Dynamic response of bank-intermediated trade finance to a global liquidity shock

ABSTRACT

This paper investigates the bank-level responses of a bank-intermediated instrument of trade finance to a negative global liquidity shock in Korea. Using a factor-augmented vector autoregression approach, the results show that there exists significant heterogeneity in bank-level trade-finance responses to a global liquidity shock. In addition, we find that the source of the heterogeneity is the bank-level foreign currency liquidity condition; banks with a better foreign currency liquidity condition may dampen the negative impact of a global liquidity shock on trade-finance.

KEYWORDS:

• Trade finance
• favar
• documentary bill
• liquidity shock

JEL CLASSIFICATION:

• F10
• F40
• G10
• G21

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ See Chamberlain and Rothschild (1983), Bai and Ng (2003), and Stock and Watson (1998, 2002, 2016) regarding an approximate factor model.

² This approach is applied in Boivin, Giannoni, and Mihov (2009), Buch, Eickmeier, and Prieto (2014), and Metiu (2016). We modify and use the Matlab codes provided by Boivin, Giannoni, and Mihov (2009) and Cesa-Bianchi at sites.google.com/site/ambropo/.

³ Based on the criterion applied by Boivin, Giannoni, and Mihov (2009) and Buch, Eickmeier, and Prieto (2014), the number of the latent factor K is determined to be 5.

⁴ See Boivin and Giannoni (2007), Boivin, Giannoni, and Mihov (2009), and Buch, Eickmeier, and Prieto (2014) for more details regarding the implementation of the iterative procedure. We apply a convergence criterion in the iteration procedure; no significant change in the sum of squared residual in the repeated regression estimation in iteration, which is implemented by Buch, Eickmeier, and Prieto (2014). A sum of squared difference of less than or equal to 0.000001 in the iteration procedure is regarded as no significant change.

⁵ Bijapur (2010) also maintains that the TED spread can serve as a proxy for identifying shifts in the supply of credit. Similar identification schemes in trade-financing context were used in Hwang and Im (2013, 2017).

⁶ Dedola and Lippi (2005), Buch, Eickmeier, and Prieto (2014), and Metiu (2016) also use a similar approach.

⁷ We thank an anonymous referee for suggesting the idea of controlling a crisis effect.

⁸ The other popular bank-intermediated trade-finance instrument is foreign trade loans which is extended to exporters before shipments. However, the data of the foreign trade loan is not available at bank level. See Hwang and Im (2013, 2017) for further details.

⁹ Documentary collection also involves documentary bills. See Hwang and Im (2013, 2017) for further details regarding the typical trade-finance instruments in Korea.

¹⁰ These include following banks: Six nationwide commercial banks; Citibank Korea (CT), KEB Hana Bank (HN), KB Kookmin Bank (KB), Standard Chartered Korea (SC), Shinhan Bank (SH), Woori Bank (WR). Five local commercial banks; Busan Bank (BS), Daegu Bank (DG), Jeonbuk Bank (JB), Kwangju Bank (KJ), Kyongnam Bank (KN). Four specialized banks; Industrial Bank of Korea (IBK), Korea Development Bank (KDB), Nonghyup Bank (NH), Suhyup Bank (SU). KDB and IBK are the government-owned banks. NH and SU are the cooperative banks. Government-owned Korea Eximbank and a local commercial bank, Jeju bank are excluded for data availability reason.

¹¹ In the cross-sectional regression estimation, the bank-level characteristics as explanatory variables are averaged over the sample period (2000–2016).

¹² The bank-type dummy variable takes 1 for general commercial banks (nationwide and local banks), and 0 otherwise (specialized banks).

¹³ We use the TED spread as a measure of the global liquidity constraint following previous literature. (see Bijapur 2010; Dungey, Dwyer, and Flavin 2013; Mancini, Ranaldo, and Wrampelmeyer 2013; Hwang and Im 2013, 2017; Vermeulen et al. 2015).

¹⁴ Recently, Fischer, Hossfeld, and Radeck (2018) show that the world trade data such as CPB world trade volume is a better demand measure for exports than world GDP or real imports-based measures.

¹⁵ Since the data of documentary bills purchased for a few banks (NH, IBK, SU, KDB) are available starting from 1999q2, and log-differenced, we use the data from 2000q2 to 2016q1 in the actual estimation.

¹⁶ The impulse response functions in this section are generated in the following FAVAR(1) model; no constant and deterministic time trend, five unobserved common factors, Cholesky decomposition as an identification scheme. The impulse responses are qualitatively similar for models with more than or equal to five latent factors but the results change significantly when we decrease the number of factors below five. For this reason and sparse parameterization purpose, we use five unobserved factors.

¹⁷ This exercise corresponds to the cross-sectional regression estimation between the bank-level impulse responses and the time-average bank-level characteristics. Therefore, the cross-sectional sample size is fifteen which is the number of a bank.

¹⁸ Again, note that this argument is somewhat inconclusive since the statistical significance becomes weak when we control the crisis effect in extraction of the impulse response functions.

Additional information

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea [NRF-2016S1A5A8018939];