Abstract
Combining Data Envelopment Analysis and dynamic panel data methods, we find that adoption of digital payment technologies by Indian banks has helped enhance their cost efficiency. Instead of directly reducing the inputs used in intermediation, the gain in efficiency may be on account of cheaper availability of such inputs when banks go digital. These gains may stem from assimilation into the entire digital payments ecosystem, as opposed to piecemeal adoption of technology. We find both cost and technical efficiencies exhibiting persistence. Banks’ relative asset holdings in the industry, non-performing assets, cost of deposits and returns on advances and equity are other important variables that drive cost efficiency.
Keywords: Payment systems, Banks and depository institutions, Panel data, Technological change
Introduction
The banking sector is a crucial engine of economic growth as it channelises financial resources for productive activities. With rapid advances in the technology space globally, the financial sector has witnessed radical transformation and restructuring in businesses. In India, the financial sector reforms under the New Economic Policy regime from the 1990s have paved the way for technology-intensive banking with high priority accorded to modern payment and settlement systems [45]. Accordingly, India’s digital payment ecosystem has witnessed rapid growth momentum, with increased onboarding of new users into digital payments post COVID-19 [47]. The innovations in the payment space provide customers with easy and hassle-free access to banking services. For banks, technology-intensive products can lead to improvements in internal processes, competitiveness, and cost and profitability advantages through network, scope, and scale economies leading to effective financial intermediation and risk management.
This study attempts empirical investigation into the channels through which payments technology impacts Indian commercial banks, with the goal of deriving useful policy perspectives. Despite sizeable literature on measurement of banking sector’s efficiency, the empirical literature associating digital payments and bank efficiency is scarce for emerging market economies like India. We use Data Envelopment Analysis (DEA) to compute efficiency scores of banks and dynamic panel data methods to gauge the impact of technological innovations through digital payments on cost and technical efficiency scores over the period 2011–12 to 2018–19. Our empirical findings suggest that greater adoption of payments technology is associated with improvement in cost efficiency but not technical efficiency. This differentiated impact in the Indian context could be attributable to banks realising efficiency gains by economising on the expenditure on loanable funds, labour, and fixed assets rather than merely reducing the quantities of inputs. We find evidence of stickiness of efficiency scores over time. Higher share of non-performing loans and increased cost of deposits are negatively associated with bank efficiency, whereas, the rate of return on advances, return on equity and capital adequacy ratio may improve cost efficiency.
The remainder of the paper is presented in five sections including review of literature (Sect. 2), methodology and data (Sect. 3), stylised facts (Sect. 4), empirical findings (Sect. 5) and conclusion (Sect. 6).
Review of literature
Globally, studies provide evidence on the deployment of self-service technologies like Automated Teller Machines (ATMs) leading to a reduction in operating costs [31, 53], and enhancement in cost efficiency of the banking system [26, 42]. The adoption of innovative technologies by banks can lead to scale and experience economies [22], enhanced capacity to lend and benefits to customers [9]. Kumar [38] used DEA to compute the total factor productivity (TFP) of Indian banks to show electronic transactions leading to enhanced productivity levels. More recent studies show that electronic payments impact cost efficiency and financial performance [2, 23]. Further, the emergence of newer ‘disruptive’ technologies has entailed an overhaul of the traditional banking model owing to the potential of Fintech to augment operational efficiency, increase profitability, improve service delivery and strengthen risk control capabilities [55]. Mor and Gupta [40] find that the deployment of Artificial Intelligence (AI) in chatbots, virtual assistants and ATMs can reduce technical inefficiency of Indian banks. Advancements in payments technology also have a positive effect on profitability of banks through a reduction in labour and transaction costs [30, 32, 39]. Arora and Arora [4] find evidence of information technology (IT) investments positively impacting operating profits per employee. Research also underscores the importance of innovative payment channels like internet banking, point of sale (PoS) machines, and telephone banking in improving bank’s market share [41]. Further, there exists association between customer satisfaction and cost-effective technological innovations in banks [1].
Contrary to the findings of these studies, another strand of literature does not find sufficient impact of digital innovations on banks’ efficiency and financial performance [35]. The number of ATMs do not have statistically significant impact on efficiency [43]. Owing to the large investment needed in IT systems and the inability to curb labour expenditure given the manual nature of certain banking operations, there exists a negative association between ATM intensity and technical efficiency scores in India [49]. Thus, a rise in ATM deployment needs to be complemented with other electronic channels such as PoS, automated clearinghouse, remote banking system, and internet banking [34]. Moreover, a mere increase in the number of devices like ATMs, cards, and PoS machines does not necessarily result in improved efficiency. They need to be supported by direct usage by customers, as measured by the transaction values and volumes processed under these channels [37]. This limited association between banks’ IT investments and performance indicates a ‘profitability paradox’ [8].
Methodology and data
Studies on banks’ efficiency and its various determinants utilise DEA and Stochastic Frontier Analysis (SFA). DEA, owing to Farrel [24] and Charnes et al. [16], is a non-parametric linear mathematical programming and deterministic approach that gauges efficiency of banks against an efficiency frontier- a benchmark based on observed data. SFA, on the other hand, is a parametric method that attributes variations in the efficiency scores to two components: symmetric random shocks, and managerial errors and coordination failures [12]. DEA is preferred to other efficiency measurement techniques due to various advantages: ease of use and ability to incorporate multiple inputs and outputs in estimating efficiency scores [33], no requirement of a pre-specified production function, thereby obviating the functional form misspecification problem [20], assessment of efficiency over time [15], and no a priori assumptions regarding the distribution of efficiency [43]. In India, several studies have adopted DEA [13, 20, 48]. However, a major shortcoming of DEA is that it assumes away the existence of random shocks to the production function. Nevertheless, research also points out similar results obtained under both parametric and non-parametric approaches to efficiency measurement [27].
We use DEA to calculate cost (CE) and technical (TE) efficiency scores of Indian banks. A digital index is constructed using technology variables: ATM, PoS, National Electronic Funds Transfer (NEFT), Debit Cards, and Credit Cards. Dynamic panel regression is used to ascertain the impact of the digital index on the efficiency levels of sample banks.
Computation of technical efficiency scores
The DEA model by Charnes et al. [16], popularly the CCR model, assumes constant returns to scale and the efficiency score obtained is known as overall technical efficiency (OTE). Depending upon the model’s orientation (input or output), OTE can be interpreted as a firm’s ability to obtain maximum possible outputs for a set of inputs or minimum wastage of inputs for a given level of output. Banker et al. [6] proposed the BCC model, a modified CCR model, with variable return to scale through a convexity constraint. The efficiency scores computed are termed pure technical efficiency (PTE) scores. For our research we use the output-oriented BCC model to compute the technical efficiency scores as the technology used by banks is not restricted to constant returns to scale.
Cost efficiency scores
The cost efficiency model (CEM) is an extension of basic DEA models, introduced by Farrel [24] and developed by Grosskopf et al. [28]. The calculation of cost efficiency of a decision-making unit (DMU) needs data on inputs, input prices and outputs. The CEM shows how far a bank's cost is from the best-practice bank's cost for producing the same bundle of output [56]. For our study, the new CEM by Tone [52] is used as it addresses certain shortcomings of the traditional method. In the standard CEM, two DMUs with the same level of inputs and outputs have the same CE, despite significant difference in input prices. In Tone’s model, however, the product of price and input quantity is minimised, to reflect the ‘true’ CE of a firm relative to other firms. We used the DEA max solver program to compute CE (Type II) scores.
Panel data model specification
We compute TE and CE scores and the digital payment (volume) index (DIvol) for sample banks from 2011 to 2018. To explore the impact of payment technology on bank efficiency, the dynamic panel data estimation techniques proposed by Arellano and Bond [3] and Blundell and Bond [14] are employed. Since the dependent variables for the analysis are DEA efficiency scores that tend to exhibit persistence over a period [54], we use this methodology to capture the dynamic aspects of CE and TE. In the Arellano–Bond method, first difference of the regression equation is taken to eliminate the individual effects. Following this, deeper lags of the dependent variable are used as instruments for differenced lags of the dependent variable (which are endogenous). Blundell and Bond [14] derived a condition under which it is possible to use an additional set of moment conditions to improve the small-sample performance of the Arellano-Bond estimator. We use linear efficient two-step Generalised Method of Moments (GMM) estimators, as they are found to be more efficient in the literature. We use the vce (robust) option to estimate the Huber-White Sandwich Variance Covariance Estimator (VCE). This technique assumes that the errors are independently distributed across observations and allows the errors to be heteroskedastic.
Data
We use a balanced panel1 comprising 328 bank-level observations (41 banks for 8 years: 2011–12 to 2018–19), comprising 21 public sector banks (PSBs), 17 domestic private banks (PBs), and 3 foreign banks (FBs), out of total 94 scheduled commercial banks (SCBs). For 2018–19, our sample banks accounted for 94.7 per cent of the deposits of the entire banking sector. Balance sheet data on assets, liabilities, incomes, and expenses were taken from Reserve Bank of India’s annual publications (‘Statistical Tables Relating to Banks in India’, ‘Branch Banking Statistics’) and payment systems data (NEFT, ATM, PoS and Cards) are from the Reserve Bank’s monthly data releases.
Input and output variables for efficiency score
The literature provides no consensus on specification of inputs and outputs for a banking firm. Broadly, two approaches are followed: the production or value-added approach and the intermediation approach. For the first, banks are considered as primary providers of financial services (loans and deposits), while capital and labour are treated as inputs. For the second, banks as financial intermediaries collect deposits from savers and channelise these funds into loans and investments. Banks are liable to pay interest to the depositors (interest expenditure) and earn interest from the borrowers (interest income). Since this approach includes variables concerned with the overall cost of the banking sector, it is considered appropriate to reflect on the economic viability of commercial banks [25]. Berger and Humphrey [10] point out the suitability of the intermediation approach for measuring bank-level efficiency, and the production approach for assessing efficiency at the branch level. We selected input and output variables, as explained in Appendix Table 3, based on the intermediation approach, and utilised three input variables (loanable funds,2 fixed assets, and total employees), their corresponding prices, and two output variables (net interest income and non-interest income) for computation of CE scores. The same set of variables was used for TE computation, which does not account for input prices. Thus, TE scores denote the ability of the observed banks to maximise net interest and non-interest incomes given the inputs of deposits, borrowings, physical capital, and labour. CE scores denote how the banks minimise the expenditure on these inputs, given the level of interest and non-interest earnings.
Table 3.
Variables description for cost and technical efficiency calculation.
Source: Author's calculations
| Variables | Description | |
|---|---|---|
| Inputs | ||
| 1 | Loanable funds | Deposits + borrowing |
| 2 | Fixed assets | Fixed assets |
| 3 | Total employees |
Calculated using business per employee ratio Business = Deposits + Advances Number of Employees = Business/Business per employee |
| Outputs | ||
| 1 | Net interest income | Interest earned − Interest Spent |
| 2 | Non-interest income | Other Income |
| Prices | ||
| 1 | Price of loanable fund | (Interest on deposit + interest on borrowing)/Loanable fund |
| 2 | Price of fixed assets | (Rent, taxes and lighting + Printing and stationer + depreciation + repair + Insurance)/Value of FA |
| 3 | Price of total employees | Payment & provision to employees/ Total Employees |
Variables for regression analysis
We construct the digital payments index comprising transaction volumes of key digital payment indicators. The volume index included NEFT transactions, card transactions (credit and debit cards) and the number of ATM and PoS machines deployed by the bank over the financial year. Each indicator is calculated as the percentage share of the individual bank in the sample total for that year. Following this, the mean of these five shares is computed to arrive at the index value for the bank, implying equal weighing of the parameters. The final values thus calculated are rescaled such that they range between 0 and 100 each year.
Taking cues from the literature, other bank-specific covariates are also used for better model specification and insights. The number of bank offices measures not only the bank 'size' but also the degree of dependence on offline infrastructure for operations. Higher the number of branches, greater may be the level of inefficiency of the bank owing to subpar managerial control [20], and higher personnel and administrative costs [43]. Banks with larger asset holdings are expected to benefit from scale economies and exhibit cost efficiency [36, 46]. Five variables impacting the profitability of the bank are used: the net Non-Performing Asset (NPA) ratio indicating the real loss taken by the bank on bad debts, the cost of deposits for the bank (COD, i.e., interest paid on deposits divided by average deposits), the return on advances (ROA, which equals the interest or discount on advances and bills divided by the average advances), the capital adequacy ratio (CAR, a proxy for the bank's capacity to meet time liabilities and manage risks such as credit, operational and market risks), and the return on equity ratio (ROE, a ratio relating net profit or net income to shareholders' equity). Research shows that higher levels of NPAs tend to be detrimental to efficiency scores and bank performance [7, 20, 29]. Further, we expect higher ROA to lead to more efficiency, as supported by Berger and Mester [11], and Chen and Lin [17]. There is an alternative perspective pertaining to the impact of capital adequacy norms on banks’ efficiency. For regulators, higher CAR is indicative of higher bank efficiency on account of low cost of funds and improved financial health [20]. From the commercial bank's side, CAR hinders bank's performance [18] as it can lead to overcautious lending, heightened risk aversion [13] and reduced bank profitability [39]. ROE puts pressure on banks to perform well as they are listed in the stock market, therefore, it is expected to be positively associated with banking efficiency. ROE is highly dependent on the effective use of resources by banks [51].
Stylised facts
Summary statistics of input, output and control variables used in DEA are provided in Appendix Tables 4 and 5. Electronic payments in India have witnessed a manifold increase. The annual average (across sample banks) NEFT transaction volumes increased by more than ten times between 2011 and 2018 (Appendix Table 6). Similarly, transactions through credit cards and PoS machines surged by nearly five times. In contrast, growth of debit card transactions and ATMs was lower, perhaps due to earlier arrival and adoption in India. There is evidence of rapid growth in transactions through ATMs and debit cards in the previous decade [50]. Table 1 presents summary statistics for technical and cost efficiency scores.
Table 4.
Summary statistics of input and output variables for DEA.
Source: Author's calculations
| Years | Statistics | Fixed assets | Loanable funds | Total employees | Other income | Net Interest income |
|---|---|---|---|---|---|---|
| 2011 | Mean | 12,971 | 1,633,432 | 24,785 | 34,211 | 52,044 |
| SD | 13,015 | 2,000,331 | 38,781 | 64,078 | 69,685 | |
| Min | 589 | 59,379 | 1462 | 852 | 1930 | |
| Max | 54,665 | 11,700,000 | 239,376 | 354,609 | 423,421 | |
| 2012 | Mean | 14,452 | 1,888,267 | 25,664 | 38,757 | 57,734 |
| SD | 14,352 | 2,321,719 | 38,937 | 76,145 | 72,848 | |
| Min | 943 | 110,779 | 1991 | 946 | 2656 | |
| Max | 70,050 | 13,700,000 | 238,201 | 446,007 | 430,955 | |
| 2013 | Mean | 17,426 | 2,167,107 | 27,444 | 20,630 | 61,343 |
| SD | 18,338 | 2,694,922 | 40,388 | 28,223 | 82,522 | |
| Min | 1169 | 134,688 | 2592 | 186 | 2801 | |
| Max | 80,022 | 15,800,000 | 244,817 | 158,246 | 491,230 | |
| 2014 | Mean | 18,472 | 2,385,322 | 28,183 | 22,637 | 65,271 |
| SD | 19,656 | 3,035,049 | 39,205 | 27,747 | 88,376 | |
| Min | 1273 | 133,230 | 2229 | 671 | 2988 | |
| Max | 93,292 | 17,800,000 | 233,130 | 143,515 | 519,650 | |
| 2015 | Mean | 25,839 | 2,593,027 | 29,052 | 24,303 | 69,897 |
| SD | 26,110 | 3,422,939 | 38,759 | 28,969 | 92,655 | |
| Min | 1564 | 116,056 | 2190 | 1047 | 2954 | |
| Max | 103,893 | 20,500,000 | 226,394 | 160,368 | 531,618 | |
| 2016 | Mean | 34,350 | 2,791,291 | 29,866 | 31,043 | 74,087 |
| SD | 67,226 | 3,888,531 | 38,989 | 35,054 | 101,485 | |
| Min | 1544 | 114,389 | 2005 | 2036 | 3136 | |
| Max | 429,189 | 23,600,000 | 222,650 | 185,529 | 573,258 | |
| 2017 | Mean | 33,730 | 3,165,763 | 31,423 | 32,997 | 84,000 |
| SD | 63,001 | 4,940,190 | 46,324 | 40,040 | 125,417 | |
| Min | 1403 | 113,118 | 1853 | 1254 | 3848 | |
| Max | 399,923 | 30,700,000 | 277,918 | 225,759 | 724,996 | |
| 2018 | Mean | 35,385 | 3,407,067 | 3138 | 42,130 | 98,391 |
| SD | 62,244 | 5,354,891 | 4579 | 64,759 | 153,261 | |
| Min | 1384 | 108,037 | 174 | 527 | 3468 | |
| Max | 391,976 | 33,100,000 | 27,156 | 367,749 | 883,489 |
Table 5.
Summary statistics for controls.
Source: Author's calculations
| Years | Statistics | NPA | Return on advances | Cost of deposits | CAR | No of branches | Return on equity | Asset sharea |
|---|---|---|---|---|---|---|---|---|
| 2011 | Mean | 1.04 | 11.16 | 6.72 | 13.86 | 1861 | 14.38 | 2.44 |
| SD | 0.68 | 1.30 | 1.13 | 2.42 | 2469 | 6.81 | 2.99 | |
| Min | 0.05 | 8.67 | 3.21 | 9.49 | 49 | − 14.70 | 0.09 | |
| Max | 3.09 | 14.23 | 8.40 | 23.20 | 14,821 | 24.91 | 17.58 | |
| 2012 | Mean | 1.46 | 11.22 | 7.00 | 13.56 | 2025 | 13.85 | 2.44 |
| SD | 0.97 | 1.35 | 1.06 | 2.11 | 2599 | 5.64 | 3.02 | |
| Min | 0.01 | 8.40 | 3.85 | 11.02 | 49 | 0.35 | 0.15 | |
| Max | 3.36 | 14.13 | 8.74 | 18.74 | 15,513 | 24.81 | 17.93 | |
| 2013 | Mean | 1.93 | 10.92 | 6.86 | 12.65 | 2235 | 8.98 | 2.44 |
| SD | 1.42 | 1.39 | 1.12 | 2.45 | 2759 | 10.38 | 3.04 | |
| Min | 0.05 | 7.69 | 3.85 | 8.67 | 50 | − 33.54 | 0.15 | |
| Max | 7.18 | 13.35 | 8.42 | 18.83 | 16,359 | 25.02 | 17.87 | |
| 2014 | Mean | 2.29 | 10.64 | 6.79 | 12.57 | 2443 | 8.44 | 2.44 |
| SD | 1.56 | 1.23 | 1.06 | 2.12 | 2928 | 8.85 | 3.14 | |
| Min | 0.12 | 7.47 | 4.26 | 9.59 | 50 | − 33.07 | 0.13 | |
| Max | 6.22 | 12.85 | 8.17 | 17.17 | 17,199 | 21.33 | 18.55 | |
| 2015 | Mean | 4.08 | 10.21 | 6.46 | 12.44 | 2598 | 1.28 | 2.44 |
| SD | 3.02 | 1.24 | 0.96 | 2.20 | 2997 | 13.14 | 3.26 | |
| Min | 0.28 | 7.34 | 3.93 | 7.51 | 50 | − 34.01 | 0.10 | |
| Max | 11.89 | 13.49 | 7.86 | 16.64 | 17,578 | 19.94 | 19.64 | |
| 2016 | Mean | 5.08 | 9.55 | 5.88 | 12.95 | 2712 | 3.05 | 2.44 |
| SD | 3.90 | 1.15 | 0.85 | 2.20 | 3089 | 11.59 | 3.43 | |
| Min | 0.18 | 7.18 | 3.24 | 10.26 | 36 | − 26.98 | 0.10 | |
| Max | 13.99 | 11.54 | 7.07 | 18.76 | 18,146 | 18.58 | 20.87 | |
| 2017 | Mean | 6.09 | 8.91 | 5.25 | 13.01 | 2933 | − 8.02 | 2.44 |
| SD | 4.72 | 1.11 | 0.71 | 2.84 | 3937 | 19.37 | 3.82 | |
| Min | 0.28 | 7.03 | 2.64 | 8.69 | 26 | − 46.63 | 0.08 | |
| Max | 16.69 | 11.03 | 6.28 | 18.48 | 24,492 | 17.87 | 23.71 | |
| 2018 | Mean | 4.38 | 9.01 | 5.22 | 13.22 | 2937 | − 11.35 | 2.44 |
| SD | 2.90 | 0.98 | 0.72 | 3.01 | 3811 | 27.60 | 3.80 | |
| Min | 0.19 | 7.48 | 2.64 | 2.00 | 26 | − 103.27 | 0.08 | |
| Max | 10.81 | 11.02 | 6.72 | 17.89 | 23,560 | 17.94 | 23.40 |
aAsset share is calculated by dividing total assets of individual banks with aggregate total assets of all 41 banks, multiplied by 100
Table 6.
Summary statistics for digital payment variables.
Source: Author's calculations
| Year | Statistics | Credit cards | Debit cards | NEFT | ATM | PoS |
|---|---|---|---|---|---|---|
| 2011 | Mean | 7.28 | 122 | 4.99 | 1,911 | 14,545 |
| Std. Dev | 18.50 | 313 | 7.76 | 3,653 | 44,517 | |
| Min | 0 | 0.51 | 0.04 | 44 | 0 | |
| Max | 79.30 | 1,920 | 35.09 | 21,313 | 192,084 | |
| 2012 | Mean | 9.23 | 130 | 8.75 | 2,370 | 17,856 |
| Std. Dev | 23.60 | 343 | 13.29 | 4,317 | 52,849 | |
| Min | 0 | 0.95 | 0.09 | 128 | 0 | |
| Max | 107 | 2,170 | 64.32 | 23,823 | 214,203 | |
| 2013 | Mean | 12 | 150 | 14.66 | 3,090 | 23,464 |
| Std. Dev | 31.80 | 407 | 22.55 | 5,783 | 64,681 | |
| Min | 0 | 1.82 | 0.28 | 144 | 0 | |
| Max | 161 | 2,590 | 122.90 | 34,155 | 264,119 | |
| 2014 | Mean | 14.40 | 175 | 20.63 | 3,980 | 25,599 |
| Std. Dev | 38.70 | 482 | 31.79 | 7,369 | 67,388 | |
| Min | 0 | 2.64 | 0.78 | 139 | 0 | |
| Max | 198 | 3,080 | 179.23 | 45,206 | 270,922 | |
| 2015 | Mean | 18.40 | 208 | 27.21 | 4,421 | 29,325 |
| Std. Dev | 48.80 | 575 | 43.87 | 7,791 | 71,859 | |
| Min | 0 | 3.95 | 0.91 | 125 | 0 | |
| Max | 251 | 3,690 | 253.86 | 47,705 | 266,504 | |
| 2016 | Mean | 25.30 | 246 | 34.47 | 4,738 | 40,063 |
| Std. Dev | 65.60 | 672 | 50.86 | 8,188 | 92,522 | |
| Min | 0.00 | 5.52 | 1.18 | 110 | 0 | |
| Max | 333.00 | 4,310 | 271.81 | 49,977 | 367,334 | |
| 2017 | Mean | 32.90 | 289 | 43.75 | 5,035 | 69,722 |
| Std. Dev | 82.10 | 795 | 67.00 | 9,473 | 143,113 | |
| Min | 0.00 | 7.70 | 1.47 | 98 | 0 | |
| Max | 405.00 | 5,110 | 368.00 | 59,080 | 624,171 | |
| 2018 | Mean | 41.30 | 343 | 51.25 | 4,928 | 82,226 |
| Std. Dev | 99.80 | 928 | 75.63 | 9,416 | 161,290 | |
| Min | 0.00 | 9.57 | 1.79 | 90 | 0 | |
| Max | 486.00 | 5,970 | 401.69 | 58,973 | 582,451 |
Table 1.
Summary Statistics for Efficiency Scores
Source: Author’s calculations
| Year | TE | CE | ||||||
|---|---|---|---|---|---|---|---|---|
| Mean | Std. Dev | Min | Max | Mean | Std. Dev | Min | Max | |
| 2011 | 0.698 | 0.208 | 0.334 | 1 | 0.508 | 0.229 | 0.227 | 1 |
| 2012 | 0.652 | 0.215 | 0.303 | 1 | 0.554 | 0.250 | 0.266 | 1 |
| 2013 | 0.804 | 0.190 | 0.393 | 1 | 0.548 | 0.268 | 0.235 | 1 |
| 2014 | 0.738 | 0.198 | 0.342 | 1 | 0.546 | 0.269 | 0.214 | 1 |
| 2015 | 0.701 | 0.220 | 0.332 | 1 | 0.561 | 0.285 | 0.218 | 1 |
| 2016 | 0.685 | 0.238 | 0.308 | 1 | 0.539 | 0.281 | 0.190 | 1 |
| 2017 | 0.695 | 0.237 | 0.284 | 1 | 0.535 | 0.279 | 0.201 | 1 |
| 2018 | 0.709 | 0.225 | 0.409 | 1 | 0.496 | 0.250 | 0.234 | 1 |
Empirical results
We estimate models for both CE and TE with the digital volume index as well as its individual components separately as explanatory variables. Note that all model specifications presented pass the post-estimation check for autocorrelation (Arellano-Bond). The key findings are as follows.
Persistence effect
The first-order autoregressive term of CE and TE is positive and statistically significant in all specifications (Table 2), justifying the use of dynamic panel data modelling. This implies the stickiness of relative CE and TE in the Indian banking industry. Between CE and TE models, the persistence effect reduces in magnitude for the former, as controls are added to the model, while it magnifies for the latter. TE is almost exclusively explained by its persistence, along with autoregression of higher order (two).
Table 2.
Impact of digital index on cost and technical efficiency of Indian SCBs
Source: Author's calculations
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Variables | lnCE | lnCE | lnCE | lnTE | lnTE | lnTE |
| L.lnCE |
0.838*** (0.0943) |
0.683*** (0.151) |
0.468*** (0.150) |
|||
| L.lnTE |
0.345*** (0.0973) |
0.539*** (0.0772) |
0.346*** (0.0720) |
|||
| L2.lnTE |
− 0.169** (0.0794) |
|||||
| Constant |
− 0.133* (0.0702) |
− 0.196 (0.120) |
− 0.566** (0.281) |
− 0.434*** (0.0881) |
− 0.149 (0.0912) |
0.166 (0.277) |
| Digital index |
0.00192 (0.00117) |
0.00580* (0.00326) |
0.00732* (0.00392) |
0.0139 (0.00941) |
0.00503 (0.00546) |
0.00143 (0.00476) |
| Net NPA |
− 0.0161* (0.00825) |
− 0.0381*** (0.00910) |
||||
| Cost of deposits |
− 0.0886** (0.0398) |
− 0.105* (0.0574) |
||||
| Log of asset share |
− 0.101 (0.0619) |
− 0.213*** (0.0733) |
− 0.0313 (0.0716) |
− 0.139 (0.130) |
||
| No. of branches |
− 2.24e-05 (3.46e-05) |
3.00e-05 (3.43e-05) |
− 2.24e-05 (3.64e-05) |
3.36e-05 (4.23e-05) |
||
| Return on Advances |
0.0647* (0.0338) |
0.0423 (0.0426) |
||||
| Return on Equity |
0.00142* (0.000856) |
0.000805 (0.00144) |
||||
| Capital Adequacy Ratio |
0.00343 (0.00912) |
-0.00833 (0.0122) |
||||
| Observations | 287 | 287 | 287 | 246 | 287 | 287 |
| Number of banks | 41 | 41 | 41 | 41 | 41 | 41 |
Standard errors in parentheses
***p < 0.01, ** p < 0.05, * p < 0.1
Effect of payments technology adoption on cost efficiency
Controlling for bank size and business-side determinants, the coefficient for digital payments technology adoption emerges positive and statistically significant for banks’ CE. All other covariates except return on advances and return on equity adversely affect CE. Thus, better adoption of technology makes business sense for banks, as it provides a pathway to improve cost efficiency of operations relative to other banks [21]. Payments technology such as ATMs, PoS and cards may help improve CE by (a) reduction in man-hours otherwise dedicated towards cash withdrawals, balance checking, passbook printing, etc. and routing employee's time towards more productive types of work; (b) formalised and automated accounting of transactions; (c) network effects- banks having a wider card and ATM network may be able to procure more and better customers, especially among the young demographic; (d) reduction in costs stemming from human error; and (e) reduction in costs of regulatory compliance.
Unlike CE, there is no evidence that banks’ TE is affected by technological innovation. This implies that rather than help banks economise on inputs (labour, fixed capital, and loanable funds) to generate the same income from advances and investments, digital payment technologies help by economising on the cost of these inputs. These benefits may stem from a reduction in cost of loanable funds, since banks that are highly digital may be able to mobilize deposits more readily, as well as raise loans cheaply due to ready availability of accurate and electronic ledgers. It may be due to reduction in expenditure needed on fixed assets. Illustratively, increased penetration of debit cards, ATMs and PoS machines can directly help in reducing the need for opening brick and mortar branches. Digital technologies may also have helped banks economise on labour costs. On the other hand, the efficiency of processes through which a bank converts inputs and deposits to its income stream may still be governed by factors such as quality of management and the institutional and regulatory framework of the industry.
To elucidate if individual payment technologies affect CE differently, we use the components of the digital index as standalone predictors (Appendix Table 11). Here, we find no statistically significant impact of the payment modes when introduced separately. Hence, banks may need to deploy the entire gamut of digital payment modes to reap efficiency gains, rather than relying only on a select few modes. These gains may stem from inclusion in the entire digital ecosystem, as opposed to piecemeal adoption of technology. As another robustness check, we include an indicator for demonetisation3 in the model (Appendix Table 12). We find evidence that demonetisation affected banks’ CE. The number of bank offices variable also showed a statistically significant and positive impact on CE.
Table 11.
Impact of individual digital parameters on CE.
Source: Author's calculations
| Variables | (1) | (2) | (3) | (4) | (5) |
|---|---|---|---|---|---|
| lnCE | lnCE | lnCE | lnCE | lnCE | |
| L.lnCE |
0.545*** (0.134) |
0.498*** (0.154) |
0.543*** (0.138) |
0.552*** (0.136) |
0.557*** (0.132) |
| Constant |
− 0.587** (0.295) |
− 0.592** (0.297) |
− 0.597** (0.301) |
− 0.587** (0.297) |
− 0.554** (0.282) |
| NEFTTransactions |
− 0.000123 (0.000505) |
||||
| No. of ATMs |
1.57e-05 (2.17e-05) |
||||
| No. of PoS |
3.00e-07 (3.52e-07) |
||||
| CC transaction volume |
− 4.40e-10 (4.24e-10) |
||||
| DC transaction volume |
− 1.83e-11 (4.16e-11) |
||||
| Net NPA |
− 0.0130* (0.00790) |
− 0.0158* (0.00916) |
− 0.0148* (0.00834) |
− 0.0132* (0.00793) |
-0.0131 (0.00809) |
| Cost on Deposits |
− 0.0978** (0.0400) |
− 0.104** (0.0441) |
− 0.0894** (0.0412) |
− 0.102** (0.0399) |
− 0.0993** (0.0404) |
| Log of asset share |
− 0.157** (0.0727) |
− 0.166** (0.0747) |
− 0.155** (0.0785) |
− 0.161** (0.0727) |
− 0.145** (0.0704) |
| No. of branches |
2.53e-05 (3.18e-05) |
3.06e-06 (3.19e-05) |
2.16e-05 (3.85e-05) |
3.10e-05 (3.06e-05) |
1.97e-05 (2.50e-05) |
| Return on Advances |
0.0798** (0.0365) |
0.0807** (0.0381) |
0.0742** (0.0371) |
0.0822** (0.0363) |
0.0793** (0.0366) |
| Return on Equity |
0.00101 (0.000826) |
0.00129 (0.000872) |
0.000992 (0.000866) |
0.00111 (0.000819) |
0.000982 (0.000826) |
| Capital adequacy Ratio |
0.00708 (0.00913) |
0.00707 (0.00943) |
0.00786 (0.00993) |
0.00712 (0.00903) |
0.00728 (0.00908) |
| Observations | 287 | 287 | 287 | 287 | 287 |
| Number of banks | 41 | 41 | 41 | 41 | 41 |
Standard errors in parentheses
***p < 0.01, **p < 0.05, *p < 0.1
Table 12.
Results with demonetisation.
Source: Author's calculations
| Variables | (1) | (2) |
|---|---|---|
| lnCE | lnTE | |
| L.lnCE |
0.465*** (0.135) |
|
| L.lnTE |
0.337*** (0.0729) |
|
| Constant |
− 0.347 (0.307) |
0.233 (0.282) |
| Digital Index |
0.00727* (0.00393) |
0.00160 (0.00415) |
| Net NPA |
− 0.0183** (0.00823) |
− 0.0393*** (0.00912) |
| Cost on Deposits | − 0.120*** | − 0.116* |
| (0.0457) | (0.0650) | |
| Log of asset shares |
− 0.231*** (0.0593) |
− 0.148 (0.115) |
| No. of branches |
2.80e-05** (1.28e-05) |
3.64e-05 (3.43e-05) |
| Return on Advances |
0.0654* (0.0353) |
0.0439 (0.0421) |
| Return on Equity |
0.000967 (0.00104) |
0.000643 (0.00151) |
| Capital Adequacy Ratio |
0.00318 (0.0104) |
− 0.00927 (0.0131) |
| Demonetisation dummy |
− 0.0661** (0.0320) |
− 0.0225 (0.0301) |
| Observations | 287 | 287 |
| Number of banks | 41 | 41 |
Standard errors in parentheses
***p < 0.01, **p < 0.05, *p < 0.1
Effect of bank size
Bank size emerges as an important determinant of CE (Appendix Table 13). The number of bank offices had no statistically significant impact. However, the percentage share of the asset holdings showed detrimental impact on CE, attributable to larger banks’ relative inefficiency due to complexity of operations, management, and duplication of efforts [5]. It is also possible for larger banks to be more efficient, owing to advantages generated by their market power and ability to diversify credit risk. Larger banks can afford to have a diverse portfolio of loans and hence, better manage risks and costs associated with simple portfolios. Xiaogang et al. [57] showed that medium size banks are the most inefficient. Smaller banks are relatively more efficient than the medium sized banks, whereas large banks appear to be the most efficient ones. As the bank size changes from small to medium the efficiency reduces but even larger banks lead to higher efficiency attainment, directing ‘U’ shape relationship of bank size and efficiency scores. This could be the possible reason that the coefficient of asset share displays a negative sign in the analysis. To confirm the ‘U’ shape relationship, we introduced non-linearity by including the square of the total asset share in the model. The coefficient is positive but statistically insignificant. Initially, relative cost efficiency dwindles with an increase in bank size, but after a certain size threshold, it may start increasing again.
Table 13.
Results with square of asset share.
Source: Author's calculations
| Variables | (1) |
|---|---|
| lnCE | |
| L.lnCE |
0.487*** (0.172) |
| Constant |
− 0.507* (0.278) |
| Digital Index |
0.00584 (0.00436) |
| Net NPA |
− 0.0160** (0.00793) |
| Cost on Deposits |
− 0.103** (0.0435) |
| Log of asset shares |
− 0.182** (0.0815) |
| Log of square of asset shares |
0.00148 (0.0608) |
| No. of branches |
2.95e-05 (3.43e-05) |
| Return on Advances |
0.0728** (0.0364) |
| Return on Equity |
0.00144* (0.000802) |
| Capital Adequacy Ratio |
0.00162 (0.00882) |
| Observations | 287 |
| Number of banks | 41 |
Standard errors in parentheses
***p < 0.01, **p < 0.05, *p < 0.1
Profitability variables
Our empirical findings validate the role of business side variables including net NPA ratio, ROA, COD, RoE and the CAR. COD is a significant variable that constrains banks from attaining higher CE, in addition to the losses incurred by the bank, as captured by the net NPA ratio. Lower cost of funds indicates improved resource mobilisation and higher efficiency levels of banks [44]. A significant negative association of NNPA with CE directs that higher problem loans lead to lower bank efficiency [7, 19, 20]. Higher COD and Net NPA also negatively impact the banks' technical efficiency, indicating that they hinder cost reduction and constrain banks from efficiently converting inputs into earnings. CAR also exhibits negative relation with both CE and TE scores, even though the coefficient is statistically insignificant. Banks with high CAR are considered safe and more likely to meet financial obligation, making CAR an important variable for financial sustainability of banks. This variable, however, is associated with lower profitability of banks in accordance with the risk-return trade-off theory (Hersugondo et. al, 2021). ROA and RoE positively impact CE of Indian commercial banks. Since banks listed in the stock market are under pressure to provide high returns to shareholders, RoE induces positive impact on cost efficiency scores of the banks. On the other hand, ROA directly impacts the output variable of our model i.e., interest earnings and is therefore expected to improve banks’ TE. Our results show that it has a significant positive impact on banks’ CE in line with Berger and Mester [11] and Chen and Lin [17].
Conclusion
This paper employs a dynamic panel data estimation approach on bank-level cost and technical efficiency estimates for the period 2011–2018, to capture the impact of payment technology innovation on banking sector efficiency. We find that intensified use of digital payment technologies (viz NEFT, cards, ATMs and PoS machines) makes good business sense for the banks, by helping them improve the cost efficiency of their operations. Interestingly, we are unable to establish a significant effect of digital payment technology use on the technical efficiency of banking operations, hinting that payments technology enhances efficiency of banks by helping them reduce their expenditure on inputs rather than the quantity of inputs directly. We augment the banking efficiency literature by establishing persistence or stickiness of both cost and technical efficiency, indicating that empirical analysis of the industry in the future should consider the dynamic aspect of DEA-derived efficiency measures. While innovations in the payment system demonstrably provide customers with easy, convenient, and hassle-free access to banking services, lead to effective financial intermediation, and risk management, our paper establishes empirical backing for the fact that adoption of these technologies is also a good business decision.
Appendix
See Tables 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13.
Table 7.
Ownership wise statistics of mean of TE and CE scores.
Source: Author's calculations
| TE | CE | |||||||
|---|---|---|---|---|---|---|---|---|
| Foreign | Private | Public | All Banks | Foreign | Private | Public | All Banks | |
| Average | 0.975 | 0.811 | 0.590 | 0.710 | 0.892 | 0.676 | 0.372 | 0.536 |
| S.D | 0.039 | 0.063 | 0.066 | 0.045 | 0.070 | 0.067 | 0.027 | 0.022 |
| Min | 0.882 | 0.703 | 0.527 | 0.652 | 0.774 | 0.550 | 0.342 | 0.496 |
| Max | 1.000 | 0.878 | 0.716 | 0.804 | 1.000 | 0.755 | 0.409 | 0.561 |
Table 8.
Summary statistics for public sector banks scores
| Year | Obs | TE | CE | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Mean | Std. Dev | Min | Max | Mean | Std. Dev | Min | Max | ||
| 2011 | 21 | 0.631 | 0.170 | 0.392 | 1 | 0.409 | 0.182 | 0.227 | 1 |
| 2012 | 21 | 0.577 | 0.195 | 0.303 | 1 | 0.408 | 0.180 | 0.266 | 1 |
| 2013 | 21 | 0.716 | 0.191 | 0.393 | 1 | 0.380 | 0.183 | 0.235 | 1 |
| 2014 | 21 | 0.638 | 0.161 | 0.342 | 1 | 0.374 | 0.177 | 0.214 | 1 |
| 2015 | 21 | 0.541 | 0.145 | 0.332 | 1 | 0.354 | 0.167 | 0.218 | 1 |
| 2016 | 21 | 0.527 | 0.157 | 0.308 | 1 | 0.343 | 0.174 | 0.190 | 1 |
| 2017 | 21 | 0.536 | 0.163 | 0.284 | 1 | 0.342 | 0.174 | 0.201 | 1 |
| 2018 | 21 | 0.557 | 0.134 | 0.409 | 1 | 0.361 | 0.167 | 0.234 | 1 |
Table 9.
Summary statistics for private sector banks scores
| Year | Obs | TE | CE | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Mean | Std. Dev | Min | Max | Mean | Std. Dev | Min | Max | ||
| 2011 | 17 | 0.726 | 0.220 | 0.334 | 1 | 0.550 | 0.187 | 0.302 | 1 |
| 2012 | 17 | 0.703 | 0.207 | 0.436 | 1 | 0.656 | 0.200 | 0.356 | 1 |
| 2013 | 17 | 0.878 | 0.147 | 0.601 | 1 | 0.698 | 0.235 | 0.386 | 1 |
| 2014 | 17 | 0.818 | 0.183 | 0.505 | 1 | 0.696 | 0.230 | 0.412 | 1 |
| 2015 | 17 | 0.849 | 0.152 | 0.613 | 1 | 0.755 | 0.215 | 0.440 | 1 |
| 2016 | 17 | 0.829 | 0.197 | 0.437 | 1 | 0.724 | 0.227 | 0.431 | 1 |
| 2017 | 17 | 0.837 | 0.187 | 0.503 | 1 | 0.716 | 0.230 | 0.444 | 1 |
| 2018 | 17 | 0.851 | 0.200 | 0.421 | 1 | 0.614 | 0.251 | 0.321 | 1 |
Table 10.
Summary statistics for foreign banks scores
| Year | Obs | TE | CE | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Mean | Std. Dev | Min | Max | Mean | Std. Dev | Min | Max | ||
| 2011 | 3 | 1 | 0 | 1 | 1 | 0.968 | 0.056 | 0.904 | 1 |
| 2012 | 3 | 0.882 | 0.204 | 0.647 | 1 | 1 | 0.000 | 1 | 1 |
| 2013 | 3 | 1.000 | 0.000 | 1 | 1 | 0.872 | 0.141 | 0.721 | 1 |
| 2014 | 3 | 0.988 | 0.021 | 0.963 | 1 | 0.891 | 0.189 | 0.672 | 1 |
| 2015 | 3 | 0.983 | 0.029 | 0.950 | 1 | 0.909 | 0.157 | 0.728 | 1 |
| 2016 | 3 | 0.982 | 0.031 | 0.947 | 1 | 0.867 | 0.127 | 0.747 | 1 |
| 2017 | 3 | 1 | 0 | 1 | 1 | 0.852 | 0.133 | 0.742 | 1 |
| 2018 | 3 | 0.966 | 0.057 | 0.900 | 1 | 0.774 | 0.198 | 0.627 | 1 |
Data availability
Banks’ balance sheet data were accessed from the Reserve Bank of India’s (RBI) ‘Statistical Tables Relating to Banks in India’. Data on number of bank branches were from the RBI’s Branch Banking Statistics’. Both are annual publications available on the RBI website (https://rbi.org.in/Scripts/publications.aspx). Data on payment systems (NEFT, ATM, PoS and Cards) are from the monthly Data Releases titled ‘Bankwise Volumes in NEFT/RTGS/Mobile Transactions/Internet Banking Transactions’ (https://rbi.org.in/scripts/NEFTView.aspx) and Bankwise ATM/PoS/Card Statistics (https://rbi.org.in/scripts/ATMView.aspx) available on the RBI website.
Footnotes
Our dataset avoids biases at both the beginning and end of the time period considered. The former may arise due to the impacts of the Global Financial Crisis in 2008 as well as the Payment and Settlement Systems Act, 2007, whose provisions were implemented in the successive years and decisively shaped the digital payments industry in India. We limit our dataset up to FY2019, given that the impact of COVID in the years after could lead to biased results.
Loanable funds comprise deposits and borrowings after deducting the balances with RBI. For efficiency score calculations, we take loanable funds as sum of deposits and borrowings only. For robustness, we also calculated the efficiency scores using loanable funds (Deposit + Borrowing – Balances with RBI) and found that the scores obtained were the same.
The variable takes value one for financial year 2016–2017 (the year of withdrawal of specified bank notes as legal tender in India) onwards and zero for all preceding years.
The views expressed in the paper are those of the authors and do not reflect the views of the institutions they are affiliated to.
References
- 1.Ameme B, Wireko J. Impact of technological innovations on customers in the banking industry in developing countries. Bus Manag Rev. 2016;7(3):388–397. [Google Scholar]
- 2.Ardizzi G, Crudu F, Petraglia C. Innovation and cost efficiency in the banking industry: the role of electronic payments. Econ Notes. 2019;48(1):12121. doi: 10.1111/ecno.12121. [DOI] [Google Scholar]
- 3.Arellano M, Bond S. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev Econ Stud. 1991;58(2):277–297. doi: 10.2307/2297968. [DOI] [Google Scholar]
- 4.Arora H, Arora P. Effect of investments in information technology on bank performance: empirical evidence from Indian public sector banks. Int J Bus Inf Syst. 2013;13(4):400–417. [Google Scholar]
- 5.Ataullah A, Cockerill T, Le H. Financial liberalization and bank efficiency: a comparative analysis of India and Pakistan. Appl Econ. 2004;36(17):1915–1924. doi: 10.1080/000368404200068638. [DOI] [Google Scholar]
- 6.Banker RD, Charnes A, Cooper WW. Some models for estimating technical and scale inefficiencies in data envelopment analysis. Manag Sci. 1984;30(9):1078–1092. doi: 10.1287/mnsc.30.9.1078. [DOI] [Google Scholar]
- 7.Barros CP, Managi S, Matousek R. The technical efficiency of the Japanese banks: non-radial directional performance measurement with undesirable output. Omega. 2012;40(1):1–8. doi: 10.1016/j.omega.2011.02.005. [DOI] [Google Scholar]
- 8.Beccalli E. Does IT investment improve bank performance? Evidence from Europe. J Bank Finance. 2007;31(7):2205–2230. doi: 10.1016/j.jbankfin.2006.10.022. [DOI] [Google Scholar]
- 9.Berger AN. The economic effects of technological progress: evidence from the banking industry. J Money Credit Bank. 2003;35(2):141–176. doi: 10.1353/mcb.2003.0009. [DOI] [Google Scholar]
- 10.Berger AN, Humphrey DB. Efficiency of financial institutions: international survey and directions for future research. Eur J Oper Res. 1997;98(2):175–212. doi: 10.1016/S0377-2217(96)00342-6. [DOI] [Google Scholar]
- 11.Berger AN, Mester LJ. Inside the black box: What explains differences in the efficiencies of financial institutions? J Bank Finance. 1997;21(7):895–947. doi: 10.1016/S0378-4266(97)00010-1. [DOI] [Google Scholar]
- 12.Bhattacharyya A, Pal S. Financial reforms and technical efficiency in Indian commercial banking: a generalized stochastic frontier analysis. Rev Financ Econ. 2013;22(3):109–117. doi: 10.1016/j.rfe.2013.04.002. [DOI] [Google Scholar]
- 13.Bhattacharyya A, Lovell CK, Sahay P. The impact of liberalization on the productive efficiency of Indian commercial banks. Eur J Oper Res. 1997;98(2):332–345. doi: 10.1016/S0377-2217(96)00351-7. [DOI] [Google Scholar]
- 14.Blundell R, Bond S. Initial conditions and moment restrictions in dynamic panel data models. J Econom. 1998;87(1):115–143. doi: 10.1016/S0304-4076(98)00009-8. [DOI] [Google Scholar]
- 15.Camanho AS, Dyson RG. Cost efficiency measurement with price uncertainty: a DEA application to bank branch assessments. Eur J Oper Res. 2005;161(2):432–446. doi: 10.1016/j.ejor.2003.07.018. [DOI] [Google Scholar]
- 16.Charnes A, Cooper W, Rhodes E. Measuring the efficiency of decision-making units. Eur J Oper Res. 1978;2(6):429–444. doi: 10.1016/0377-2217(78)90138-8. [DOI] [Google Scholar]
- 17.Chen CW, Lin JB, Yi B. CEO duality and firm performance: an endogenous issue. Corp Ownersh Control. 2008;6(1):58–65. doi: 10.22495/cocv6i1p6. [DOI] [Google Scholar]
- 18.Dao BTT, Nguyen KA. Bank capital adequacy ratio and bank performance in Vietnam: a simultaneous equations framework. J Asian Finance Econ Bus. 2020;7(6):39–46. doi: 10.13106/jafeb.2020.vol7.no6.039. [DOI] [Google Scholar]
- 19.Dar AH, Mathur SK, Mishra S. The efficiency of Indian banks: a DEA, malmquist and SFA analysis with bad output. J Quant Econ. 2021;19(4):653–701. doi: 10.1007/s40953-021-00247-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Das A, Ghosh S. Financial deregulation and efficiency: an empirical analysis of Indian banks during the post reform period. Rev Financ Econ. 2006;15(3):193–221. doi: 10.1016/j.rfe.2005.06.002. [DOI] [Google Scholar]
- 21.Delgado J, Hernando I, Nieto MJ. Do European primarily Internet banks show scale and experience efficiencies? Eur Financ Manag. 2007;13(4):643–671. doi: 10.1111/j.1468-036X.2007.00377.x. [DOI] [Google Scholar]
- 22.DeYoung R. The performance of internet- based business models: evidence from the banking industry. J Bus. 2005;78(3):893–947. doi: 10.1086/429648. [DOI] [Google Scholar]
- 23.Ekinci R. The impact on digitalization on financial sector performance. In: Kahyaoğlu SB, editor. The impact of artificial intelligence on governance, economics and finance. I. Singapore: Springer; 2021. pp. 99–119. [Google Scholar]
- 24.Farrel M. The measurement of productive efficiency. J Roy Stat Soc. 1957;120(3):253–281. doi: 10.2307/2343100. [DOI] [Google Scholar]
- 25.Ferrier GD, Lovell CK. Measuring cost efficiency in banking: econometric and linear programming evidence. J Econom. 1990;46(1–2):229–245. doi: 10.1016/0304-4076(90)90057-Z. [DOI] [Google Scholar]
- 26.Floros CH, Giordani G (2008) ATM and banking efficiency: the case of Greece.***
- 27.Goddard JA, Molyneux P, Wilson JO. European banking: efficiency, technology and growth. Wiley; 2001. [Google Scholar]
- 28.Grosskopf S, Hayes K, Yaisawarang S. Measuring economies of diversification: a frontier approach. J Bus Econ Stat. 1992;10(4):453–459. [Google Scholar]
- 29.Hersugondo H, Anjani N, Pamungkas ID. The role of non-performing asset, capital, adequacy and insolvency risk on bank performance: a case study in Indonesia. J Asian Finance Econ Bus. 2021;8(3):319–329. [Google Scholar]
- 30.Holden K, El-Bannany M. Investment in information technology systems and other determinants of bank profitability in the UK. Appl Financ Econ. 2004;14(5):361–365. doi: 10.1080/0960310042000211623. [DOI] [Google Scholar]
- 31.Humphrey DB, Kim M, Vale B. Realizing the gains from electronic payments: Costs, pricing, and payment choice. J Money Credit Bank. 2001;33:216–234. doi: 10.2307/2673882. [DOI] [Google Scholar]
- 32.Hung CS, Yen DC, Ou CS. An empirical study of the relationship between a self-service technology investment and firm financial performance. J Eng Tech Manag. 2012;29(1):62–70. doi: 10.1016/j.jengtecman.2011.09.006. [DOI] [Google Scholar]
- 33.Jacobs R. Alternative methods to examine hospital efficiency: data envelopment analysis and stochastic frontier analysis. Health Care Manag Sci. 2001;4(2):103–115. doi: 10.1023/A:1011453526849. [DOI] [PubMed] [Google Scholar]
- 34.Jui-Chu L, Jin-Li H, Kang-Liang S. The effect of electronic banking on the cost efficiency of commercial banks: an empirical study. Int J Manag. 2005;22(4):605. [Google Scholar]
- 35.Kahveci E, Wolfs B. Digital banking impact on Turkish deposit banks performance. Banks Bank Syst. 2018;13(3):57–58. [Google Scholar]
- 36.Karray SC, Eddine Chichti J. Bank size and efficiency in developing countries: intermediation approach versus value added approach and impact of non-traditional activities. Asian Econ Financ Rev. 2013;3(5):593–613. [Google Scholar]
- 37.Kocisova K, Gavurova B, Sopko J. Do more cards and terminals guarantee higher efficiency? The case of European Union banking. J Int Stud. 2018;11(2):49–62. doi: 10.14254/2071-8330.2018/11-2/4. [DOI] [Google Scholar]
- 38.Kumar SS. Total factor productivity of Indian banking sector-impact of information technology. Reserve Bank oIndia Occas Pap. 2013;34(1):66–86. [Google Scholar]
- 39.Le TD, Ngo T. The determinants of bank profitability: a cross-country analysis. Central Bank Rev. 2020;20(2):65–73. doi: 10.1016/j.cbrev.2020.04.001. [DOI] [Google Scholar]
- 40.Mor S, Gupta G. Artificial intelligence and technical efficiency: the case of Indian commercial banks. Strateg Chang. 2021;30(3):235–245. doi: 10.1002/jsc.2406. [DOI] [Google Scholar]
- 41.Nazaritehrani A, Mashali B. Development of E-banking channels and market share in developing countries. Financ Innov. 2020 doi: 10.1186/s40854-020-0171-z. [DOI] [Google Scholar]
- 42.Ou CS, Hung SY, Yen DC, Liu FC. Impact of ATM intensity on cost efficiency: An empirical evaluation in Taiwan. Inf Manag. 2009;46(8):442–447. doi: 10.1016/j.im.2009.08.005. [DOI] [Google Scholar]
- 43.Pasiouras F. Estimating the technical and scale efficiency of Greek commercial banks: the impact of credit risk, off-balance sheet activities, and international operations. Res Int Bus Financ. 2008;22(3):301–318. doi: 10.1016/j.ribaf.2007.09.002. [DOI] [Google Scholar]
- 44.Rakhe PB. Profitability of foreign banks vis-à-vis other bank groups in India–a panel data analysis. Reserve Bank India Occas Pap. 2010;31(2):49–67. [Google Scholar]
- 45.RBI . RBI Annual report. Department of economic policy and research. Reserve Bank of India; 2001. [Google Scholar]
- 46.Sapci A, Miles B. Bank size, returns to scale, and cost efficiency. J Econ Bus. 2019;105:105842. doi: 10.1016/j.jeconbus.2019.04.003. [DOI] [Google Scholar]
- 47.Saroy R, Awasthy S, Singh NK, Adki S, Dhal S. Impact of COVID-19 on Digital Payment Habits of Indian Households. BEMP. 2022 doi: 10.21098/bemp.v25i0.1823. [DOI] [Google Scholar]
- 48.Sathye M. Efficiency of banks in a developing economy: the case of India. Eur J Oper Res. 2003;148(3):662–671. doi: 10.1016/S0377-2217(02)00471-X. [DOI] [Google Scholar]
- 49.Sathye S, Sathye M. Do ATMs increase technical efficiency of banks in a developing countries? Evidence from the Indian banks. Aust Account Rev. 2017;27(1):101. doi: 10.1111/auar.12110. [DOI] [Google Scholar]
- 50.Sawant BS. Technological developments in Indian banking sector. Indian Streams Res J. 2011;1(9):1–4. [Google Scholar]
- 51.Sayani H, Kishore P, Kumar V. Internal determinants of return on equity: case of the UAE commercial banks. Bank Financ Rev. 2017;9(1):47–74. [Google Scholar]
- 52.Tone K. A strange case of the cost and allocative efficiencies in DEA. J Oper Res Soc. 2002;53:1225–1231. doi: 10.1057/palgrave.jors.2601438. [DOI] [Google Scholar]
- 53.Valverde SC, Humphrey DB, Del Paso RL (2004) Electronic payments and ATMs: changing technology and cost efficiency in banking. In: 25th SUERF Colloquium Volume.
- 54.Wang MH, Huang TH. A study on the persistence of Farrell's efficiency measure under a dynamic framework. Eur J Oper Res. 2007;180(3):1302–1316. doi: 10.1016/j.ejor.2006.04.043. [DOI] [Google Scholar]
- 55.Wang Y, Xiuping S, Zhang Q. Can fintech improve the efficiency of commercial banks? An analysis based on big data. Res Int Bus Financ. 2021;55:101338. doi: 10.1016/j.ribaf.2020.101338. [DOI] [Google Scholar]
- 56.Weill L. Measuring cost efficiency in European banking: a comparison of frontier techniques. J Prod Anal. 2004;21:133–152. doi: 10.1023/B:PROD.0000016869.09423.0c. [DOI] [Google Scholar]
- 57.Xiaogang C, Skully M, Brown K. Banking efficiency in China: application of DEA to pre-and post-deregulation eras: 1993–2000. China Econ Rev. 2005;16(3):229–245. doi: 10.1016/j.chieco.2005.02.001. [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Banks’ balance sheet data were accessed from the Reserve Bank of India’s (RBI) ‘Statistical Tables Relating to Banks in India’. Data on number of bank branches were from the RBI’s Branch Banking Statistics’. Both are annual publications available on the RBI website (https://rbi.org.in/Scripts/publications.aspx). Data on payment systems (NEFT, ATM, PoS and Cards) are from the monthly Data Releases titled ‘Bankwise Volumes in NEFT/RTGS/Mobile Transactions/Internet Banking Transactions’ (https://rbi.org.in/scripts/NEFTView.aspx) and Bankwise ATM/PoS/Card Statistics (https://rbi.org.in/scripts/ATMView.aspx) available on the RBI website.