INTRODUCTION

One of the strongest regularities in American elections is the midterm cycle: the party holding the White House loses votes and seats in each midterm congressional election, relative to the previous, on-year, election. Figure 4.1 displays this remarkable regularity for the period since 1918. The president's party lost vote share in all of the 18 midterm elections shown in the figure. Very few phenomena in politics appear so regularly and consistently: the midterm cycle calls for an explanation that goes beyond specific personalities, or specific events such as, for example, Watergate or the Vietnam war.

In this chapter we argue that the midterm effect is part of the balancing effort of the electorate, who face two polarized parties. We study an important element of uncertainty concerning the distribution of voters' preferences: the model of chapter 3 is extended by the assumption that the ideological position of the median voter is not known with certainty, by anybody. Even though the parties' ideal policies continue to be known by everybody, this uncertainty concerning voter preferences makes the electoral results, and, in particular, the results of the presidential elections uncertain. The voters do not know for sure which party is going to win the presidency; they can only evaluate the probability of the two possible outcomes, as in chapter 2. In midterm elections, instead, they know which party holds the presidency.

The difference between the on-years, in which legislative elections are held under uncertainty about the identity of the president, and midterms, in which the uncertainty is removed, is the key to the midterm cycle.