The duration analysis
Table 2 presents the estimates of Eq. (1) above. The first specification (reported in the first column of Table 2) includes only the control variables taken from the existing literature (vector X) and does not include the decentralization variable, which facilitates comparisons of our results with those published in earlier analyses. The estimates in the second and third columns include expenditure decentralization (edec), which is, in fact, our main variable of interest.Footnote 11
Table 2 Baseline results of the duration analysis
The negative and significant effect of edec (see both columns 2 and 3 in Table 2) indicates that more decentralized countries have a lower hazard rate, i.e., that the probability of a consolidation period lasting for an additional year increases with higher degrees of decentralization. Figure 2 illustrates this survival function for different levels of decentralization. The survival function for countries with a 50% degree of decentralization (the line farthest to the right) is significantly shifted to the right compared to countries with lower degrees of decentralization. Therefore, ceteris paribus, more decentralized countries consolidate for longer periods.
The rest of the independent variables in the model return the expected signs, but many are not statistically significant. On the macroeconomic side, it emerges that an increase in interest payments makes it more likely that a consolidation period will be terminated sooner. On the political side, we find that right-wing governments tend to consolidate for longer periods, ceteris paribus. While the political cycle does not appear to affect the duration of the consolidation episodes (the coefficient associated with the election year dummy is not statistically significant), the coefficient for type of government shows a significant and negative impact. Another interesting result is that consolidations based on tax hikes are more likely to result in shorter consolidation episodes.
Thus, our initial analysis suggests that fiscal decentralization has a significant effect on the duration of fiscal consolidation. We seek now to identify the channel responsible for this effect. On the one hand, more decentralized countries may find it easier to consolidate for longer periods simply because central governments can shift the adjustment burden down to sub-national sectors. On the other hand, the positive impact on duration could be explained by the fact that sub-national governments contribute substantially to the stabilization function of economies making it easier to sustain longer lasting consolidation efforts. The first effect would be more pronounced if fewer legal powers have been transferred to sub-national sectors: if sub-central governments are responsible for large shares of expenditure, but do not possess corresponding revenue-raising powers, it is easy for the central government to cut transfers to the detriment of lower governmental layers. We test this hypothesis by augmenting the model with the respective indicators of real autonomy and their interaction terms with decentralization. The results are presented in Table 3.
Table 3 Duration analysis with interaction effects
In general, the negative coefficient on expenditure decentralization holds across all specifications. However, the baseline effect is significant only when the indicator for executive influence is entered in the model, which might indicate a problem of collinearity, given that in some instances expenditure decentralization might evolve in line with the real autonomy variables. Including an interaction term in model (1) between expenditure decentralization and a real autonomy indicator (any one of the three included separately in the model) results in a positive and statistically significant coefficient associated with the main effect of fiscal autonomy and a negative coefficient on the interaction term. This implies that decentralized countries consolidate for longer periods but that the fiscal autonomy of sub-national governmental tiers can dampen this to the point that the result is offset if sub-national governments have enough real autonomy over their fiscal policies. To interpret the negative interaction term and to illustrate the overall effect, we plot the survival functions in Figs. 3, 4 and 5 for different combinations of the two variables, introducing the three real autonomy variables separately.
In Fig. 3, the line furthest to the left represents the effect obtained with a high degree of fiscal autonomy at sub-national levels and a low degree of expenditure decentralization (20% of the sample). Countries with these characteristics tend to consolidate for shorter periods of time. With the same level of expenditure decentralization, but with no fiscal autonomy, the consolidation spells have, on average, a slightly longer duration as the survival curve shifts to the right. This indicates that, in countries that have quite a high degree of centralization, greater real fiscal autonomy might avoid the adjustment burden being shifted to sub-national sectors (whether this indeed happens is analyzed in the next step after discussing the other effects of real autonomy on duration). Turning to the countries characterized by high degrees of decentralization (around 50% of the sample), the curves shift as expected to the right, as greater decentralization increases survival probability, that is, the probability of having longer consolidation spells. In fact, these two survival functions furthest to the right almost overlap. With no fiscal autonomy, the main effect of decentralization is found to prevail, as the interaction term and the main effect of fiscal autonomy both are equal to zero When switching to a high autonomy regime, we would expect a shift to the left, as such regimes were found to consolidate for shorter periods of time. However, this effect is largely offset by the negative interaction term, which acquires considerable relative importance.Footnote 12 At this level of decentralization, it seems that switching from a high- to low-fiscal autonomy model of decentralization no longer affects the duration of the consolidation episodes.
The results for real executive influence, illustrated in Fig. 4, are in line with those obtained for fiscal autonomy. This finding suggests that decentralization does not operate solely through the interaction term, but it also shifts the survival function to the right by itself. While the order of the lines is the same as in Fig. 3, the impact of decentralization now more than offsets the reduction in the survival probability owing to executive influence. Figure 5 completes the analysis by showing the results for the indicator for the co-determination of national decision-making (shared ruling). Again, the results are similar to those obtained for the other two indicators, suggesting that the most important component in the institutional arrangements is the fiscal element and highlighting the importance of the proper design of the fiscal institutions shaping intergovernmental relations.
Thus, the results so far suggest that countries with high degrees of decentralization tend to consolidate for longer periods; however, this effect is mitigated by greater real autonomy in the hands of sub-central governments. In contrast, if countries are decentralized to a lesser degree (here 20%), central governments can shift the burden of longer consolidation periods to those lower tiers of government, unless, that is, they are shielded by some element of real autonomy. If sub-national governments possess a sufficiently high degree of real autonomy, the duration of consolidation periods tends not to be so long. However, with a high degree of decentralization with respect to expenditure, real sub-national autonomy essentially makes little difference.
The effects on sub-national public finances
The second step in the analysis, the estimation of Eq. (2), seeks to determine whether the fact that more decentralized economies consolidate for longer periods is driven by cuts in transfers to sub-national tiers of government. Table 4 suggests that implementing consolidation measures and the duration of such measures (consolidation and consolidation time, respectively) both affect the change in intergovernmental transfers negatively. In other words, when statistically significant, the effect of consolidation periods on sub-national public finances is negative. This suggests that fiscal consolidation in the past has reduced the intergovernmental transfers received by local governments. In addition, the negative baseline effect of consolidation is even larger in magnitude when controlling for real sub-central autonomy (mainly fiscal autonomy and executive influence, as observed in columns 2, 3 and 5).
Table 4 The effects of consolidation on the change in intergovernmental transfers received by sub-central governments
An examination of the effect of real sub-central autonomy seems to show that only shared ruling significantly and positively affects the change in intergovernmental transfers (columns 4 and 8). This means that if a regional government or its representatives can exercise authority over the institutional set-up by co-determining national legislation and policy through intergovernmental conferences (e.g., by either directly participating in making national law or by sharing executive responsibilities with the national government for implementing policy either in the region or in the country as a whole),Footnote 13 central government transfers may increase. This finding suggests that the more regions are able to write the rules of the game, the more they are likely to obtain in terms of resources from a bargaining process with the central authority.
In order to analyze the joint effect of consolidation and real local autonomy on the change in transfers, it therefore seems worth considering the interaction between the two. In this respect, Fig. 6 plots the marginal effect of a consolidation period on the log-difference of transfers for all the possible values of the fiscal autonomy variable.
The slope of the interaction terms proves not to be significant in the regression. Nevertheless, the figure provides a number of interesting insights. First, during consolidation periods, transfers to lower levels of government are substantially reduced as the impact of consolidation is negative and different from zero. This suggests that consolidation—at least partially—has been achieved in the past by reducing transfers to sub-national levels, confirming the stylized facts presented in Vammalle and Hulbert (2013). However, the effect is different from zero only as long as real autonomy, measured here in terms of fiscal autonomy, is not sufficiently large. Sub-national sectors with a high degree of fiscal autonomy can use this additional power to prevent central government cuts in their transfer revenue.
Similar evidence is offered when repeating the exercise with the other two indicators of real sub-central autonomy, that is, executive influence and shared ruling (Figs. 7, 8).
While the slope becomes flatter, the estimated averages remain negative, indicating transfer cuts during consolidation periods. Here again, sub-national sectors that enjoy substantial influence in policy making do not experience any significant reductions in transfers during consolidation periods, given that for higher values of this indicator, consolidation periods are not characterized by any significant cuts in transfers to lower levels of government.
The success of consolidation
The last step in our analysis examines the success of the consolidation episodes in terms of the reduction achieved in the public debt/GDP ratio and the improvement attained in the budget balance/GDP ratio. This step builds on previous studies that have adopted similar empirical strategies for identifying successful fiscal consolidations (e.g., Alesina and Perotti 1995; Heylen and Everaert 2000; Alesina and Ardagna 2010).Footnote 14 Table 5 shows the average changes in the budget balance/GDP and public debt/GDP ratios during the consolidation episodes in our sample. Statistics are presented for the full sample as well as for the two sub-samples of countries grouped by their degree of decentralization, i.e., considering expenditure decentralization below and above its average.
Table 5 Average changes in deficit/GDP and debt/GDP during consolidation episodes
On average, the primary balance/GDP ratio improves during periods of consolidation, with annual values between +0.44 and +2.01% points of GDP. Long consolidation episodes achieve substantial cumulative increases in this ratio, with a maximum of +16.11, which is the average total change for the 2 eight-year periods in the sample. In most cases, the average yearly improvement in the primary balance/GDP ratio is greater for less decentralized countries (i.e., with values of expenditure decentralization below the edec mean), particularly for very short consolidation episodes. For example, in the nine one-year consolidations in countries with low degrees of decentralization, the primary balance/GDP ratio improved by an average of 1.54% points, while in more decentralized countries it increased by just 0.55% points of GDP. A further finding, which is unsurprising given the results of our analysis above, is that the majority of the short-lived episodes of consolidation are concentrated in the sub-sample of countries with low levels of expenditure decentralization.
Table 5 also reports the impact on the public debt/GDP ratio, where the total change is calculated as the value of debt/GDP at the end of the consolidation period minus its value in the year immediately preceding the start of the consolidation episode. On average, the debt/GDP ratio increases when governments reduce the primary deficit, something that can only be explained in terms of the adverse effects on GDP of the consolidation efforts and high interest payments in these periods. In fact, Perotti (2012) convincingly argues that the four cases normally presented as proof that austerity measures can be expansionary should be considered exceptions rather than the rule. In our sample, the debt/GDP ratio increases on average during consolidation episodes, in many cases quite notably (for instance, the average increase for the three five-year episodes is equal to +21.32). Interestingly, when episodes of the same length appear in both sub-samples, the highest average increases in the debt-to-GDP are concentrated in the sub-sample of the less decentralized countries (the one exception being the debt increase during the two-year long consolidation episodes).
Thus, our evidence shows not only that consolidation episodes are of longer duration when expenditure decentralization is higher, but it also seems to indicate that improvements in the primary balance/GDP ratio are, on average, less pronounced. Our results are in line with those reported by Darby et al. (2005), who use a smaller unbalanced panel of 15 OECD countries for the 1970–1999 period,Footnote 15 and find that high levels of expenditure decentralization may result in a fiscal environment that is not conducive to successful consolidation attempts.
While the previous exercise was carried out for the general government, the following (and final) analysis of the success of consolidation seeks to disentangle the consequences of consolidation for central and local public finances, respectively. We estimate simple econometric models with the change in the central and local government debt/GDP ratios (introduced separately) as dependent variables, and with the variables characterizing the consolidation episodes on the right-hand side. The results are reported in Table 6 and they show a positive correlation between both consolidation events (i.e. consolidation) and their duration (i.e. consolidation time) with increases in the central governments’ debt/GDP ratios, although the estimated coefficients are not statistically significant (see columns 1 and 6). In the case of the change in sub-national debt, the correlation with consolidation is positive while that with duration is negative, but once again the coefficients are not statistically significantly different from zero (see columns 2 and 7).
Table 6 The effects of consolidation on the change of central and sub-central debt/GDP ratio
However, some statistically significant evidence of consolidation emerges when controlling for the level of real sub-national autonomy (see columns 4 and 5). Indeed, the coefficient on fiscal consolidation is positive and becomes significant, suggesting that the change in sub-national debt-to-GDP ratio increases when consolidation packages are implemented at the country level. This “shift” effect from central to local government public finances during consolidation seems in fact to be mitigated by the real authority exerted by sub-national governments, as demonstrated by the negative (and significant) coefficients on the interaction terms. Hence, in countries with enough real sub-national autonomy, the increase in the variation of the debt/GDP ratio is smaller during a consolidation episode. This confirms the hypothesis that sub-national governments with sufficient political power and influence in politics at the center are shielded against this type of intervention, and is in line with the results reported by Foremny and von Hagen (2013), who find a similar effect for federal versus unitary countries.
Similar results hold in the case of consolidation time as shown in column 9, where the interaction between consolidation time and executive influence is negative and statistically significant, while the coefficient on consolidation time alone is positive. Ultimately, consolidation efforts at the national level can worsen local public sectors’ fiscal stances—i.e., increasing their change in debt-to-GDP ratios—if the latter do not have sufficient authority, responsibility and participation with respect to the central government and its fiscal decisions.