Alistair Smith has written, no doubt, the definitive book on the timing of elections in the United Kingdom. This, together with two party swings, is a subject that greatly concerns politicians and commentators in Britain. The conventional wisdom is that the choice of election date is a crucial decision for the incumbent prime minister, and that this choice gives the incumbent a considerable advantage. This follows from the ability of the incumbent to engage in “political surfing” (holding election when the government's popularity is at its highest) or to actively manage the political business cycle to achieve favorable economic conditions at the time of the election. Smith argues, however, that this advantage is overstated. The very act of calling an early election sends a signal to voters that the government has information about future performance that it wishes to hide from the voters. As a result, when governments call early elections, their actual vote share tends to be considerably less than the opinion polls before the announcement would suggest. For example, when Margaret Thatcher called early elections in 1982, she got 42% of the vote, as opposed to preannouncement polls of 48%.

The author constructs a Bayesian game theoretic model to demonstrate this logic. He shows that it is only rational for a government that anticipates a decline in performance to call early elections and sacrifice the remainder of its term. Realizing this, voters will lower their assessment of the government, making an early election even less desirable to a government confident of future performance. Indeed, one simple version of the model works too well—it predicts that it never makes sense to call an early election. However, when certain refinements are added (not all voters are sophisticated enough too process the signal sent by early elections; early elections catch the opposition off-guard and not prepared for a campaign), we get more plausible results. In this case, early elections will pit incompetent incumbents against ill-prepared challengers.

Of course, the model is susceptible to the criticism that few voters are sufficiently sophisticated to reason in such terms. However, Smith does not need to show that many individuals follow this reasoning. It is possible that many individuals follow cues from relatively few sources, and indeed, he goes to considerable lengths to show that governments that call early elections are accused by the media and the opposition of “cutting and running” and having something to hide. The burden of proof on the author is to show that the electorate as a whole behaves as if it used the information provided by the government's choice of election date.

The remainder of the book tests this hypothesis. Chapters 3 and 4 provide econometric evidence, while Chapter 5 provides case studies. The model predicts that when elections are earlier than expected, government support will be less than it was prior to the announcement. Furthermore, we would expect that early elections would be followed by poor economic performance, in terms of inflation, unemployment, and growth. If the government is trying to catch the opposition off-guard, then we would expect that early elections are followed by short campaigns to maximize this advantage. In spite of the small number of cases (14 elections), the author obtains statistically significant results for all these predictions, with the exception of the effect on economic growth. (Even this would show a significant effect in the expected direction if the exceptional case of the February 1974 election that led to the end of the miners' strike is excluded.)

It should be noted that the independent variable in the model is the timing of elections relative to expectations. A great deal of the methodological ingenuity of the book goes into estimating expectations of elections. This is done in two ways. The first involves a hazards model, whereby the cumulative probability of an election is calculated using such factors as the government's majority and the remaining length of its term. This is then compared to the cumulative probability of an election estimated using these variables, plus future economic performance (which the government is assumed to know). If the election would only be expected knowing what the government knows, then it is a surprise election. The second method relies on counts of newspaper stories in the six months preceding the election. The first method poses some endogeneity problems—future economic performance is used to measure the independent variable, as well as being one of the dependent variables. Fortunately, the second method does not have this problem.

The author does not make a normative case about the relative merit of fixed and variable election cycles, but it is clear where such a case would lead. He refutes the main argument made in favor of fixed electoral cycles (that variable election dates give an advantage to incumbents). Furthermore, he documents various advantages of flexibility in election dates—it allows unviable governments to put themselves out of their misery, and it reduces the incentive for governments to manipulate the political business cycle.

In Chapter 6, the author extends the analysis to other countries, such as Australia and Canada. However, the analysis in its present form can only be applied to countries where a single agent chooses the election date. It is notable that the 1979 election is excluded from the analysis because the government was brought down involuntarily after losing a vote of no confidence. In many multiparty parliamentary democracies, this is the usual way in which governments fall. However, Election Timing has a broader significance in another sense. There has been a great deal of theoretical literature suggesting that elites can manipulate democratic institutions, by exploiting legislative voting cycles, the provision of information, or the political business cycle. However, recently there has been a substantial body of work that suggests that this manipulation is far harder to achieve than was once thought. This book adds to this growing consensus.