Temporal Aggregation in Political Budget Cycles

Electoral cycles have been widely debated since the pioneering studies of the 1970s on monetary and fiscal policy.¹ Drazen sums up the evidence in terms of active fiscal policy and passive monetary policy: while expansionary fiscal policy is the main impulse behind electoral cycles, monetary policy has an accommodating role.² Recently, there has been a flurry of empirical work on political budget cycles (PBCs) using cross-country panels. The most influential works concentrate on the budget surplus because it can capture both increases in expenditures and tax cuts before elections and is thus the most sensitive indicator of aggregate PBCs.³

A drawback of these extant studies is their reliance on annual observations. Since elections take place between January and December, annual data do not allow one to identify the election year precisely. To get around this problem, other schemes have been proposed. For example, by the rule of the semester, the previous year is counted as the election year if elections take place before July.⁴ What are the consequences of temporal aggregation for the measurement of PBCs? Instead of being concentrated in the election year, pre-electoral effects may be spread out over the two years leading up to elections. More important, if the sign of policies is reversed after elections, lower frequency data may mask PBCs because the effects in the election year cancel out, as Akhmedov and Zhuravskaya point out in their country study of Russia.⁵

Table 1.

Simulation of the Effects of Temporal Aggregation on the Budget Surplus^a

To illustrate the differences between using annual and quarterly data to identify the election year, imagine that elections take place in June. Furthermore, assume there is a pre-electoral deterioration in the budget surplus (surplus = -1) during the four quarters running up to elections, followed by an improvement (surplus = 1) the next four quarters. In the calendar year in which elections take place, the net effect on the budget surplus is zero. Elections are not always in June, of course. What happens if elections are evenly spread out over the year? Table 1 presents a simulation where Y(0) is the calendar year of elections, Y(-1) is the year before, and Y(1) is the year after.

For elections in the first quarter of the election year, namely, Y(0)_ Q(1), our textbook example of a pre-electoral expansion in the deficit during the four quarters prior to elections, followed by a contraction the next four quarters, implies that the surplus equals -1 from Y(-1)_Q(2) to Y(0)_Q(1), it equals 1 from Y(0)_Q(2) to Y(1)_Q(1), and it equals 0 otherwise, when the budget surplus is at its average value, normalized here at zero. The same procedure is followed for elections in the other three quarters. If we aggregate the effects by calendar year, there is a serious underestimation of PBCs in the election year, since pre-electoral expansions are almost cancelled out by post-electoral contractions, leaving a net expansive effect that is only 25 percent of the total stimulus before elections. The least affected by temporal aggregation are post-electoral fiscal adjustments, since 62.5 percent of the total effect is reflected the year after elections. On the other hand, if there were no post-electoral fiscal adjustments, the problem of temporal aggregation would be much less severe, since 62.5 percent of the total expansive effect would be reflected the calendar year of elections and 37.5 percent the year before.

This paper tackles the issue of temporal aggregation in PBCs head-on. We build a cross-country panel with annual and quarterly data from nineteen Latin American countries and twenty member states of the Organization for Economic Cooperation and Development (OECD) over the 1980-2005 period.⁶ Quarterly data allow us to identify the election year more precisely: with annual data, the election year is the calendar year in which elections take...