Fiscal impacts of electoral abstention: a study on the electorate biometric update in Brazilian municipalities

ABSTRACT

Because electoral abstention may generate a difference between the preferences of general voters, i.e., those who are eligible to vote, and the preferences of effective voters, i.e., those who do vote, policies adopted by incumbents may differ according to differences in turnout rates across the electorate. The Brazilian biometric electorate update offers an innovative opportunity to explore exogenous variations in abstention rates, allowing us to verify its impact on public policies, especially local public expenditures. By combining propensity score matching, differences-in-differences and instrumental variables models, we find that the electorate biometric update decreased abstention rates in local elections in Brazil, which, in turn, changed local public spending composition towards expenditures on education. The remaining categories of public expenditures explored in this study, however, seem not to be affected by the change in the electorate composition.

KEYWORDS:

• Electorate biometric registration
• voter turnout
• fiscal policy
• Brazilian municipalities
• Propensity score matching
• Differences-in-differences
• Instrumental variables

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Reinforcing the positive self-selection of migrants, there is also extensive literature on the migration of human capital to places where it is already abundant, a phenomenon known in the literature as brain drain (Beine, Docquier, and Rapoport 2001, 2008).

2. The CPF is an important personal document equivalent to the Social Security Number in the USA or the National Insurance Number in England.

3. Reais (R$) is the Brazilian local currency, currently equivalent to 1 US$ = R$ 5.3.

4. Like an electricity bill, for instance.

5. The PSM based on separate cross sectional logit estimations for each year between 2009 and 2012 was chosen due to the better pairing quality – see Tables 2, A3 and Figure 1. As an alternative procedure (with a lower pairing quality, however), we formed new treatment and control groups matched according to estimations provided by panel fixed effects (or conditional) logit estimations. See results presented in section 7.

6. This variable is measured as index as of 2008. Therefore, for each Brazilian municipality, the electorate size is divided by the 2008 electorate size such that for 2008 this variable is normalised to one.

7. According to Table A2 in the Appendix, it is possible to observe that all categories of local public spending (measured as % of total local spending) have stable means and standard deviations over the sample years.

8. In particular, it is possible to observe that the two groups of municipalities have similar proportion of households served with sewage collection and piped potable water, similar geographic area and population size. similar indicators of child health and individuals with similar monthly income.

9. The null hypothesis of the Hausman test is that both fixed and random effects estimations are consistent but the last is efficient. The alternative hypothesis is that only fixed effects estimation is consistent. The test statistic is distributed according to a χ² distribution function.

10. For OLS and fixed effects, we report the F statistic for the overall significance of coefficients. For random effects estimations, we report the χ² statistic (See Cameron and Trivedi 2005).

11. Descriptive statistics regarding the dependent and independent variables of the second stage for the matched sample are presented in Table A1 in the Appendix.

12. Therefore, for each Brazilian municipality, 2008 and 2012 values were both divided by 2008 values such that numbers for 2008 are normalised to one.

13. According to this model, the estimation is conditional on those sectional unities whose dependent (binary) variable changes at least once over time, i.e., sectional unities whose dependent variable is always zero or always one are excluded from the estimations (Cameron and Trivedi 2005).

14. For this matching, nontreated and treated municipalities were matched according to the least propensity score difference conditional on belonging to the same year, i.e., a municipality treated in 2009 was paired to a nontreated municipality of the same year with the closest propensity score.

Additional information

Funding

The work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico Coordenação de Aperfeiçoamento de Pessoal de Nível Superior.