Not for the first time, psephologists got the election result wrong. Ivor Crewe explains why he and his friends failed to predict a landslide

The Conservative party was not the only casualty of the 1997 election. So were Britain's academic election specialists - the psephologists - whose forecasts of the result were outmatched by the pollsters and the City.

The final forecast polls uniformly pointed to a Labour landslide majority of well above 100. The spread-betting prices in the City on election day, which reflected the weight of money invested, assumed a Labour majority of 107. But the panel of 20 academic election specialists appointed by the Reuters press agency predicted a Labour majority of, on average, 92. (My Reuters forecast of 119 was bolder than most but still way off.) We psephologists performed relatively poorly because we lacked confidence in our own analysis of electoral behaviour in the 1992-97 parliament. Many of us pointed out that the familiar postwar parliamentary electoral cycle for the government of a honeymoon, followed by a mid-term slump, followed by a pre-election recovery, had clearly failed to materialise after 1992.

Opinion polls, by-elections and local elections alike showed that the Conservatives never recovered from a massive haemorrhage of support after Britain was forced out of the Exchange Rate Mechanism on "Black Wednesday" in September 1992 and that Labour enjoyed a permanent boost, mainly at the Liberal Democrats' expense, after Tony Blair's election as leader.

³ÉÈËVRÊÓƵ

Ignoring the evidence before our eyes we anticipated a late but strong Conservative recovery, citing models and conventional wisdoms which fitted our instincts but proved to be wrong. The Conservatives, we assumed, were bound to benefit from the rekindling of deep-seated party loyalty; the booming economy, "feel good" factor and voters' secret fear of tax rises. Moreover, the polls were probably exaggerating the true Labour lead, as in 1992.

We should have placed more faith in our own research. The weakening of party loyalties and growth of electoral volatility since the 1950s has been well documented.

³ÉÈËVRÊÓƵ

But instead of recognising that New Labour's relocation to the ideological centre could dislodge millions of loosely aligned Conservatives and Liberal Democrats to trigger a landslide, many of us fell back on the fallacy of historical precedent. Memo for next time: an electoral record lasts for just as long as it lasts.

The most serious casualty of the election is the economic model of voting. "It's the economy, stupid" was the academics' mantra for 20 years before the Clinton campaign adopted it. The central idea is simple: people vote with their wallets. Elections are verdicts on the state of the economy.

The Conservative government presided over an exceptionally strong economy; historically low rates of inflation and (nominal) interest rates; steadily falling unemployment; a strong pound; an above-average growth rate and the Footsie 100 at an all-time high.

Its reward? Not just defeat but humiliation: the smallest number of Conservative MPs since 1906; the lowest Conservative share of the vote since 1832.

Moreover, the largest anti-Conservative swings took p1ace in the boom regions of Greater London (13.3 per cent) and the South East (12.2 per cent) where property prices and job vacancies were rising fastest.

Radical revisions will be needed to salvage the economic model. It cannot be saved by reliance on the "subjective economy'' or "feel good factor". The majority of voters acknowledged that the economy had improved.

As for personal economic pros-pects, optimists outnumbered pessimists by 5 percentage points. David Sanders's economic model of government support, in which personal economic optimism is central, worked perfectly in 1987 and 1992, but this time its forecast of a 37 per cent Conservative vote was 6 points adrift.

What turned the 1995-97 period into Britain's first voteless recovery was Black Wednesday, when overnight the Conservatives squandered their reputation for economic competence. From that point the electorate refused to credit the government for the boom and, for the first time since polls began, consistently regarded Labour as the more reliable guardian of the economy.

³ÉÈËVRÊÓƵ

An Anglia TV campaign poll in Basildon (which Labour gained on a 15 per cent swing) found that 54 per cent agreed that the economy was improving but only 23 per cent of them gave the credit to "John Major's government". Revised economic theories of voting will need to model the attribution of blame and credit to the government. When Sanders's model was refined in this way its forecast of the Conservative vote fell to 33 per cent, only two points adrift.

³ÉÈËVRÊÓƵ

The 1997 election also put paid to the fiscal variant of the economic model: low-tax parties win elections, high-tax parties lose them. After 1992 New Labour became a convinced subscriber to this simplistic idea. The belief that John Smith's tax-raising shadow budget lost Labour the 1992 election is almost certainly a myth. Labour lost because it was not trusted to manage the economy. According to a post-election Gallup poll, 86 per cent of voters expect taxes to rise under a Labour government, but this did not prevent the largest swing to Labour since 1945.

The role of the tax issue in voting decisions is more complicated than a crude "selfish voter" model assumes: what appears to matter is the total tax take, not just income tax, and voters' trust in a party to keep tax rises limited and to spend tax revenue wisely.

The failure of the economic model in 1997 is also a reminder that voters judge parties on their overall competence to govern, of which management of the economy is only one, albeit important, measure. Weak leadership, party disunity, policy failure and scandal are others and they can override economic success. Psephologists need to model voters' trust in a party's general governing competence, not merely its ability to manage the economy.

In underestimating the Labour majority, our main error was to assume a uniform swing across all constituencies. In the event the magnitude - but not direction - of constituency swings was unusually non-uniform, delivering an additional 25 Conservative seats to Labour and 15 Conservative seats to the Liberal Democrats and thus a Labour majority of 179 rather than 129. We should have known better: recent studies - including ones by Reuters' panellists - had shown that constituency campaigning and the strength of local parties could make an appreciable difference to the local vote.

The election proved them right. The Labour and Liberal Democrat parties' ruthless concentration of resources on Conservative-held marginals paid off. The swing to Labour was 10.5 per cent nationally but 12.l per cent in the 80 most marginal seats. The impact of anti-Conservative tactical voting was patchy and complex. The Liberal Democrat vote did fall by more than average in Conservative-Labour marginals, but only slightly. The Liberal Democrats' unexpected gains occasionally arose from squeezing the Labour vote (as in Sheffield Hallam) but more from restraining the Labour surge to below average (as in Colchester). Yet elsewhere, such as in Bristol West, the Liberal Democrats came a good second in 1992, but Labour leapfrogged from third to first place. Psephologists will have a field day teasing out why tactical voting flourished in some but not other promising constituencies and why it worked to the Liberal Democrats' benefit in some cases but Labour's in others.

One other puzzle cries out for the psephologists' attention. All the conditions for high turnout were present at the election: a relatively new register, a lengthy campaign, a record number of nationally campaigning parties and candidates and fine spring weather on polling day.

Yet turnout fell sharply by six points to its lowest level - 7l.3 per cent - since 1935. In many inner- city seats barely more than half the registered electorate bothered to vote.

The above-average falls in turnout in many safe seats suggest that people were less motivated to vote where the outcome was certain. But turnout fell everywhere, including highly marginal and intensely targeted seats, so nationwide factors must have been at play.

Disillusion with politicians and the political process may have been one: a sense that less was at stake than in the 1980s was almost certainly the other. Shortly before the 1992 election Gallup reported that 58 per cent perceived "really important differences between the parties"; shortly before this election the proportion was only 46 per cent.

Labour's ideological dilution and relocation to the centre helped it to win handsomely. Psephologists will want to investigate whether it also contributed to an alarming public disengagement from the democratic process.

This article draws on an electoral database formed for a forthcoming special issue of Parliamentary Affairs, edited by Pippa Norris. Ivor Crewe is vice chancellor and professor of government at the University of Essex.

³ÉÈËVRÊÓƵ