2005年-世界发展银行全球_Poverty_Traps_and_Nonlinear_Income_Dynamics_with_Measurement_Error_and_Individual_Heterogeneity_32页_765kb

Summary of "Poverty Traps and Nonlinear Income Dynamics with Measurement Error and Individual Heterogeneity"

Core Content

This paper investigates the existence of poverty traps in Mexico using a dynamic pseudo-panel approach to estimate nonlinear income dynamics. It addresses the challenges of empirical detection of poverty traps, such as limited panel data, measurement error, and attrition, by employing pseudo-panel methods that allow for the analysis of longer-term trends and individual heterogeneity.

Main Viewpoints

• Poverty traps are theoretical constructs suggesting that current poverty can perpetuate future poverty due to structural constraints, such as borrowing limits or indivisible investments.
• The American Dream effect posits that individuals can escape poverty through hard work and thrift, implying that initial poverty does not necessarily trap individuals.
• Empirical challenges in detecting poverty traps include:
  • Lack of long panels
  • Measurement error
  • Attrition bias
• Pseudo-panels are introduced as a solution, tracking cohorts of individuals over repeated cross-sectional surveys to overcome these limitations.

Key Information

Data and Methodology

• Data Source: The study uses the Mexican Urban Labor Force Survey (ENEU), a quarterly rotating panel from 1987 to 2001, covering 39 cities.
• Sample Restrictions: The analysis focuses on households with heads aged 25–49, as these are more likely to rely on labor income. Only 2% of the sample has zero labor income.
• Measurement Error: The paper acknowledges that measurement error can lead to misclassification of households as poor or non-poor, affecting estimates.
• Estimation Approach: The authors estimate nonlinear income dynamics using a cubic function of lagged income, allowing for individual and cohort-level heterogeneity.

Model Specification

• The true income dynamics model is given by:

  $$
  Y_{i,t}^* = \beta_1 Y_{i,t-1}^* + \beta_2 (Y_{i,t-1}^)^2 + \beta_3 (Y_{i,t-1}^)^3 + \alpha_i + u_{i,t}
  $$

• Observed income is:

  $$
  Y_{i,t} = Y_{i,t}^* + \varepsilon_{i,t}
  $$

• The paper derives a correction factor for the intercept term in the cubic model due to measurement error, assuming symmetric and stationary measurement errors.

Estimation with Pseudo-panels

• Pseudo-panel methods allow for the estimation of nonlinear income dynamics and provide consistent estimates of slope parameters.

• The paper introduces a term $\lambda_{c(t),t}$ to account for differences in the means of lagged income terms between cohorts, which converges to zero as the number of individuals increases.

• The authors estimate the intercept $\alpha_c$ by adjusting for the impact of measurement error on the intercept term, using the formula:

  $$
  \widetilde{\alpha}_c = \widehat{\alpha}c + \widehat{\beta}2 \widehat{\sigma}\varepsilon^2 + 3 \widehat{\beta}3 \bar{Y}{c(t-1),t-1} \widehat{\sigma}\varepsilon^2
  $$

Results

• Short panel analysis of labor income does not find evidence of poverty traps, suggesting high income mobility.
• Pseudo-panel analysis reveals a greater influence of past income on current income, but the nonlinear terms are small in magnitude, and the income mapping is close to linear.
• Even after accounting for measurement error and individual heterogeneity, no evidence of poverty traps is found for any group in the sample.
• The results hold when considering alternative measures of household resources (e.g., full income and expenditure) and when allowing for slope parameter heterogeneity across education groups.

Conclusion

The study concludes that while nonlinear income dynamics exist, there is no empirical evidence supporting the existence of poverty traps in Mexico. The use of pseudo-panels significantly reduces bias from measurement error and attrition, making them a more reliable tool for analyzing income dynamics and poverty traps compared to traditional short panels.

Methodological Contributions

• Introduces a correction factor for the intercept in nonlinear models due to measurement error.
• Allows for individual and cohort-level heterogeneity in the estimation of income dynamics.
• Demonstrates that pseudo-panels can provide more accurate estimates of income persistence and non-convexities than genuine panels, especially in the presence of measurement error.