THEORIES OF POLITICAL BUSINESS CYCLES predict that the quadrennial election cycle in the United States should affect the timing of the peaks and troughs of United States business cycles. Various political business cycles theories have different implications concerning this relationship. The opportunistic political business cycle model (Nordhaus 1975, Lindbeck 1976, Tufte 1978) suggests that a business cycle trough (that is, the beginning of an expansion phase of the business cycle) is likely in the period before an election as an incumbent attempts to increase the chance of reelection. This model also suggests that a business cycle peak (that is, the beginning of the contraction phase of a business cycle) follows soon after an election as the preelection stimulus is reversed. Alternatively, partisan political business cycle theories suggest that the likelihood of a peak or a trough following a presidential election depends upon which party was victorious. Rational partisan theory (Alesina 1987) suggests that a business cycle peak marking the end of an expansion is more likely in the wake of a Republican presidential victory than at other times and less likely after a Democratic presidential victory than at other times. Conversely, a business cycle trough is less likely after a Republican has won a presidential election than at other times and more likely after a Democrat has won than at other times.(1)
The predictions of these political business cycle theories correspond to the popular view that in politics timing is all. Previous empirical research on these theories, however, only addresses the timing issue indirectly by focusing on the amplitude of macroeconomic variables before and after elections or across the tenure of different parties (for example, Alesina and Roubini 1992). In this paper we provide a more direct test of the temporal links between political and economic events. We use duration analysis to test whether the likelihood of the occurrence of a business cycle turning point in the United States (that is, either the end of a contraction or the end of an expansion) is significantly affected by the occurrence and the outcome of an election.
Duration analysis is particularly well suited for analyzing the temporal links between elections and business cycle turning points. Duration analysis allows for directly testing the determinants of the likelihood of the end of a business cycle phase in any period conditional upon the phase lasting up until that period.(2) The determinants of the timing of peaks and troughs that we focus on in this paper are the occurrence and the outcome of presidential elections. Duration analysis enables an estimate of the effect of elections on the likelihood of the end of a business cycle phase holding constant other factors. In particular, duration analysis controls for duration dependence that arises when there is a changing probability of the end of a business cycle as the cycle itself progress.(3)
The empirical results presented in this paper do not support the prediction from opportunistic political business cycle theory that a contraction is more likely to end in the period before an election than in other periods regardless of the political party of the president. We do find, however, a significant increase in the likelihood of the end of a contraction ceterus paribus in the two-year period before an election when there is a Democratic president. There is also evidence that expansions are more likely to end soon after an election than in other periods. This is consistent with predictions of opportunistic political business cycle theory.
This result for postelection business cycle effects is further examined in the context of rational partisan theory by disaggregating postelection periods according to which party won the election. Consistent with rational political business cycle theory we find a significant increase in the likelihood of the end of an expansion and a significant decrease in the likelihood of the end of a contraction following the election of a Republican president. We find less significant evidence, however, of predictions from rational partisan theory consistent with the behavior of the economy in the period following the election of a Democratic president.
In the next section of the paper we discuss the manner in which we use duration analysis to test for the effect of the U.S. quadrennial election cycle on the turning points of its business cycle over the period 1855 to the present as well as for different subsamples. Results follow in the second section of the paper. Concluding comments are offered in section 3.
1. ELECTIONS AND THE DURATION OF BUSINESS CYCLES
Theories of political business cycles differ in their assumptions across several dimensions.(4) One dimension concerns the nature of the economy itself. For example, the early literature, such as that by Nordhaus (1975), Lindbeck (1976) and Hibbs (1977), assumes that the economy is characterized by a stable inflation-output trade-off, inflation is directly controlled by policymakers, and expectations of inflation are adaptive. More recent work reflects the rational expectations critique of these assumptions. The basic assumptions of Persson and Tabellini (1990), Rogoff and Sibert (1988), Rogoff (1990) and Alesina (1987), for example, are that people are forward-looking and make decisions based upon all information available to them at the time. The link between attempted political manipulation and phases of the business cycle is more tenuous under these assumptions than under the assumption of a stable Phillips curve. In particular, there is little scope for a preelection stimulation of the aggregate economy (though there is a scope for political budget cycles) since the dates of quadrennial elections are known in advance. Postelection effects of elections are more short-lived when people are rational and forward-looking than when there is a stable Phillips curve.
The motivation of policymakers represents another important dimension along which models of political business cycles can be categorized. Opportunistic political business cycle models, such as Nordhaus (1975), assume that the goal of all policymakers is to be reelected and policy is used toward this end. Incorporating forward-looking behavior into this model, as in the rational opportunistic political business cycle model of Persson and Tabellini (1990), mitigates the extent to which the economy can be moved by policy and makes the voters' goal to elect the most "competent" candidate regardless of ideology.
The goal pursued by policymakers in partisan political business cycle models is not reelection but instead realizing ends commensurate with their ideology. In the work of Hibbs (1977, 1987), in which politicians can exploit a stable output-inflation tradeoff, this leads to differences across the tenure of left-wing and right-wing governments. The rational partisan theory of Alesina (1987) retains the assumption of policymakers pursuing ideological motives but tempers their ability to realize their goals by modeling an economy characterized by rational wage-setters who are temporarily bound by nominal contracts. In this model, wages are set equal to expected inflation. In the period before an election, the expected inflation rate is a weighted average of the likelihood of the election of the party more sensitive to the costs of inflation (in this case, the Republicans) and the party less sensitive to inflation's costs (the Democrats). The outcome of the election determines the actual inflation rate and therefore whether real wages are unexpectedly high (due to the victory of the Republicans) and there is a contraction or whether real wages are unexpectedly low (due to the victory of the Democrats) and there is an expansion. The length of the deviation of output from its natural rate in this model is the length of the wage contract, not, as in Hibbs' model, the entire tenure of the administration.
These theories present different testable implications concerning the temporal relationship between elections and business cycle turning points.(5) Opportunistic political business cycle theory predicts a higher likelihood of a business cycle trough (that is, the end of a contraction) with the coming of an election. This theory also predicts that the onset of a contraction (that is, a business cycle peak) to offset the preelection stimulative policy is more likely following an election than at other times. These predictions stand regardless of the party in power. Alternatively, the party in power is central to the predictions of the timing of business cycle turning points drawn from the insights of the rational partisan theory. Rational partisan theory predicts that the likelihood of a business cycle peak marking the end of an expansion is higher after the election of a Republican president than at other times and is lower after the election of a Democratic president than at other times. This theory also predicts that the likelihood of a business cycle trough marking the end of an expansion is lower after the election of a Republican president than at other times and higher after the election of a Democratic president than at other times.
The timing of business cycle turning points relative to the quadrennial United States presidential election cycle implied by these theories lends itself to an empirical investigation using duration analysis. The data used in duration analysis consist of spells. In our data, a spell represents the number of months in either a contraction or an expansion. The focus of duration analysis is the hazard function. The hazard function at time t, h(t, x(t)), is an estimate of the probability of the completion of a spell during the time (t, t + dt), given that the spell has lasted up until time t. We estimate the hazard function for the probability of a peak (trough) in the business cycle during the next month given that the economy has been in an expansion (contraction) up until the beginning of that month.(6) The hazard function exhibits duration dependence if its value at any moment is a function of the time already spent in a spell. The hazard function may shift due to exogenous factors, represented by the vector x(t), which are called covariates. In a continuous time framework the hazard function is defined as
[MATHEMATICAL EXPRESSION OMITTED]
The hazard function can be interpreted as the probability of a turning point in the short interval dt after t, …