The Effect of Bureaucrat Performance Scorecards on Service Delivery and Bribes in Bangladesh

Researchers:
Martin Mattsson
Location:
Bangladesh
Sample:
1,034,688 applications made in 311 sub-district land offices, as well as interviews with 2,869 applicants
Timeline:
2018 - 2020
Initiative(s):
Governance Initiative
Target group:
  • Civil servants
Outcome of interest:
  • Social service delivery
  • Transparency and accountability
  • Corruption and Leakages
  • Productivity
  • Service provider performance
Intervention type:
  • Digital and mobile
  • Information
  • Administrative reform
AEA RCT registration number:
AEARCTR-0003232
Partners:

Slow government service delivery and corruption are common problems in many low- and middle-income countries. Given that paying bribes for faster service delivery is common, it has been suggested that improving the average delivery speed could reduce bribes. In Bangladesh, researchers conducted a randomized evaluation to measure the effect of performance scorecards—providing information about delays in public service delivery to the responsible bureaucrats and their supervisors—on the speed of service delivery and the payment of bribes. Results show that the intervention increased on-time service delivery, but did not decrease bribe payments on average; bribes among high-performers increased.

Policy issue

Slow government service delivery is a common problem in many low-and middle-income countries. A commonly stated reason for paying bribes for public services is to get those services completed in a timelier manner. For example, among households that paid a bribe for a public service in Bangladesh, 44.6 percent stated that "timely service" was one of the reasons for the bribe.1 Therefore, many existing theories of corruption and delays in public service delivery suggest that if delays are reduced, corruption may also decrease.

Better monitoring of bureaucrats’ performance might be a way to reduce delays in service delivery and thereby reduce corruption. To improve the delivery of their social programs, many governments in low-and middle-income countries have incorporated digital innovations. One benefit of these digital systems is that they generate administrative data about the timeliness of public service provision by individual government bureaucrats. This data makes it feasible and inexpensive to systematically monitor an individual bureaucrat’s service delivery. 

Can using the data from an e-governance system to track the speed of service delivery increase bureaucrats’ processing times? Furthermore, if processing times improve, does this reduce bribe payments for service delivery? 

Context of the evaluation

In Bangladesh, corruption is a widespread phenomenon in service delivery, with 70.8 percent of households experiencing some form of corruption when receiving services from public and private sector institutions in 2021.2  Among the households in this study that stated they had to pay a bribe while changing their official land records, a service provided by the government, most reported these bribes were to “to get the work done” or “to avoid hassle.” 

New owners of plots of land in Bangladesh must change the official land records in order to receive the official rights to their new plots. The government official in charge of this “mutation” process is the Assistant Commissioner Land (ACL), who heads a local land office. ACLs are supervised by Upazila Nirbahi Officers (UNOs), who conduct an annual performance review that form the basis for the ACLs’ career progression.

Land record changes are meant to be completed within a 45 working day time limit. However, applications are often delayed: 56 percent of applications in the comparison group were processed within this time period, with the average processing time lasting 64 working days. Additionally, ACLs and other land office staff are known for requiring bribes. Among surveyed applicants, 73 percent said that it is normal for a person like themselves to pay a bribe to get their application processed. The average bribe amount was BDT 6,718 (approximately USD 80).

Starting in February 2017, a new digital system for initiating and processing applications for land record changes was gradually implemented across all of Bangladesh. The system was designed to make the process simpler and quicker for both the applicant and the bureaucrat. The system also stores data on when applications were made and when they were processed. Although this data could be used to assess the ACLs’ performance in processing land record changes on time, it had not been used for this purpose before this study. 

A Bangladeshi bureaucrat uses the digital land record system
A Bangladeshi bureaucrat uses the new digital system for initiating and processing applications for land record changes.
Photo credit: Martin Mattsson, National University of Singapore

Details of the intervention

The researcher, in partnership with the Ministry of Land and the Access to Information Initiative (a2i) of the Government of Bangladesh, conducted a randomized evaluation to measure the impact of monthly scorecards containing information about ACLs’ performance on processing speeds and bribe payments.

The study took place in 311 land offices that had installed the e-governance system, representing 60 percent of all land offices in Bangladesh. 155 land offices were randomly selected to receive the scorecard intervention; a comparison group of 156 land offices received no intervention. Because the e-governance system was rolled out gradually, the study took place across two waves, which began in September 2018 and April 2019. Scorecards were sent monthly until March 2020, when the Covid-19 pandemic caused the program to end. 

In the land offices in the intervention group, ACLs and their supervisors received monthly scorecards displaying how many applications for land record changes were processed within the 45-working-day time limit, how many applications were overdue, and the ranking of each ACL relative to others in the experiment along these two measures. The researcher used administrative data generated by the e-governance system to obtain the information presented in the scorecards.

Using administrative data, the researcher tested if the performance scorecards improved the speed of service delivery by measuring the number of land record change applications processed in time and the average processing time, and combining these two outcomes to create a time index. In the offices in the first wave of the evaluation, the researcher also interviewed existing applicants for land record changes about whether they had paid bribes for their service, their expectations for processing time, their willingness to pay for faster services, and their overall satisfaction with the service. Additionally, on randomly selected days, applicants in both the control and treatment groups were provided information about improvements in processing times. 

Results and policy lessons

The intervention improved application processing time, with the effect driven by offices that were underperforming prior to the intervention. However, the performance scorecards lead to a moderate, though statistically insignificant, increase in bribe payments. This increase was driven by offices overperforming at the start of the experiment, for whom this increase was statistically significant.

Effect on Processing Times: The performance scorecards increased the number of applications processed within the government time limit and improved processing times overall. The scorecards increased the share of applications processed within the 45-working-day limit by 6 percentage points (11 percent) on a base of 56 percent in the comparison group. They also led to a 13 percent reduction in overall processing times, from an average of 64 days in the comparison group at the intervention’s start. There was no consistent pattern of an increase or decrease in the effect size throughout the 16 months of the intervention.

The effect on processing times was driven by offices that were underperforming relative to the median before the intervention. For these offices, receiving scorecards led to a 0.24 standard deviation (SD) increase on the time index compared to a 0.03 SD increase among offices overperforming before the intervention.

The researcher suggests this improvement in performance may be due to two main channels. First, the scorecard information may have improved the ability of supervisors to incentivize employees, potentially via better promotions and job postings for those with good scorecards. Second, bureaucrats may have changed their behavior when they received their own scorecards, as their performance on the job became more salient.

Effect on Bribe Payments: The performance scorecards led to an increase in bribe payments on average, although this effect is not statistically significant. This is also true for the randomly selected group of applicants who were informed of improvements in processing times. Among offices that overperformed at baseline, bribe payments increased by BDT 2,069 (US$24 or 38 percent). Meanwhile, there was no impact for underperforming offices. The researcher suggests that bureaucrats may refrain from taking more bribes in the first place because they are concerned about their reputation or feel a sense of pride/shame in taking fewer/more bribes. When the overperforming bureaucrats then receive positive feedback, this improves these bureaucrats’ reputation, and they become more willing to take bribes as their original reputational concerns are reduced. These results are inconsistent with the common prediction that faster average service delivery may reduce corruption, at least in the context of this intervention.

Tentative cost-benefit analysis and policy implications: It is hard to compare the overall costs and benefits of the intervention. The improvements in processing times are valued at about USD 9.7 million per year in the 155 treatment offices based on user-reported valuations. However, the increase in bribe payments is estimated to be approximately USD 6.6 million per year. This comparison, combined with the lack of impact on stated overall satisfaction among applicants, leave us with no strong evidence that the scorecards had either a positive or a negative effect on average applicant welfare.

More broadly, these results suggest that policies to reduce corruption should not rely on improving the speed of public service delivery. While bureaucrats are often recognized for outstanding performance, it may be more effective to provide feedback to their underperforming counterparts. Finally, providing positive feedback on one domain of bureaucrats’ performance may, by bolstering their reputations, create an incentive for them to underperform in another domain—pointing to the need to measure performance across multiple dimensions when evaluating such interventions.

The scorecards were discontinued at the outset of the Covid-19 pandemic in Bangladesh. Given the ambiguous impact on average applicant welfare, the research team did not pursue a scale-up of the scorecard system. However, the codes and systems used to generate the performance scorecards from the administrative data were transferred to a2i, building their capacity to put other similar systems in place in the future.

Mattsson, Martin. “Information Systems, Service Delivery, and Corruption: Evidence from the Bangladesh Civil Service.” Working paper, August 2022, Available at SSRN: https://ssrn.com/abstract=3986989

1.

Transparency International Bangladesh. “Corruption in Service Sectors: National Household Survey 2021”. August 31, 2022. https://ti-bangladesh.org/beta3/images/2022/nhs/2021/extended-executive-summary-english.pdf

2.
Transparency International Bangladesh. “Corruption in Service Sectors: National Household Survey 2021”. August 31, 2022. https://ti-bangladesh.org/beta3/images/2022/nhs/2021/extended-executive-summary-english.pdf