This paper contributes to the literature of public finance and the impact of transfers on local tax revenue raising and spending patterns. We find causal evidence of a Flypaper Effect in Brazilian municipal finance, whereby intergovernmental transfers result in increased public spending at the local level. Furthermore, increased transfers lead to greater local tax revenue collection, dis-
in school resources, as well as the re-allocation of resources towards schools offering higher levels of instruction. In this section we will discuss how these results relate to several recent papers on similar topics.
In their study of municipal finance in Sweden, Dalhberg et al. (2008 [12]) find results con- sistent with a 100% pass-through rate of federal grants to municipal spending. They find no evidence, however, of changes in taxes in response to changes in transfers. Our results are par- ticularly striking in comparison due to the positive effect of transfers on taxes: this increase in taxes ‘crowds-in’ the transfer, so that a 1 R$ increase in transfers raises revenues by 1.47 R$. Spending increases as well, with a 1 R$ increase in the transfer leading to a 1.25 R$ increase in spending.25
Our finding of a substantial Flypaper Effect in Brazilian municipalities is consistent with Gardner’s (2013) [17] study of municipal finance in the Northeastern region. Although the variation in revenue-sharing she exploits is different from our own, she finds no evidence for the crowd-out municipal revenue-raising following increases in federal transfers. Gardner does not find any causal evidence for an increase in taxes due to the transfer, although she does detect a positive relation in the cross-section.
In contrast with our finding of a small but positive increase in education infrastructure and teaching staff, Gadenne (2014) [16] finds no change in classroom number or quality in response to changes in the FPM. In addition to methodological considerations, two differences between our
approaches should be noted.26 First, Gadenne’s results are based on classrooms in use, rather
than the physical stock of available classrooms. While we find an overall negative effect of the transfer on classrooms in use, this effect weakens and becomes statistically insignificant when the interaction with initial municipal GDP per capita is added. Second, while we include all
25
These back-of-the-envelope calculations are based on mean values from Table 4 and the coefficients reported in Table 12. A 1% increase of transfers at the mean is equal to 60 000 R$. This leads to a 0.4% increase in revenues, which translates into an increase of 88 000 R$ at the mean, or a unit increase of 1:1.47. Capital spending increases by 53 700 R$, while current spending increases by 21 500 R$, for a total spending increase of 75 000 R$, or a spending increase of 1:1.25.
26
While both papers exploit the FPM rule for identification, Gadenne (2014) uses a regression discontinuity design based on own-municipality population, while we derive an instrument based on the population of other municipalities in the same state. Furthermore, although the variables used overlap, we measure the log of number of classrooms, while Gadenne reports the number of classrooms per school-aged child. Finally, our panels span slightly different time periods (1996-2006, in our case, versus 1998-2009).
in theory, be the most responsive to changes in municipal funding, they are also more likely to be primary schools (see Table 8). Since the introduction of the FUNDEF in 1998, funding to primary schools is largely determined by revenue-sharing rules which equalise funding on a per- capita basis across the state. Although municipalities were free to contribute additional resources on top of this amount, to the extent that the funding rules were binding, increases in the FPM which resulted from changes in the population coefficients should not affect funding to primary
schools.27 In states where the funding rule did not bind we would indeed expect changes in the
transfer amount to affect local spending, even at the primary level; however, these could be very small in aggregate.
Nevertheless, even our most comparable estimates do not line up with those of Gadenne. In Table 26, columns 5 and 6 report on results of a regression of logged classrooms in use in municipality-run schools only. Both overall, and when an interaction of the transfer with GDP per capita is included, we find a negative effect of transfers on classrooms in use at municipal schools. Ultimately, this negative effect – and indeed, the small fall in the number of classrooms in use overall – should not overshadow the substantial re-allocation we find across schools. Returning to Table 26, for example, we can see that the fall in municipal-school classrooms is compensated for by an increase in state-school classroom. More importantly, it is also associated with a much larger increase in the number of classroom in use in secondary schools (see Table 25).
We cannot comment extensively on the relation of our results to those of Bastos & Straume (2013) [4]. While their finding that increases in municipal transfers cause an expansion of municipality-run preschools appears to contrast with our own results – that is, that preschool classrooms decrease when transfers increase – a direct comparison is not warranted. Our hier- archical classification of schools according to the levels of instruction offered is designed primarily to capture expansions into upper levels of instruction. It therefore classifies a significant fraction
27
To understand why this is the case, consider the funding thresholds imposed by the FUNDEF. In essence, the FUNDEF stipulated that 15% of the primary municipal transfers (including the FPM) must be contributed to a state-wide fund. This fund was then redistributed to public school districts based on the number of students enrolled in primary schools. Exogenous variation in a municipality’s FPM transfer comes from changes in the population thresholds (either own population, as in Gadenne (2014) [16] and Litschig & Morrison (2013) [25], or other municipalities’ populations, as in our own analysis). Since this does not affect the overall FPM but merely its distribution within the state, it should have no impact on the funds collected by the state FUNDEF. Exogenous changes in FPM should therefore not affect the resources a municipality receives from the FUNDEF: resources which are destined toward primary schools.
ondary school, are classified as senior primary schools; see Appendix A.2). For this reason, we do not give much weight to our findings with respect to preschools.