The quality and quantity of bank intermediation and economic growth: evidence from Asia Pacific

ABSTRACT

We investigate the impact of the quantity and quality of bank intermediation on economic growth across 14 Asia-Pacific economies over 2003–2015. Measures of bank shareholder value efficiency as well as profit and cost efficiency are used as indicators of intermediation quality. We also employ measures of liquidity creation (fat and nonfat) as a proxy for the quantity of bank intermediation. Our main finding is that the quality of bank intermediation (enhanced credit allocation) is a driver of economic growth in developed Asia-Pacific economies, whereas it is the quantity of bank intermediation (capital accumulation) that positively influences growth in developing nations. From a policy perspective, our findings suggest that policymakers in developed nations should concentrate their efforts on reforms that enhance bank efficiency. Second, reforms that stimulate capital accumulation should be encouraged in developing economies because this is the main channel that spurs economic growth.

KEYWORDS:

• Bank efficiency
• liquidity creation
• regional growth
• banks in Asia Pacific

JEL CLASSIFICATION:

• G21
• O16
• O47
• O52

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ For instance, Levine (1998), Rajan and Zingales (1998), Beck and Levine (2002), Calderon and Liu (2003), Rioja and Valev (2004), Aghion, Howitt, and Mayer-Foulkes (2005), Rousseau and Wachtel (2011), Zhang, Wang, and Wang (2012), Pradhan et al. (2016), among others.

² They use three alternative measures of capital accumulation, including the ratio between loans disbursed in the region by banks and special credit institutions and the regional GDP, the share of bank loans granted to the private sector as a fraction of total loans, and the share of loans by cooperative banks on credit provided by all of the commercial banks in the region.

³ All but Berger, Hasan, and Klapper (2004) focus on Europe and even here, of the 49 nations studied only 7 are in the Asia-Pacific region.

⁴ We refer to ‘most’ economies because, given the IMF’s country classification, the Asia-Pacific region includes Australia, Japan, Singapore, and Hong Kong, which have advanced financial systems.

⁵ See IMF (2013) for information regarding the recent GDP growth of Asia-Pacific countries and other economies, and see Klapper, Martinez-Peria, and Zia (2014) for a detailed account of banking in developing Asia and a table that summarizes the features of individual countries’ banking systems.

⁶ With the exception of Japan.

⁷ Economic conditions may also influence bank efficiency and liquidity creation. For instance, during economic recessions, both the quality and quantity of credit demand and supply can decrease.

⁸ We do not include other country-level control variables such as the bank concentration ratio and economic freedom because they are already included in our efficiency estimation models.

⁹ For instance, one cannot control for the many differences in terms of regulatory structures, markets, and culture (Berger, Hasan, and Klapper 2004).

¹⁰ In detail, the first stage focuses on the prediction of the technical inefficiency effects by estimating the stochastic frontier production function, assuming that these inefficiency effects are identically distributed; the second stage involves the estimation of the determinants of the estimated technical inefficiency effects, which contradicts the assumption of identically distributed inefficiency effects in the first stage (Battese and Coelli 1995). So the one-stage approach is a preferred.

¹¹ In order to calculate the natural logarithm, we find the maximum losses among banks and then add the absolute value of these losses plus 1 to all banks.

¹² Following Heffernan and Fu (2010), EVA is normalized by factor inputs to minimize possible heteroscedasticity and scale effects in the model and to ensure its comparability with Tobin’s Q.

¹³ Following common practice, we calculate a 1-year period local CAPM.

¹⁴ As indicated in Grabowski (2009), cost of capital estimates derived from typical CAPM models may be biased downward during crisis periods, and such estimates may also be subject to ‘significant estimation and data input problems’ (32). For example, T-bond yields are a typical benchmark used in the CAPM model to estimate the cost of capital. However, these yields were temporarily very low for several months around the 2008–2009 crisis, boosting EVA estimates for this period. Therefore, we adjust the CAPM model by using the market risk premium (MRP) developed by Fernández, Aguirreamalloa, and Corres (2011). As these authors do not provide the MRP for Sri Lanka, we use the average MRP for India and Pakistan as a proxy for Sri Lanka’s MRP.

¹⁵ Models with NFLC as the quantity measure for bank intermediation are estimated as robustness checks, and the results are reported in Section 4.2.

¹⁶ Berger and Bouwman (2009) classify loans as liquid, semi-liquid, or illiquid based on category or maturity. We classify loans by category because Bankscope does not provide maturity information for loans that are issued by banks in the Asia-Pacific region. Moreover, according to Berger and Bouwman (2009), classification by category is better than classification by maturity because the ease, cost, and timeliness with which banks obtain liquid funds to satisfy their obligations are more important than the time to self-liquidation.

¹⁷ Berger and Bouwman (2009, 3794) note that the intuition for liquidity creation is that ‘banks create liquidity because they hold illiquid items in place of the nonbank public and give the public liquid items’.

¹⁸ Our final sample includes unbalanced panel data for 14 Asia-Pacific economies with 6474 observations for 822 banks, representing over 89% of all commercial bank assets in the Asia Pacific region.

¹⁹ These liquidity ratios are similar to those reported for banks in the United States (39%; Berger and Bouwman 2009), in Russia (27–30%; Fungacova, Weill, and Zhou 2010), and in the Czech Republic (15–33%; Horvath, Jakub, and Weill 2014).

²⁰ The correlation matrix is reported in Appendix 4.

²¹ 0.0425 + 0.0065 = 0.049; 0.0425/0.049 = 0.87.

²² 0.0425 × 0.1370 = 0.0058; 0.0065 x 0.2016 = 0.0013.

²³ The developed economies in Asia Pacific include Australia, Japan, Hong Kong, South Korea, Singapore, and Taiwan.

²⁴ Hasan, Koetter, and Wedow (2009) find that an increase in PE by one standard deviation yields 48 basis points of additional economic growth (they find that CE has no significant impact on growth).

²⁵ The developing economies in Asia Pacific include China, India, Indonesia, Malaysia, Pakistan, Philippines, Sri Lanka, and Thailand.

²⁶ Please refer to OECD (2015) for details.

²⁷ To address potential multicollinearity problems between the institutional variables and the quantity/quality indicators, we follow Klock, Mansi, and Maxwell (2005) and orthogonalize the potentially correlated variables to delineate the incremental effects of the institutional variables.

²⁸ The 10 factors include: property rights, freedom from corruption, fiscal freedom, government spending, business freedom, labour freedom, monetary freedom, trade freedom, investment freedom, and financial freedom. For the sake of brevity, we only report those with significant coefficients. The remaining results are available upon request.

²⁹ In this one-step model, following Lozano-Vivas and Pasiouras (2010, 2014), and Radić, Fiordelisi, and Girardone (2012), we also include some environmental variables to model the inefficiency distribution, including the real GDP growth rate, inflation rate, 3-bank asset concentration ratio, the minimum regulatory capital-to-assets ratio, and economic freedom which measures the degree of freedom from government interference afforded to businesses and individuals. A higher value indicates greater freedom. This approach allows us to account for heterogeneity across banks and still benchmark different banks against an identical frontier (Bos et al. 2008).

³⁰ To estimate CE using bank-level data, we divide the entire sample into three groups according to the IMF’s definition of similar regional blocs, namely, Industrialized Asia, Newly Industrialized Economies, and Developing Asia. Consequently, three dummy variables are employed in the CE model to control for different levels of economic development in the Asia-Pacific region. However, when we examine the effects of bank development on economic growth, we have to use country level rather than bank-level data. Given there are only 22 observations for Industrialized Asia, we follow the World Bank’s practice to combine Industrialized Asia with Newly Industrialized Economies to form ‘developed economies’.

³¹ See, among others, Fu and Heffernan (2007, 2009) and Fiordelisi and Molyneux (2010).

³² See, among others, Stiroh (2000).

Additional information

Funding

The authors are grateful for funding from the University of Macau (Grant reference no.: MYRG085) and Macau University of Science and Technology Faculty Research Grant (Grant reference no.: FRG-17-035-MSB). We appreciate the comments from participants of the 2016 BOFIT Research Seminar. All errors of course rest with the authors.