The Positive Effect of Local Tax Capacity on Provision of Water Service

The results of the OLS estimation presented in column 1 of Table 3.4 show that neither the local taxes effort nor royalties have any significant correlation with water coverage, whereas central government water transfers do. In fact, according to the OLS model an increase of 1 percent in the central government water transfers during the period 1994-2005 saw an increase in the proportion of municipal water coverage by 0.054 points. The measure of the number of years following the reform or the existence of an EICEs in the municipality is also not statistically significant in the coverage equation. The existence of an ESP, on the contrary, is associated with greater coverage.

The IV estimations are in columns 2 and 3 of Table 3.4. The first stage of the model indicates that the average undervaluation of properties, as expected, negatively affects the average per capita taxes. The poverty rate, also as expected, is negatively correlated with tax capacity, while the GINI coefficients for land ownership and population exhibit a positive correlation. The second stage in column 3 shows that an exogenous variation of the fiscal effort positively and significantly impacts water coverage. Thus, if per capita taxes increase by 1 percent (over 1994-2005), then the proportion of water coverage increases by a sizable 0.15 points. In the IV model, the water transfers did not turn statistically significant, as the existence of the public or private character of the providers did not explain the increase in water coverage. In conclusion, the differences in the fiscal effort are quite

(continued)

Table 3.4 (continued)

Source: See appendix

*,**,***: Coefficients significant at the 10%, 5%, and 1% levels, respectively. Panel regressions with robust standard errors. Standard errors in brackets

Socio-economic variables such as Poverty Rate, Gini of Land Value, and Population(ln) were included as controls relevant when it comes to explaining the disparities in water coverage across Colombian municipalities. Again, as seen in the previous models for education, the impact of local politics cannot be observed with the variables that we measured. Although we know from qualitative accounts and more detailed fieldwork that local politics does matter, the variables—as presented here— do not capture the impact directly. Indirectly, however, one could interpret the efficiency differences in the impact between national and local resources available to municipal administrations as evidence of there being a virtuous cycle between fiscal autonomy and the outcome of public service delivery.

The variables for quality water are also explained by municipal fiscal effort as shown in Table 3.5. The dependent variable used is the compliance of local drinking water with the potable parameters during the period 2006-2009. In this case, the dependent variable is Wi_t where i is the municipality and t the year.

Column 1 of Table 3.5 presents the results of the OLS fixed effect panel model, showing that per capita taxes are not statistically related to the quality of water, while per capita transfers in fact are. Column 2 displays the first stage of the regression, revealing—as expected—that the undervaluation of the properties negatively impacts taxes. The second stage, in column 3, shows that an exogenous variation of per capita taxes leads to improved water quality. Both per capita transfer and per capita royalties are not significant in explaining water quality. The model also gave estimations using random effects (last three columns), and although the coefficient for

Source: See appendix

******. Coefficients significant at the 10%, 5%, and 1% levels, respectively. Panel regressions with robust standard errors. Standard errors in brackets Socio-economic variables such as Poverty Rate, Gini of Land Value, and Population(ln) were included as controls per capita taxes is lower than with fixed effects it is greatly significant. As in the IV fixed effects model, the coefficients for per capita transfers and royalties are not statistically different from zero. While the political variables have no clear impact in the fixed effects model, some of them turn significant in the random effects model. The Squared Effective number of parties has a negative impact suggesting that a limited or fragmented local party system could negatively affect the provision of quality water. Thus, we find convincing evidence that greater local fiscal effort is linked to greater coverage and better quality of water. As with the education sector, both national transfers and royalties seem to have a limited impact on the provision of the goods, for different reasons. National transfers, on the one hand, may be too strict and rigid, while royalties may be too flexible.