Policy makers' preferences, party ideology, and the political business cycle.

1. Introduction

It is generally accepted that partisan interests and the beliefs of policy makers subject to a democratic electoral process are important determinants of actual macroeconomic policies. More disagreement exists, however, about the nature of the/rue behavior of policy makers. In this respect, most of the attention within the political business-cycles literature over the last 30 years has been captured by the notions of opportunistic and partisan cycles (see Drazen (2000) for a literature review).

Initially formalized by Nordhaus (1975), opportunistic cycles are the result of preelection manipulative policies. Given an electoral cycle, opportunistic policy makers choose macroeconomic policies likely to generate lower unemployment rates and higher income levels right before election time, with the intention of gaining the voter's favor. The cycle is then completed when both inflationary pressures and contractionary tendencies appear after the elections.

As pointed out by Hibbs (1977) and later by Alesina (1987), political business cycles can also arise from ideological differences across political parties that alternate power. (1) One party, traditionally the left-wing party, appeals more to a labor base and promotes expansionary policies that minimize the output gap at the cost of eventual inflationary pressures. The other party, in contrast, appeals more to capital owners and is more concerned with keeping inflation in check at the cost of some unemployment. The cycle is generated as the two parties alternate power and the opposing preferences translate into opposing policies.

Whether opportunistic or partisan behavior actually takes place is an empirical question yet to be resolved. The most common test of both theories is to run an econometric autoregression of a macroeconomic variable (such as unemployment, output growth, or inflation) on itself, other economic variables, and a political dummy (for electoral years or the type of party in power) and then look for a link between the political events and the dependent macroeconomic variable.

The results of these empirical tests change noticeably with the measure of economic activity that is chosen as the dependent variable. Studies that use GDP or unemployment measures as dependent variables generally support partisan cycles theories but do not find evidence of opportunism (Beck 1987; Paldman 1991: and Alesina, Cohen, and Roubini 1992). In contrast, studies that use inflation or monetary instruments as dependent variables tend to reject partisan cycles and favor claims of opportunistic behavior (Sheffrin 1989; Alesina, Roubini, and Cohen 1997: and Faust and Irons 1999). (2)]

It is not surprising that the results of these tests change as the choice of macroeconomic variable studied changes. Figure 1 illustrates this point graphically by plotting time-series data on inflation, output growth, and electoral dates for a sample of countries. As can be seen for all these countries, inflation and output growth sometimes move in opposite directions. Thus, the analysis of a sustained period of economic growth and low inflation could naturally lead to different conclusions regarding the opportunistic behavior of the government.

[FIGURE 1 OMITTED]

This paper reexamines how policy makers and political parties behave by using a new approach. Instead of looking at macroeconomic variables and their relation to political variables, this paper focuses on measuring policy makers' revealed preferences toward stabilizing inflation and output growth directly. We argue that this method has several advantages over the traditional approaches: First, by generating a one-dimensional measure for preferences, we eliminate the complications associated with choosing among several macroeconomic variables as the dependent variables in our study.

Second, it is possible that some parties are more efficient than others in achieving policy targets or that some parties have a different perception of the output-inflation trade off in the economy. In this respect, authors like Cecchetti, Flores-Lagunes, and Krause (2006) have argued that policy makers become more efficient over time. Given that most of these differences are unobservable, studies that compare the behavior of macroeconomic variables across political parties with different ideologies or across electoral cycles are potentially biased. Our study of preferences, in contrast, overcomes this problem because the estimated preference parameters are independent of policy efficiency considerations.

Finally, our method directly deals with some of the endogeneity concerns of past studies. If the political dummy variables included in typical econometric regressions are determined by some omitted variables that also affect macroeconomic outcomes or if the timing of the elections is chosen strategically by the incumbent when macroeconomic conditions are good, then the results obtained in such regressions are likely to be biased (see Faust and Irons 1999 for a more detailed explanation).

We isolate this bias by comparing our results for the total sample with the results obtained from a subsample consisting only of electoral cycles that are predetermined by either law or custom. (3) Furthermore, as shown in the next section, our measure of political preferences is independent of the level of macroeconomic variables and omitted variables that create temporary shocks to the economy, thus avoiding any possible reverse causality problems.

Using a standard loss function and a series of macroeconomic variables, we estimate policy makers' preferences toward inflation and output growth stability for a sample of 24 countries during the period 1974-2000. We then combine these estimations with political records about electoral dates in order to answer three basic questions: (1) How do preferences toward stabilizing the main macroeconomic variables (inflation and output) change along the electoral cycle? (2) How do these preferences change as the ideology of the party in power changes? (3) Does the incumbent party try to resemble the behavior of a rival party as an election year approaches?

The results of our estimations show that political parties with a leftist ideology have a stronger preference toward stabilizing output growth than right-wing political parties for the majority of countries studied; that is, there exists an ideological gap between parties of the kind proposed by Alesina (1987) and Hibbs (1977). At the same time, our results also show that, in some countries, the incumbent's party acts opportunistically either by stimulating the economy before the elections or by making their economic policies similar to those of opposing parties. The presence of party resemblance strategies has been neglected in the past, but it emerges as an important form of opportunism in our results.

The remainder of the article is organized as follows: Section 2 explains the method we employ to estimate the relative weight that authorities place on inflation and output stability. Section 3 discusses our main findings regarding the behavior of incumbent political parties, and Section 4 presents our conclusions and possible directions for future research.

2. Measuring Policy Maker's Preferences

Deriving the Preference Parameters

In order to obtain our measures of policy makers' preferences and study its behavior over the political business cycle, we begin by assuming that the primary concern of the incumbent government is to achieve stabilization of the economy through the reduction in the variability of inflation and output growth. In doing this, we abstract from other policy goals, such as stabilizing exchange rates and interest rates, as well as achieving more equity in income distribution, for we consider that these serve rather as intermediate goals toward achieving domestic macroeconomic performance, measured by price and output stability.

Also, at this point, we declare ourselves agnostic as to which policy instrument the authorities will use (e.g., monetary policy, fiscal policy, exchange rate policy, or any other demand-side policy), and we simply represent the control variable by r, to which, from here on, we will refer simply as the interest rate for expositional convenience. We do not include in our analysis policies that may have effects on the supply side of the economy because most of them are associated with longer term goals that go beyond the scope of our study.

It is also important to note that many of the countries included in our sample have independent central banks, the primary objective of which is short-run economic stabilization. Therefore, if the monetary authorities in these countries, as we should expect, have relatively stable preferences, any observed variation in policy maker's behavior would capture the influence of other policies that are under the control of the incumbent party.

Consistent with most contemporary analyses of government policy and the theory of political business cycles, we summarize the policy maker's objective through the following standard quadratic loss function: (4)

(1) L = [E.sub.t][[lambda][([[pi].sub.t] - [[pi].sup.T.sub.t]).sup.2] + (1- [lambda])[([y.sub.t] - [y.sup.T.sub.t]).sup.2];

where E, is the expectation operator at time t, [pi] is inflation, y is (log) aggregate output; [[pi].sub.t] and [y.sup.T] are the target levels of inflation and output, (5) and [lambda] is the relative weight given to squared deviations of inflation relative to deviations of output from their respective desired levels.

Minimization of the loss function in Equation 1 requires knowledge of the determinants of deviations of inflation and output from their respective targets. We assume that two random shocks push y and [pi] away from [y.sup.T] and [[pi].sub.T]. First. an aggregate demand shock (d) moves inflation and output in the same direction, while an aggregate supply shock (s) moves inflation and output in opposite directions. (6) Because policy is only capable of moving inflation and output in the same direction, its effect is analogous to that of an aggregate demand shock.

We define aggregate demand (AD) as the negative relationship between and (y - [y.sup.T]) and ([pi] - [[pi].sub.T]) that is shifted by the demand shock and the deviations of the policy instrument from its equilibrium value (r), (7)

(2) y - [y.sup.T] = [omega]([pi] - [[pi].sub.T] - [phi](r - d); [omega]>0, [phi] > 0,

where [omega] is the inverse of the slope of the aggregate demand function and -[phi] is the response of output to changes in the policy instrument. (8) Analogously, aggregate supply (AS) is the positive relationship between inflation deviations and output deviations that is shifted by the supply shock,

(3) [pi] - [[pi].sub.T] = [gamma](y - [y.sup.T]) - s; [gamma] > 0,

where [gamma] is the slope of the aggregate supply function. The aggregate disturbances d and s have been normalized to yield the simple representation of the AD-AS model. (9)

Combining Equations 2 and 3. we obtain expressions for (y - [y.sup.T]) and [pi] - [[pi].sub.T] as a function of the structural parameters, the aggregate shocks and the policy instrument,

(4) y - [y.sup.T] = -[phi](r-d) + [omega]s/(1+[omega][gamma]),

(5) [pi] - [[pi].sub.T] = -[phi][gamma](r-d)-s/(1+[omega][gamma]).

Minimizing the quadratic loss function, subject to the constraints imposed by the structure of the economy, yields a simple linear policy rule of the form

(6) r = ad + bs,

where the expression for the coefficient b (which we later use for computing the preference parameter [lambda] is given by

(7) b = -[lambda][gamma] + (1-[lambda])[omega]/ [phi][[[gamma].sup.2] + (1-[lambda])]/

Equation 7 is useful to verify the response to supply shocks as a function of the preferences: If the policy maker only cares about stabilizing output ([lambda], = 0), the reaction to a positive supply shock (which raises the output gap), is to increase the interest rate by the factor of [omega]/[phi] (in order to reduce the output gap). Analogously, for a strict inflation targeter ([lambda]- 1), the reaction to an inflation-reducing positive supply shock should be to lower the interest rate by a factor of 1/([phi][lambda], in order to stabilize inflation around its target.

We are interested in the parameters of the observed or actual policy rule pursued by authorities, for this will allow us to estimate their actual preferences, regardless of whether or not they are behaving optimally. The procedure to obtain the relevant coefficients is as follows: Starting from the reduced form representation of the economy, given by Equations 4 and 5, we substitute the linear policy rule of Equation 6. By construction, we can define the aggregate supply and demand shocks in such a way that they will be uncorrelated ([[sigma].sub.d,s] = 0). Hence, the observed variances of output and inflation around their target levels can be given by the following expressions:

(8) var(y) [equivalent to] E[([y.sub.t] - [y.sup.T.sub.t]).sup.2] = [(1+[omega][gamma]).sup.-2][[phi].sup.2][(a-1).sup.2][[sigma].sup.2.sub.d] + [([omega]-[phi]b).sup.2][[sigma].sup.2.sub.s],

(9) var([pi]) [equivalent to] E[([[pi].sub.t] - [[pi].sup.T.sub.t]).sup.2] = [(1+[omega][gamma]).sup.- 2][[gamma].sup.2][phi].sup.2][(a-1).sup.2][[sigma].sup.2.sub.d] + [(1 + [phi]b).sup.2][[sigma].sup.2.sub.s].

Combining Equations 8 and 9, we can solve for the parameter h of the actual policy rule as

(10) b = (1+[omega][gamma])[[sigma].sup.2.sub.[pi]]-[[gamma].sup.2] [[sigma].sup.2.sub.y] - (1+[omega][gamma])[[sigma].sup.2.sub.s]/ 2[gamma][phi][[sigma].sup.2.sub.s].

Merging Equations 7 and 10, we can derive the coefficient of preference for inflation stability, [lambda], as a function of the structural parameters,

(11) [lambda] = ([omega]-[phi]b)/([omega]-[phi]b)+[gamma](1+[phi]b).

Note that, because L is a function of the structural parameters of the economy and the reaction of policy to supply shocks (parameter b), policy maker's preferences will not be a function of the levels of inflation and output but only of the relationship between these variables given by the structural model. This allows us to overcome potential endogeneity problems associated with using the prevailing levels of macroeconomic variables to analyze the political business cycle and avoid reverse causality problems.

Many authors before us have undertaken the task of estimating policy intentions, mostly for analyzing the behavior of monetary policy, (10) The main advantage of our procedure, as we show above, is that we do not need to assume optimal policy in order to estimate these preference parameters. Still, the method we propose is indirect by nature, and some could argue that the best approach is to directly survey policy makers in order to find out which are the relative importances they place on certain key macroeconomic variables. This direct method, however, has quite a few shortcomings:

(1) The decision-making process may not be centralized because several institutions and individuals may be responsible for policy. So even if we could survey all of them, how do we achieve a single measure for preferences?

(2) Some policy makers may not be willing to publicly disclose their intentions if they believe that this type of information would adversely affect the outcome and effectiveness of policy decision.

(3) Finally, even if we find a centralized entity in charge of policy that is completely transparent, situations out of the policy maker's control could affect his/her relative preference toward a particular objective (e.g.. having to intervene in order to bring the economy out of an unexpectedly sharp recession).

For all of the above reasons, we feel that looking at policy makers' revealed preferences through estimating a model for the economies of interest is more practical and easier to implement. Furthermore, because our analysis will be only from a historical perspective without, at this point, making any policy recommendations for future action, our proposed method is not subject to the Lucas (1976) critique. We indicate how we estimate the structural parameters needed to obtain our measures of [lambda] next.

Estimating the Structural Parameters

Let us revisit the stylized model in Equations 2 and 3:

(2') [y.sub.t] = [omega][[pi].sub.t] - [phi]([r.sub.t]-[d.sub.t]),

(3') [[pi].sub.t] = [gamma][y.sub.t] - [S.sub.t],

where we have defined y = y - [y.sup.T] and [pi] = [pi] - [[pi].sup.T]. We measure inflation by using the annualized change in the Consumer Price Index (CPI), while our measure of output is given the log of industrial production. For estimation purposes, we assume that the target level for inflation is given by its linear trend, whereas the target for (log) output is obtained by applying the Hodrick-Prescott filter to the data. As a robustness check for the above choices, we also considered alternative targets for inflation (average inflation) and output (log-linear trend), without any major differences in the outcomes. (11)

As such, estimating the system only allows us to identity the parameter [gamma]. Hence, in order to achieve the identification of [omega] and [phi], we make the operational assumption that the aggregate supply shock can be decomposed into a domestic and a foreign component, namely,

(12) [S.sub.t] = [h.sub.t] - [psi][f.sub.t],

where h represents the domestic (home) component of the shock, while f represents the foreign disturbance. The underlying assumption is that f affects domestic prices directly, while its impact on output arises indirectly through its effect on inflation. To be consistent with this description, we will use external price inflation as a proxy for f in the estimation, as we detail below.

The stylized model in Equations 2' and 3' can be reformulated to take into account the dynamic behavior of the economy, a feature present in the data. To accomplish this, we assume that the demand disturbance and the domestic component of the supply disturbance have persistent effects on the economy and model [d.sub.t] and [h.sub.t], as AR(2) processes, that is, (12)

(13) [d.sub.t] = [[phi].sub.1][d.sub.t-1] + [[phi].sub.2][d.sub.t-2] + [k.sub.d,t]; [E.sub.t]([k.sub.d,t]) = 0,

(14) [h.sub.t] = [[chi].sub.1][h.sub.t-1] + [[chi].sub.2][h.sub.t-2] + [k.sub.h,t]; [E.sub.t]([k.sub.h,t]) = 0.

Using Equations 2', 3', and 12, we can represent the aggregate shocks as

(15) [d.sub.t] = [y.sub.t] + [omega][[pi].sub.t]/[phi] + [r.sub.t],

(16) [h.sub.t] = [s.sub.t] -[psi][f.sub.t] = [gamma][t.sub.t] - [psi][f.sub.t]/

Substituting Equation 15 into the right-hand side of 13 and the solution into 2' yields

(17) [y.sub.t] = -[omega][[pi].sub.t] - [phi]([r.sub.t] + [[phi].sub.1][r.sub.t-1] + [[phi].sub.2][r.sub.t-2]) + [[phi].sub.1] [y.sub.t-1] + [[phi].sub.2][r.sub.t-2] + [omega][[phi].sub.1][[pi].sub.t-1] + [omega][[phi].sub.2][[pi].sub.t-2] + [phi][k.sub.d,t].

Analogously, substituting Equation 16 into the right-hand side of Equation 14 and the solution into Equation 3' results in the following:

(18) [[pi].sub.t] + [gamma][y.sub.t] + [psi]([f.sub.t] + [[chi].sub.1] [f.sub.t-1] + [[chi].sub.2][f.sub.t-2]) - [gamma][[chi].sub.1][y.sub.t-1] - [gamma] [[chi].sub.2][y.sub.t-2] + [[chi].sub.1][[pi].sub.t-1] + [[chi].sub.2] [[pi].sub.t-2] - [k.sub.h,t].

The system of Equations 17 and 18 represents a dynamic aggregate demand-aggregate supply model. To make its estimation operational, we proxy the term [r.sub.t] + [[phi].sub.1][r.sub.t-1] + [phi].sub.2][r.sub.t-2] with the lagged demeaned expost real interest rate ([i.sub.t-1] - [[pi].sub.t-1]), where i is the short-term nominal interest rate. It is important to note that, while it is true that we are using a mainly monetary variable (namely the short-term real interest rate) as the policy maker's instrument, we are not making any claims as to how this interest rate is determined. Therefore, fiscal and/or exchange rate policy could indeed play a role in determining the level of the instrument, and the extent of that role visa vis the relative importance of monetary policy is an empirical issue and, therefore, country specific.

Finally, we proxy the expression [f.sub.t] + [[chi].sub.1][f.sub.t-1] + [[chi].sub.2][f.sub.t-2] with one lag of demeaned external price inflation ([f.sub.t] + [[pi].sup.x.sub.t-1]), where e is nominal exchange rate devaluation and [[pi].sup.x] is foreign inflation. Taking this into account, we estimate the dynamic behavior of output and inflation through the following system:

(19) [y.sub.t] = -[omega][[pi].sub.t] - [pi]([l.sub.t-1] - [[pi].sub.t-1]) + [2.summation over l=1][[alpha].sub.1l] [y.sub.t-l] + [2.summation over l=1][[alpha].sub.1](l+2)[[pi].sub.t-l] + [u.sub.yt],

(20) [[pi].sub.t] = [gamma][y.sub.t] + [psi]([e.sub.t-1] + [[pi].sup.x.sub.t-1]) + [2.summation over l=1][[alpha].sub.2l][y.sub.t-l] + [2.summation over l=1][[alpha].sub.2](l+2) [[pi].sub.t-l] + [u.sub.[pi]t].

There are two crucial assumptions for estimating the system. First, in the aggregate demand equation, the (lagged) real interest rate has only a direct effect on output, and its outcome on prices arises through the effect on output. Second, (lagged) external price inflation only affects domestic inflation contemporaneously, with an indirect effect on output.

While the first identification assumption is often found in the literature (see, e.g., Rudebusch and Svensson 1999 and the references in Taylor 2000), the use of the second one can be justified if, as mentioned above, a change in external prices has its direct effect on domestic inflation immediately and only causes a change in domestic output after a period. (13) It is important to note that we are claiming neither that external price inflation does not affect output nor that a change in the real interest rate has no effect on inflation. Instead, our assumptions are that the interest rate affects first production and then prices and that external inflation affects domestic prices first and the adjustment in output takes place later, both of which are consistent with recent empirical findings.

Because the real interest rate only enters the dynamic aggregate demand, Equation 19, and the external price inflation only enters the dynamic aggregate supply, Equation 20, we can identify the parameters [omega], [gamma], and [phi] through the estimation of the vector autoregression that Equations 21 and 22 describe. (14)

(21) [y.sub.t] = [[beta].sub.1r]([l.sub.t-1] - [[pi].sub.t-1]) + [[beta].sub.1f]([e.sub.t-1] + [[pi].sup.x.sub.t-1]) + [2.summation over l=1)[[beta].sub.1l][y.sub.t-l] + [2.summation over l=1] [[beta].sub.1(l+2)][[pi].sub.t-1] + [[epsilon].sub.yt],

(22) [[pi].sub.t] = [[beta].sub.2r]([l.sub.t-1] - [[pi].sub.t-1]) + [[beta].sub.2f]([e.sub.t-1] + [[pi].sup.x.sub.t-1]) + [2.summation over l=1)[[beta].sub.2l][y.sub.t-l] + [2.summation over l=1] [[beta].sub.2(l+2)][[pi].sub.t-1] + [[epsilon].sub.[pi]t].

It is straightforward to show that the estimates for [omega], [gamma], and [phi] can be obtained from the VAR estimation as follows:

(23) [omega] = [[beta].sub.1f]/[[beta].sub.2f]

(24) [gamma] = [[beta].sub.2r]/[[beta].sub.1r]

(25) [phi] = [[beta].sub.1r](1 + [omega][gamma]).

Because the preference measure obtained in Equation 11 is a function of the structural parameters estimated by Equations 23-25 and 10, we can compute Z directly from these estimates, as shown in Equation 11'.

(11') [lambda] = ([omega] - [phi]b)/([omega] - [phi]b) + [gamma][phi]b).

The data set used for the estimation was obtained from the 2002 IMF International Financial Statistics and from other country-specific sources, as described in the Data Appendix. We use rolling regressions of 20 quarters each to obtain quarterly results for [lambda] for the 24 countries in the sample, countryspecific data availability determined the maximum span of each time series for [lambda]. To avoid any problems of seasonality in the estimate of policy makers" revealed preferences and given that the data we employ for studying the political cycle is available in annual frequency, we compute a yearly estimate for [lambda] by taking a simple four-quarter average from the time series. We turn to our main findings in she next section.

3. Results: Patterns of Political Behavior

We now present the analysis of policy maker's preferences as summarized by the parameter [lambda] for our sample of countries. In particular, we examine the following three questions:

(1) How do preferences toward the main macroeconomic variables (inflation and output stability) change along the electoral cycle?

(2) How do these preferences change as the ideology of the party in power changes?

(3) Does the incumbent party try to resemble the behavior of a rival party as an election year approaches?

The analysis spans the period 1974-2000 for most countries; the most notable exceptions are the European Monetary Union countries, for which we do not analyze the years after their adoption of the Euro, and some developing economies, for which some data did not become available on a quarterly frequency until the late 1970s--early 1980s.

We attempted to incorporate as many countries as possible in trying to answer the three questions above. The sample size, however, was restricted by the nature of the inquiries. The study of party ideologies, on the one hand, requires the presence of at least two ideologies that effectively compete for power, a rare characteristic in many countries. The study of preferences along the electoral cycle, on the other hand, is complicated by unstable electoral cycles, military occupations, and changes in the electoral and/or political system.

Information of political parties and electoral dates was obtained mainly from the Database on Political Institutions (DPI) in Beck et al. (200l), which contains a sizeable wealth of information, and was also partly completed through direct inquiries to individual government sources. Throughout the analysis, a year is considered to be an electoral year if democratic elections took place during or after March of that year. Although rare, years in which the electoral process takes place either in January or February are unlikely to reflect policy changes related exclusively to the elections on that same year because, for the most part of it, the newly elected government will be in power.

As pointed out by Ginsburgh and Michel (1983), data points for years where the electoral cycle was interrupted or extended are likely to misrepresent any pattern of political behavior, as the policy makers might not be capable of predicting any surprise elections or surprise extensions of their tenure. Thus, as a robustness check, we also constructed a subsample of observations from years within electoral cycles of normal length, where a normal length of the electoral cycle (normal cycle) is defined by the historical mode.

It must also be noted that using normal cycles alone mitigates the endogeneity problems that arise from the possibility that the timing of the elections and the political preferences themselves are correlated with unobserved economic factors (e.g., whenever the incumbent chooses to call for elections when macroeconomic conditions are good or the opposing party does so when economic conditions are bad). By using normal cycles, we deal only with elections that are predetermined either by law or by custom. A similar argument is used also by Shi and Svensson (2002).

Whenever appropriate, we will report the results for both the actual cycles (entire sample) and the normal cycles subsample. The latter becomes slightly smaller as countries such as France, Spain, and Denmark, which have no historical mode for their electoral cycles, were excluded.

Evidence on Opportunistic Behavior

With respect to the relationship between policy preferences and the electoral cycle, we start by focusing on the possibility of opportunistic behavior. That is, we look for evidence that preferences for economic policy become more expansionary as the election day approaches, regardless of the ideology of the party in power or the number of political parties that effectively alternate power.

In order to study the patterns of policy preferences throughout the electoral cycle, we classified the values of the relative preference for inflation stability ([lambda]) into groups according to their position in the electoral cycle. That is, for each country, we formed a group for electoral years only, another group for the preelectoral years, other groups for the years before that, and a final group for the years after a previous election. The average [lambda] within each group was then calculated and used to characterize the bahavior of policy makers' preferences throughout the cycle.

The policy makers" behavior was classified as opportunistic or not opportunistic for all countries where data were available, as presented in Table 1, where additional information about the individual political systems is provided. Using a slightly different definition than Nordhaus' (1975), we classify revealed preferences as opportunistic according to the following two criteria.

Cycle     Opportunistic Criterion 1

3-year    Electoral year's average value of [lambda]. is lower than for
          both preelectoral years.

4-year    Average value of [lambda] is lower for last two years of the
          cycle than for first two years and the
          electoral year's value of [lambda], is lower than for the
          preelectoral year.

5-year    Average value of [lambda] is lower [lambda] last two years of
          the cycle than for first three years and the
          electoral year's value of [lambda] is lower than all of the
          three first years.

6-year    Average value of [lambda] is lower for last three years of
          the cycle than for first three years and the
          electoral year's value of [lambda], is lower than all of the
          three first years.

Cycle Opportunistic Criterion 2

All       Electoral year's value of [lambda] is lower than for
          immediate preelectoral year.

These two criteria intend to capture trends toward expansionary policies as the election year approaches as well as expansionary policy shocks during the electoral year. As shown in the 3rd and 4th columns of Table 1, under Criterion 1, only 6 out of 24 countries exhibited policy patterns compatible with opportunistic behavior. Similarly, when the sample is restricted to only normal cycles, 7 out of 21 countries show opportunistic patterns.

Nonetheless, as columns 5 and 6 of Table 1 show, when using Criterion 2, the average value of [lambda] during electoral years is lower than in the immediate preelectoral year for 15 out of 24 countries; and in 13 out of 21 countries, the value of [lambda] was lower at the end of the normal cycle than at the start. As a result, we cannot determine unambiguously whether policy makers, in general, are behaving in a Nordhaus-opportunistic manner.

Finally, we note that governments with shorter political cycles show more evident opportunistic behavior. The possibility that shorter cycles give incentive to opportunism could be explained by the public officials' desire to show results or by more intense political battles. The specific elements that shape the bahavior of government is an interesting topic that is beyond the objectives of this article.

Evidence on Party Ideology

In order to study the policy preferences of political parties representing different ideologies, we separated the parties into two categories: Left and right; the distinction made here mimics the classification of the DPI. We were able to gather information on 16 countries with at least two different ideologies alternating power, where we included only countries for which data was available for at least one full cycle for each ideology. The average country-specific [lambda] values for both ideologies are presented in columns 7 and 8 of Table l.

As shown there, parties associated with a left ideology have lower values of [lambda] in 11 out of 16 countries. Only for Australia, France, Germany, Norway, and Switzerland do we find that the leftwing party is more concerned about inflation than the right-wing party, a result that is not too surprising in the case of the four European countries, where the ideological differences between the parties are less marked than in most of the rest of the world.

Still, if we look at the overall average for all 16 countries, the value for [lambda] is 0.66 and 0.71 for let1 and right parties, respectively, and the difference is significantly different than zero at the 1% level. When the sample is restricted to normal cycles only, the results are almost identical, with the only change being that the difference between the average [lambda]'s of left and right ideologies becomes even larger.

The above result supports the claim made by Alesina (1987) and Hibbs (1977) that party ideology will have an incidence in how the government pursues policy goals: The right-wing party will be mostly concerned about low and stable inflation, whereas the left-wing party will have more interest in stable and sustainable economic growth. Does that necessarily mean that each individual party is not behaving in an opportunist manner? We further examine this issue next in our analysis of party resemblance.

Evidence on Party Resemblance

We now turn our attention toward what we consider is another important aspect of political opportunism: party resemblance. In countries with strong bipartisan systems with clearly separated ideologies, the incumbent's party could benefit from public policies that emulate those of a rival party, as these policies are likely to attract voters in the middle of the ideological spectrum. If such a kind of opportunism exists, then we should be able to observe it with our data.

In order to collect party-specific information with more than one data point, we selected countries with a stable and distinguishable bipartisan system where each party has reached power for at least two periods within our sample data. We than calculated the patterns of revealed preferences throughout the electoral cycle for each political party separately. Altogether, eight countries meet these criteria: Australia, Canada, Costa Rica, Germany, New Zealand, Sweden, Switzerland, and the United States. We graph the average values for [lambda] for each party during the entire cycle in Figure 2.

Noticeably, only about half of these countries were considered opportunistic by our previous analysis (three according to criterion 1 and six when using criterion 2) and only four of them showed [lambda] values that were significantly higher for the right-wing party. Still, for all eight countries, we find party resemblance, measured by convergence of preferences or the reduction in [lambda] difference prior to the elections, as shown in the last column of Table 1.

For all these countries, we find that the difference among political parties of their revealed preferences toward inflation stabilization becomes smaller as the election date arrives. On average, for the eight countries, this difference is reduced by 50% between the preelectoral year to the electoral year. The reduction in [lambda] difference on election year ranges from almost 7% (Sweden) all the way to >96% (Australia). This behavior is consistent with the median-voter models described by Buchanan and Tullock (1965), Mueller (1976), Caplin and Nalebuff (199l), among many others.

Taking a close look at the behavior of the right-wing parties in Figure 2, we find a striking resemblance in seven out of the eight countries: On the electoral year, the right either reduces its preference toward inflation stability or maintains it basically unchanged. This observation shows evidence that the right ideology tends to behave opportunistically according to Nordhaus" standards. The only exception is Switzerland, but in this country, the left has, on average, a higher [lambda] than the right, so the increase in [lambda] on election year is consistent with party resemblance.

At the same time, we cannot find any general pattern of a Nordhaus-opportunistic conduct for the left; in Costa Rica, Sweden, and Switzerland, the left tends to fight inflation on the election year; in Australia and Germany, they try to expand the economy: while in Canada, New Zealand, and the United States, there is no significant change in [lambda] during the latter part of the administration.

Summarizing, our empirical findings, on the one hand, support the presence of a partisan cycle in policy makers' intentions, consistent with Hibbs (1977) and Alesina (1987), with the right-wing party being, on average, mostly concerned about inflation stability and the left-wing party relatively more interested in output growth stability. On the other hand, we cannot unambiguously conclude from our results that countries are opportunistic in the Nordhaus" (1975) sense; if we look at average preferences, only 25% of the countries (or 33%, when limiting our focus to normal cycles) become more expansive in the second part of the administration: while, if we focus solely on what happens in the election year, roughly 62% of the countries follow an opportunistic behavior.

Yet, when we separate party ideologies for the eight countries that meet the required criteria and analyze the political cycle, we find strong evidence of a Nordhaus-opportunistic behavior of the rightwing parties and overwhelming support of a convergence of preferences or party resemblance in the electoral year, which should also be viewed as an opportunistic conduct of policy makers, consistent with the median-voter literature.

4. Conclusions

In this article, we generate a time series for the revealed preferences of policy makers toward inflation stability (vis a vis output growth stability) for a sample of 24 economies in order to study the behavior of political parties. Such behavior is essential in both the partisan cycle models developed by Hibbs (1977) and Alesina (1987) and the opportunistic political-cycle analysis first introduced by Nordhaus (1975). Our evidence supports Hibbs and Alesina's claim over Nordhaus' view; still, for at least eight countries, we find that both approaches can explain the incumbent party's behavior at election time, and for these countries, we find strong support of a different type of opportunistic conduct, namely party resemblance.

This study leaves open several questions that should be addressed by future research. First, one could study if the strength or variability of the political cycle can be explained by country-specific factors, such as the political system (parliamentary vs. presidential), the existence of reelection, government size, economic development, and more. Also, it would be interesting to analyze whether or not changes in the economic system and institutions (central bank structure, trade unification or monetary unions, e.g.) have contributed to making the cycle less volatile. Finally, one could use the proposed measure of preferences and their changes to find similarities between countries and establish whether or not the political cycle is being exported. We hope to address all these important questions soon.

Appendix: Data Sources

All macroeconomic data for Australia, Austria, Barbados, Belgium, Canada, Costa Rica, Denmark, Finland, France, Germany, Israel, Japan, Mexico, Netherlands, New Zealand, Norway, Peru, Portugal, Spain, Sweden, Switzerland, Trinidad and Tobago, the United Kingdom, and the United States are from International Financial Statistics CD ROM (September 2002), except for Costa Rica, where the data for output were obtained from the Banco Central de Costa Rica. For Austria, Belgium, France, Germany, Netherlands, and Spain, the data are for 1974:I-1988:IV; for Denmark, Norway, and Sweden, 1974:I-2000:IV; Australia, Japan, the United Kingdom, and the United States, 1974:I-2001:IV; Canada, 1975:1-2001:IV; Switzerland, 1976:I-2000:IV; Costa Rica, 1976:I-2001:IV; Barbados, 1977:I-2000:IV; Finland and Trinidad and Tobago, 1978:I-1998:IV; New Zealand, 1978:I-2000:IV; Mexico, 1978:I-2001:IV; Peru, 1979:I-2001:IV; Portugal, 1981:I-1998:IV; and Israel, 1982:I-2000:IV. Output (y) is given by seasonally adjusted industrial production, except for Costa Rica (Indice Mensual de Actividad Economica), and Peru (GDP-volume, seasonally adjusted). Inflation ([pi]) is given by the annualized CPI inflation rate for all countries. The nominal interest rate (i) is given by the money-market rate except for Israel and Mexico (deposit rate), and Barbados, Costa Rica, Peru, and Trinidad and Tobago (discount rate). Devaluation (e) is given by the annualized percentage change of the exchange rate to the U.S. dollar, except for the U.S. (U.S.$ALK.[pounds sterling] exchange rate). External inflation ([[pi].sub.x]) is given by annualized U.S. CPI inflation, except for the United States (annualized U.K. CPI inflation).

Information of political parties and electoral dates was obtained mainly from the Database on Political Institutions in Beck et al. (2001) and also partly completed through direct inquiries to individual government sources.

Table 1. Preferences According to Party Ideology and Opportunistic
Criteria

                               Opportunistic     Opportunistic
                                Criterion 1       Criterion 2

                    Length    Actual   Normal   Actual   Normal
Country            of Cycle   Cycle    Cycle    Cycle    Cycle

Australia          3 years     Yes      Yes      Yes      Yes
Austria            4 years      No       No       No       No
Barbados           Variable     No       No       No       No
Belgium            4 years      No       No       No       No
Canada             4 years      No       No       No       No
Costa Rica         4 years      No       No      Yes      Yes
Denmark            Variable     No     -- (b)     No     -- (b)
Finland            4 years      No       No      Yes      Yes
France             Variable    Yes       b       Yes     -- (b)
Germany            4 years     Yes      Yes      Yes      Yes
Israel             4 years      No       No      Yes      Yes
Japan              3 years     Yes      Yes      Yes      Yes
Mexico             6 years      No       No      Yes      Yes
Netherlands        Variable     No       No       No       No
New Zealand        3 years     Yes      Yes      Yes      Yes
Norway             4 years      No       No      Yes      Yes
Peru               5 years      No       No       No       No
Portugal           4 years      No      Yes      Yes      Yes
Spain              Variable     No     -- (b)     No     -- (b)
Sweden             3 years      No      Yes      Yes      Yes
Switzerland        4 years      No       No       No       No
Trinidad and
  Tobago           Variable    Yes      Yes      Yes      Yes
UK                 4 years      No       No      Yes       No
USA                4 years      No       No      Yes      Yes
Average/fraction              25.00%   33.33%   62.50%   61.90%

                       Preferences

                   Left-Wing   Right-Wing   Convergence of
Country              Party       Party      Preferences (a)

Australia           0.7158       0.5379         96.49%
Austria             -- (b)       -- (b)         -- (b)
Barbados            0.7074       0.8189         -- (b)
Belgium             -- (b)       -- (b)         -- (b)
Canada              0.8295       0.9223         36.46%
Costa Rica          0.1332       0.2087         64.75%
Denmark             0.8098       0.8314         -- (b)
Finland             0.7779       0.9409         -- (b)
France              0.8065       0.7028         -- (b)
Germany             0.9024       0.7307         70.21%
Israel              -- (b)       -- (b)         -- (b)
Japan               -- (b)       -- (b)         -- (b)
Mexico              -- (b)       -- (b)         -- (b)
Netherlands         -- (b)       -- (b)         -- (b)
New Zealand         0.4576       0.8112         27.80%
Norway              0.8380       0.7645         -- (b)
Peru                0.0102       0.2007         -- (b)
Portugal            -- (b)       -- (b)         -- (b)
Spain               -- (b)       -- (b)         -- (b)
Sweden              0.7514       0.7529         6.76%
Switzerland         0.8171       0.7548         48.70%
Trinidad and
  Tobago            0.8434       0.8560         -- (b)
UK                  0.3923       0.6328         -- (b)
USA                 0.8183       0.8810         47.70%
Average/fraction    0.6632       0.7092         49.86%

(a) Reduction in the difference of preferences between political
parties on election year (party resemblance).

(b) Computation not available/applicable.

(1) The specific adjustments are different in Alesina's (1187) and Hibbs' (1977) models, as their assumptions about the rationality of the economic agents vary.

(2) Other studies examining the political business cycle include McCallum (1978), Hibbs (1987, 1994), Alesina and Sachs (1988). Keil (1989), Nordhaus, Alesina, and Schultze (1989), Haynes and Stone (1990), and Castro and Veiga (2004).

(3) A similar approach is followed by Shi and Svensson (2002).

(4) Due to the assumption of a quadratic loss function, the analysis we undertake restricts expansionary and contractionary policy preferences In be symetric, which may not always be the case (e.g., policy makers may be more concerned with a fall in output than an equally proportional increase in output). For example, Ruge-Murcia (2003) performs a derivation and estimation of a game-theoretic model with asymmetric preferences around the target levels of inflation and the unemployment rate. Extending the analysis to allow for this asymmetry is an interesting exercise, which goes beyond the scope of the present study.

(5) Note that the target levels for inflation and output can be a function of time: we come back to this issue later in the article.

(6) In the formal representation that follows, we assume that u positive demand shock increases both inflation and output, whereas a positive supply shock lowers inflation and rises the level of output.

(7) The equilibrium value of the interest rate is defined as the value needed such that output would equal its potential (or target) level.

(8) Romer (2000) provides a good description on how to derive an analogous version of this model. See also Krause (2003a) for a theoretical derivation using a rational expectations optimization process in the presence of imperfect information.

(9) Krause (2003b) shows how the AD-AS model and the one-period loss function yields the exact same reduced from representations for optimal inflation and output as the forward-looking new Keynesian model developed by Roberts (1995) and employed by Clarida. Gall and Gertler (1999) and others, whenever supply shocks are not autocorrelated. However, once we perform the estimation of the parameters in the next subsection, we relax this assumption and employ an AR(2) process for both supply and demand disturbances.

(10) See, for example, recent studies by Rudebusch (2001). Cecchetti and Ehmmnn (2001). Dennis (2001). Cecchetti. McConnell. and Perez-Quiros (2002). Favero and Rovelli (2003), and Castelnuovo and Surico (2004).

(11) The only exception is the case of Portugal, for which a linear trend lot inflation yields substantially higher values for [lambda] during the mid- to late-1980s as compared with the estimates we obtain when using average inflation as the policy maker's goal. Nevertheless, because this affects the level of [lambda] across all periods and not so much the direction of its change, we have opted to ignore this issue.

(12) The assumption about the autoregressive structure of the shocks is only crucial in terms of determining the order of the vector autoregression in Equations 19 and 20. Specifically, for the current specification of the AD AS model, and AR(n) process for the disturbances will result in the estimation of an n order var.

(13) This would clearly be the case if the source of the external price change were an oil shock or a modification in the terms of trade that the economy faces.

(14) The var specification is similar to the one proposed by Mojon and Peersman (20D1) for measuring the effects of monetary policy in countries of the Euro area. The most important differences are that we do not include U.S. real GDP and nominal interest rate as controls and that we estimate the same basic model for all countries.

References

Alesina, Alberto. 1987. Macroeconomic policy in a two party system as a repeated game. Quarterly Journal of Economics 102:651-78.

Alesina, Alberto, Gerald Cohen, and Nouriel Roubini. 1992. Macroeconomic policy and elections in OECD democracies. Economics and Politics 4:1-30.

Alesina, Alberto, Nouriel Roubini, and Gerald Cohen. 1997. Political cycles and the macroeconomy. Cambridge, MA: MIT Press.

Alesina, Alberto, and Jeffrey Sachs. 1988. Political parties and the business cycle in the United States, 1948-1984. Journal of Money Credit and Banking 20:63-82.

Beck, Nathaniel L. 1987. Elections and the fed: Is there a political monetary cycle? American Journal of Political Science 31:194-216.

Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh. 2001. New tools in comparative political economy: The database of political institutions. Worm Bank Economic Review 15:165-76.

Buchanan, James M., and Gordon Tullock. 1965. The calculus of consent. Ann Arbor, MI: University of Michigan Press, pp. 131-45.

Caplin, Andrew, and Barry Nalebuff. 1991. Aggregation and social choice: A mean voter theorem. Econometrica 59:1-23.

Castelnuovo, Efrem, and Paolo Surico. 2004. Model uncertainty, optimal monetary policy and the preferences of the Fed. Scottish Journal of Political Economy 51:105-26.

Castro, Vitor, and Francisco J. Veiga. 2004. Political business cycles and inflation stabilization. Economic Letters 83:1-6.

Cecchetti, Stephen G., and Michael Ehrmann. 2001. Does inflation targeting increase output volatility? An international comparison of policymakers' preferences and outcomes. In Monetary policy: Rules and transmission mechanisms, Proceedings of the Third Annual Conference of the Central Bank of Chile, edited by Klaus Schmidt-Hebbel. Santiago, Chile, pp. 247-74.

Cecchetti, Stephen G., Alfonso Flores-Lagunes, and Stefan Krause. 2004. Has monetary policy become more efficient? A cross-country analysis. Mimeo.

Cecchetti, Stephen G., Margaret M. McConnell, and Gabriel Perez-Quiros. 2002. Policymakers' revealed preferences and the output-inflation variability trade-off: Implications for the European system of central banks. The Manchester School 70:596-618.

Clarida, Richard, Jordi Gali, and Mark Gertler. 1999. The science of monetary policy: A new Keynesian perspective. Journal of Economic Literature 37:1661-707.

Dennis, Richard. 2001. The policy preferences of the U.S. Federal Reserve. Federal Reserve Bank of San Francisco Working Paper No. 01-08, (revised, June 2003).

Drazen, Allan. 2000. The political business cycle after 25 years. NBER Macroeconomics Annual 2000. Cambridge, MA: MIT Press.

Faust, Jon, and John Irons. 1999. Money, politics and the post-war business cycle. Journal of Monetary Economics 43:61-89. Favero, Carlo, and Ricardo Rovelli. 2003. Macroeconomic stability and the preferences of the Fed: A formal analysis. 1961-1998. Journal of Money, Credit, and Banking 35:545-56.

Ginsburgh, Victor, and Philippe Michel. 1983. Random timing of elections and the political business cycle. Public Choice 40:155-64.

Haynes, Stephen E., and Joe A. Stone. 1990. Political models of the business cycle should be revived. Economic Inquiry 28: 442-65.

Hibbs, Douglas A., Jr. 1977. Political parties and macroeconomic policy. American Political Science Review 71:1467-87.

Hibbs, Douglas A., Jr. 1987. The American political economy. Cambridge, MA: Harvard University Press.

Hibbs, Douglas A., Jr. 1994. The partisan model of macroeconomic cycles: More theory and evidence for the United States. Economics and Politics 6:1-23.

Keil, Manfred W. 1988. Is the political business cycle really dead'? Southern Economic Journal 55:86 99.

Krause, Stefan. 2003a. Why should policy makers care about inflation and output variability? The role of monetary policy stabilization under asymmetric information. Department of Economics, Emory University Working Paper No. 03-13.

Krause, Stefan. 2003b. Optimal monetary policy and the equivalency between the one period AD-AS model and the forward-looking new Keynesian model. Department of Economics, Emory University Working Paper No. 03-17.

Lucas, Robert E. 1976. Econometric policy evaluation: A critique. Carnegie-Rochester Conference Series on Public Policy 1:19-46.

McCallum, Bennett T. 1978. The political business cycle: An empirical test. Southern Economic Journal 42:504-15.

Mojon, Benjoit, and Gert Peersman. 2001. A VAR description of the effects of the monetary policy in the countries of the Euro area. ECB Working Paper No. 92.

Mueller, Dennis C. 1976. Public choice: A survey. Journal of Economic Literature 14:395-5433.

Nordhaus, William D. 1975. The political business cycle. The Review of Economic Studies 42:169-90.

Nordhaus, William D., Alberto Alesina. and Charles L. Schultze. 1989. Alternative approaches to the political business cycle. Brooking Papers on Economic Activity 2:1-68.

Paldam, Martin. 1991, Politics matter after all: Testing Alesina's theory of RE partisan cycles on data for seventeen countries. IEA Conferences Series 97:369-98.

Roberts, John. 1995. New Keynesian economics and the Phillips curve. Journal of Money, Credit, and Banking 27:975-84.

Romer, David. 2000. Keynesian macroeconomics without the LM curve. Journal of Economicx Perspectives 14:149-69.

Rudebusch, Glenn D. 2001. Is the Fed too timid? Monetary policy in an uncertain world. Review of Economies and Statistics 83:203-17.

Rudebusch, Glenn D., and Lars E. O. Svensson. 1999. Policy rules for inflation targeting. In Monetary Policy Rules, edited by John B. Taylor. Chicago, IL: University of Chicago Press, pp. 203-46.

Ruge-Murcia, Francisco J. 2003. Inflation targeting under asymmetric preferences. Journal of Money, Credit, and Banking 35:763-85.

Sheffrin, Steven M. 1989. Evaluating rational partisan business cycle theory. Economics and Politics 1:239-59.

Shi, Min, and Jakob Svensson. 2002. Political budget cycles in developed and developing countries. Mimeo.

Taylor, John B. 2000. Alternative views of the monetary transmission mechanism: What difference do they make for monetary policy? Oxford Review of Economic Policy 16:60-73.

Stefan Krause * and Fabio Mendez ([dagger])

* Department of Economics, Emery University, Atlanta, GA 30322. USA; E-mail [e-mail].

([dagger]) Department of Economics, University of Arkansas, Business Building, Room 402, Fayetteville, AR 72701. USA; E-mail [e-mail]; corresponding author.

We thank Stephen G. Ceccheiti, Chris Curran, Roisin O'Sullivan and Jerry Thursby for their useful suggestions.

Received June 2004: accepted October 2004