Mayoral Public Sector Work Experience and Tax Collection Performance in Colombian Local Governments

ABSTRACT

Local governments in Latin America tend to have limited capacity to collect taxes. Existing theory explains tax collection performance as a function of political, socio-economic, and cultural factors. Yet across Latin America, tax collection performance varies greatly between localities that are similar along these dimensions. We suggest that organizational capacity, represented by managers' human capital, explains variation in local tax collection performance. We test this theoretical idea against panel data on most Colombian local governments, where elected mayors are the top managers determining the local tax collection strategy. We examine two indicators of tax collection performance: (1) the per-capita amount of property tax collected and (2) property tax collected as a percentage of total property valuation. Controlling for important confounders, the length of the mayor's prior public sector work experience is positively associated with the per-capita amount of property tax collected but not the percentage of total property valuation collected in property tax. Our results suggest that Colombian mayors gain relevant skills over their career. Our study has implications for other developing countries with a similar property tax collection system.

ACKNOWLEDGEMENTS

We thank the anonymous reviewer and the editor for helpful suggestions. Special thanks to Giovanni Cortés and others staff members in the Office of Public Finances of the National Planning Department as well as to Diana Marcela Cuellar and others staff members in the Geographic Institute Agustín Codazzi for providing data on and detailed description of the public finances of the Colombian municipalities.

Notes

Notes: We report ordinary least squares coefficient estimates, with Huber-White t-statistics (adjusted for clustering on municipality) in parentheses.

*Significant at 10%; **Significant at 5%; ***Significant at 1%.

This even holds true to some extent in England, an area of very high government capacity. The collection rate of council tax, a modified property tax, varied between about 80% and 99% across English local governments for financial years 1999/2000 to 2004/2005, with approximately 10% of local governments failing to collect 10% or more of tax due (Boyne et al. 2009).

According to the Leviathan model (Brennan and Buchanan 1980, 184), fiscal decentralization limits the tendency of central government to grow to its constitutionally possible maximum size by generating local competition, assuming citizens are mobile. The evidence is mixed, however. For instance, Oates (1985) found no systematic relationship between the degree of decentralization of the public sector and the size of government as a proxy for revenue maximization.

In this study, instead of focusing on local tax capacity (capability to finance public services), we focus on tax collection performance in terms of local tax effort (a local government's effort to utilize its tax capacity). See Berry and Fording (1997) on the measurement of state capacity and effort in the U.S.

The Laffer curve can provide interesting insights for understanding revenue collection. While it has been discussed widely, the basic idea is a simple truism: If the tax rate is zero, zero tax will be collected. If the tax rate is 100%, productive activity will cease, and again zero tax will be collected. The revenue-maximizing tax rate must therefore lie somewhere strictly between zero and 100%. Buchanan and Lee (1982) suggest a mechanism to explain why some jurisdictions find themselves using a tax rate higher than the one that would yield the maximum tax revenue. The mechanism is based on decision makers' short-term interest as opposite to the politically irrelevant long-term perspective.

On the effect of grants on taxes, another strand of literature, termed the “Friedman-Sanford hypothesis,” suggests that temporary federal grants create future ratchets in state taxes because recipient states feel obligated to continue the programs once financed by the grant, pushing them to offset the grant monies with taxes (Sobel and Crowley 2010).

That was the case of Bolivia's National Revolutionary Movement (MNR), Colombia's Liberal and Conservative Party, and Venezuela's Democratic Action (AD) and the Christian Democratic Party (Copei). On the contrary, in Ecuador and Peru, multi, short-lived parties failed to devolve power to sub-regional level due to their inability to gather legislature consensus (Barr 2006).

Within Latin America, Colombia exhibits high levels of decentralization. According to the Inter-American Development Bank, Colombia ranks third in the region after the federal states of Argentina and Brazil (Inter-American Development Bank 1997).

In the U.S., these are council-manager and mayor-council; in the UK, cabinet-leader and elected mayor with a cabinet. In the U.S., there are also a third and a fourth form. The third form is the commission government. About 1%–2% of local governments still use this form. They are mostly small cities. The fourth form consists of town-meeting and representative town-meeting governments. They are found in Connecticut, Maine, Massachusetts, New Hampshire, and Vermont (Dye and McManus 2003, 341–345).

During the time period covered by our study, immediate re-election of the mayor was prohibited, which helps us by ensuring a lot of overtime variation. This prohibition has since been lifted. In October 2009—after 18 years—the Colombian Senate approved immediate re-election to a second term for mayors and governors, which went into effect for those elected in October 2011 for the administrative period of 2012–2015. These mayors and governors will be eligible to be re-elected once in 2015 to serve another four-year term.

The SGP reserves 4% for special assignations and the remaining 96% is distributed among all the Colombian municipalities (1,102), allocating 24.5% for health, 58.5% for education, and 17% to cover unsatisfied needs as well as to finance urban and rural infrastructure.

In readjusting property valuation, Colombia employs a split system in which the central government, through a parastatal agency—the Instituto Geografico Agustín Codazzi (IGAC)—conducts the valuations but the local government collects (Dillinger 1991).

The tax rate for urban land without construction can be up to 33 per thousand Colombian pesos.

Cf. Registraduría Nacional del Estado Civil (2008).

We used the GDP deflator provided by the World Bank to deflate all currency variables. The base year of this variable for Colombia is 2005.

The index is developed by the Colombian Chapter of Transparency International (http://www.transparenciacolombia.org.co/).

We use pooled ordinary least squares with a lagged dependent variable instead of one of the more sophisticated dynamic panel data estimators (such as Arellano-Bond), as the latter would require more time periods than we have in our panel, which has only four time periods. A dynamic panel data estimator would have the advantage of removing time-invariant differences between municipalities. In our models, much of this variation is absorbed by the coefficient on the lagged dependent variable, which is consequently larger than it should be (e.g., see Bond 2002). Pooled ordinary least squares has been widely used in public administration research on similar “short” panels, i.e., many cross-sectional units, few time periods. As one example, see Meier and O'Toole (2003), which uses a panel of Texas school districts.

Interestingly, a third dependent variable, the percentage of billed property tax that is actually collected is a random variable in every sense of the word, i.e., that specification was not statistically different from a constant-only model. In Colombia, the property tax rates published in municipal ordinances have no bearing on what is actually collected, and they vary unpredictably from year to year. This is very different from the North American or European experience, where tax rates do determine total revenue, adjusting for the overall size of the tax base and the amount of exemptions.

Percentage point change = 100 × (exp(estimated coefficient) − 1).

100 × (exp(0.0315) − 1) ≈ 3.2.

We considered the possibility that there might be an optimal number of middle managers per capita, such that exceeding this number would lead to a reduction instead of an increase in tax collection (implying an inverse U-shaped relationship). Therefore, we also estimated both models from Table 2 with an additional variable included: the squared value of the number of middle managers per 1,000 population. In these two specifications, neither the base nor the squared term of middle managers per 1,000 population was statistically different from zero, although the null hypothesis of a joint test of both variables was rejected in both specifications. The results are in accordance with the idea that the majority of local governments in our data have too few rather than too many middle managers.

100 × (exp(0.0055) − 1) ≈ 0.6.

Looking at Table 1, one standard deviation of the transparency index is about 8.5 percentage points.

8.5 × 100 × (exp(0.0038) − 1) ≈ 3.2.

100 × (exp(0.1317) − 1) ≈ 14.

100 × (exp(0.0628) − 1) ≈ 6.5.

Tax expenditures refer to exclusions, exemptions, allowance, deductions, credits, preferential tax rates, or tax deferrals (Glenday and Swift 2006, 1). To identify tax expenditures, Glenday and Swift (2006) recommend establishing a normative or a benchmark tax structure. Nevertheless, in Latin American countries, determining their benchmark tax structures seems difficult due to the extensive number of informal jobs in agriculture and services (Glenday and Swift 2006).

Additional information

Notes on contributors

Nicolai Petrovsky

Nicolai Petrovsky ([email protected]) is an Assistant Professor in the Martin School of Public Policy & Administration at the University of Kentucky. His PhD is from Texas A&M University. Prior to joining the Martin School, he worked at Cardiff University. His research is concerned with determinants of governmental and organizational performance as well as consequences of performance measures. He is currently researching aspects of these topics in projects on British executive agencies and Colombian local governments.

Claudia N. Avellaneda

Claudia N. Avellaneda ([email protected]) is an Associate Professor in the School of Public and Environmental Affairs (SPEA) at Indiana University. She received her PhD in Political Science at Texas A&M University. She is interested in public management, public policy, and comparative politics with a regional focus on Latin America. Her current research examines the determinants of local government performance and highlights the role of mayors' background, networking activities, and decision making.