The impact of expenditure rules on budgetary discipline over the cycle

Abstract

We study the impact of expenditure rules on the propensity for governments to deviate from their expenditure plans in response to surprising cyclical developments. Theoretical considerations suggest that due to political fragmentation in the budgetary process expenditure policy might be prone to a procyclical bias. However, this tendency may be mitigated by strictly enforced expenditure rules. These hypotheses are tested against data from a panel of EU Member States. Our key findings are that (1) deviations between actual and planned government expenditure tend to be positively related to output gap surprises, and (2) expenditure rules reduce this procyclical bias. These results are particularly pronounced when the analysis is confined to spending items with a high degree of budgetary flexibility.

Keywords:

• expenditure rules
• fiscal discipline
• stabilisation
• spending bias

JEL Classification:

• C23
• E62
• H50

Acknowledgements

We are grateful to Antonio Afonso, Thiess Büttner, Cristina Checherita, Jacopo Cimadomo, Geert Langenus, Philipp Mohl, Christiane Nickel, Ad van Riet, Jürgen von Hagen, Peter Wierts, participants of the ESCB Working Group on Public Finance 2009 Workshop in Cyprus and an anonymous referee for useful comments. The views expressed in this article are those of the authors and do not necessarily reflect those of the European Central Bank.

Notes

¹ For models establishing a formal link between the common pool problem in budgetary processes and fiscal pro-cyclicality, see Tornell and Lane (1999) and Talvi and Végh (2005).

² Focussing on the overall size of public spending, Debrun et al. (2008) find that those EU countries with strong expenditure rules tend to have slightly lower primary expenditure-to-GDP ratios. Badinger (2009) finds that after the introduction of fiscal rules in several OECD countries since the 1990s, the volatility of government spending has fallen significantly in these countries. For a sample of 15 EU countries, Wierts (2008) finds that government spending tends to be less responsive to revenue windfalls or shortfalls in countries with stringent expenditure rules. By contrast, Büttner and Wildasin (2009), find that the presence of formal borrowing restrictions for US municipalities intensifies fluctuations in investment spending in response to revenue shocks.

³ Actual revenues tend to be slightly above the levels targeted in the programmes thus exerting a positive impact on budget balances.

⁴ While stability and convergence programmes contain growth rate forecasts, they do not always provide the information needed to derive the GDP forecast in levels. Hence, we use the respective Commission forecast to adjust denominators and control for differences between stability and convergence programme and Commission growth rate forecasts in the empirical analysis (Section III).

⁵ Since 2001, EU Member States have been required to provide detailed projections in their stability and convergence programmes for collective consumption, social transfers in kind, social transfers other than in kind, interest payments, subsidies, gross fixed capital formation and other expenditure.

⁶ Most notably, the variable collective consumption typically includes compensation of employees and intermediate consumption. Given the transaction cost associated with changes in the public workforce and the elaborate negotiation processes surrounding public sector wage setting, the flexibility for compensation of employees to respond to surprises in cyclical conditions appears limited. By contrast, intermediate consumption which includes a broad range of operating expenses leaves significantly more scope for adjustments at the margin.

⁷ For example, Germany only reports social transfers other than in kind as a separate variable whereas social transfers in kind is generally reported along with collective consumption. By contrast, in some years, the French stability programmes only report total social transfers.

⁸ For an alternative approach to isolating the discretionary aspect of fiscal policy based on statistical properties of time-series, see Badinger (2009) and Afonso et al. (2010).

⁹ Belgium, Malta and France do not report sufficiently disaggregated spending plans in the 2001 vintage, the 2005 vintage, and 2001 to 2004 programme vintages, respectively, to compute discretionary spending deviations thus reducing the sample size for this variable by six observations. Since Spain only reports gross fixed capital formation in the period 2001 to 2005 stability programme vintages, we use this variable exclusively to compute the discretionary spending deviations. Since expenditure plans for Bulgaria and Romania are only available starting in 2007, they are not considered in the analysis.

¹⁰ For a detailed description of the computation of this index, see European Commission (2006) and Debrun et al. (2008). The index is normalized to have a zero mean and unit variance. It also displays some time-variability, especially in the 1990's. However, for the time-period considered in this study, the index only varies across countries but not across time, except for a single increase in the expenditure rule index in Italy in 2004 and in France in 2005.

¹¹ Note that since the expenditure rules index is almost perfectly collinear with the fixed effects (note 9), it is not included in the regression equation.

¹² The European Commission computes trend GDP based on a Hodrick–Prescott filter; in general, trend values derived by this method tend to be strongly influenced by actual values at the end of the sample. However, the Commission's methodology corrects for the end point bias, thus supporting our identification strategy; see Röger and Ongena (1999).

¹³ Throughout the analysis, the relevant specification tests support our identification strategy. Based on the Sargan/Hansen statistic, the instrumental variables pass the over identification test. The null hypothesis of weak instruments is clearly rejected. Finally, in many specifications, we reject the null hypothesis that output gap surprises are exogenous or the p-values of the corresponding test statistic are only slightly above 10% thus suggesting the use of instrumental variable estimation.

¹⁴ Several robustness checks of the baseline specification were conducted. For example, we included the lagged spending item under consideration (i.e. lagged primary, discretionary or interest spending) to allow for the possibility that the ambition of fiscal plans depends on the initial size of the respective budgetary aggregate. Moreover, we explicitly controlled for factors giving rise to a ‘denominator effect’ as described in Section II. To this end, we added inflation and the difference between the nominal GDP growth forecast from period t − 1 by the European Commission and the one reported in the stability and convergence programmes as additional regressors. Our main results are unaffected by these variations in the basic model. Results for the robustness checks are available upon request.

¹⁵ For example, Buti and van den Noord (2004), Cimadomo (2008) and Afonso and Hauptmeier (2009) find for EMU and OECD countries, respectively, that governments tend to adopt a looser fiscal stance in election years. Focussing on tax policy, Andrikopolous et al. (2006) do not find evidence for an electoral cycle which is consistent with the commonly held perception that political pressures for budgetary expansion are mainly focussed on the expenditure rather than the revenue side. Interestingly, for the case of the German regional governments (Länder) Jochimsen and Nuscheler (2010) find the opposite pattern with the increase in debt being lower in pre-election years. The authors interpret this finding as reflecting German voters’ preferences for fiscal prudence.

¹⁶ Recall that the overall effect of a marginal change in the output gap surprise is and thus depends on the coefficient on the interaction term and the value of the expenditure rules index.

¹⁷ The charts are constructed according to the methodology suggested by Brambor et al. (2006).