The impact of government deficits on growth
has been, for long, the object of a controversy between the Keynesians, for whom
an ex ante reduction of the government deficit has negative impacts on growth
(at least in the medium-term), and neo-classical economists, for whom activity
is entirely determined by supply, so that government deficits do not have any
effect on activity. The Danish experience of 1983 - 1986 and the Irish experience
of 1986 - 1989, during which drastic reductions in the government deficits did
not involve significant output losses, have led to theoretical models being proposed
which build a bridge between these two antagonistic approaches. According to these
models, the economy would be rather Keynesian in normal times, but "anti-Keynesian"
in a period of budgetary crisis.
The goal of this paper is to test systematically whether such an anti-Keynesian behaviour does occur during the large-scale retrenchments and expansions experienced in 17 OECD countries since the early 1970s. The selected fiscal episodes are periods of continuous improvement (or deterioration) of the primary structural surplus, which include a sub-period when change was large. 37 episodes are thus listed, of which a careful descriptive analysis is then carried out, to look for factors that might explain the great diversity observed in their effects on activity. The very question of the measurement of these effects (based on growth observed ex post) is discussed, and several indicators are constructed. Four main categories of factors are investigated: (i) the overall context in which the episodes were undertaken (the fiscal situation, the initial position in the business cycle); (ii) some intrinsic features of the country (such as its size, openness); (iii) the composition of the fiscal adjustment (or expansion); (iv) the monetary and exchange rate policies during the episodes. Moreover, in order to examine whether large scale episodes exhibit specific features, the same analysis is carried out for more standard episodes of fiscal policy. The conclusions of this descriptive part of the study are: (i) large-scale retrenchments were generally undertaken in response to a large deterioration in the public finances, while large-scale expansions most often responded to a slowing-down of activity; (ii) large-scale adjustments frequently resulted in smaller proportional effects on the activity than standard ones, once accounting for the initial positions in the business cycles; in particular, a non-negligible number of such episodes did not lead to output losses; (iii) while large-scale retrenchments were indeed undertaken in smaller, more open countries on average, no regularity in changes in interest rates and exchange rates nor in the composition of retrenchments can be observed; (iv) changes in the household savings ratio appear to be clearly correlated to the effect of fiscal policy on the activity: 'anti-Keynesian' fiscal retrenchments were in general followed by a drop in household savings ratios.
The second part of the paper makes an attempt at reconciling the results of the descriptive analysis with Keynesian mechanisms. For this purpose, two possible channels that may account for the divergence between reality and the neo-Keynesian approach are studied. First, the risk of a bias in the selection of the episodes towards 'successful' adjustments in terms of growth (which could occur if, for instance, fiscal restrictions were immediately interrupted in the case of a slowing of the activity, so that the large-scale retrenchments selected would only be those which had smaller output costs) is investigated and eliminated. Then, the noise due to the combination of small sample size and to shocks other than the fiscal policy over the episode, as well as the uncertainty surrounding the Keynesian multiplier, as embodied in simulations with Keynesian macro-econometric models, are taken into consideration to test whether the average observed multiplier (which is close to zero) really contradicts the predictions of Keynesian models. The average observed multiplier does not appear significatively different from the theoretical multiplier, which does not formally contradict the results of Keynesian macro-econometric models, but the empirical data are not normally distributed around this average, and probably include too many negative multipliers to be conform to the results of these models in particular.
The last part provides an econometric study of the link between the success of an adjustment and consumer behaviour. To this end, cross-section episode consumption regressions are run for the sample countries, using as common explanatory variables, fiscal policy both during large-scale episodes, and outside such episodes, and by setting other coefficients to be identical for all countries or not. The major results are the following: (i) fiscal balance variables are significant during large-scale episodes only, and their effect on consumption is higher when the coefficients of other variables are constrained; (ii) the fiscal balance variable is significant only for anti-Keynesian large-scale retrenchments when an additional distinction is made between 'Keynesian' and 'anti-Keynesian' episodes. Thus, anti-Keynesian fiscal retrenchments would indeed have been associated with a greater decline in the household savings ratio than would be explained by traditional, consumption-smoothing behaviour. |
Abstract |
