Can Measuring Performance Improve Local Government Quality in Nondemocracies? Evidence from Vietnam

Read the paper


Subnational Performance Assessments (SPAs) gather and publicize information on local government performance to facilitate monitoring, generate competition among officials, and improve quality. Their effectiveness is well-studied in democracies, but what happens when they are used in countries governed by single-party regimes?

To find out, researchers examined a naturally occurring experiment in Vietnam. There, in 2010, the United Nations Development Programme (UNDP) collaborated with the Vietnam Fatherland Front (an umbrella for organizations such as Vietnam’s Labor Confederation, Women’s Union, Peasant Union, Youth Union, and Communist Party, among others) to create the Vietnam Provincial Governance and Public Administration Index (PAPI). The goal of this survey is to measure local governance along six dimensions: (1) citizen participation at local levels, (2) transparency, (3) vertical accountability, (4) control of corruption, (5) public administrative procedures, and (6) public service delivery.

Due to budget constraints in 2010, the survey was administered in only 30 of Vietnam’s 63 provinces, giving researchers an opportunity to compare results between provinces where performance had and had not been measured. While performance did improve across all areas being measured, the authors found that built-in incentive structures in Vietnam’s centralized single-party regimes distorted the positive impact of this SPA.


The staggered rollout of the PAPI subnational assessment index in 2010-2011 is the only known case of a SPA being implemented as a randomized control experiment. In 2010, the PAPI team used matching procedures to identify statistical twins before randomly selecting one from each pair to create a representative sample of roughly half (30) of Vietnam’s provinces. In 2011, all provinces were surveyed, and the researchers used randomization inference1 to compare the outcomes of the control and treatment groups. While the researchers found that provinces that had been evaluated saw better scores in all areas (participation, transparency, accountability, corruption, public administrative procedures, and public service delivery) compared to those that had not been evaluated, the only area that saw any lasting gains was improved administrative procedures, and even that gain eventually faded.

To understand this failure of the PAPI Index to generate more persistent, rapid change, the researchers next looked at Government priorities. While democratic governments are accountable to their people (downward accountability), in authoritarian regimes, officials are incentivized to impress higher-ups—they want to be promoted (upward accountability). Analyzing Google search data, the researchers found that between 2008 and 2011, the Vietnamese government was intensely focused on improving administrative procedures. Then, between 2012 and 2020, the central government dropped its focus on administrative procedures, and the scores related to administrative procedures slowed to their typical, glacial rate of improvement. The temporary improvement in administrative procedures, and subsequent retreat, demonstrates the power of prioritization bias (making decisions based on the priorities of the central government) to affect change, a finding that illustrates the difficulty in using subnational performance assessments to stimulate competition and reform efforts in governance areas that are not emphasized by national leaders.


  1. SPAs are a good way to improve local governance in nondemocracies—if you also understand that the clear long-term priorities of the central government are what local governments improve most.
  2. While such priorities can appear to be narrow, if the central government is reform-oriented and interested in promoting change, SPAs can have a dramatic effect on the lives of citizens. In the specific case of PAPI, the citizens welcomed the reduction in administrative procedures, which saved them valuable time and reduced their chances of being asked for petty bribes by officials.

1Randomization inference is a statistical technique used to analyze experimental data by creating a distribution of possible outcomes. It involves randomly shuffling the treatment assignments multiple times to create a distribution of outcomes that could have occurred by chance. By comparing the actual outcome to the distribution of outcomes under random assignment, researchers can better assess the statistical significance of their results.



Edmund Malesky (Duke University), Tuan-Ngoc Phan (Duke University), Anh Quoc Le (Duke University)

Related Content

Related Links


Economic Governance, Governance, Citizen Participation, Corruption, Southeast Asia, Vietnam