Consistent growth may not be enough for voters

Kenneth Clarke cried yesterday: "If we fight the election on the economy, we will win." Well what else would you expect from a Conservative Chancellor at his last party conference before a general election, determined to rally the troops?

But with voter-bribing tax cuts still eluding him, Mr Clarke is indeed hoping that the past few years of consistent economic growth will pay off in votes. And he has a long tradition of analysis and punditry to boost his optimism. Economic success wins elections for incumbents, so the received wisdom goes. Or, in the words of the incorrigible Clinton adviser, James Carville: "It's the economy, stupid."

Moreover, the received wisdom has statistics to back it up, in the US at least. For more than 20 years, economists in the US have shown that a relatively simple econometric relationship holds between economic growth and the political parties' share of the Presidential vote. Roughly speaking, if you push up growth by 1 per cent in the last year of a Democratic President's term of office, then you push the Democratic candidate's vote up by around 1 per cent in the subsequent election (see graph).

However, the implications for Mr Clarke are not as cheerful as they may seem. In the 1992 US Presidential election, the model broke down. Economic growth simply did not deliver votes for George Bush. Moreover, the several possible reasons why growth wasn't enough for the Republicans in 1992 will all be worrying for Mr Clarke too.

The original economists' models of the links between growth and votes in the US emerged in the Seventies. Yale Professor Ray Fair considered Presidential elections from 1916 to 1976 and developed a model which proved accurate (more or less) at predicting the 1980, 1984 and 1988 elections. The share of the vote, according to Professor Fair's model, depends on whether the candidate is already the incumbent President, the level of economic growth that year and very slightly on the level of inflation. Professor Fair also found that voters seemed to have short memories. Only growth in the last year before the election was significant.

If Professor Fair's original model held true today in Britain, Mr Clarke could relax. Forget the recession, forget Black Wednesday, forget all those tax increases; the important thing for voters is the growth record for the next six months, which everyone agrees will be good.

However, even in the US, the original Fair model no longer works. It predicted that Mr Bush would win the 1992 election comfortably. He was the incumbent, inflation was low, and his recent economic growth record wasn't bad. Instead, Bill Clinton won with 53 per cent of the vote.

Amused by how seriously the world took his original model, Professor Fair has offered several revisions to the theory to explain what happened in 1992.*

For a start, it could be lack of good news. Although economic growth in the US in 1992 was reasonable enough, over the previous three years growth remained consistently below average. At no point was there even a blip of good economic news. Taking this into account, Professor Fair found that the 1992 election result seemed far more consistent with its predecessors after all.

But if economic good news does matter, and if British voters are affected in the same way, then the implications for Mr Clarke are not good. The UK economy has not exhibited any "good news" blips in the past three years either. Steady, unexciting rates of growth may be good for the economy over the long term, but if they don't enthuse voters, Mr Clarke will be seeking a new job next year.

The second new factor considered by Professor Fair to explain the 1992 election result is boredom. He calls it the duration effect. When a party has been in power too long, he suggests, voters are so bored and fed up they are likely to turf the incumbent party out regardless of the state of the economy. And the statistics support his theory. Again the implications for Mr Clarke, if the results cross the Atlantic, are not good.

But even Professor Fair's revised model of the connection between economies and election results may miss out the most important factors. What about real wage growth for a start?

When Mr Carville talked about the economy, he talked about the stagnant real wage growth of average workers in the US, rather than overall levels of economic growth. Plug that into the equation, and the 1992 result may make more sense.

Real wage growth hasn't been wonderful for the Conservatives in the UK either. For all Mr Clarke's claims that voters are now better off, wage growth has been relatively sluggish in the past couple of years, and the tax cuts that have contributed to rising living standards this year come so hot on the heels of tax increases in the past that voters may well be suspicious.

In fact, that suspicion could well be the most important factor that breaks the model both here and in the US. Voters may be influenced not so much by economic variables but by politicians lying about economic variables. In 1992, Mr Bush went to the polls having broken his promise not to raise taxes. John Major and Mr Clarke are about to do the same thing.

As the graph from Goldman Sachs shows, consumer confidence and government popularity moved roughly together over 20 years until 1993, when tax rises started to kick in. Could it be that breaking such politically sensitive promises has decoupled voting intentions from the state of the economy?

Unfortunately Professor Fair's econometric model can't give us the answer, because it is so hard to measure political trust.

As Professor Fair himself cheerfully admits, there are huge weaknesses in plotting statistical relationships to explain such complex matters - not least the difficulties in measuring important variables. We have so little data to make such sweeping claims.

Even going back to 1916, Professor Fair is only able to consider 20 Presidential elections. Search a small data set for a long time, throw in enough variables and some kind of statistical relationship is bound to emerge. But there is no guarantee that the relationship you actually find explains anything at all.

Professor Fair, with his revised model, predicts that Mr Clinton should win 49.5 per cent of the vote in November this year.

If he is accurate, we should give his model another look. In the meantime, however, the many other possible explanations for Mr Bush's 1992 defeat seem far more persuasive, and more worrying for Conservatives, who rest their electoral hopes on economic growth.

* "Econometrics and Presidential Elections" by Ray C Fair, Journal of Economic Perspectives, Summer 1996.