2011年-世界发展银行全球_School_Inputs_Household_Substitution_and_Test_Scores_44页_1mb

Summary of "School Inputs, Household Substitution, and Test Scores"

Core Content

This paper investigates the relationship between school inputs and student test scores, with a focus on how households respond to changes in school funding. It presents a dynamic household optimization model that accounts for the substitutability between school and household educational inputs. The model is tested in two distinct low-income country settings: Zambia and India, using both cross-sectional and experimental data.

Main Findings

• Household Responses to School Funding: Households tend to offset anticipated school grants more than unanticipated grants. This suggests that when households expect school funding, they reduce their own educational spending in anticipation.
• Impact on Test Scores: Unanticipated school grants lead to significant improvements in student test scores, whereas anticipated grants have no effect. This indicates that the effect of school funding on learning outcomes depends on whether the funding is anticipated or not.
• Robustness Across Settings: The findings are consistent across Zambia and India, despite the different contexts and implementation agencies (government in Zambia, a non-profit in India). This supports the generalizability of the results.
• No Heterogeneity in Responses: There is no significant difference in household responses between asset-poor and asset-rich households, suggesting that school grants for learning materials are perceived as income transfers rather than targeted investments.
• Implications for Education Policy: The results imply that the long-term impact of school grants on learning is unlikely to exceed the income elasticity of test scores. This has important implications for the effectiveness of education programs in developing countries.

Key Predictions of the Model

• Anticipated vs. Unanticipated Inputs: If school and household inputs are substitutes, anticipated increases in school inputs will reduce the growth of test scores, while unanticipated increases will lead to greater improvements.
• User-Cost Framework: The model uses a user-cost framework to derive the optimal path of educational outcomes. The user cost reflects the cost of increasing test scores in one period, adjusted for future costs and benefits.
• Euler Equation: The model predicts that the marginal effect of anticipated funds is lower than that of unanticipated funds due to household re-optimization.

Methodology

• Zambia Case:
  • The study uses cross-sectional data from 172 schools in 4 provinces, covering 58% of the population.
  • Schools received fixed rule-based grants, which were predictable and anticipated by households.
  • Discretionary grants were unpredictable and unanticipated.
  • The paper uses variation in per-student grant amounts, based on school enrollment, to examine the crowding-out effect of household expenditures.
• India Case:
  • The study is based on a randomized experiment in the Indian state of Andhra Pradesh.
  • 200 government-run schools were randomly assigned to receive a school grant of about $3 per pupil.
  • The first year of the grant was unanticipated, while the second year was anticipated.
  • The results show that household spending declined in the second year, and test scores improved only in the first year.

Theoretical Model

• The model assumes that households maximize an inter-temporal utility function subject to budget and production function constraints.
• The production function for test scores is defined as $TS_t = F(TS_{t-1}, w_t, z_t, \mu, \eta)$, where $TS_t$ is test score at time $t$, $w_t$ is school input, $z_t$ is household input, $\mu$ and $\eta$ are child and school characteristics, respectively.
• The model incorporates the concept of user cost and derives the Euler equation to characterize the optimal growth path of test scores.

Policy Implications

• The findings suggest that naive estimates of the impact of school spending on learning outcomes may be biased if they do not account for household responses.
• The model highlights the importance of distinguishing between the policy effect and the production function parameters when estimating the impact of education programs.
• The results have direct implications for the design and evaluation of education programs in developing countries, emphasizing the need to consider household behavior when assessing the effectiveness of school inputs.

Conclusion

• The paper demonstrates that the impact of school inputs on test scores is influenced by whether the inputs are anticipated or unanticipated.
• The model and empirical results together suggest that household substitution effects significantly reduce the effectiveness of school grants, especially when they are anticipated.
• The study contributes to the broader understanding of how education policies interact with household behavior and the importance of considering these interactions in impact evaluations.