2009年-世界发展银行全球_Neighborhood_Growth_Effects___An_Annual_Panel_Data_Approach_39页_424kb

Summary of "Neighbourhood Growth Effects: An Annual Panel Data Approach"

Core Content

This paper investigates neighbourhood growth effects using an annual panel data approach. It extends previous studies by distinguishing between neighbourhood, regional, and global effects, allowing for more nuanced analysis of how economic performance in one country can influence its neighbors and the broader world.

The study uses a static model with spatial effects, where the impact of a country's economic conditions is modeled through spatial lags. This model is based on a weighted distance matrix, with weights reflecting proximity and shared borders, and it includes country-specific fixed effects and time trends to control for unobserved heterogeneity.

Main Views and Key Information

1. Neighbourhood Effects Exist

• There are small but significant net neighbourhood effects over and above regional and global trends.
• Neighbourhood effects are defined as the influence of nearby countries, while regional effects refer to the influence of the broader region, and global effects to the rest of the world.
• Neighbourhood effects are more pronounced in Asia and the Americas, while Europe and Africa show small or no neighbourhood effects, but large regional effects.

2. Asymmetry in Effects

• Falls in per capita GDP in the neighbourhood are more correlated than rises, suggesting asymmetries in how economic shocks are transmitted.
• This asymmetry may reflect the greater influence of negative shocks in close proximity.

3. Channels of Spillovers

• The paper identifies several potential channels for spillovers:
  • Bilateral trade and investment
  • Technology diffusion
  • Remittance income
  • Migration and knowledge flows
• Trade and investment are highlighted as important mechanisms, with bilateral exports and domestic investment contributing to spillover effects.

4. Model Specification

• The model is specified as:

$$
\ln y_{it} = \gamma \mathbf{X}{it} + f_i + g_i * t + \rho_N \ln \Omega{it}^N + \rho_R \ln \Omega_{it}^R + \rho_G \ln \Omega_{it}^G
$$

• This allows for the separate estimation of neighbourhood, regional, and global effects.
• A revised specification (equation 3b) further decomposes these effects, distinguishing between net and gross effects.

5. Empirical Strategy

• The paper uses annual panel data covering 1980–2005 (full sample) and 1980–2004 (restricted sample), with up to 134 countries.
• It includes trade openness, gross fixed capital formation, and secondary school enrolment as control variables.
• Missing data is addressed by omitting extreme percentiles of observations and using year dummies to account for world shocks.

6. Robustness and Limitations

• The results are robust to different neighbourhood definitions and samples.
• The paper acknowledges the difficulty in distinguishing between genuine spillovers and common shocks.
• It also notes that feedback effects and cross-sectional dependence may affect the reliability of estimates, but these are mitigated through difference-based estimators and instrumental variables.

Key Findings

6.1 Cross-sectional Results

• A border coefficient of 0.29 indicates that a 1% faster growth in neighboring countries is associated with 0.29% faster growth in the focal country.
• This is higher than the 0.15 coefficient found in Moreno & Trehan (1997), suggesting increased integration or greater spillover effects in recent years.
• Control variables reduce the magnitude of the coefficients, indicating that trade and investment are important channels for spillovers.

6.2 World-wide Neighbourhood and Regional Effects

• Gross effects (the sum of all three coefficients) suggest a 0.577% increase in growth for a 1% rise in global per capita GDP.
• Net effects (the difference between global and regional effects) show that neighbourhood effects can be greater than regional effects, especially in Asia and the Americas.
• South Africa has a strong regional effect due to its size, but no additional effect beyond the region.

6.3 Regional Disparities

• Asia and the Americas show strong neighbourhood effects.
• Europe and Africa show small or no neighbourhood effects, but significant regional effects.
• For Africa, resource-rich countries (especially oil exporters) have a strong influence on regional growth.

6.4 Asymmetries

• Negative shocks in the neighbourhood are more correlated than positive shocks, indicating asymmetry in the transmission of economic performance.
• This suggests that economic downturns in nearby countries have a greater impact on the focal country than economic upturns.

Summary and Discussion

The paper concludes that neighbourhood effects exist and are more significant than regional or global effects in certain regions. It emphasizes the importance of trade and investment in spillover transmission and highlights the asymmetry in how economic shocks are felt. While the results are robust, the paper cautions that distinguishing between spillovers and common shocks remains a challenge. The model's reliability depends on the removal of common correlations in the error terms, which is achieved through difference-based estimators and year dummies.

The implications of the findings suggest that economic policies in one country can have spillover effects on its neighbors, and that regional integration and trade linkages are key to understanding these effects.