«TAX FARMING REDUX: EXPERIMENTAL EVIDENCE ON PERFORMANCE PAY FOR TAX COLLECTORS Adnan Q. Khan Asim I. Khwaja Benjamin A. Olken Working Paper 20627 ...»
NBER WORKING PAPER SERIES
TAX FARMING REDUX:
EXPERIMENTAL EVIDENCE ON PERFORMANCE PAY FOR TAX COLLECTORS
Adnan Q. Khan
Asim I. Khwaja
Benjamin A. Olken
Working Paper 20627
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue Cambridge, MA 02138 October 2014 This project is the results of collaboration among many people. We thank Jon Hill, Donghee Jo, Kunal Mangal, Wayne Sandholtz, Mahvish Shaukat, Gabriel Tourek, He Yang, and Gabriel Zucker for outstanding research assistance in Cambridge and Zahir Ali, Osman Haq, Turab Hassan, Zahra Mansoor, Obeid Rahman, Shahrukh Raja, Adeel Shafqat, and Sadaqat Shah for outstanding research assistance in Lahore.
We thank all the Secretaries, Director Generals, Directors, the two Project Directors from the Punjab Department of Excise and Taxation, the Punjab Finance, Planning and Development departments and the Chief Secretary and Chief Minister's offices for their support over the many years of this project.
Financial support for the evaluation came from 3ie, the IGC, and the NSF (under grant SES-1124134), and financial support for the incentive payments described here came from the Government of the Punjab, Pakistan. This RCT was registered in the American Economic Association Registry for randomized control trials under Trial number AEARCTR-0000252. The views expressed here are those of the authors and do not necessarily reflect those of the many individuals or organizations acknowledged here, nor those of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer- reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.
© 2014 by Adnan Q. Khan, Asim I. Khwaja, and Benjamin A. Olken. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.
Tax Farming Redux: Experimental Evidence on Performance Pay for Tax Collectors Adnan Q. Khan, Asim I. Khwaja, and Benjamin A. Olken NBER Working Paper No. 20627 October 2014 JEL No. D73,H26
ABSTRACTPerformance pay for tax collectors has the potential to raise revenues, but might come at a cost if taxpayers face undue pressure from collectors. We report the first large-scale field experiment on these issues, where we experimentally allocated 482 property tax units in Punjab, Pakistan into one of three performance-pay schemes or a control. After two years, incentivized units had 9.3 log points higher revenue than controls, which translates to a 46 percent higher growth rate. The scheme that rewarded purely on revenue did best, increasing revenue by 12.8 log points (62 percent higher growth rate), with little penalty for customer satisfaction and assessment accuracy compared to the two other schemes that explicitly also rewarded these dimensions. Further analysis reveals that these revenue gains accrue from a small number of properties becoming taxed at their true value, which is substantially more than they had been taxed at previously. The majority of properties in incentivized areas in fact pay no more taxes, but do report higher bribes. The results are consistent with a collusive setting in which performance pay increases collector's bargaining power over taxpayers, who either have to pay higher bribes to avoid being reassessed, or pay substantially higher taxes if collusion breaks down.
A randomized controlled trials registry entry is available at:
An Online appendix is available at:
http://goo.gl/1xifaf 1 Introduction Tax systems throughout the developing world collect substantially less revenue as a share of GDP than their counterparts in developed countries (Gordon and Li 2009, Kleven et al. 2014).
While there are many sources of diﬀerences, an important one is the substantial role played by the tax oﬃcials in assessing, enforcing, and auditing taxes. Combined with relatively low wages and limited performance rewards, the temptations for tax inspectors to collude with taxpayers to reduce tax receipts are great.
Historically, the state addressed this problem by rewarding tax collectors on the basis of tax collected. From the Roman Empire through the French Monarchy (Bartlett 1994, White 2004), regimes sold the rights to collect taxes to “tax farmers,” who then kept a fraction (or in some cases all) of the tax revenue they collected from a given area. US states similarly experimented with highly incentivized “tax ferrets” to collect property taxes in the 19th century. But the empowered tax oﬃcials in these regimes were unpopular, and over the past several hundred years the world has moved increasingly to salaried tax oﬃcials (Parrillo 2013).
In developed countries with well functioning civil service systems, reliable ﬁnancial and administrative data, and third party veriﬁcation methods (Gordon and Li 2009, Kleven et al. 2011, Kleven forthcoming), there has been relatively little appetite for revisiting this question of compensation. But in developing countries, where tax inspectors are poorly compensated, corruption is considered rampant, and there is limited and only partial administrative and third party data to draw on (e.g., Carillo et al. 2014), the tradeoﬀ between the expected beneﬁts and costs of empowering tax oﬃcials is not so obvious. Therefore in such contexts with limited tax capacity mechanisms that may no longer be as desirable in developed countries may still oﬀer a viable and eﬀective solution. Not surprisingly, countries such as Brazil, Peru, Pakistan, and others have begun to reconsider incentives for tax staﬀ (Das-Gupta and Mookherjee 1998, Kahn et al. 2001). Yet there is little rigorous evidence of the tradeoﬀs developing country governments face between increased revenue from incentives and the costs that may occur in the form of taxpayer dissatisfaction and over-taxation as a result of empowering tax collectors in this way.
In this paper, we provide what is to the best of our knowledge the ﬁrst experimental evidence on this question. Working with the Punjab, Pakistan provincial government, we randomly allocated tax oﬃcials in the entire provincial urban property tax department, which consists of 482 property tax units (known as circles), into one of three versions of performance-based pay schemes or a control group. A total of 218 circles, consisting of about 550 tax personnel, were randomly allocated to one of the three treatment groups, for two ﬁscal years. The incentives were large: the three-person tax team in each treated circle was collectively given an average of 30 percent of all tax revenues it collected above a historically-predicted benchmark.1 Many personnel in treated areas were able to 1 The team of three tax personnel in each circle (an “inspector”, “constable”, and “clerk”) together received either 20 percent, 30 percent, or 40 percent of all revenues collected above a benchmark. For equity reasons the percentage 1 double their baseline salaries or more through these incentives.
Given concerns about potential negative impacts of high-powered incentives on taxpayer satisfaction and assessment accuracy, the three schemes varied in both the extent to which they based performance pay explicitly on these non-revenue outcomes and the extent to which they allowed for subjective evaluation on the part of the tax department. The “Revenue” scheme provided incentives based solely on revenue collected above a benchmark predicted from historical data. To address multi-tasking concerns (Holmstrom and Milgrom 1991), the “Revenue Plus” scheme provided incentives exactly as in the Revenue scheme, but made adjustments (plus/minus three-fourths of baseline salary) based on whether the circle ranked in the top, middle, or bottom third of circles in terms of taxpayer satisfaction and accuracy of tax assessments, as determined by an independent survey of taxpayers. To allow for more subjective assessments rather than purely formulaic criteria (Baker et al. 1994, MacLeod 2003), the third scheme, “Flexible Bonus,” took this a step further by both rewarding collectors for a much wider set of pre-speciﬁed criteria set by the tax department, and by allowing for subjective adjustments based on period-end overall performance.
It is important to note that the impact of even the revenue only scheme on tax revenues and corruption is not, ex-ante, obvious. Consider a simple bargaining setting in which a tax collector colludes with a taxpayer to reduce the tax assessment in exchange for a bribe. If there is no cost to either party from reducing tax liability, then performance pay for tax collectors will simply raise the bribe paid with no impact on revenue, as the taxpayer now has to compensate the collector’s foregone incentive payment with a higher bribe. In more realistic settings, where there is some cost to either party from reduced tax liabilities, there will be two diﬀerent eﬀects: some taxpayers will continue in the collusive, low-tax equilibrium but have to pay higher bribes, while others will end up paying higher taxes and lower bribes as they switch from the collusive, low-tax equilibrium to a non-collusive, high-tax equilibrium. As we subsequently show, performance pay could thus have heterogeneous eﬀects on tax revenue and bribes among taxpayers.
We evaluate the impact of the schemes using multiple sources of data. For tax revenue outcomes, we obtained administrative data which we veriﬁed by conducting random spot-checks against the tax department’s bank records. For outcomes such as perceived corruption and satisfaction with the tax department we conducted a survey of over 16,000 taxpayers and their properties throughout the province. For estimating assessment accuracy, the surveyors also directly observed and recorded the property characteristics used in the tax calculation. We then manually matched surveyed properties to the tax rolls to obtain the corresponding tax records for each property. Since tax assessment is determined formulaically from these property characteristics this allowed us to determine the accuracy of assessments by comparing our survey measurements to those on the oﬃcial tax rolls.
We ﬁnd that, on average across the three schemes, by the end of the two years performance pay led to an increase in tax revenue of about 9.3 log points based on the administrative data.
was pre-speciﬁed to be lower in larger circles, and the payments were divided among inspectors, constables, and clerks in a ﬁxed proportion (40:30:30). Details are given in Section 4.1 below.
2 This translates to a 46 percent higher growth rate in revenues compared to control areas. We show that this came predominantly through an increase in the reported tax base (i.e. the total assessed value of properties) rather than through increased recovery or changes in exemptions granted. On average, we ﬁnd little impact of the schemes on taxpayer satisfaction. Speciﬁcally, the increased revenue generated as a result of the schemes is not accompanied by a decline in the typical taxpayers’ perceptions of the quality of service from the tax oﬃce or in their satisfaction with their dealings with the tax oﬃce. We also ﬁnd no overall change in the accuracy of tax assessments. Thus, on average, we ﬁnd that the incentives increase revenue with little obvious downside in terms of overall perception of the tax department in the eyes of the typical taxpayer.
Comparing the three schemes, we ﬁnd that they diﬀer substantially in terms of their impact on revenue, with relatively small diﬀerences on taxpayer satisfaction and perception of the tax department. Speciﬁcally, the Revenue scheme, which provided incentives purely based on revenue collected, showed 15.2 log points higher current-year revenues relative to controls (57 percent higher growth rate) by the second year. In comparison the Revenue Plus scheme achieved only 8.1 log points, and the Flexible Bonus scheme only a statistically insigniﬁcant 3.5 log point increase in current-year revenue. While the Revenue Plus scheme did improve perceived customer satisfaction and quality perceptions relative to to the Revenue and Flexible Bonus schemes, the diﬀerences were small, and the substantially lower revenue collected meant that this scheme had a substantially lower rate of return. The Flexible Bonus scheme did not do better on any dimension we can measure in our data, and in fact did worse compared to the control group on perception of the department’s quality. Thus, adding multiple dimensions to performance pay substantially diluted the impact on revenue without a substantial corresponding increase in non-monetary outcomes.
Our survey data allows us to dig deeper by examining how being under a performance pay regime impacts taxpayers interactions with the tax department. We ﬁnd heterogeneity in response among taxpayers that is consistent with there being collusion between tax inspectors and taxpayers.
For most properties in performance pay tax circles, taxpayers were not reassessed and reported no change in tax paid. However, relative to the control group, they reported a Rs. 600 (about US $6) increase in the going rate for a bribe paid to property tax oﬃcers for properties similar to theirs, which represents a roughly 33% increase. While this does not necessarily imply that every household paid these higher bribes, respondents also indicated that bribe payments were more frequent.