2016年-世界发展银行全球_Decentralization_of_Health_and_Education_in_Developing_Countries___A_Quality-Adjusted_Review_of_the_Empirical_Literature_43页_352kb

Summary of "Decentralization of Health and Education in Developing Countries: A Quality-Adjusted Review of the Empirical Literature"

Core Content

This review evaluates the empirical impact of decentralization on the delivery of health and education services in developing countries. It emphasizes the importance of using high-quality identification strategies to assess the causal effects of decentralization and identifies patterns in the evidence based on thematic and methodological criteria.

Main Substantive Themes

The review focuses on two primary themes:

1. Preference Matching: The extent to which local governments align public goods with citizens' preferences.
2. Technical Efficiency: The ability of decentralized governments to produce more or better public goods with the same inputs.

Key Findings

Preference Matching

• Education: Decentralization significantly improves preference matching. Local governments are more responsive to citizens' needs, as evidenced by increased investment in education in regions with high illiteracy.
• Health: There is less evidence for preference matching in health. While some studies suggest improvements, the results are not as robust or consistent as in education.

Technical Efficiency

• Health: Higher-quality evidence shows that decentralization improves technical efficiency, as seen in lower infant mortality rates (IMR) associated with increased local government expenditure.
• Education: Decentralization also enhances technical efficiency, particularly when it involves the transfer of authority, responsibility, and resources to lower tiers of government or schools.
• School-Based Management (SBM): SBM reforms, which devolve authority to the school level, show mixed results but generally support the idea that decentralization can improve efficiency.

Quality of Evidence

The review categorizes studies based on the credibility of their identification strategies:

Rank 1: Very Strongly Credible

• Characteristics: Valid comparison groups, minimal endogeneity concerns, and use of randomized control trials (RCTs).
• Examples:
  • Duflo et al. (2007) – Kenya
  • Glewwe and Maiga (2011) – Madagascar

Rank 2: Strongly Credible

• Characteristics: Reasonable comparison groups, efforts to address endogeneity, and use of quasi-experimental designs like difference-in-differences (DiD) and instrumental variables (IV).
• Examples:
  • Faguet (2004) – Bolivia
  • Faguet and Sánchez (2008) – Bolivia
  • Skoufias et al. (2011) – Indonesia

Rank 3: Somewhat Credible

• Characteristics: Limited success in constructing comparison groups, potential endogeneity biases, and use of panel data with random or between effects models.
• Examples:
  • Arze del Granado et al. (2005) – Cross-country
  • Akin et al. (2005) – Uganda
  • Freinkman and Plekhanov (2009) – Russia

Rank 4: Less Credible

• Characteristics: Self-selected populations, serious endogeneity challenges, and use of cross-sectional data without advanced methodologies.
• Examples:
  • Schwartz et al. (2002) – Philippines
  • Hasnain (2008) – Pakistan
  • Di Gropello (2002) – Philippines

Important Considerations

• Partial Reforms: Many studies examine partial decentralizations, which may not fully transfer authority, resources, and responsibility. These partial reforms often lead to ambiguous or inconclusive results.
• Equity Concerns: Decentralization can exacerbate inequality between richer and poorer subnational units unless accompanied by redistributive fiscal transfers.
• Methodological Improvements: Recent studies have improved identification strategies, moving away from traditional surveys and towards quasi-experimental and observational methods to approximate the gold standard of randomized experiments.

Conclusion and Policy Implications

• Firmer Conclusions: The review underscores the need for better research design and more rigorous identification strategies to draw definitive conclusions about decentralization's impact.
• Policy Focus: Decentralization can improve both preference matching and technical efficiency in health and education, but only when it is implemented comprehensively and with adequate safeguards.
• Future Research: The paper suggests a need for more in-depth examination of the mechanisms and prerequisites of successful decentralization reforms.

Summary Table

Methodology

• Scope: Focus on quantitative studies from the last 20 years on health and education decentralization in developing countries.
• Inclusion Criteria: Studies that evaluate the causal effects of decentralization, excluding qualitative work due to methodological and space constraints.
• Organizing Principle: Grouping studies by theme and quality, rather than by country or time, to identify clearer patterns and causal effects.

Limitations

• The review acknowledges that many studies are based on partial decentralizations and do not meet all criteria for effective devolution.
• It does not address the political motivations behind decentralization, which is a significant area for future research.

Final Remarks

While the literature on decentralization is vast and often contradictory, the review highlights that high-quality evidence supports the idea that decentralization can enhance both preference matching in education and technical efficiency in health and education. However, the effectiveness of decentralization depends on comprehensive implementation and adequate institutional safeguards to prevent inequality and inefficiency.