THE SECRET OF CLINTON'S POLL LEAD

It's that it's not based on anything scientific, says James Srodes. Even the pollsters admit it (privately)

Washington, DC SEVEN weeks of hard campaigning are still left in the American presidential cam- paign. Yet most political journalists are convinced that Bill Clinton is so certain of re-election that they prefer to spend their idle hours debating who will succeed to power when the inevitable pre-Christmas cabinet reshuffle occurs. The consensus is so firm because the much-touted opinion polls conducted by the journalists' respective organs declare with- out equivocation that President Clinton holds a 'double-digit' lead over Bob Dole in the standard question: 'If the election were today, would you vote for A or B?' That lead is variously described as being in the low 50s per cent among those who prefer Clinton; voters for Dole amount to the high 30s, while 8 or 9 per cent prefer Ross Perot, and a dozen or so claim to be undecided. Yet there is an edginess about the jour- nalists when the question turns to the sys- tem of polling itself. Is that data accurate? Is a fair interpretation made of this vital news which drives the decisions of candi- dates and their political strategists and, at the end, influences even the voters' deci- sion on 5 November?

'Oh, the polls. Well, that is our dirty little secret,' sighed a New York Times writer, who was relaxing at a Georgetown dinner party after a week on the mad with one of the can- didates. In partnership with the CBS televi- sion broadcasting network, the New York Times poll is one of the fiye dominant sam- plings of opinion driving this election, along with the consortium of Gallup-CNN (the cable network) and USA Today newspaper, the Washington Post-ABC alliance, the Los Angeles Times, and the NBC-Wall Street Jour- nal. These five are the reports most often quoted by other neWs media (and by each other) in the race to tell the American people what they are thinking about the election and how it is likely to finish.

Why the sigh? What's the 'dirty little secret'? The Times man fidgets a bit and then explains: 'You see, the budget on the polls is really tight and we don't really use senior people either to conduct the poll or to interpret the results. Basically, what hap- pens is that we farm the actual polling out to CBS, they hire casuals to come in and make the telephone calls, and then we both share the data and pass it up to the news- rooms to make of it what they will. Frankly, I don't pay any attention to it.'

It is worth noting that at this point in the

summer of 1994 none of the major media polling alliances picked up the groundswell that carried the Republicans to control of both chambers of Congress. Calls to the Washington Post reveal that it has only two reporters who write polling stories. The actual data collection is overseen by Ken- neth Winneg, an executive with the Chilton Research Service, a commercial data collector. Are the big media polls accurate?

'I've been doing polling for the last ten years and, statistically, the job is done bet- ter than ever. If there is a weak spot, it is how the data is used after we pass it on. I'm not referring to our own clients, mind you. But some of the other journalists don't do as well in interpreting the data they get,' he said.

Kathleen Francovic, the CBS executive who directs the joint poll done also on behalf of the New York Times, goes further: 'It's scary. Over the last ten years or so jour- nalists, and even academics — who should know better — have taken what we produce and treated it as gospel. We have a situation now where our data measuring the Presi- dent's approval rating is now viewed as the real public support for the President. Jour- nalists who are so sceptical of other things are not at all cautious about our data.

'You can see this in stories which begin,

"Polls show ... " or, "Polls indicate . ". It has become so much a part of our cul- ture now that it is unfortunate. We have our own problems on the statistical side. We have competition from telemarketing firms who call many of our same people; we have problems in asking questions the appropriate way; we worry that we may be doing something that in some way makes for biased data. But we can control those things. What we can't do much about is how the data is used when it gets out of our hands,' Francovic added.

'What I don't want is to wake up on 6 November and have to explain how we got it wrong again,' she laments.

Most of the big media polls are conduct- ed the same way, with small variations. On polling day, part-time casuals (perhaps a hundred of them) arrive at the phone banks late in the afternoon. Each is paid an hourly wage and ten to 15 calls are assigned. The so-called horse-race format is followed: 'If the election were held today .. . ?' Respondents are limited, for data-collating purposes, to fixed-response choices, roughly, Yes, No, Don't Know, or some variation such as Agree Strongly, Disagree Strongly, Agree/Disagree Some- what, and the ever present Don't Know.

If one does one's sums, it is obvious that a survey of nationwide American opinion thus involves actually talking to perhaps as few as 1,500 respondents at a time. Statisticians do confirm that for statistical purposes that is enough, as long as the sample is unbiased. Yet the samples conducted by the major media groups are hardly 'random'.

First of all, the samplings are universally done by telephone, which rules out many poor people and young house-sharers. Then again, 'random' in poll-talk does not mean strictly at random. Telephone area codes, which pick states, regions, cities and even neighbourhoods, are picked by a closely guarded weighted formula which the statisticians believe produces the most accurate picture of American opinion. Moreover, the 1,500 or so people talked to are sometimes broken down further into registered voters and those who are not only registered but are determined to cast their ballot in the coming election. A third weighted formula selects responses by the respondent's race, gender, age, income level and education level.

But America is not just a narrow con- struct of rigid data determined by race, gen- der and income. Can one really argue that black men who are high school graduates, earn $25,000 a year and live in the Mid- West all think alike? Some permutations of the American electorate are wildly unstable. In more detailed academic studies that fol- lowed the 1994 upheavals at the polls, the most dramatic shift in voting allegiance occurred among a group described as 'women, some college'. This refers to a part of the electorate that is mostly young, single, often with at least one child, and with one or two years of post-high school education. In 1992, these women responded to Mr Clin- ton's boyish charm and promises of social programmes for working mothers by voting almost 85 per cent in his favour. In 1994, just two years later, these same women — now disgusted with Clinton's personal scandals and his failure to deliver on his campaign promises to them — turned out and voted for Republican Congressional candidates by an identical percentage. Polls misread their views in both elections.

It is no secret that most of the major American news media, their reporters and editors have a vested interest in Clinton's re-election. There has been a minimising on the news pages of public revulsion at the continuing sex scandal involving the former White House strategist Dick Mor- ris. Doubts about the coming Bosnia elec- tion and Mr Clinton's pledge to have US troops home by Christmas are largely ignored. So are the suspicions that Saddam Hussein has emerged in a stronger position at home as a result of the American bomb- ings. If Mr Clinton's poll results are above 50 per cent, his disapproval rating is firmly among the highest of any president: nearly 40 per cent of the electorate don't trust the man, no matter what.

By 5 November, American voters will have seen how Bob Dole and Jack Kemp define themselves on policies that most voters already strongly share: the size of government, welfare reform, lower taxes. There will have been televised debates. There is still time to think. And, as they did two years ago, the voters may astonish the people who poll them.