Can governments sleep more soundly when holding international reserves? A banking and financial vulnerabilities perspective*

ABSTRACT

We use a sample of 40 developing and emerging countries over the period 1995–2015 to assess the effectiveness of international reserve holding as a crisis mitigator. We test the relevance of the reserve accumulation decreasing returns assumption by estimating the most recent version of the PSTR model. We find that increasing stocks of international reserves allows domestic authorities to mitigate the negative impacts of financial and banking vulnerabilities on GDP growth rates leading to reject the decreasing returns assumption. This evidence is robust to sensitivity checks.

KEYWORDS:

• Official reserves
• emerging and developing economies
• banking and financial vulnerabilities
• panel smooth transition regression model

JEL CLASSIFICATION:

• E52
• E58
• F30
• F41

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00036846.2023.2177597

Notes

¹ See appendix A for the list of selected countries.

² See appendix B for a description of the datA. See Supplementary material (Tables 1–2) for summary statistics.

Table 1. Homogeneity test (1995Q1-2015Q4): resm2 as transition variable.

Table 2. Nonlinear/Asymmetric panel unit root test with linearity tests in alternative hypothesis.

³ The worldwide influence of U.S. business cycles is the reason why we include the U.S. index. The correlation between the U.S. economic policy uncertainty index and the world index is 0.8. For details, see Baker, Bloom, and Davis (2016).

⁴ Between the two extreme cases, i.e. m > 1 and $\mathit{\gamma }\to \mathrm{\infty }$, the number of identical regimes remains 2 but the function switches between zero and 1 at $c_1,\dots ,c_{\mathit{m}.}$.

⁵ In bootstrapping the LM statistic, Gonzalez et al. (2017) made use of the warp-speed method proposed by Giacomini, Politis, and White (2013).

⁶ The LM statistic remains oversized.

⁷ Results for the other transition variables (resdebt and resgdp) are available upon request. We find also one transition i-e m = 1.

⁸ Available upon request.

⁹ For details, see Supplementary material A.

¹⁰ We also preformed Ucar and Omay’s (2009) test. Results are available upon request.

¹¹ The results for the remaining variables are available upon request.

¹² For Tables in the rest of the paper, (*): significant at the 10% level; (**): significant at the 5% level and (***): significant at the 1% level. Regime 2 denotes the coefficients of each variable above the threshold parameter.

¹³ See Supplementary material for tests with alternative threshold variables. Results of weak cross-sectional dependence test (Tables 5–6) reject the null hypothesis of no cross-sectional dependence. Results of homogeneity test and non-remaining heterogeneity test (Tables 7–8–9–10) reject the linearity. The number of thresholds is established to 1. Results of Ucar and Omay’s (2009) andEmirmahmutoglu and Omay’s (2014) are available in (Tables 1–12). Conclusions are similar for the resm2 variable.

Table 7. Full sample 1995Q1-2015Q4.

Table 8. Float sample 1995Q1-2015Q4.

Table 9. Non OECD sample 1995Q1-2015Q4.

Table 10. Peg sample 1995Q1-2015Q4.

Table 11. Full sample 1995Q1-2015q4_tot_kilian.

Table 12. Float sample 1995Q1-2015q4_tot_kilian.

¹⁴ Data are unavailable for Czech Republic, Estonia, Israel, and Sri Lanka.

¹⁵ For details, see Hatzius and Stehn (2018).

¹⁶ All results are available in the Supplementary material (Tables 13-16).

¹⁷ Table 4 in Supplementary material.

¹⁸ See Supplementary material for the subsamples, Tables 17–19.

¹⁹ See Supplementary material (Tables 20–23 for resgdp and 24–27 for resdebt).