Methods used in the Present Study

propriate method in the estimation of productivity, Ataullah et al. (2004) mentioned that parametric study, which uses econometric estimation is based on pre-specified functional form. Efficiency measure in parametric method is influenced by the pre- specified func- tional form. In the non-parametric method, for example DEA does not require to specify functional form apriori. However, DEA assumes the measurement error or statistical noise in the estimation as constant which according to Ataullah et al. (2004) is a major disad- vantage of DEA method. They also argued that compilation of input and output prices is difficult or such prices may be distorted because of long regime of regulation in banking industry. Other market imperfections may also distort price mechanism in banking indus- try in developing countries. According to Sanyal and Shankar (2011), one of the major drawbacks of DEA is that it does not correct for the endogeneity of inputs and ignores possible heterogeneity across banks.

Following the above discussion which has highlighted advantages and disadvantages of various methods, non-parametric DEA method has been regarded as more appropriate in measuring productivity in banking industry. Therefore, in the present study attempt has been made to estimate TFP growth of the banking industry using input oriented DEA based Malmquist Productivity Index (MPI). The selection of Malmquist Productivity In- dex is based on the argument that it does not require price data and allows for the use of multiple inputs and outputs in the estimation (Kumar and Gulati, 2014). An MPI higher than 1 implies growth in the productivity whereas an MPI below 1 indicates a decline in the total factor productivity. Statistical package ‘R’ is used in the estimation of MPI.

For the second part of our analysis i.e. to identify important determinants of productiv- ity growth of banking services in India, the present study carries out a regression analysis under panel data framework. Detail discussion about the regression analysis has been pre- sented in the respective section.

4.4.1 Defining Input and Output in Banking Services

In a multi input-output service like banking, it is also difficult to specifically define input and output. Precise definition of input and output of banks is difficult because banks are typically a multi-input and multi-output firm (Mohan, 2005b). Discussing issues related to banking theory, Frexias and Rochet (1997) highlighted three main approaches of mea- surement of different activities of banks. These are production approach, intermediation

approach, and modern approach. According to the author, the first two approaches are based on classical microeconomic theory while the third one modifies the classical theory and incorporated risk management into the measurement of activities in banking services.

Production approach considers banks as producers of deposits and loans using capital, labour and other material factors as inputs. In the production approach, output is measured in terms of number of accounts and outstanding loans. Total cost of banks in this approach is measured as sum of all operating costs (Ferrier and Lovell, 1990).

Intermediation approach defines financial institutions as intermediating funds between savers and investors (Das and Ghosh, 2006). In this approach, banks collect funds and intermediate these funds to loans and other assets. In intermediation approach, output is measured in terms of dollar value of deposit accounts and loans. Input cost in intermedi- ation approach is measured in terms of operating cost and interest cost. Mohan (2005b) mentions three variants of intermediation approach, these are asset approach, user cost ap- proach and value added approach. The asset approach focuses exclusively on the role of banks as financial intermediaries, depositors and final uses of bank asset. Deposits and other liabilities along with labour and physical capital are termed as inputs. Loans and investments constitute banks’ output in asset approach. The user cost approach determines bank’s inputs and output on the basis of their contribution to bank revenue. If financial returns on an asset exceeds opportunity costs they are considered as output and otherwise taken as inputs. Value added approach identifies those categories as output that contributes into the value addition of bank. This approach considers deposits and loans as outputs as they form a significant portion of value added.

Finally, the modern approach measures output integrating risk, agency costs and quality of bank services (Mohan, 2005b). This approach also incorporates quality of bank assets and the probability of bank failure in the measurement of costs.

Considering differences in defining the input and output in banking services, it is very crucial to select input and output carefully. In a review of estimation methodology of banking intermediation services, Barman and Samanta (2007) have pointed out that major part of revenue in the banking sector comes from the financial intermediation services rendered by banks to their customers. System of National Accounts 1993 (SNA 1993) used the term Financial Intermediary Services Indirectly Measured (FISIM) instead of Banking Services as in SNA 1968. The FISIM is measured as difference between property income

received and interest paid. The authors focus on measurement of output and prices of intermediation service. In most of the previous studies deposit and loan or investments has been used as one of outputs in banking services, which falls under intermediation approach (Rezvanian et al., 2008; Ataullah et al., 2004; Keshari and Paul, 1994; Bhattacharyya et al., 1997a). Following the notion that banking is an intermediation services, the present study considers input and output combination according to intermediation approach. Hence, the present study has considered advances and investments as outputs of bank while demand deposits, saving deposits, fixed deposits, capital related operating expenses and employee expenses are regarded as inputs of bank.¹³

Required data for the productivity analysis are obtained from various issues of Statis- tical Tables Relating to Banks in India (STRB) published annually by Reserve Bank of India.¹⁴ The present study is carried out for the period of 1990-91 to 2014-15. Selec- tion of 1990-91 as the beginning of our sample period is based on the intuition that prior to 1990-91, the banking sector was less competitive as there was restriction on entry of new banks. According to Reserve Bank of India (2008), the restriction on the entry of new banks resulted in lack of competitiveness in the entire banking sector in India which resulted in poor performance of the Indian banking sector. Moreover, restrictive policies such as stringent regulation on interest rate also had adverse impact on the competitive environment. Since 1991, many new banks have started operation in India while many others have ceased their operation. Due to this entry and exit of banks, number of banks operating in India differs year to year. In addition, merger of some banks with other banks also have affected the number of operating banks in different years. Therefore, the present dataset is inconsistent in number of banks (panels) over the years which is not suitable for estimation of Malmquist Productivity Index over the year continuously. In order to over- come the problem, this study computes Malmquist productivity index comparing data for two years at a time starting from 1990-91 to 1991-92. Furthermore, such breakups made it possible to include or exclude any new entry or exit of bank. Table 4.1 presents the year wise break up of sample banks by category of ownership.