The Cash Nexus: Money and Politics in Modern History, 1700-2000

by Niall Ferguson

  This chapter shows that, over the long run, economic factors have in fact played a minor role in election outcomes, in the British case at least. Indeed, there was only a relatively short period – the heyday of ‘Keynesian’ demand management in the 1960s – when the relationship was even close to the one implied by the ‘feelgood factor’ theory. Otherwise, there has seldom been a stable causal link from economic to political success, for two reasons. First, political attempts to manipulate the economic cycle have generally had unanticipated negative consequences: this was painfully obvious in the 1970s, when (as we saw in Chapter 5) fiscal and monetary policies aimed at boosting unemployment were blamed for accelerating inflation. Secondly, voters do not simply reward incumbents when the economy has grown and punish them when it has not: their responses to economic change are far more complex.

  THE BRITISH CASE

  In order to take the longest possible view of the role of economics in British politics, it is worth going back to December 1832, the first general election held after the Great Reform Act. As Table 11 shows, governments have not always been made or unmade by general election results. There have been approximately fifty-one governments since 1832: slightly more if we regard every election as marking a change of government; slightly fewer if we only count changes of prime minister. And there have been forty-one general elections. Of all the fifty changes of government identified here, no more than eighteen can be said to have happened as a result of election defeats. A total of twenty-two incumbent governments have won elections. (The other two elections had quirky results: the Conservatives lost the 1852 election, but did not fall from power for some months, while the election of November 1885 was a draw, in the sense that the balance of power in a hung parliament was won by the Irish nationalists.) The other reasons for changes of government are summarized in the table.

  It will be seen at once that economic questions can only be directly blamed for four changes of government: the fall of Peel over the repeal of the Corn Laws in 1846 (though strictly speaking he was defeated over an Irish Coercion Bill); the fall of the Derby ministry over its budget in 1852; the collapse of Arthur Balfour’s government over Tariff Reform in 1905; and the disintegration of Ramsay MacDonald’s Labour government over unemployment benefit cuts in 1931. By comparison, the Irish question has been just as important in bringing down governments, and questions of foreign policy and defence significantly more important (nine changes of government in all). No fewer than eleven changes of prime minister have been due to ill health, though none of these led to a change in the party complexion of the government.

  Table 11. Reasons for changes of government or prime minister, 1832–1997

  * * *

  Election

  18

  Economic question

  4

  Constitutional question

  1

  Irish question

  4

  Foreign affairs question

  4

  Military question

  5

  Monarch

  2

  Death or illness

  11

  Other

  1

  * * *

  Is it possible to be any more precise than this about the role of economic factors in British elections over the long run? Table 12 provides some stark evidence on the role of economic factors in British elections since 1922. In the absence of opinion polls for the pre-1945 period, the most that can be done is to look at the increase or decrease between elections of the vote of the party in power before the election, and compare this with changes in two key economic indicators over the same period, inflation and unemployment. The figures suggest that there was little or no correlation between economic success and election success in twentieth-century Britain. In only ten out of twenty-one elections did a rise in inflation coincide with fall in a government’s share of the popular vote. For unemployment (assuming that a rise would lead to a fall in popularity) the figure is just nine. Simple correlations reveal a negative relationship between government performance and inflation, suggesting the possibility that governments did better if they reduced inflation; however, regression analysis reveals this as spurious. The only statistically robust relationship is a positive one between government popularity and the unemployment rate.

  Table 12. British economic indicators and election results, 1918–1997: change since previous election

  Note: In the case of coalition governments, only Conservative votes have been counted, as in each case the Conservatives were the dominant partner.

  A further counter-intuitive inference may be drawn from these figures. Big government defeats in the past have tended to come at times of relative economic prosperity. This seems at least plausible in December 1923 and May 1929, as well as in May 1997. It also applies to the election of 1906, the result of which (a Liberal landslide) bears comparison with the Labour victory in 1997. Although an economic policy issue is usually seen as having played an important part in the election – namely the divisions within the Conservative government over Tariff Reform – the economy itself was in good health: unemployment had been falling since 1904 and exports had risen sharply since the previous election.45 Conversely, a number of major government victories have coincided with economic deterioration: the classic pre-war example is October 1931, in the depths of the Depression. The Major government’s victory in April 1992 also came in the wake of a recession which (if the feelgood factor had been decisive) should have seen Labour sweep to victory.

  It is possible to take a closer look at the post-1945 era, using monthly opinion poll, unemployment and inflation data. Figure 21 compares the government’s lead in the opinion polls with a ‘misery index’ which simply adds together the current unemployment and inflation rates. Superficially, there are traces of the inverse relationship predicted by the feelgood model, namely that rises in unemployment and inflation would make governments less popular. When the misery index rose sharply in 1951, the popularity of the Attlee government fell, leading to its defeat in the election that October. The subsequent troughs of the misery index seem to coincide closely with the peaks of Conservative popularity in the summer of 1955 and the autumn of 1960; while the rise of the misery index to above 7 in May 1962 saw a slump in the Tory lead. But from the mid-1960s onwards the relationship is less distinct. At a Cabinet meeting on 14 July 1966 Wilson commented revealingly on a poll that showed Labour 16 points ahead of the Conservatives: ‘He couldn’t understand how or why! It appeared the more unpopular the measures we took, the more popular we became.’46 Yet when the government’s lead collapsed almost immediately thereafter, its loss of popularity was out of all proportion to the rise in misery captured by the index. The comparable collapse between 1974 and 1977 seems more readily explicable in economic terms; but comparably high misery had much less of an effect on the Thatcher government’s popularity in 1980 and 1981. Thereafter, the government lead appears to move more or less independently of the misery index.

  Using more sophisticated statistical methods to distinguish the effects of unemployment, inflation and interest rates,47 we can be quite precise about the decline of the feelgood factor (see Appendix B). It was certainly present at the beginning of the period, as Figure 21 suggests. Under the Conservative governments of the 1950s, a 1 per cent rise in unemployment was associated with a fall in the government lead of nearly 5 per cent; the effect of rising inflation and interest rates was also negative, though less damaging. For the first Wilson government, the effect of rising unemployment was even more serious: a rise in unemployment of just 1 per cent could reduce government popularity by nearly 10 per cent. Rising interest rates were also associated with falls in government popularity; though higher inflation seems to have had no significant effect. Interestingly, the figures under Heath were extremely close to those under his Conservative predecessors. And unemployment was associated with bigger falls in support under Wilson and Callaghan between 1974 and 1979.

  Figur
e 21. Government lead (left-hand axis) and the ‘misery index’ (right-hand axis), 1948–2000

  Sources: Butler and Butler, British Political Facts; Central Statistical Office, Monthly Digest of Statistics; HMSO, Ministry of Labour Gazette; Central Statistical Office, Retail Prices 1914–1990. Recent data are from the following websites: Bank of England, HM Treasury, statistics.gov.uk, except for the opinion poll data, which are from the Gallup Organisation (as published in the Daily Telegraph).

  Notes: The indicators used in the analysis are as follows: government lead: percent age of respondents who would vote for governing party if an election were held tomorrow less percentage who would vote for principal Opposition party.

  RPI: retail price index, percentage change over the previous year.

  Unemployment figures (defined from January 1971 as the claimant count) are given without seasonal adjustment.

  However, only unemployment had a statistically significant relationship with the popularity of the Wilson–Callaghan governments. Indeed, for all subsequent governments only one of the three indicators is significant: interest rates in the case of the Conservatives, unemployment in the case of Labour. This suggests that Mrs Thatcher may have been successful in shifting the attention of voters away from unemployment, which had a slightly positive but insignificant relationship with government popularity between 1979 and 1992. It is perhaps more surprising, given Thatcher’s anti-inflation rhetoric, that the correlation between inflation and government unpopularity was not stronger. But Ridley’s ‘law’ – ‘the higher the interest rate, the less popular has a government been in the opinion polls’ – does appear to have held: every 1 per cent increase in base rates correlated with a 3 per cent drop in government popularity.

  What of the period leading up to the 1997 election? Given John Major’s public commitment to price stability as a ‘gain’ worth substantial ‘pain’, it is not surprising that the correlation between inflation and government popularity was strongly negative between 1992 and 1997. The most striking feature of the years 1992–7, however, was the perversity of the correlation between interest rates and the government’s position in the polls. Bizarrely, a 1 per cent rise in the Bank of England base rate was associated with an increase in government popularity of around 8 per cent; or, to be exact, falling interest rates after September 1992 coincided with a collapse in government support. It was this inversion of the ‘feelgood’ model – on which they had faithfully based their campaign – which condemned the Conservatives to defeat. Even more perplexing is the positive relationship between unemployment and government popularity since 1997, which shows that the popularity of the Blair government has actually declined as unemployment has continued to fall – an unprecedented phenomenon in the history of the modern Labour party.

  THE POLITICAL BUSINESS CYCLE

  There are two ways of explaining why electoral outcomes might not – or at least not always – be determined by economics. One is that politicians simply lack the skill to manipulate the economy successfully.

  In his Economic Theory of Democracy published in 1957, Anthony Downs proposed that ‘parties formulate policies in order to win elections, rather than win elections in order to formulate policies’ and that, once elected, ‘democratic governments act rationally to maximise political support’.48 This formed the basis of William Nordhaus’s theory of the ‘political business cycle’, which suggested that governments would tend to manipulate the economy so that the economic cycle would peak shortly before they came up for re-election. As the Benn diaries and other sources quoted above show, there is little doubt that they try to. And there is some empirical evidence for such behaviour. In some countries at least – the United States, Germany and New Zealand, for example – unemployment has appeared to follow a political cycle, rising in the first two years of government and falling in the last two years.49 As we have seen, the British data do not show the same pattern. However, it has been demonstrated, using slightly different methods, that in two-thirds of election years up until the 1970s, the increase in disposable income rose above the mean in the pre-election year.50

  The trouble with the expansionary policies used to get down unemployment was, of course, that they generated higher than anticipated inflation. As politicians primed the pump with increasing frequency, the Phillips curve – the apparently close relationship between employment and inflation – began to steepen. The conclusion many commentators drew was that the political business cycle might after all be unsustainable because, in the words of Samuel Brittan, it gave rise to ‘the politics of excessive expectations’.51 To Peter Jay, writing in the mid-1970s, ‘a crisis of political economy’ seemed imminent.52 In Britain and America that crisis took the form of a counter-inflationary backlash under Thatcher and Reagan.

  The polarization of politics which occurred in the late 1970s prompted a ‘partisan’ modification to the political business cycle theory. Perhaps different parties had different policy preferences: left-wing politicians worrying more about unemployment because of their working-class constituents, while conservatives worried more about inflation because of their rentier supporters. One influential study done in 1977 calculated that, on average, post-war unemployment had been higher in Britain under the Conservatives than under Labour.53 Democrats too were likely to aim for lower unemployment and higher inflation than Republicans.54 The election of conservative governments in many countries therefore did not represent the end of political manipulation of the economy so much as a realization that the benefits of inflation had been overtaken by the costs. As Brittan constantly lamented during the 1980s, and as the politicians’ memoirs confirm, the political business cycle lived on under the Conservatives, with interest rate changes carefully timed with one eye (or both) on their political impact. There seems little doubt that in the period 1983–7 the Tories did ‘manipulate the money supply … in order to influence public opinion’, responding to evidence of their own unpopularity by relaxing monetary policy.55 A further theoretical explanation for the well-timed profligacy of some conservative administrations is that conservatives may raise fiscal deficits (through tax cuts) precisely in order to constrain left-wing rivals if the latter are likely to come to power, by forcing them to limit or even cut public spending on the other side of the budget.56

  However, such political tinkering with nominal indicators might have less visible real effects if voters could see through the politicians’ intentions. In that case, the political business cycle would be more likely to show up in the budget or monetary policy than in growth, employment or inflation data. Studies comparing all the OECD countries appear to bear this out.57 One possible explanation of the decline of the feelgood factor after the 1970s might therefore simply be public disillusionment as the policies of one government after another generated new economic grievances. To put it another way, the more a government targeted one particular variable, the more likely it was to cease to correlate closely with its popularity, as other problems developed elsewhere in the economy – a variation on Goodhart’s famous law that the very act of basing policy on one indicator may undermine its predictive power.

  There is, however, a second explanation for the non-existence or decline of the political business cycle. This relates to the other – and much more complex – human variable in the electoral equation, namely the voters.

  THE VOTE FUNCTION

  The political scientist Helmut Norpoth has written: ‘The economy is a concern that almost everywhere bonds electorates and governments as tightly as Siamese twins joined at the hip…. Economic voting … is hard-wired into the brain of citizens in democracies.’58 Yet the evidence suggests that the circuitry is highly complex and may occasionally blow a fuse. Indeed, any idea of a simple causal link from prosperity to popularity must be abandoned in the face of a mass of empirical research from around the democratic world.59

  Using data from thirty-eight countries, but comparing only the vote for the major party in office in the most recent e
lections and the poll and pre-poll figures for real GDP growth, Norpoth has found (using a simple regression) that ‘for every percentage point that real GDP grows in the election year, the major incumbent party stands to gain roughly 1½ per cent of the vote above its normal share’.60 Another study found similar evidence of ‘economic voting’ in Europe and the US in the early 1980s.61 However, Paldam’s 1991 survey of the ‘vote function’ in seventeen OECD countries over four decades found only superficial relationships between votes for parties in power and a variety of economic indicators (the change in unemployment, price rises and GDP growth). When subjected to more rigorous statistical tests, these relationships turned out to be weak, even when allowance was made for such variables as the political complexion of the government, the number of parties, and the size of the country.62 Only by making adjustments to take account of differences in the degree of the responsibility of government parties for economic performance (given differing political systems), and by considering inflation and unemployment in comparative rather than absolute terms, can significant links from the economy to elections be found.63

  More detailed work on specific countries (a great deal of which, it should be stressed, focuses on the United States) has raised eight questions about the way the economy and voting behaviour are related:

  Do voters care about inflation, unemployment or some other measure of economic well-being?

  Are voters motivated by individual self-interest or do they have regard for the common good?

  Do voters view the economy differently according to which party they identify with?