Working backwards, the election night fiasco can be laid at the door of the networks and their polling arm, the Voter News Service. Everyone knew that Florida was going to be a critical state, and the networks said in advance that they would be extra-cautious in “calling” Florida. But in the end they were anything but cautious. Starting at 7:49 pm Eastern Time all the networks gave Florida to Gore even though parts of the state (in a different time zone) were still voting. Two hours later they began to backtrack, and from 2:16 am gave Florida to Bush. The second Florida call for Bush came despite the fact that returns were available on the Florida website which showed the race tightening, with a Bush lead of only 2,100 votes as the last counties came in. Finally, at 2:59 am the networks began reversing their call on Florida and returning the state to the “too close too call” column. But the second misguided Florida call and declaration of Bush as President helped create a sense that Bush really had won the election, and was just waiting to have it confirmed rather than its being truly undecided.

The use of exit polls to predict elections began in 1980, but over the next two decades there were no really close races to test their accuracy, lulling many into a false sense of security. For the first decade the networks conducted their own individual exit polls but then, in 1990, they pooled resources to save money, creating the Voter News Service (VNS). The next crucial development was in 1994 when ABC decided to set up its own decision desk staffed with experts to interpret the VNS data. Before that VNS would make a single call and pass it to all the networks. In 1994 ABC proceeded to beat all the other networks and so, by 1996, the others had followed and all the networks began to analyse the VNS data themselves, introducing a new degree of competition. So two possible avenues for errors were introduced. The first was that the VNS data on which everyone relied could be wrong. The second was that in their haste to be first the networks could make their own mistakes. Both of these eventualities occurred on November 7.

Faulty exit poll data, mistakes in input and problems in the models through which the exit poll data was passed all helped corrupt the Florida result. Traditionally if there is a wide margin in an exit poll (a 10 per cent lead or more for a candidate) then the networks will instantly call the state. If it is closer they will start to look at data from key precincts which have been identified as good indicators and will compare this and other results information with the exit poll. One of the key problems is what to do when exit poll results don’t match up with the first real results. Which set of results do you trust? This was the problem faced by the BBC in Britain in 1992. The models used were also based on demographic realities which may have changed since the last election (particularly the case in Florida where there is an even higher degree of geographical mobility than in most of the US). A precinct which may have been a good indicator of the trend of the state in 1996 may not be so accurate in 2000. Few realise the importance of models in exit polls on both sides of the Atlantic.

Exit poll interviews are done in waves through the day, and each time slot may have its own unique demographic features (white collar workers vote early, blue collar after work, retirees vote midday and so on). Unexpectedly large turnouts can also throw the poll and its predictions, as well as the increasing number of people who vote by mail or absentee ballot (23 per cent in California). In a leaked memo, VNS cited absentee ballots as the key source of error. But VNS may not get the chance to right their mistakes since a number of networks have threatened to pull out of the consortium.

All the networks have set up investigations (CNN have appointed an independent advisory commission) and a number of proposals are likely to emerge, such as a call for a common poll closing time. ABC has issued new guidelines saying it will not call any state until all the polls have closed and will spend more time doing independent analysis of VNS data and insulate its decision makers from what other networks are saying to avoid competitive pressures (by not having TVs in their rooms). Others have followed on some of these points, and NBC and FOX have both said they may pull out of the VNS consortium; but competitive pressures mean that mistakes are still likely to be made in the future.

On the night, the networks failed to explain fully on what basis that they were “calling” a state and how the whole process worked. Republicans have also claimed that the networks were biased in their decisions. There is anecdotal evidence that the networks proved happy to give states to Gore very quickly when his lead was tiny in the exit poll (Florida), but did not call states for Bush even when he had a healthy five point lead in the exit poll (Georgia). There has been no explanation of this, although some might be forthcoming in January when executives are called before the House Commerce Subcommittee on Telecommunications, chaired by Billy Tauzin, a Republican critic of the networks. Tauzin has claimed that on CNN there were delays in calling nine states that Bush won by at least six points but no delays in any state Gore won by six points. There is no way in which Congress can restrict either exit polls or what the networks say on air during election night, since this would breach the First Amendment. Any changes will have to voluntary.

The dimpled chads of Florida may also have obscured deeper trends in the whole election campaign. This is especially dangerous on the media front, where the year 2000 saw a number of ominous signs for those worried about the level of political debate. For a start there was simply less coverage of the campaign. Until the last week, there was less than in any previous election, 53 per cent below the 1992 level and lower even than the 1996 race which was considered pretty dull. Eight years ago, the three main networks gave about 25 minutes a night of election news or eight minutes each. This time, according to Stephen Hess of the Brookings Institution, the figure was about 12 minutes, or four minutes each.

The average length of the candidates’ sound bites has shrunk by 30 per cent since 1988, falling from 10 seconds to a measly seven seconds (barely enough time to utter an inanity). What coverage there was also focused more on the “horse-race” – who was up and who down, rather than on issues and substance and on the candidates’ qualifications and what they would do in office. If a candidate was ahead in the latest poll everything he was doing was interpreted as right, and vice-versa. Cable news networks would report the latest gyrations of the daily tracking poll (which was notoriously inaccurate) as straightforward, and build their coverage around it. Different polling organisations showed huge differences in voting intentions (often double digit). In most cases this was caused by the way in which they reweighted the sample, and there appeared to be a major lack of understanding over how polling actually works and how there is no real way of knowing which poll is “right”.

Targeted

As well as the changes in the way news was produced, there were linked changes in the way it was consumed by the public. Perhaps the most significant development was the fragmentation of the news market and the arrival of narrow-casting (as opposed to broad-casting). There have been significant changes over the last three elections in terms of where people get their political information from, according to the Pew Research Center for the People and the Press. Network news has massively declined as a source of campaign news. In the 1992 campaign 55 per cent of people said network news was their primary source – in 2000 it was 22 per cent. For the first time in a presidential campaign in 2000, more voters cited cable news (36 per cent) as their primary source rather than either network (22 per cent) or local TV (21 per cent). On the networks, political news has also been pushed to early hours of the morning shows. But overall, even though it was fragmenting, television remained the main source for 70 per cent of voters (down from 82 per cent in 1992). There was an even greater decline in newspapers’ influence, with 39 per cent of voters saying they got most of their campaign news from newspapers compared to 57 per cent in 1992.

In place of traditional media sources, the internet is growing fast, with 11 per cent of voters saying it was their main source of campaign news in 2000, nearly four times the three per cent figure just four years ago. On election night 5.8 million people tuned in to CNN, while CNN’s website had four million visitors, four times more than a week before and third more than MSNBC.com, its nearest rival.

Last year, for the first time, not all the national networks carried the presidential debates but scheduled other programmes, significantly reducing the audience (which was well down from past peaks like 1992). Their argument was that if people wanted to watch the debates they could find them on the 24-hour cable news coverage – this has become a general refrain to excuse declining coverage of “hard news” like politics in network news programmes. The danger of leaving “serious” political coverage to cable TV is that a quarter of Americans do not have cable. Moreover, cable TV news gets a two million audience compared to 20 million for network evening news, and the network news is the only national medium in the US that reaches a significant audience. Its declining importance therefore marks a major shift in American political campaigning.

Related to this decline, the 2000 election saw the coming of age of TV comedy and talk shows as a power in US elections, a trend which began with Bill Clinton playing his saxophone on the Arsenio Hall show in 1992. Half of all adults say they regularly or sometimes get their information on political candidates from non-hard news sources like MTV News or comedy programmes like Saturday Night Live. This rises to 80 per cent for those under 30. The David Letterman and Jay Leno shows have a particularly strong hold. When George W. Bush went on Letterman he got more speaking airtime (13 minutes) in one go than he received in the entire month of October on all three network newscasts combined. On election night, the satirical Daily Show’s coverage on Comedy Central was watched by nearly half a million 18-34 year olds – as many as watched FOX News.

These changes have had a profound impact on campaigns and elections. The stereotypes created by these shows have a huge impact in popular perceptions, and the audiences of these shows are exactly the people who don’t watch normal news much and are normally very hard to reach (the so-called Heineken factor). This makes them very attractive to politicians, especially when faced with diminishing and negative coverage on the main network news. Linked to this was the arrival of narrowcasting or microtargeting – the honing of both a message and its delivery to target a specific niche in the market rather than the country as a whole. The emergence of one-to-one marketing is a general trend in American consumerism and politics, and has followed commercial marketing in both substance and technique.

Marketing

If politics are marginalised, trivialised and individualised then everyone suffers – motivating voters to participate and turn out on polling day requires a sense of the importance of the issues at stake rather than an exploration of which candidate people would rather sit next to on a long journey. This was a major topic for the media during the US campaign, which reflects the influence of TV talk shows and the degree to which politics (and news in general) in America has converged with entertainment with the Presidency (especially in the Clinton years) becoming a soap opera with the President as entertainer-in-chief. Having gone on more trivial media and talked about trivial things, politicians are then notably apt to decry the trivialisation of politics and the failure of the mainstream media to take them seriously.

Some of these trends are unique to America, but many also have relevance for the UK media. While in Britain there is no “calling” of seats before real results come in, the problems with exit polls failing to give an accurate forecast were apparent in 1992 (another example of how little polls can be relied on in close races). More significant are the dangers of the media and politicians together trivialising political debate, whilst also pushing it out of mainstream news. Politicians trying to bypass national media and hard political debate for softer formats (both print and broadcast) can easily compound the media’s sense that, in a competitive environment, serious coverage which involves stretching both politicians and voters rather than just presenting sound-bites is an audience-loser, lowering the level of political discourse to the detriment and frustration of all involved. And the danger of failing to treat campaign news seriously is all too evident from the American experience. According to a December Gallup poll, 65 per cent of Americans – more than at any time in the previous 15 years – say that news organisations’ stories and reports are inaccurate. Only one-third say that news organisations get the facts straight and more than half felt that the news organisations exerted too much influence on the outcome of the election. American politicians, journalists and the public may already be locked into a vicious circle that we would do well to avoid on this side of the Atlantic.