Fiscal rules, discretionary fiscal policy and macroeconomic stability: an empirical assessment for OECD countries

Abstract

Does aggressive use of discretionary fiscal policy induce macroeconomic instability in terms of higher output and inflation volatility? Three main conclusions arise from our cross section and panel analysis for a sample of 20 OECD countries: first, discretionary fiscal policy has a significant and sizeable effect on volatility of GDP (per capita) and all of its components. Second, there is no direct effect on inflation volatility; since output volatility is an important determinant of inflation volatility, discretionary fiscal policy indirectly exacerbates inflation volatility. These results turn out robust with respect to alternative fiscal policy measures and endogeneity concerns. Finally, many of the fiscal rules introduced since 1990 appear to have reduced the use of discretionary fiscal policy.

Notes

¹ As mentioned by Gali and Perotti, equation 2 can also be interpreted as reduced form of a structural model, where governments have a target level of the debt-GDP ratio and there are costs of changing the structural deficit over time (see Ballabriga and Martinez-Mongay (2002)).

² Since Equation 2 is interpreted as fiscal rule, and the first-stage regression as forecast function of the policy maker for the output gap, we use the second-stage residuals (rather than the structural residuals) of the IV estimation as measure of fiscal shocks. This is also required to obtain orthogonality between the endogenous structural component (i.e. the second-stage predicted values from Equation 2) and the exogenous structural component (i.e. the second-stage residuals from Equation 2).

³ For total primary spending (EXP), total primary receipts (REC), and the primary deficit (NL) the GDP deflator was used to convert the nominal figures into real terms.

⁴ Notice that the time periods from which the country-specific standard deviations are calculated differ somewhat across countries as a result of data availability (see Appendix A1). The results are hardly affected, if overlapping time periods are used.

⁵ For the revenue-based measure of fiscal policy, fiscal decentralization in terms of revenues was used (FD^REV ); for the deficit measures, both FD^EXP and FD^REV were included as instruments.

⁶ A similar point, though referring to counter-cyclical fiscal policy, was already made by Friedman (1953).

⁷ Another requirement for the application of the GMM system estimator is that the series are stationary. We checked for stochastic trends using panel unit root tests, allowing for individual root processes: the null of a unit root is rejected for both output volatility and our fiscal measures. Regarding the presence of deterministic trends, it should be borne in mind that all models include time-specific effects; this is equivalent to transforming the data into deviations from time means, which implicitly controls for the presence of a common trend.

⁸ As an alternative measure, we used the CPI deflator, and obtained qualitatively identical results.

⁹ This holds true if the volatility of the output gap is omitted.

¹⁰ The similarity of the results when the volatility of the output gap is replaced by that of output growth (per capita) is plausible; if the trend growth is constant the two variables by measure exactly the same thing.

¹¹ Since e = My = Mε, where M = I-X(X′X)⁻¹X′, it follows that Var(e) = Var(Mε) = σ²MM′ = σ²M, which is not diagonal (even if Var(e) = σ²I).

¹² Note that there are no instruments for the first observation available; as a result of differencing, T_i-1 equations remain.

¹³ This requires the first moment of to be stationary (which is fulfilled here).