Optimal size of currency swap between central banks: evidence from China

Abstract

Currency swaps are used by central banks to provide short-term liquidity to help enhance financial stability for both counterparts. We apply the newsboy inventory optimization to interpret the factors related to the optimal size of a currency swap, and we find that the mean value of foreign exchange demand, its volatility and the distribution form are important for the optimal swap size. Empirical studies regarding swap arrangements between China and its trading partners show that total trade volume and its long-term SD are robust explanatory variables.

Keywords:

• currency swap
• export and import
• foreign exchange
• optimal size

JEL Classification:

• E58
• F31
• F33

Acknowledgement

This research is supported by National Natural Science Foundation of China with grant no. 70831001.

Notes

¹ For instance, swap agreements totalling USD 250 bn have been signed among China, South Korea, Hong Kong, Singapore, Malaysia, Indonesia, Mongolia and other countries since the financial crisis in 2008.

² The logic here is to follow the ‘newsboy problem’, one of the most important models in the field of inventory and supply chain management; see, for example, Scarf (1958) and Gallego and Moon (1993).

³ We assume a normal distribution for payment demands to provide a baseline analysis. Other forms of distribution are beyond the scope of this article and should be considered for further study.

⁴ Foreign Direct Investment (FDI) size is relatively small compared with trade volume. For example, China's total goods trade is USD 2207 bn but the FDI flowing into China is USD 90 bn and China's overseas investment is USD 56.5 bn in 2009. In 2010, China's total trade is USD 2973 bn, FDI to China is USD 105 bn and China's overseas investment is USD 50 bn.

⁵ For example, for the swap agreements signed in 2009, we use corresponding trade data from 2008. For the swap agreements signed in 2010, we use corresponding trade data from 2009.

⁶ We also calculate the average value and SDs for 2, 4 and 6 years, respectively. No significant difference is detected from the regression results.