A study of the 1980 US Presidential election showed that among many voters heard projections of the outcome before the local polls closed and of those who had not yet voted, those who heard the results were less likely to vote. (Paraphrased from [1]). This paper cites a study from 1981 in which they concluded that the 'decline' in voter turnout on the West coast wasn't attributable to the projection but rather to the large number of factors which influence a voter in a given election year. Another study cited in [1] shows the opposite. The paper claims that both of the other studies make unrealistic assumptions. Issues such as election-to-election campaign changes, salient issues, weather, etc are not controlled for properly in those studies. This implies that their conclusions are invalid.
Another study from 1984, also about the 1980 elections says that this election night problem has been approached from many angles with various measurements. The conclusion that they draw is that the early broadcasts did have a small yet measurable impact on the election turnout. They conclude that it was not enough to impact the outcome of the presidential elections but the margin of victory for about 14 congressional elections was smaller than the impact. (Paraphrased from [2]).
The situation today is much worse than it was 20 years ago.
Our 24-hour news cycle allows voters to be more tuned-in to the current election outcomes and projects. The most severe example of a problem with early projections and reporting of outcomes happened in 2000. The election was not decided until 36 days after election day. On election night each of the major networks was very eager to call states for a certain candidate. Often when one network had enough poll data to call a state other networks would pick this up and call the state themselves. This need to be the first to call the election only served to amplify the underlying problem. All of the major networks made two mistakes on the night of the election. They first called Florida for Al Gore, then after retracting that, they called it for George W. Bush, which they also retracted.
A now disbanded Voter News Service existed in 2000 which was created by the major broadcast networks and the Associated Press to conduct election night exit polls and provide the networks and AP with data on when to "call" a state. The problems from the 2000 election arose from the fact that VNS wanted to quickly "call" states. Because of failures in 2000 and 2002, the service was disbanded in January of 2003. This service was responsible for the information that the major broadcast networks and AP had during that night. The data received depends largely on unreliable exit polls, statistics, models and NOT actual counted votes. CNN did not disclose where it was getting the information to "call" a state.
In [3], where most of this information is covered, a general recommendation is made about changes to election night coverage. The questions asked of the networks are, What are the benefits? What are the costs? During most of the year(s) we would like our news coverage to be accurate, quick and reported when known. However this "journalistic ethic" might need some re-thinking during election night. (Again heavily paraphrased from [3])
This report about CNN is quite interesting. I read the preamble, introduction, summary findings and skimmed over the recommendations. I didn't read very much of the actual report.
The current practice of projecting the presidential election outcome before all polls across the country have closed must be stopped. I believe that this practice creates an unfair influence on the election.
The major networks can wait until the polls have closed across the country, hopefully including Hawaii and Alaska, so that eligible potential voters cannot be influenced. By this time many actual vote counts from the contiguous states will be known. They can now call states based on actual vote counts. It's likely that few would stay up this late to receive the result. Most will just read/hear about it the next morning when they are getting ready for work, school, etc. It's a lot like Christmas, at least for Americans. Santa Claus comes over night and the presents are there in the morning. The votes are counted over night and the results are available the next morning.
I do realize that this takes a lot of the "fun" out of election night but that's okay. I believe it's worth the sacrifice. There are still US and local house and senate races which can be called and discussed. These cannot be influenced by voters on the west coast, and they might be equally contentious as the presidential race.
I invite you to comment below about this issue. I'd like to hear your thoughts.
Sources:
[1]
Election Night Reporting and Voter Turnout
Author(s): John E. Jackson
Source: American Journal of Political Science, Vol. 27, No. 4 (Nov., 1983), pp. 615-635
Published by: Midwest Political Science Association
Stable URL: http://www.jstor.org/stable/2110886
[2]
Scooping the Voters? The Consequences of the Networks' Early Call of the 1980 Presidential Race
Author: Michael X. Delli Carpini
Source: The Journal of Politics, Vol. 46, No.3 (Aug., 1984),866-885.
Stable URL: http://links.jstor.org/sici?sici=0022-3816%28198408%2946%3A3%3C866%3ASTVTCO%3E2.0.CO%3B2-Y
[3]
Television's Performance on Election Night 2000
- A Report for CNN
Authors: Joan Konner, James Risser, and Ben Wattenberg
January 29, 2001
Stable URL: http://archives.cnn.com/2001/ALLPOLITICS/stories/02/02/cnn.report/cnn.pdf
Interesting side-story:
While composing this entry I had to re-open the CNN report to determine which parts I had read and because of some issues with my window manager I was re-sizing the PDF viewer window as well as my browser window, alternatingly. This eventually led me to accidentally close the browser. It wasn't until I re-opened it that I realized my blog entry was likely gone. Then I recalled that LJ auto-saves the entries so I was hopeful. However my entry was gone. I had managed to open/close a few instances of my browser ... probably because it's 3:15am. I was very very sad for a while. As I was brushing my teeth I wondered if a cookie or something had stored my draft. I started poking around in the Firefox directory and went to <default profile>/sessions, and did a grep for "election". It produced a lot of output and luckily it had my complete blog post included. I randomly picked the correct folder to check on my first try, woo. I would have checked all the others until I was convinced that it wasn't saved locally. I'm very glad that it was saved though because I put good amount of time into this post, and it's still here!!