Budget deficit and interest rates: empirical evidence for Spain

Abstract

Evidence, whether international or national, of the significance of the links between budget deficits and interest rates, is in general terms inconclusive. The aim of the present work is to contribute new findings for the Spanish economy using annual data for the period 1964–2000. It is found that budget deficits did not appear to raise long-run nominal interest rates during our sample period.

Notes

 See Equation 1 in Evans (1985). In the present application, it is not chosen to model a reduced form for the inflation expectations; instead, a proxy is used in the estimated equation.

 Mauleón and Pérez (1984), Mauleón (1987), Raymond and Palet (1990), and Esteve and Tamarit (1995) all reject the Ricardian hypothesis. Only Ballabriga and Sebastián (1992) do not reject the REH.

 In short, the statistical data were obtained from the National Statitical Institute (INE), the Bank of Spain, the Ministry of Finance and other indirect sources.

 To complement the analysis the ARDL procedure proposed by Pesaran and Shin (1999) and Pesaran et al. (2001) was employed to examine the cointegration properties of the estimated equations. The main advantage of this testing and estimation approach lies in the fact that it can be applied irrespective of whether the regressors are stationary or integrated, and this avoids the pretesting problems associated with standard cointegration analysis which requires the stationarity properties of the variables.

The ARDL procedure involves two stages. At the first stage the existence of the long-run relation between the variables under investigation is tested computing the F-statistic for testing the significance of the lagged levels of the variables in the error correction form of the underlying ARDL model; Pesaran et al. (2001) have tabulated the appropriate critical values because the distribution of the F-statistic is non-standard. In the present case, these computed F-statistics were 14.47, 12.92, 9.16 and 4.59 for regressions 1–4, respectively. In all cases these values fall well outside the 99% band covering all the possible classifications of the variables into I(0) and I(1), or even fractionally integrated one and, therefore, the null hypothesis of no cointegration can be rejected at 99%.

The second stage of the ARDL procedure is to estimate the coefficients of the non-spurious long-run relations selected in the first stage. In the present case all the four equations analysed are not in fact spurious and the point estimates arrived at were similar to that obtained with the 2SLS procedure. In particular, neither of the coefficients associated to the budget deficit variable was significant. The only remarkable differences were the insignificance of the public debt variable in Equation 3 and the significance of annual real GDP growth in Equation 4.

 Lagged values of the regressors (except for the i* variable, which is considered to be exogenous) and the dependent variable were used as instruments.

 Moreover, this method allows to correct for the bias produced by the use of expectations variables in the proposed models.

 Because the interpretation of the stock of public debt as a proxy for country risk, investors will demand a higher risk premium if the country's debt–income ratio is high and, thus, long-term interest rates will be a positive function of this ratio. Seater (1993, p. 176) finds two possible explanations for a statistically significant negative relationship between government debt and interest rates: problems of future taxation and the absence of any marginal tax rate variable in the estimated regressions.

 In all cases the estimated coefficient of the expected inflation variable has a value less than unity. Bajo and Esteve (1998) also find a partial Fisher effect for the Spanish economy in the long term that is explained by the existence of some form of monetary illusion in the Spanish financial market.