2009年-世界发展银行全球_When_Do_Sudden_Stops_Really_Hurt__20页_327kb

Summary of "When Do Sudden Stops Really Hurt?"

Core Content

This paper investigates the drivers and consequences of sudden stops of capital flows, with a particular emphasis on the role of external vulnerability in determining the depth and length of such crises. The study analyzes 43 countries (26 emerging, 17 industrialized) over the period 1993–2006, providing a broader country and time coverage than previous literature. It introduces methodological improvements, including the use of cluster-robust standard errors and formal testing of exogeneity, to enhance the reliability of findings.

Main Findings

• External Vulnerability and Sudden Stops: External vulnerability significantly increases the probability of sudden stop crises. Countries with higher external vulnerability experience more prolonged recovery periods after a sudden stop, even though the immediate output loss is not necessarily more severe.
• Cumulative Output Effect: While external vulnerability does not affect the size of the instantaneous output effect, it reduces the speed of adjustment of GDP growth to its long-term trend, leading to a cumulative output effect over time.
• Liability Dollarization: Countries with high liability dollarization (original sin) are more prone to sudden stops due to financial fragility and exposure to currency and maturity mismatches.
• FDI Inflows: FDI inflows are associated with a lower probability of sudden stops, suggesting they provide more stability compared to other forms of capital flows.
• Terms of Trade Shocks: Negative terms of trade shocks reduce GDP growth, and the effect is mean-reverting, meaning growth tends to return to its long-term trend over time.

Key Variables and Definitions

• Sudden Stop Variable: Defined as a dummy variable capturing large and unexpected declines in capital flows, identified using monthly data and based on the methodology from Calvo et al. (2004).
• External Vulnerability: Measured as the ratio of the current account deficit to domestic absorption of tradable goods. Higher external vulnerability implies greater dependence on foreign financing.
• Liability Dollarization: Proxied by the original sin index, which measures the extent to which a country borrows in foreign currency. It is also estimated using BIS data for domestic liability dollarization.
• FDI Inflows: Represented by the ratio of net FDI inflows to GDP, with a negative association with sudden stops.
• Terms of Trade Shocks: Measured as changes in the terms of trade, with a positive coefficient indicating that a worsening of terms of trade reduces GDP growth.

Methodology

• The authors use a treatment effects model that jointly estimates the outcome equation (describing GDP growth dynamics) and the treatment equation (describing the likelihood of sudden stops).
• The outcome equation includes the error-correction term, interaction terms between growth deviation and external vulnerability, and a dummy for sudden stops.
• The treatment equation incorporates variables such as liability dollarization, FDI, and external vulnerability to predict the probability of sudden stops.
• The model accounts for time-series autocorrelation and unobserved country-specific effects using cluster-robust standard errors.
• The authors perform robustness checks using both random effects and fixed effects models, and find that results are consistent across different specifications.

Results Overview

• Joint Estimation:
  • A sudden stop leads to a 5.3 percentage point drop in per capita GDP growth.
  • The speed of adjustment is captured by the coefficient on the error-correction term, indicating how quickly growth returns to its long-term trend.
  • The interaction term between sudden stop and external vulnerability (λ₁) is statistically insignificant, suggesting no immediate impact on output, but a cumulative effect over time.
  • Terms of trade shocks have a negative and significant effect on GDP growth, with a 1 percentage point decline in terms of trade reducing GDP growth by 0.12 percentage points.
• Robustness Checks:
  • Results from separate equation estimations (random and fixed effects) are similar to those from the joint estimation, reinforcing the validity of the findings.
  • The exogeneity of sudden stops is supported by a high p-value from the Wald test, indicating no significant correlation between the error terms in the outcome and treatment equations.

Conclusion

The paper contributes to the understanding of sudden stop crises by highlighting the long-term adjustment dynamics and the role of external vulnerability. It suggests that while countries with high external vulnerability may not experience more severe output losses during a sudden stop, they are more likely to face protracted recovery periods. This insight is crucial for policy design, especially in the context of fiscal stimulus and crisis management. The study also underscores the importance of diversifying capital inflows and reducing external vulnerability to mitigate the effects of sudden stops.