Political parties and the variable duration of business cycles.

• Author: Haynes, Stephen E.

1. Introduction

   A number of recent studies explore how political factors influence real economic activity. With electoral models, economic performance peaks near the election, as in the retrospective model of Nordhaus. |30~. Electoral cycles also emerge (at least in fiscal policy) in the rational-expectations model of Rogoff and Sibert |34~ and Rogoff |33~. With partisan models, economic activity differs systematically by party, as in both the ideological model of Hibbs |26~ and the rational-expectations model of Alesina |1~. Economic activity surges after election of Democratic presidents and recedes after election of Republican presidents. Although recent tests of the electoral approach are supportive |21; 22; 24; 25~, the partisan approach enjoys the more compelling support |2; 3; 9; 10; 23; 26; 27~. The electoral and partisan views are not mutually exclusive, as the possibility of conflict between ideology and the desire to be reelected as the election approaches is well recognized. Indeed, our evidence here is consistent with a combined electoral-partisan model.(1)

   The macroeconomic literature at large, however, has been slow to absorb accumulating evidence that the party affiliation of the incumbent president, together with the timing of the electoral period, are linked to real economic activity. In this study, we gauge the significance of these links for the United States and relate them to two important strands of the macroeconomic literature. One strand, the "time-series" literature, investigates the statistical properties of measures of real economic activity, finding that these measures often exhibit nonconstant means and heteroscedastic variances. In autoregressive, integrated, moving-average (ARIMA) models, the conditional mean is a function of past innovations, while the conditional variance is constant. Applications of ARIMA and related models, beginning with Nelson and Plosser |29~, suggest that real economic activity for the United States exhibits near random-walk behavior. In autoregressive, conditional heteroscedastic (ARCH) models, the conditional variance is a function of past (squared) innovations, while the conditional mean is constant. Applications of ARCH and related models, beginning with Engle |17~, suggest that real activity exhibits heteroscedastic variances.

   The second strand, the "structural business-cycle" literature, examines not only heteroscedastic variances (that is, variable amplitudes) in cyclical movements of real activity but also variable durations. Blanchard and Watson |7~, for example, investigate several questions: is the macroeconomy influenced by one type or many types of shocks; are these shocks occasional and large or numerous and small; and is the resulting structure of real output constant or changing? Their analysis rejects a constant duration for business cycles, and in fact "cast(s) doubt on the usefulness of using 'the business cycle' as a reference frame in the analysis of economic time series |7, 146~.(2)

   Few studies have attempted to link directly these three attributes of real activity--nonconstant means, heteroscedastic variances, and variable durations--to observed changes in structural or fundamental determinants.(3) In this study we investigate one specific explanation for these attributes: changes in policy regime, identified as shifts in the party affiliation of the incumbent President, along with the timing of the electoral period. Thus, we gauge the extent to which shifts in the means and variances and the variable duration of business cycles are attributable to political factors. In this way, we attempt to link evidence from the time-series and structural business-cycle literatures with evidence on political business cycles. We do not, however, attempt a formal evaluation of alternative theoretical interpretations of the political factors.

   Our analysis begins with an examination of the anivariate behavior of real activity (as reflected in estimates of means, autocovariances, autospectra, and autoregressive models) for evidence of periodic behavior that coincides with the length of the presidential term, but distinguished by political party. Finding significant evidence for periodic behavior that differs by party affiliation of the president, we then estimate the timing and magnitude of the political components. Finally, we explore the implications of our findings for understanding the variable duration of business cycles. Our conclusion is that the data are consistent with a switching model for Republican and Democratic presidential administrations--where switches occur stochastically at integer multiples of four years, and the duration is bounded between four and eight years. Indeed, an information set containing only the party of the incumbent president and the quarter of the electoral term explains over one-third of the variation in the first-difference of measures of real economic activity during the period from 1951 to 1990.

2. Univariate Behavior of Real Economic Activity

   In this section, we examine the univariate behavior of three real U.S. variables--the civilian unemployment rate, real gross national product, and the industrial production index.(4) The sample extends from 1951:I, which is the end of the first postwar boom, to 1990:IV. For each series, the mean, the autocovariance and autospectrum functions, and autoregressive models are computed for the full sample and Republican and Democratic subsamples. In order to ensure stationarity, each series is expressed as the first difference of its natural logarithm.(5) The abbreviations UN, GNP, and PROD represent the first difference in the logarithm of the original data (multiplied by 100).

   Table I. Summary Statistics Statistic Sample UN GNP PROD Dickey-Fuller Pooled -4.14(*) -2.43 -3.05 Republican -3.72(*) -2.28 -2.41 Democrat -2.41 -2.31 -1.82 Mean Pooled 0.41 2.90 3.28 Republican 1.45 2.26 2.26 Democrat -1.49 3.92 5.37 t-difference 2.38(**) 2.50(**) 1.89(**) Variance Pooled 53.98 15.98 96.38 Republican 68.18 17.00 105.45 Democrat 23.99 12.35 75.10 F -ratio 2.84(**) 1.38(*) 1.40(*) Notes: The Dickey-Fuller statistic is the t-statistic of the coefficient (minus one) on |X.sub.t-1~ of a regression of |X.sup.t~ on |X.sub.t-i~, ||Delta~X.sub.t-i~, time, and a constant; see Dickey and Fuller |14~. The t-difference is the t-statistic on the difference of means by political party; and F-ratio is the F-statistic on the ratio of variances by political party. Statistics identified with ** are significant at five percent; and * significant at ten percent. Mean and Autocovariance Estimates

   Table I summarizes the mean of each series for the full sample and the Republican and Democratic subsamples. Consistent with the partisan view, UN is significantly higher and GNP and PROD significantly lower in Republican administrations. Furthermore, the level of the unemployment rate (not included in the table) is also significantly higher for Republicans. Table I also reports the variance, i.e., the autocovariance at lag zero, in UN, GNP, and PROD. For all three variables, the variance during the Republican subsample significantly exceeds the variance during the Democratic subsample.(6)

   Figure 1a is a plot of the autocovariance estimates for UN from lag one to lag twenty for the full sample (solid line), the Republican subsample (dashed line), and the Democratic subsample (solid-dashed line), with the standard error identified for each series. The patterns for both the full sample and the Republican subsample suggest a dominant sixteen-quarter cyclical pattern, as evidenced by positive coefficients at lags one and two, a series of negative coefficients, and finally several positive coefficients with a significant peak occurring at lag sixteen.(7) Estimates for the Democratic subsample, however, are not statistically different from white noise.

   Figure 1b is an analogous plot of the autocovariance estimates for GNP. Although noisier than the pattern for UN, the Republican subsample estimates for GNP also suggest a dominant cycle of about sixteen quarters. The near-random behavior for the Democratic subsample, however, tends to obscure this pattern in the full sample.(8) Figure 1c plots the estimates for PROD. The pattern for Republicans continues to show a dominant sixteen-quarter cycle, but the pattern is again obscured in the full sample by the apparent random behavior of PROD during the Democratic subsample.(9)

   In summary, the autocovariance estimates for all three variables appear dominated in the Republican subsample by a sixteen-quarter cycle, but this pattern is weak in the pooled data because of the near-random pattern in the Democratic subsample. However, this evidence does not provide either a formal test of the significance of the sixteen-quarter cycle or a direct comparison of the variation at this cycle to variations associated with other periodicities.

   Autospectrum

   The autospectrum, the Fourier transform of the autocovariance function, is a statistic that decomposes the variance of a series into frequency components of differing periodicities (the mean of the...