x

We are happy to announce that all trial registrations will now be issued DOIs (digital object identifiers). For more information, see here.
The Political Economy of Public Employee Absence: Experimental Evidence from Pakistan
Last registered on July 01, 2016

Pre-Trial

Trial Information
General Information
Title
The Political Economy of Public Employee Absence: Experimental Evidence from Pakistan
RCT ID
AEARCTR-0001363
Initial registration date
July 01, 2016
Last updated
July 01, 2016 9:08 AM EDT
Location(s)
Region
Primary Investigator
Affiliation
UC San Diego
Other Primary Investigator(s)
PI Affiliation
New York University
PI Affiliation
Lahore University of Management Sciences
PI Affiliation
International Growth Centre - Pakistan
Additional Trial Information
Status
Completed
Start date
2011-08-01
End date
2012-10-31
Secondary IDs
Abstract
In many developing countries, public sector absence is both common and resistant to reform. One explanation for this is that politicians provide public jobs with limited work requirements as patronage. We test this patronage hypothesis in Pakistan using: (i) a randomized controlled evaluation of a novel smartphone absence monitoring technology; (ii) data on election outcomes in the 240 constituencies where the experiment took place; (iii) attendance recorded during unannounced visits and; (iv) surveys of connections between local politicians and health staff. Four results support this view. First, while doctors are present at 42 percent of clinics in competitive constituencies, they are present at only 13 percent of clinics in uncompetitive constituencies. Second, doctors who know their local parliamentarian personally are present at an average of 0.727 of three unannounced visits, while doctors without this connection are present at 1.309 of the three visits. Third, around 40 percent of inspectors and health administrators report interference by politicians when they try to sanction doctors. Fourth, the effect of the smartphone monitoring technology, which almost doubled inspection rates, is highly localized to competitive constituencies. Last, we find evidence that program impact is in part due to the transmission of information to senior officers. We test this by manipulating the salience of staff absence in data presented to officials using an online dashboard. These effects are also largest in politically competitive constituencies. Our results have implications for the study of bureaucratic incentives in fragile states and are potentially actionable for policymakers trying to build state capacity.
External Link(s)
Registration Citation
Citation
Callen, Michael et al. 2016. "The Political Economy of Public Employee Absence: Experimental Evidence from Pakistan." AEA RCT Registry. July 01. https://doi.org/10.1257/rct.1363-1.0.
Former Citation
Callen, Michael et al. 2016. "The Political Economy of Public Employee Absence: Experimental Evidence from Pakistan." AEA RCT Registry. July 01. https://www.socialscienceregistry.org/trials/1363/history/9204.
Experimental Details
Interventions
Intervention(s)
We evaluated the effectiveness of a smartphone absence monitoring technology in reducing doctor absenteeism at clinics in Punjab, Pakistan. In partnership with the Government of Punjab, we conducted a randomized evaluation of a smartphone attendance monitoring application to determine a) whether technology-based monitoring can improve health worker attendance and b) whether the impact of monitoring depends on the degree of political competition in a given area.

In 2011, the Government of Punjab introduced the "Monitoring the Monitors" program, which replaced the paper-based monitoring system with an android-based smartphone application that could transmit data on facility utilization, resource availability, and worker absence to a central database in real time. The application ensured accuracy by requiring all data to be geotagged, time-stamped, and accompanied by photographs of the health staff and inspector. Clinics where three or more health staff were absent were highlighted in red on an online dashboard, and we measured whether senior health officials responded to this information.

In order to measure the relationship between doctor attendance and political patronage, we constructed an index of political competition using data from the 2008 Punjab Provincial Assembly elections. The index classified constituencies as having low, medium, or high degrees of political competition depending on the vote share for each political party. From a representative sample of 850 health clinics in 35 districts of Punjab, we randomly selected 427 clinics in 18 districts to receive the smartphone monitoring program. The rest of the clinics served as the comparison group and continued to use the paper-based monitoring system. We ensured that treatment and comparison districts had a comparable number of constituencies classified as having high, medium, or low levels of political competition. We collected clinic-level data through three unannounced visits in November 2011, June 2012, and October 2012.
Intervention Start Date
2011-08-01
Intervention End Date
2012-10-31
Primary Outcomes
Primary Outcomes (end points)
- Clinic inspection rates
- Health worker attendance
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
Our project attempts to explore the use of audits by government monitors as a solution to the problem of absence. We implement a smartphone-based solution that allows health system inspectors to upload the results of their assigned visit to a basic health facility to an aggregating website (dashboard), which instantly updates reports at different levels of aggregation (zonal and provincial) with the information captured by this most recent visit.
The “Monitoring the Monitors" program replaced the traditional paper-based monitoring system, which collects data on facility utilization, resource availability, and worker absence, with an android-based smartphone application. Data are transmitted to a central database using a General Packet Radio Service (GPRS) in real time. Data are then aggregated and summary statistics, charts, and graphs are presented in a format designed in collaboration with senior health officials. That data are: (i) aggregated in the province in real time; (ii) geo-tagged, time-stamped, and complemented with facility staff photos to check for reliability; and (iii) available in real time to district and provincial officers through an online dashboard. Application development started in August 2011. After developing the application and linking it to a beta version of the online dashboard, the system was piloted in the district of Khanewal. We remove Khanewal district from the experimental sample. Health administration staff were provided with smartphones and trained to use the application. The main purpose of the pilot was to ensure that the technology was working and to refine the application and the dashboard. During the pilot, several inspectors requested that the program require pictures of all staff in attendance, not just the inspector because they thought it might reduce pressure from health staff to falsify attendance.

Our experimental sample comprised all health facilities in the district of Punjab, which has a population of 100 million. Tens of millions of public sector health users therefore stood to benefit from the program. While we have administrative data for all facilities, we monitor a subsample of 850 clinics, drawn to be representative of facilities in the province, using independent inspections. We randomly implemented the program in 18 of the 35 districts in our experimental sample. In assigning treatment we stratified on baseline attendance and the number of clinics in a district to ensure a roughly even number of treatments and controls. We randomized at the district level. The intervention channels information about inspections to district health officials; randomization at a finer level is therefore very likely to generate externalities. The Department of Health also determined that sub-district randomization was not administratively feasible. Cluster randomization also allays some concerns about externalities generated by interactions between inspectors in the same district. All inspectors in a district are required to attend monthly meetings. While they typically have frequent interactions within districts, these relations are much weaker across districts.
Experimental Design Details
Randomization Method
Computer.
Randomization Unit
District
Was the treatment clustered?
Yes
Experiment Characteristics
Sample size: planned number of clusters
850 health clinics in 35 districts of Punjab, Pakistan
Sample size: planned number of observations
850 Basic Health Units (clinics).
Sample size (or number of clusters) by treatment arms
treatment group (smartphone monitoring program): 427 clinics (in 18 districts)
control group (paper-based monitoring system): 423 clinics (in 17 districts)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
IRB Approval Date
IRB Approval Number
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
Yes
Intervention Completion Date
October 31, 2012, 12:00 AM +00:00
Is data collection complete?
Yes
Data Collection Completion Date
October 31, 2012, 12:00 AM +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
846 health clinics
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
treatment group (smartphone monitoring program): 427 clinics (in 18 districts) control group (paper-based monitoring system): 419 clinics (in 17 districts)
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers
Abstract
In many developing countries, public sector absence is both common and resistant to reform. One explanation for this is that politicians provide public jobs with limited work requirements as patronage. We test this patronage hypothesis in Pakistan using: (i) a randomized controlled evaluation of a novel smartphone absence monitoring technology; (ii) data on election outcomes in the 240 constituencies where the experiment took place; (iii) attendance recorded during unannounced visits and; (iv) surveys of connections between local politicians and health staff. Four results support this view. First, while doctors are present at 42 percent of clinics in competitive constituencies, they are present at only 13 percent of clinics in uncompetitive constituencies. Second, doctors who know their local parliamentarian personally are present at an average of 0.727 of three unannounced visits, while doctors without this connection are present at 1.309 of the three visits. Third, around 40 percent of inspectors and health administrators report interference by politicians when they try to sanction doctors. Fourth, the effect of the smartphone monitoring technology, which almost doubled inspection rates, is highly localized to competitive constituencies. Last, we find evidence that program impact is in part due to the transmission of information to senior officers. We test this by manipulating the salience of staff absence in data presented to officials using an online dashboard. These effects are also largest in politically competitive constituencies. Our results have implications for the study of bureaucratic incentives in fragile states and are potentially actionable for policymakers trying to build state capacity.
Citation
Callen, Michael, Saad Gulzar, Ali Hasanain, and Yasir Khan. "The Political Economy of Public Employee Absence: Experimental Evidence from Pakistan." Working Paper, 16 February 2014.