Many prior studies on Indian banking efficiency have typically regressed nonparametric estimates of production efficiency on environmental variables in a two-stage process. However, Simar and Wilson (2007, 2011) have demonstrated that the studies that use such conventional approaches are invalid due to complicated and unknown serial correlation among estimated efficiencies. Using the data envelopment analysis bootstrap procedure suggested by these authors, for the first time, we analyse the technical efficiency of Indian banks and regress the bootstrap scores on a set of environmental variables using a truncated regression. Banks that are on efficiency frontier as per conventional analysis are actually away from the frontier when bootstrap scores are used. Contrary to many prior studies, state ownership was found to have significant negative impact on efficiency.
Key words: Indian banks, efficiency, truncated regression, bootstrap
The objectives of this study are (a) to assess the production efficiency of Indian banks using the bootstrap approach to data envelopment analysis and (b) to examine the impact of loan quality and ownership on bias-corrected bootstrap efficiency scores. It explores these issues by addressing five related questions:
(i) What is the production efficiency of Indian banks using the unbiased bootstrap approach?
(ii) What is the effect of state ownership on bank inefficiency?
(iii) What is the effect of bank soundness on bank efficiency?
(iv) What is the effect of size on bank efficiency?
(v) What is the effect of loan quality on bank efficiency?
The paper also considers whether there was a change in bootstrap efficiency scores of Indian banks during the three periods: pre-GFC, during the GFC and post GFC.
The immediate motivation for the paper is the passage in December 2012 of the Banking Liberalisation Bill in the Indian Parliament that raises the foreign investment limits in Indian banks to 26 per cent from the present 10 per cent and liberalizes the licensing regime for banks (FT 2012). The liberalization is intended to improve the efficiency of the banking system, which is tipped to become the third largest in the world, next only to China and the United States, by 2025 (FT 2012). Further, the extant studies on Indian banking efficiency have used the nonparametric data envelopment analysis and the two-stage regression approach without bootstrapping the efficiency scores. The Reserve Bank of India (2008), for example, found that 17 out of 81 banks were on the efficiency frontier using the data envelopment approach. However, the efficiency scores were not bootstrapped. Simar and Wilson (2007) have demonstrated that these studies are invalid due to complicated and unknown serial correlation among estimated efficiencies.
Further, Dyson and Shale (2010), state that the true efficient frontier lies within the confidence limits that are produced by bootstrap procedures. This removes the main drawback that statistical inference can't be conducted with DEA efficiency scores (Halkos and Tzeremes, 2010).
Using the data envelopment analysis bootstrap procedure suggested by these authors, for the first time, we analyze the technical efficiency of Indian banks and regress the bootstrap scores on a set of environmental variables using a truncated regression. Second, while the banking sector in many countries of the developed world faced enormous problems of financial stress and sustainability, the Indian banking sector came out of the global financial crisis (GFC) relatively unscathed. Barr et al. (2000) found that banks with higher efficiency are more likely to survive than those with relatively low scores. Consequently, an examination of the efficiency of Indian banks post-GFC becomes important. Podpiera and Cihak (2005) stated that regular screening of banking efficiency is important, as it can serve as an early warning system. Third, The Economist (2012) stated that the Indian banking system runs the "risk of Spanish disease" and that "India has a bigger bad-debt problem than the rather stable level of banks' official 'non-performing" loans suggests." The magnitude of the impact of such non-performing loans (loan quality) on banking efficiency is also an issue that we examine in this paper. Fourth, few studies on Indian banking efficiency have examined the impact of ownership and credit risk (loan quality) together on production efficiency in second-stage regression. Where they have, it is either multiple regression or tobit regression that has been used on non-bootstrapped efficiency scores instead of truncated regression as suggested by Simar and Wilson (2007). Finally, the results would be of interest to researchers in emerging economies like China, Brazil, Russia and other developing countries where banks continue to be publicly owned. Fry (1995) states that a key stylised fact about developing countries is financial intermediation is mostly carried by commercial banks rather than by financial markets. Ataullah and Le (2006) emphasize that it is vital for governments in developing countries to create an environment that enhances commercial banking efficiency for overall economic growth.
Furthermore, as stated by Simar and Wilson (2007), the procedure ensures the efficient estimation of the second stage estimators, a property which is not guaranteed with alternative methods. The use of truncated regression enables us to obtain more reliable evidence (Barros and Garcia-del-Barrio, 2011).
The study proceeds as follows. Section 2 provides a background of the Indian banking system in brief, section 3 reviews prior studies, section 4 provides data and analysis and section 5 provides results. Conclusions of the study are presented in section 6.
OVERVIEW OF THE INDIAN BANKING SYSTEM
India has a massive banking system that caters to the financial needs of over 1 billion people. At the top of the banking system is the Reserve Bank of India, which is the central bank of the country. Commercial banks are the major type of financial intermediary and consist of 26 public sector, 22 private sector and 41 foreign banks (see Table 1). Besides the commercial banks, cooperative banks, which are also state-partnered institutions, mainly cater to the needs of the rural sector. As per RBI (RTPB 2012, Table IV.1), total assets of the Indian banking sector are Rs 82,994 billion with total deposits of Rs 64,537 billion and total advances of Rs 50,746 billion (INR55/USD). The return on assets of 1.08 (2012) is comparable with that of other countries of the world. The figure for the net non-performing assets as a percentage of net advances at 1.4 does, however, indicate an area of concern, given the report in The Economist cited above that a huge amount of restructured loans are not included in the ratio.
Studying banking efficiency is important. Fiordelisi et al. (2010) found that reduction in efficiency increases banks' future risks and indicated bad management. For measuring bank efficiency, the frontier analysis approach is increasingly being used. The approach consists of separating institutions that are performing poorly as compared to those that are performing well using a particular standard. The separation is achieved either by applying the non-parametric or parametric frontier method. The parametric approach includes stochastic frontier analysis, the free disposal hull, thick frontier and the distribution-free approaches (DFAs), while the non-parametric approach is data envelopment analysis (DEA) (Molyneux et al. 1996).
Though many empirical studies have examined banking efficiency over the years, few have used the bootstrap DEA procedures. Consequently, the results obtained through the use of conventional DEA would need to undergo renewed scrutiny. Matthews et al. (2009) examined the Malmquist productivity (not efficiency) and non-performing loans in Chinese banks using bootstrap procedures. Curi et al. (2013) examined foreign-bank bootstrap DEA efficiency in Luxemburg. Barros and Assaf (2011) and Halkos and Tzeremes (2013) examined bootstrap efficiency of Japanese and Greek banks respectively. Chortareas et al. (2013) used bootstrap DEA to examine banking efficiency in the European Union.
As can be seen from the above the bootstrap DEA efficiency studies in banking have largely been confined to developed countries. Examining banking efficiency using bootstrap is important in the context of developing countries since they generally have predominance of public sector banks which are typically known to be saddled with inefficiency (Fry, 1995; Ataullah and He, 2006). Governments in these countries need an efficient banking sector for promoting growth. The starting point of this is to accurately assess the banking efficiency in these countries. We address this gap in the literature using data of Indian banks.
Bhattacharya et al.'s paper (1997) was the first to apply frontier analysis (both DEA and stochastic frontier analysis) to assess the efficiency of 86 Indian banks during the early liberalization period (1986-1991). The study found publicly-owned banks to be most efficient, followed by foreign banks and privately-owned banks. Das (1997) used DEA to examine the efficiency of 65 Indian commercial banks for the year 1995 and compared their technical and allocative efficiency and found the former to be more efficient. Mukherjee et al. (2002) examined the technical efficiency of 68 commercial banks for the period 1996-1999. The findings were similar to those of the Bhattacharya et al. (1997) study that is, the publicly-owned banks were found to be more efficient than both private and foreign banks. Sathye (2003) examined the impact of ownership on Indian banking efficiency using DEA and found that publicly-owned banks were more efficient than foreign banks and privately-owned banks. Ram Mohan and Roy (2004) also found publicly-owned banks to be more efficient than privately-owned banks,...