Independent Professional Bureaucracies and Street-Level Bribery: Comparing Changes in Civil Service Law and Implementation in Latin America

Abstract

Using cross-sectional data, studies of the relation between merit-based bureaucracies and corruption usually find that nations with more professional and politically independent bureaucracies have lower corruption. However, cross-section designs cannot test this policy claim. This study adopts a pre-test‒post-test design using lagged data comparing changes in civil service laws and implementation in eight Latin American countries from 2004 to 2012 and data reporting subsequent change in bribe requests by the bureaucrats that the laws are intended to affect. Raising questions about the validity of previous results, it is found that improved laws and implementation were associated with no or increased bribe requests.

Keywords:

• anti-corruption policy
• comparative pre-test‒post-test design
• Latin America

Acknowledgement

The authors thank Nicolas Acevedo, research assistant at the Alberto Lleras Camargo School of Government, Universidad de los Andes, for his help with this project.

Notes

1. Corruption and illegal rent seeking are widely regarded as costly transfers because they almost always lead to lower economic growth (Hillman 2009, p. 99; Bentzen 2012). Mauro (1995) provided early empirical evidence that corruption is costly, and a theoretical explanation (2004) of why corruption persists, despite its adverse effects on growth.

2. We include Venezuela in our descriptive results. We drop it from the statistical analysis because the post-test bribe request data for Venezuela are not available for any of the time periods that succeed the pre-test data on the quality and implementation of laws in Venezuela.

3. LAPOP is a public opinion survey carried out every two years by the University of Vanderbilt. Covering numerous topics of governance, its survey design incorporates a weighting scheme to accurately reflect each country’s sample size relative to the population. The survey in each country is a stratified, random sample, with different weights in each sample. Our analysis reflects the population weights in each country.

4. The LAPOP survey also asked respondents about bribe requests from policemen, but in that case it did not first ask whether the respondent “used” (i.e. had contact with) the police. In the case of police, the distinction between service “users” and non-users is particularly likely to represent a source of selection bias that needs adjustment in some way. In the absence of information about respondents who did or did not have personal contact with the police, we omit the information in LAPOP on bribe requests from policemen.

5. Global Integrity publishes a yearly assessment of governance for selected countries in selected years, comprising sub-categories of six dimensions, including public administration. The dataset extends from 2004 to 2012, with (different) gaps for the selected countries. We use responses to the questions on public administration.

6. Alpha measures reliability (absence of random error) in a measurement index. It ranges from 0 to 1, where 1 is absence of random error; 0.7 is regarded as sufficiently reliable for practical use. An eigenvalue measures validity (high correlation between measured indicators and an unmeasured underlying common concept). In this study, there are six de facto indicators. If they were perfectly correlated with the single common factor they are assumed to measure, the eigenvalue would be 100. In this study, they are correlated with than half of that space, which is regarded as acceptable for practical use.

7. The Merit Index of the Inter-American Development Bank (IADB), developed by Zuvanic and Iacovello (2010), and the GI index that we use are highly correlated, with a Cronbach’s Alpha of 0.74.

8. National laws and practice regarding the hiring, firing, promotion, and pay of independent and professional civil servants are clearly intended to affect public officials at all levels of government, yet this expectation is more likely to be fulfilled in centralized than in decentralized administrative systems. The degree of administrative centralization varies among Latin American countries, and is often weak, making it incapable of reducing interference by regional politicians. Recent empirical evidence tends to find either that centralization has no clear systematic effect on governance in Latin American countries (Tanzi 1996; Faletti 2005; Treisman 2007), or that administratively centralized government appears to have more (perceived) corruption (Treisman 2000). Our measures of how laws operate in practice should capture these variations in the implementation of national laws at the local level.

9. Because the GI and LAPOP data cover different countries in each year, the lag varies for each country, and no country in our sample has continuous data for every year. Consequently, the number of observations in our statistical analysis varies by year, making the panels unbalanced and small.

10. Despite the high correlation between country averages of de jure and de facto quality scores, collinearity is unlikely to account for these null results. Of the six regressions reported in Table 5, while three show insignificant estimates, the other three report significant coefficients for the de jure and de facto quality scores, with opposite signs.

11. Civil service quality and practice are both rated on ordinal scales from 0 to 100 by the same raters. Thus we assume that the two variables are measured on the same scale, making it possible to compare the magnitude of the two regression parameter estimates within the same model.

Additional information

Notes on contributors

Laura Langbein

Laura Langbein ([email protected]), is a professor of policy analysis in the Deptartment of Public Administration and Policy at the American University. Her research includes bureaucratic discretion, pay-for-performance, intrinsic motivation, and corruption, with applications in environment and education policy.

Pablo Sanabria

Pablo Sanabria ([email protected]) is associate professor and chair of the Department of Organizational Management at Pontificia Universidad Javeriana in Cali, Colombia. He was also associate researcher at the Alberto Lleras Camargo School of Government of Universidad de los Andes, Colombia while developing this article. His research focuses on public management, human talent in the public sector, corruption and transparency, and local government.