Satell

Works Cited

Brinkley, Douglas A. “The Unfinished Presidency: Jimmy Carter’s Journey Beyond the White House.” New York Times Boo

Creskin, Michael H., and Vander Wielen, Ryan J. 2002. “The Influence of Media Projections on Voter Turnout In Presidential Elections From 1980-2000.” Presented at Annual Meeting of the Midwest Political Science Association.

Hansen, John Mark. 2001. “Uniform Poll Closing and Uniform Reporting.” National Commission on Federal Election Reform

Jackson, John E. 1983. “Election Night Reporting and Voter Turnout.” American Journal of Political Science 27:615-635.

Konner, Joan, Risser, James, and Wattenberg, Ben. 2001. “Television’s Performance on Election Night 2000: A Report for CNN.”

Lott Jr., John R. 2005. “The impact of early media election calls on Republican voting rates in Florida’s western Panhandle counties in 2000.” Public Choice 123:349–361.

Mason, Linda, Frankovich, Kathleen, and Jamieson, Kathleen H. 2001. “CBS News Coverage of Election Night: Investigation, Analysis, Recommendation.” CBS News

Sobel, Russell S., and Lawson, Robert A. “The Effect of Media projections on presidential voting in the Florida Panhandle.”

Sudman, Seymour. 1986. “Do Exit Polls Influence Voter Behavior?” Public Opinion Quarterly 50:331-339.

Uscinski, Joseph E. 2007. “Too Close to Call? Uncertainty and Bias in Election-Night Reporting.” Social Science Quarterly 88:51-67.

Note:  This essay on exit polls in Presidential Elections was originally written  for “Campaigns and Elections” a Duke Political Science course taught by Prof. Galen Irwin in Fall 2007.  It has been updated slightly.