Impact of Liquidity on the Efficiency of Banks in India Using Panel Data Analysis

Abstract

The current study investigates the impact of the liquidity coverage ratio (LCR) on the efficiency of Indian banks for the period 2010 to 2019. The study examines the effect of internal bank elements like ownership structure, transparency and disclosure, and technological advancement on the relationship between the LCR and efficiency. Bank efficiency proxied as technical efficiency is evaluated by applying the data envelope analysis approach. Applying the panel data regression technique, the authors discover that the LCR has a positive impact on the technical efficiency at a constant return to scale of banks. The relationship between the LCR and the technical efficiency at a variable return to scale is non-linear. Initially, as liquidity increases, the efficiency of banks improves, after reaching its optimum level, efficiency starts to decline. Furthermore, liquidity tends to improve efficiency of banks with higher promoter stakes, whereas opposing results are evidenced for institutional investors and technological advancement.

1. Introduction

The last two decades evidenced a global surge of events leading to multiple economic meltdowns. This not only raised the question of having the proper governance and controls on institutions that affect the economy but also highlighted the burning issue of what aspects of these institutions can be better regulated. Regulators were called to action, and an array of global regulations started emerging. Many regulatory bodies started the arduous exercise of creating regulations that can define, control, and prevent further events leading to such meltdowns. Post the financial crisis of 2007–2008, the Basel Committee on Banking Supervision (BCBS) introduced several new macro-prudential regulations. One of the key components of this new regulatory framework is the LCR.

The LCR focuses on liquidity risk and stipulates that banks maintain sufficient liquid holdings to fulfil their thirty-day obligations if crisis circumstances arise. This allows liquidity regulations to play a fundamental role in preventing bank failures and promoting the resiliency and stability of the financial sector. However, it is equally essential to observe that a stricter liquidity regime can hinder the efficient functioning of banks. This can happen as banks may respond to the higher liquidity requirements by recoursing to riskier investment opportunities, thereby circumventing the regulation and impacting the economy’s performance (Jalilian et al. 2007).

It would be prudent to follow here that while the liquidity standard lays out thorough and exact prescriptive rules, banks may still struggle when complying with the same. The regulation may specify how much liquid assets banks need to maintain, but if these guidelines do not highlight the implicated risks, they could unknowingly cause banks to carry too much or too less liquid assets. Insufficient liquidity increases the risk of a bank’s insolvency, while excessive liquidity enforces redundant costs resulting in an inefficient banking system.

An efficient banking system is essential as it works as the backbone of a nation’s economy. The banking efficiency has its meaning in many terms and the most important one is the economy efficiency (EE). EE occurs with minimizing of the input cost to produce goods and services (Goyal et al. 2019; Sakouvogui and Shaik 2020). Technical efficiency (TE) is concerned with maximum utilization of available resources to achieve certain outputs (Goyal et al. 2019; Sakouvogui and Shaik 2020). Hence, TE is the foremost important condition to EE (Goyal et al. 2019; Sakouvogui and Shaik 2020) of banks.

Despite the important role that the new liquidity standard plays in the efficient operation of the banking system, very few studies document the relationship between the LCR and bank efficiency (Bitar et al. 2015). Another factor that adds to the gap is that these studies typically depend on the standard accounting ratios of bank efficiency (Chortareas et al. 2012; Pasiouras 2008), which as per Halkos and Salamouris (2004), may not be the most effective approach to evaluate efficiency levels of banks. Further, existing literature indicates that bank efficiency is impacted by several institution and market-specific factors (Jiménez-Hernández et al. 2019). So, it can be argued that any research conducted on bank efficiency would be incomplete unless it examined how the aforementioned components played a part. The fact that existing studies do not examine how the relationship between the LCR and bank efficiency changes in response to these variables leads to the requirement of research that deep dives into the said subject.

The current study makes an attempt to bridge the aforementioned research gaps by examining how the LCR influences banks’ technical efficiency (TE) in India. The study also envisages the changes in this association under the impact of bank-specific aspects like ownership structure, transparency and disclosure (T&D) practices followed by banks, and technological advancement initiatives of banks. The Indian banks are of particular significance in investigating the effect of the new LCR on bank efficiency. Because prior to the banking reforms of 1991, Indian banks suffered from numerous inefficiencies on account of stringent regulations in place. The reforms were introduced to enhance banks’ productivity, profitability, and efficiency through a comprehensive de-regulation program. Since the reforms, the Indian banking system has witnessed tremendous improvement in its efficiency levels, enhancing the stability and resiliency of the overall economy. However, in light of the recent stricter regulatory directives, it is essential to reassess the present efficiency situation. This study will contribute to achieving this objective by probing how effectively Indian banks have responded to the new liquidity policy.

This paper significantly contributes to the current corpus of literature in numerous ways. First, the investigation enhances the body of knowledge on bank efficiency by undertaking a thorough study that uncovers the relationship between the LCR and the efficiency of Indian banks.

Second, the research adds to the expanding literature on the BASEL III Liquidity framework by assessing the effect of the LCR on Indian banks’ technical efficiency. The efficiency measure provides more extensive evidence of bank operational efficiency than traditional financial ratios because it reviews performance in a single statistic, thereby controlling the differences amongst banks through an advanced multidimensional framework (Berger and Humphrey 1997).

Third, the study augments the literature on the effect of the BASEL III standard on the banking industry of developing economies by performing the study for India, one of the fastest-growing economies. The findings of this study, combined with other related research conducted in the field, can guide national supervisors on how global liquidity regulation can be implemented in respective economies to achieve the desired efficiency levels. Further, the study’s structure lays down the groundwork for performing similar research in other intricate geographies comparable to India.

From a policy standpoint, it is crucial to comprehend the type of impact the new LCR legislation has had on the efficiency levels of banks, given that it has just recently been introduced in India. Additionally, the study findings will assist national supervisors in understanding the impact of the LCR on the broader banking sector and direct them on how to move forward with the reforms.

The following sections make up the remainder of this paper. The hypothesis is formulated, and the literature is reviewed in Section 2. The research methodology is described in Section 3. Section 4 discusses the empirical findings. Section 5 and Section 6 focus on discussion and concluding remarks, respectively.

2. Literature Review and Hypothesis Development

While very few studies evidence the affinity between liquidity regulations and technical efficiency, we can find some literature that examines the association of liquidity with bank efficiency in general, which is discussed in this section.

In economic terms, efficiency is critical. TE and economic efficiency (EE) are two important notions that differ greatly from one another (Goyal et al. 2019; Sakouvogui and Shaik 2020). TE occurs when it is impossible to enhance output without increasing input. EE occurs when the input cost of an output is as low as possible. When it comes to TE, it means that available resources are converted into services and goods with little waste. The resources are well exploited for the production of the goods or services here (Goyal et al. 2019; Sakouvogui and Shaik 2020). Hence, TE is a prerequisite for EE. This suggests that in order to obtain EE, one must first achieve TE. Only by improving TE, the impact cost can be improved.

Akhter (2018) studies the effect of liquidity on the operating efficiency of banks in Bangladesh for the period 2011 to 2016. The author finds that an increase in liquid assets has a detrimental impact on the operational efficiency of banks. Further, the author advises that after holding a minimum level of liquid assets, banks should focus on utilizing their borrowings and customer deposits towards making a sound loan portfolio to enhance their operational efficiency. In another study, Sakouvogui and Shaik (2020) examine the significance of liquidity and solvency risk factors on efficiency measures of US commercial banks from 2005 to 2007. The study results confirm that liquidity has an adverse impact on the cost efficiency of banks. It implies that higher liquidity of banks lowers the bank’s cost efficiency and the lower liquidity improves the bank’s cost efficiency as it signals that cost should be cut down for a better liquidity position.

Another piece of literature reflects on the mixed results of liquidity on bank efficiency. Altunbas et al. (2007) investigated the link between capital, risk, and inefficiency for European banks from 1992 to 2000. They found that while the banking system liquidity negatively correlates with bank efficiency, the system cost-income ratios tend to be positively related to bank-specific cost efficiencies. Similarly, Alam (2012) examines how the new BASEL III norms influenced the efficiency of banks operating under the dual banking system. The findings show that the relationship between liquidity and efficiency is negative for Islamic banks, whereas the exact relationship is positive for conventional banks. Bitar et al. (2015) received results similar to Alam (2012) while studying the association between liquidity and TE for Islamic and conventional banks. Chupradit et al. (2021) argue that liquidity risk reduces TE of Pakistan’s banks. However, Odunga et al. (2013) and Akhter (2018) argue that the negative correlation between liquidity and efficiency in banks is not always true. They indicate that liquidity is important to have smooth business operations to improve efficiency. Hence, liquidity may positively correlate to efficiency. Algeri et al. (2022) indicate that liquidity of banks have volatility effects on TE.

The conclusions drawn from the aforementioned reasons are that liquidity reserves affect bank efficiency, though the direction of this relationship might vary across geographies or banking systems. In order to further capture the impact for banks in India, we propose the below hypothesis regarding the interrelation between the LCR and TE of banks.

Hypothesis 1 (H1).

LCR has a significant impact on the efficiency of banks.

Hypothesis 2 (H2).

LCR nonlinearly associates with the efficiency of banks.

On the ownership side, we find that most of the existing research on banking and ownership patterns concentrates on examining the impact of foreign vs. domestic-owned banks (Lin et al. 2016; Lensink et al. 2008) or private vs. public banks (Chen 1998; Figueira et al. 2009). However, studies explaining the impact of varying composition of ownership forms (promoters vs. institutional investors) on different banking parameters are relatively scarce. The study aims to fill this gap by analyzing how the ownership structure of banks is instrumental in governing the association between the LCR and TE.

Very few studies examine how ownership structure affects the manner in which liquidity and efficiency relate to one another. However, there are existing studies that document the impact of varying ownership configurations on the performance of banks, and the same would form the basis of our hypothesis formulation for efficiency.

Barry et al. (2011) highlight that the risk-averse nature of promoters helps build resilient banks by reducing banks’ participation in risky investments. Similarly, Iannotta et al. (2007), in their study on the impact of promoter holdings for European banks, find that though higher promoter holdings have no impact on the profitability of European banks, their focus on the asset-quality aids in lowering the bank’s insolvency risk. Contrarily, Rastogi et al. (2021) report the opposite findings for India. They demonstrate that the higher promoter stakes have no bearing on non-performing asset levels; however, they tend to impact banks’ profitability negatively.

The explanation above demonstrates how promoter holdings can affect the different performance parameters. Hence, the below hypothesis is developed to determine how promoter stakes affect the LCR and bank efficiency.

Hypothesis 3 (H3).

Promoters’ stakes have a significant impact on the liquidity and bank efficiency association.

Similar to research on promoter stakes, conflicting evidence supports the link between institutional investor holdings and the performance of banks. According to Saghi-Zedek (2016), banks with more considerable institutional holdings benefit from their specialized knowledge resulting in improved performance.

However, according to Barry et al. (2011), institutional investors are more likely to favor riskier investment options, which can jeopardize the stability of banks and subsequently affect the economy. Similar findings were made by Rastogi et al. (2021), who found that Indian banks often suffer from NPA problems when institutional investors control the ownership structure.

With the above stated empirical evidence, we understand that institutional investors have a crucial impact on the bank’s decisions and can accordingly influence their efficiency.

Hypothesis 4 (H4).

Institutional investor holdings have a significant impact on the liquidity and bank efficiency association.

BCBS (2006) necessitates banks disclose financial and non-financial information to the public to enable them to measure the well-being of the banking sector. Disclosure intends to boost the public and investors’ confidence in the banking system, which can have a favorable impact on the efficiency of banks. There are quite a few studies that explore the relationship between T&D and bank efficiency. Lutfi et al. (2014) exhibit that T&D improves the efficiency of Indonesian banks as the public is willing to deposit money in the banks even at the lower interest rate offered, leading to a decreased cost of capital for banks. This advances the operational efficiency of banks. In another study by Oino (2019), results highlight that T&D motivates banks to have better risk management practices, translating into a higher efficiency level. The results aligned with the findings of Farvaque et al. (2012) and Akhigbe et al. (2017), who suggest that the profit efficiency of banks witnesses an improvement under the impact of T&D.

The existing literature points out that T&D has a significant positive impact on the efficiency of banks. Accordingly, we articulate the below hypothesis for the study.

Hypothesis 5 (H5).

T&D positively impacts the association between liquidity and bank efficiency.

Technological advancement in the form of enhanced investment in information and technology brings added prominence to efficiency. These improvements act in two ways. First, they increase the productivity of banks by introducing cost-effective measures. Second, they improve the competitiveness of banks in the market. With a growing emphasis on information communication and technology (ICT) in the banking sector, various studies investigate the relationship between ICT and bank efficiency.

Stella (2010) undertakes the study of 13 Nigerian banks to understand how banks’ efficiency changes in response to an increase in ICT investment. The author finds that the efficiency of banks improves post the adoption of ICT and thus recommends that the government offers adequate assistance to local banks in acquiring the latest technology. A recent paper by Lee et al. (2021) finds similar results for China and highlights that fintech development significantly improves banks’ cost efficiency. Further, empirical evidence for Australian banks is no different from that of Nigeria and China, where Salim et al. (2010) conclude that the technical efficiency of banks shows significant improvement in response to technological progress. In another study, Casolaro and Gobbi (2007) show that both the profit and cost efficiency of Italian banks become better with IT capital accumulation.

As showcased above, study results from across geographies support the investment in ICT as it positively impacts the efficiency levels of banks.

Hypothesis 6 (H6).

Investment in ICT positively impacts the association between liquidity and bank efficiency.

3. Data and Methodology

3.1. Data

This study executes the panel data of 31 banks running in India for a ten year period (2010–2019). After the post-reform period, the Indian banking sector has experienced many reforms to have efficient banking in India. Therefore, the sample time frame is legitimate for exploring fresh evidence on efficient banking operations. The study considers only 31 banks owing to the availability of synchronized and sufficient data for reliable outcomes. The CMIE prowess database is the primary source of data retrieval. The variables used for the study are elaborated on in

Table 1. List of Variables.

SN	Variable	Type	Symbol	Definition	Citations
1	Technical efficiency	DV	TE (CRS_TE and VRS_TE)	It measures the effectiveness of input resources to get maximum output. Two variants of TE (CRS_TE and VRS_TE) are calculated using DEA. Please see Appendix A.2 for detail.	Goyal et al. (2019); Cooper et al. (2000)
2	LCR	EV	LCR	It is the ratio of HQLAs to expected Net cash flows.	(Hartlage 2012)
3	Promoters’ ownership	MV	po	It represents the share of promoters’ holdings in the bank’s ownership structure.	Kanoujiya et al. (2021); Rastogi et al. (2021)
4	Institutional investors	MV	ii	It represents the share of institutional holdings in the bank’s ownership structure.	Kanoujiya et al. (2021); Rastogi et al. (2021)
5	Transparency and disclosure (T&D)	MV	td	It shows the level of transparency and disclosure of information by a bank. A T&D index is developed for its measurement. The higher index value means a higher level of T&D. Please see Appendix A.1.	Arsov and Bucevska (2017); Kamal Hassan (2012)
6	Information, communication, and technology	MV	ICT	It is the expenditure on ICT by a bank to enhance technology. The amount is in INR (crore).	Agbolade (2011); Aliyu and Tasmin (2012)
7	Assets	CV	lasset	It indicates the size of banks. The natural log is considered for uniformity.	Jayadev (2013); Rastogi et al. (2021),
5	Sales	CV	lsales	It shows the firm’s sales value. The amount is shown in INR. The natural log is considered for uniformity.	Dias (2013); Jayadev (2013)

Note: CV is the control variable, MV is the moderating variable, EV represents the explanatory variable, and DV is the dependent variable.

.

3.2. Methodology

The paper beholds the association of the LCR with TE (bank efficiency). The panel data analysis (applying static models) is employed to examine the association as it carries both the attributes of cross-section and time series. Panel data analysis delivers more information and variability to ensure reliable results (Hsiao 2007; Baltagi 2006). These models are also a better choice to deal with endogeneity and heterogeneity issues; hence, regression results are unbiased (Hsiao 2007; Baltagi 2006). Additionally, the study also investigates these associations under different conditions incorporating linear association (base model) for Hypothesis 1, non-linear association (quadratic model) for Hypothesis 2, and moderating association (under ownership, T&D, and ICT using interaction models) for Hypothesis 3, 4, 5, and 6. There are 12 models to test six hypotheses. There are two models for each hypothesis considering two proxies of efficiency (i.e., CRS_TE and VRS_TE). The conceptual model for such an association of variables is given in Figure 1.

The models applied are as below:

Base_Models (for testing Hypothesis 1):

Model 1—

$${\mathit{CRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\gamma }_1{\mathit{lasset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(1)

Model 2—

$${\mathit{VRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\gamma }_1{\mathit{lasset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(2)

Quadratic_Models (for testing Hypothesis 2):

Model 3—

$${\mathit{CRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{dLCR}}_{\mathit{it}}+{\beta }_2{\mathit{dLCR}2}_{\mathit{it}}+{\gamma }_1{\mathit{lasset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(3)

Model 4—

$${\mathit{VRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{dLCR}}_{\mathit{it}}+{\beta }_2{\mathit{dLCR}2}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(4)

Interaction_Models (for testing Hypothesis 3, 4, 5,and 6):

Model 5 (for testing Hypothesis 3)—

2SLS specification (first_stage):

$${\widehat{i\_LCR\_po}}_{it}=\widehat{{\beta }_0}+\widehat{{\beta }_1}{LCR}_{it}+\widehat{{\beta }_2}{\mathit{po}}_{\mathit{it}}+\widehat{{\beta }_3}{\mathit{lasset}}_{\mathit{it}}+\widehat{{\beta }_4}{\mathit{lsales}}_{\mathit{it}}$$

(5)

2SLS specification (second_stage):

$${\mathit{CRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\widehat{i\_LCR\_po}}_{it}+{\beta }_3{i\_LCR\_po}_{\mathit{it}}+{\gamma }_1{\mathit{lasset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(6)

Model 6 (for testing Hypothesis 4)—

$${\mathit{CRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\mathit{ii}}_{\mathit{it}}+{\beta }_3{i\_\mathit{LCR}\_\mathit{ii}}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(7)

Model 7 (for testing Hypothesis 5)—

$${\mathit{CRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\mathit{td}}_{\mathit{it}}+{\beta }_3{i\_\mathit{LCR}\_\mathit{td}}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(8)

Model 8 (for testing Hypothesis 6)—

$${\mathit{CRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\mathit{ICT}}_{\mathit{it}}+{\beta }_3{i\_\mathit{LCR}\_\mathit{ICT}}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(9)

Model 9 (for testing Hypothesis 3)—

2SLS specification (first_stage):

$${\widehat{i\_LCR\_po}}_{it}=\widehat{{\beta }_0}+\widehat{{\beta }_1}{LCR}_{it}+\widehat{{\beta }_2}{\mathit{po}}_{\mathit{it}}+\widehat{{\beta }_3}{\mathit{lasset}}_{\mathit{it}}+\widehat{{\beta }_4}{\mathit{lsales}}_{\mathit{it}}$$

(10)

2SLS specification (second_stage):

$${\mathit{VRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\widehat{i\_LCR\_po}}_{it}+{\beta }_3{i\_LCR\_po}_{\mathit{it}}+{\gamma }_1{\mathit{lasset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(11)

Model 10 (for testing Hypothesis 4)—

$${\mathit{VRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\mathit{ii}}_{\mathit{it}}+{\beta }_3{i\_\mathit{LCR}\_\mathit{ii}}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(12)

Model 11 (for testing Hypothesis 5)—

$${\mathit{VRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\mathit{td}}_{\mathit{it}}+{\beta }_3{i\_\mathit{LCR}\_\mathit{td}}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(13)

Model 12 (for testing Hypothesis 6)—

$${\mathit{VRS}\_\mathit{TE}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{LCR}}_{\mathit{it}}+{\beta }_2{\mathit{ICT}}_{\mathit{it}}+{\beta }_3{i\_\mathit{LCR}\_\mathit{ICT}}_{\mathit{it}}+{\gamma }_1{l\mathit{asset}}_{\mathit{it}}+{\gamma }_2{\mathit{lsales}}_{\mathit{it}}+u_{\mathit{it}}$$

(14)

where TE in the above equations, the dependent variable, has two variants (CRS_TE and VRS_TE) (please see Appendix A for detail on efficiency assessment), LCR is the independent variable. dLCR2 is square_term of LCR (LCR*LCR) under the quadratic models to find nonlinear connection of LCR to efficiency. Prefix ‘d’ is an indication of demean value. i_LCR_po under cap is the first stage output in 2SLS, and that value is passed to the second stage to overcome the endogeneity issue due to i_LCR_po. Values under caps include estimated values. The interaction between the LCR and TE under the impact of moderating variables (MV) such as ownership concentration (promoter ownership (po) and institutional_investors (ii)), td (please see Appendix A for finding td proxied by T&D index), and ICT is observed through the interaction terms (IT) (i_LCR_po (LCR × po), i_LCR_ii (LCR × ii), i_LCR_td (LCR × td), i_LCR_ICT (LCR × ICT)). Demean values are taken for multiplication. Since lassets and lsales are crucial in influencing banks’ economic significance in their respective industry and can interfere with efficiency evaluation, they are included as control variables. Table 1 presents a complete list of study variables along with their description. u_it and e_it are error terms, where ‘i’ represents bank at a time ‘t’. β_j is the coefficient where β₀ is constant. γ is the coefficient for control variables. There are 12 models for six hypotheses, two each for CRS_TE and VRS_TE. Models 1 and 2 are base models. Models 3 and 4 correspond to quadratic models for nonlinear establishment. Models 5 to 12 are interaction models to test the LCR and efficiency connection under the influence of moderating variables as discussed above.

4. Empirical Results

4.1. Descriptive Statistics and Multicollinearity

Table 2. Descriptive statistics.

Variables	Mean	SD	Min	Max
CRS_TE	0.8269	0.1700	0.4440	1.0000
VRS_TE	0.8640	0.1152	0.5620	1.0000
LCR	1.3691	0.4802	0.8640	4.1973
lasset	0.2762	0.2483	−0.146	1.4344
lsales	8.8913	1.5425	3.9120	12.6792
po	0.5679	0.3026	0.0000	1.0000
ii (institutional_investors)	0.2460	0.2234	0.0000	0.9860
td	0.5020	0.096	0.0000	0.8431
ICT	332.60	669.70	0.0004	6420.00

Note: SD represents standard deviation, Min is minimum, and Max is maximum.

explains the results of the descriptive analysis. The TE of banks is found to be quite satisfactory as both CRS_TE and VRS_TE have the mean values 0.8269 and 0.8640 (nearer to the maximum). This implies that banks’ resource utilization is reasonable. LCR averages 1.3691, indicating that most Indian banks are LCR compliant. The average scores of lasset (0.2762) and lsales (8.8913) are closer to their maximum, indicating that sample banks have an adequate asset and sales base. The average score of po (promoters’ holdings) 0.5678 is inclined to its maximum, indicating that promoters’ holdings dominate the ownership structure of Indian banks. However, ii (institutional investors) has a mean value of 0.2460, closer to its minimum; hence, it depicts that institutional investors have relatively lower representation in the ownership composition of Indian banks. td has an average score of 0.5020, showing a moderate level of T&D in Indian banks. The average expenditure on ICT is INR 332.00 crore (closer to minimum), indicating that expenditure on ICT is quite low in Indian banking. The standard deviation (SD) of ICT is high in comparison to other variables, and it indicates that the expenditure on technology enhancement varies among banks. The lower SD of other variables shows that these factors remain more or less the same across different sample banks.

The correlation matrix in

Table 3. Correlation Matrix.

Variables	CRS_TE	VRS_TE	LCR	Lasset	Lsales	po	ii	td	ICT	i_LCR_po	i_LCR_ii	i_LCR_td	i_LCR_ICT
CRS_TE	1.000
VRS_TE	0.320 *	1.000
LCR	0.043	−0.073	1.000
lasset	0.054	−0.011	0.791 *	1.000
lsales	0.154 *	0.081	−0.428 *	−0.454 *	1.000
po	−0.021	0.310 *	−0.137 *	−0.370	−0.013	1.000
ii	−0.005	−0.30 *	0.391 *	0.265 *	0.052	−0.755 *	1.000
td	0.068	0.030	−0.354 *	−0.370 *	0.519 *	−0.099	0.070	1.000
ICT	0.116 *	−0.069	−0.186 *	−0.236 *	0.644 *	−0.067	0.181 *	0.311 *	1.000
i_LCR_po	−0.021	0.313 *	−0.033	0.052	−0.079	0.015	−0.71 *	−0.143 *	−0.090	1.000
i_LCR_ii	0.005	−0.305	0.222 *	0.115 *	0.152 *	−0.743 *	0.700 *	0.140 *	0.228 *	−0.737 *	1.000
i_LCR_td	0.062	0.018	−0.291 *	−0.321 *	0.509 *	−0.142 *	0.133 *	0.792 *	0.317 *	−0.179 *	0.191 *	1.000
i_LCR_ICT	0.079	0.018	0.112	0.269 *	0.357 *	0.108	−0.17 *	−0.206	−0.780 *	0.120 *	−0.203 *	−0.217 *	1.000

Note: * is for significance level at 0.05.

shows that the highest significant correlation coefficient is between i_LCR_td and td, which exhibits value of 0.792. It shows a positive correlation between i_LCR_td and td. However, since the correlation coefficient is lesser than 0.80, the problem of multicollinearity does not exist.

4.2. Outcomes of Regression Analysis in Base Models

Table 4. Base Models Result for Linear Association (Static Panel Data Analysis).

	Model 1 DV: CRS_TE (RE)		Model 2 DV: VRS_TE (FE)
	Normal	Robust	Normal	Robust
LCR	0.0511 *** (0.051)	0.0511 * (0.002)	0.0083 (0.735)	0.0083 (0.852)
lasset	−0.0370 * (0.010)	−0.0370 *** (0.071)	-0.0943 * (0.000)	−0.0943 * (0.003)
lsales	0.0525 * (0.000)	0.0525 * (0.000)	0.346 * (0.000)	0.346 * (0.002)
Cons.	0.7302 * (0.000)	0.7302 * (0.000)	1.6885 * (0.000)	1.6885 * (0.000)
F-test (Model) F-test (Fixed effect)	21.76 * (0.000) 1.99 * (0.002)		4.94 * (0.006) 6.77 * (0.000)
BP-test (Random effect)	5.19 ** (0.011)		118.92 * (0.000)
Hausman Test	8.32 (0.608)		19.35 * (0.000)
Wald test for Heteroscedasticity	245.37 * (0.000)		2750.11 * (0.000)
Wooldridge Autocorrelation Test AR (1)	20.279 * (0.000)		5.042 * (0.032)
Sigma_$u_i$	0.453		0.107
Sigma_$v_i$	0.157		0.090
rho	0.076		0.584
R-Square	0.067		0.166

Note: The null for the Wald test is that there is no heteroscedasticity. The null for the Wooldridge test is no autocorrelation (with one lag). The Breusch–Pagan test (BP test) measures the random effect. RE represents random effects models, while FE represents fixed effect models. Sigma_$u_i$ and Sigma_$v_i$, respectively, stand for the variance of the individual impact (in this case, the banks) and the error term. The rho represents the UI-related variance. Parenthesis() depicts the p-value. *, **, *** significant at 0.01, 0.05, and 0.10, respectively.

explains the results of base models (Models 1 and 2) for the linear association. Model 1 looks for the connection between TE and the LCR considering crs_te, while Model 2 considers vrs_te for such linear establishment. Both models have a significant p-value < 0.05 for the Breusch–Pagan and F-test. As a result, the Hausman test is performed to determine whether fixed or random effects would be the appropriate model to be applied. As result, for the Hausman is insignificant p-value (>0.05) for model 1, the random effect model is applied. However, model 2 is suitable with fixed-effect as the Hausman test has a significant p-value (<0.05) for this model. Additionally, utilizing the robust standard error estimates for result interpretation is advised since the results are significant (p-value of <0.05) for the Wald and Wooldridge test (Baltagi 2006).

Model 1 (Table 4) shows that the LCR coefficient is both positive (0.0511) and significant at 1% significance (p-value of 0.002 < 0.01). Hence, the LCR significantly affects efficiency and strengthens it. Both control variables (lasset and lsales) are significant with a p-value < 0.05. However, lasset and lsales are found negative (−0.0370) and positive (0.0525), respectively.

In Model 2 (Table 4), here, the LCR has a positive coefficient (0.008) with an insignificant p-value of 0.852 > 0.100. Hence, the LCR and TE are not significantly connected. Like Model 1, the lasset and lsales are significant control variables with negative and positive coefficients, respectively.

4.3. Quadratic Models for the Non-Linear Relationship

Model 3 and Model 4 examine the non-linear relationship between the LCR (using the square term dLCR2) and TE. As per

Table 5. Quadratic Models Result for Non-linear Association (Static Panel Data Analysis).

	Model 3 DV:CRS_TE (RE)		Model 4 DV: VRS_TE (FE)
	Normal	Robust	Normal	Robust
dLCR	0.0493 * (0.002)	0.0493 * (0.002)	0.681 * (0.003)	0.0681 * (0.003)
dLCR2	0.0091 (0.264)	0.0091 (0.264)	−0.0357 ** (0.014)	−0.0357 ** (0.014)
lasset	−0.0418 ** (0.040)	−0.0418 ** (0.040)	−0.0953 * (0.001)	−0.0953 * (0.001)
lsales	0.0510 * (0.000)	0.0510 * (0.000)	0.0356 * (0.001)	0.0356 * (0.001)
Cons.	0.8694 * (0.000)	0.8694 * (0.000)	1.6891 * (0.000)	1.6891 * (0.000)
F-test (Model) F-test (Fixed effect)	16.51 * (0.000) 1.77 * (0.009)		8.32 * (0.006) 6.46 * (0.000)
BP-test (Random effect)	5.19 ** (0.011)		117.40 * (0.000)
Hausman Test	4.41 (0.220)		15.73 * (0.000)
Wald test for Heteroscedasticity	232.67 * (0.000)		2585.96 * (0.000)
Wooldridge Autocorrelation Test AR (1)	20.321 *(0.000)		4.763 *(0.037)
Sigma_$u_i$	0.044		0.097
Sigma_$v_i$	0.159		0.089
rho	0.072		0.544
R-Square	0.093		0.183

Note: Same as explained in Table 4. The turning point in Model 4 for quadratic association is ‘0.953’. Turning point is calculated as (−β₁/2*β₂) (Kanoujiya et al. 2023). β₁ and β₂ are coefficients of dLCR and dLCR2, respectively.

, Models 3 and 4 follow the random-effect and fixed-effect approaches, respectively, due to the reason discussed earlier for Models 1 and 2. The robust estimates are taken in these models because both autocorrelation and heteroscedasticity exist.

Similar to Models 1 and 2, dLCR2 (square of LCR) does not have a significant coefficient, even at 10% significance in Model 3 (Table 5). It corroborates that the LCR and TE are not connected nonlinearly as well. However, both control variables are significant, as in earlier models.

On considering VRS_TE, Model 4 has a significant coefficient (p-value < 0.05) with a negative value (−0.357) for dLCR2. It means that the LCR is related nonlinearly to TE in the shape of an inverted U. It denotes that the LCR enhances efficiency to an optimum level with turning point 0.953 in the early stages after then it drops down the efficiency. Both control variables lasset and lsales are significant with a negative and a positive value, respectively.

4.4. Interaction Models for Moderating Relationship (for CRS_TE)

Models 5 to Model 8 are the interaction models that examine the impact of the LCR on TE (crs_te) under po (promoters’ ownership in Model 5), ii (institutional investors in Model 6), T&D (td) (in Model 7), and ICT (in Model 8). The 2SLS regression analysis is adapted for Model 5 due to the presence of endogeneity (Baltagi 2006). The rest of these models follow either the FE or RE per the Hausman test indication. The outcomes of these models are mentioned in

Table 6. Interaction Models Result (Static Panel Data Analysis).

	Model 5 DV: CRS_TE MV: pro (2SLS)	Model 6 DV: CRS_TE MV: ii (RE)	Model 7 DV: CRS_TE MV:td (RE)	Model 8 DV: CRS_TE MV:ICT (FE)
	Estimates	Estimates	Estimates	Estimates
i_LCR_po	0.0016 *** (0.099)	-	-	-
i_LCR_ii	-	−0.0015 ** (0.013)	-	-
i_LCR_td	-	-	0.1748 (0.529)	-
i_LCR_ICT	-	-	-	−0.0002 ** (0.069)
LCR	0.084 *** (0.089)	0.1208 * (0.001)	0.6841 ** (0.047)	0.2040 * (0.002)
MV(moderating_variable)	−0.0007 *** (0.066)	−0.0001 (0.832)	0.0181 (0.873)	−0.0000 (0.270)
lasset	-	−0.0396 * (0.004)	−0.0383 *** (0.066)	0.0028 (0.941)
lsales	-	0.0530 * (0.000)	0.0527 * (0.000)	0.0606 * (0.002)
Cons.	0.7405 (0.000)	0.6076 (0.376)	0.7233 * (0.001)	−0.0193 (0.967)
F-test (Model) F-test (Fixed effect)	6.83 (0.077) 2.50 * (0.000)	31.50 * (0.000) 1.85 * (0.005)	31.93 * (0.000) 9.23 * (0.000) 1.93 * (0.003) 2.18 * (0.000)
BP-test (Random effect)	11.85 * (0.003)	3.99 ** (0.022)	4.24 ** (0.019) 5.46 * (0.009)
Hausman Test	16.95 * (0.000)	8.02 (0.161)	9.10 (0.105) 12.73 * (0.026)
Wald test for Heteroscedasticity	-	-	247.99 * (0.000) 252.91 * (0.000)
Wooldridge Autocorrelation Test AR (1)	-	-	18.272 * (0.000) 22.416 * (0.000)
Sigma_$u_i$	0.025	0.035	0.046 0.105
Sigma_$v_i$	0.177	0.156	0.157 0.156
rho	0.019	0.048	0.080 0.311
R-Square	0.036	0.097	0.081 0.112

Note: Same as explained in Table 4.

.

The i_LCR_po exhibits a positive coefficient (0.0016) which is significant at 10% (p-value 0.099 < 0.100). This indicates that the LCR promotes efficiency under a higher degree of promoters’ holding (Table 6). Both the LCR and po individually are found significant at 10%. The LCR is positive, and po is negative for TE.

In Model 6, the coefficient of i_LCR_ii is negative (−0.0015) with a significant p-value of 0.013 < 0.05 (Table 6). This implies that the LCR is detrimental for TE when there are higher holdings of institutional investors. The LCR is significant and positive (0.1208), individually showing the LCR increases TE. However, ii is found insignificant in Model 6 (Table 6). Additionally, both control variables are significant at 5%. lasset is negative, and lsales is positive. On observing Model 7 (Table 6), the i_LCR_td does not show a significant coefficient; hence, it implies that the LCR does not affect TE under the influence of the T&D level. However, the LCR is found to be significant and positive (0.6841) at 5%. Both control variables are significant (as in Model 6). The i_LCR_ICT coefficient is negative (−0.0002) and significant at 10% significance (Model 8 in Table 6). It means that the LCR diminishes efficiency when technology enhances in Indian banks. The LCR is individually significant at 5% with a positive value of 0.2040, which signals that the LCR helps improve efficiency. The ICT is found insignificant at a 10% significance level. Hence, technology has no significant role in efficiency improvement. Only lsales is found significant and positive for TE among the control variables.

4.5. Interaction Models for Moderating Relationship (for VRS_TE)

There are four models (Model 9 to Model 12) which take TE as VRS_TE (

Table 7. Interaction Models Result (Static Panel Data Analysis).

	Model 9 DV: VRS_TE MV: po (2SLS)	Model 10 DV: VRS_TE MV: ii (FE)	Model 11 DV: VRS_TE MV:td (FE)	Model 12 DV: VRS_TE MV:ICT (FE)
	Estimates	Estimates	Estimates	Estimates
i_LCR_po	0.0000 (0.979)	-	-	-
i_LCR_ii		−0.0020 * (0.007)
i_LCR_td	-	-	0.1634 (0.409)	-
i_LCR_ICT	-	-	-	0.0002 (0.944)
LCR	0.0283 (0.326)	−0.0086 (0.808)	−0.0073 (0.872)	−0.0071 (0.904)
MV(moderating_variable)	0.0003 (0.274)	−0.0010 ** (0.046)	−0.3112 (0.216)	−0.0000 (0.432)
lasset	------	−0.0837 * (0.003)	−0.0881 * (0.004)	−0.0904 * (0.004)
lsales	------	0.0393 * (0.000)	0.0389 * (0.001)	0.0418 * (0.002)
Cons.	0.8600 * (0.000)	1.5567 * (0.000)	1.5783 * (0.000)	1.5872 * (0.000)
F-test (Model) F-test (Fixed effect)	2.89 (0.409) 3.77 * (0.000)	8.50 * (0.000) 5.11 * (0.000)	4.59 * (0.003) 4.39 * (0.004) 6.67 * (0.000) 6.49 * (0.000)
BP-test (Random effect)	50.27 * (0.000)	71.42 * (0.000)	104.99 * (0.000) 109.95 * (0.000)
Hausman Test	6.73(0.081)	18.15 * (0.002)	26.59 * (0.000) 29.79 * (0.000)
Wald test for Heteroscedasticity	-	573.43 * (0.000)	402.23 * (0.000) 361.42 * (0.000)
Wooldridge Autocorrelation Test AR (1)	-	4.68 ** (0.038)	5.005 ** (0.032) 5.254 ** (0.029)
Sigma_$u_i$	0.066	0.091	0.107 0.103
Sigma_$v_i$	0.096	0.089	0.090 0.089
rho	0.320	0.512	0.585 0.571
R-Square	0.044	0.194	0.177 0.180

Note: Same as explained in Table 4.

). Like Model 5, Model 9 also follows the 2SLS model due to the endogeneity issue. A fixed-effect model is employed as the Hausman test shows a significant p-value for Models 10, 11, and 12. The robust estimates are shown for these three models in Table 7 as either autocorrelation or heteroscedasticity, or both are present in these models.

Neither the coefficient of the LCR nor the coefficient of po is found significant at 10% (Table 7). The interaction term (i_LCR_po) is also not significant in Model 9. In Model 10, i_LCR_ii has a negative (−0.0020) and significant coefficient with a p-value of 0.007 < 0.010. Hence, it indicates that the LCR reduces efficiency under higher holdings of institutional investors. The LCR is not significant individually. However, ii is significant in isolation showing a negative impact on TE. Both control variables are significant, having a p-value < 0.05. lasset is negative (−0.0837), but lsales is positive (0.0393) for TE. The coefficient of i_LCT_td (interaction term) is insignificant. Hence, no significant interaction effect is found between the LCR and TE under T&D. The LCR and td are also insignificant individually at a 10% significance level. Both lasset and lsales are significant, having negative and positive coefficients. The i_LCR_ICT has an insignificant (10%) and positive (6.49e-06) coefficient. Therefore, the LCR does not affect efficiency under technology enhancement. Here again, in this model, the LCR and ICT are insignificant individually for TE. Both the control variables are significant, with a negative coefficient for lasset and a positive coefficient for lsales.

4.6. Endogeneity and Robustness

Table 8. Endogeneity Test (crs_te).

	LCR	dLCR2	i_LCR_po	i_LCR_ii	i_LCR_td	i_LCR_ICT	ICT	Td
Durbin Chi-2	0.5589 (0.4547)	0.2234 (0.6364)	6.1698 * (0.0130)	3.7350 (0.0533)	0.1678 (0.6820)	1.3489 (0.2455)	0.3313 (0.5649)	0.6746 (0.4114)
Wu–Hausman Test	0.5474 (0.4602)	0.2185 (0.6406)	6.2040 * (0.0135)	3.7129 (0.0553)	0.1641 (0.6858)	1.3260 (0.2508)	0.3242 (0.5697)	0.6611 (0.4171)

Note: The p-value value in () * represents a significant value at a 5% significance level.

and

Table 9. Endogeneity Test (vrs_te).

	LCR	dLCR2	i_LCR_po	i_LCR_ii	i_LCR_td	i_LCR_ICT	ICT	Td
Durbin Chi-2	0.3533 (0.5532)	0.2208 (0.6384)	6.1698 * (0.0130)	3.7350 (0.0533)	0.1678 (0.6820)	1.3489 (0.2455)	0.3313 (0.5649)	0.6746 (0.4114)
Wu–Hausman Test	0.3457 (0.5571)	0.2159 (0.6426)	6.2040 * (0.0135)	3.7129 (0.0553)	0.1641 (0.6858)	1.3260 (0.2508)	0.3242 (0.5697)	0.6611 (0.4171)

Note: The p-value in (). * represents a significant value at a 5% significance level.

report the outcome of the endogeneity test. All models except Model 5 and Model 9 do not have the endogeneity problem due to the exogenous variable (i_lcr_op) for endogenous variables (CRS_TE and VRS_TE). To check for endogeneity problems, the Durbin Chi2 and Wu–Hausman tests are used (Baltagi 2006). The results of both tests support the null hypothesis of no endogeneity with insignificant p-values. Models 5 and 9 show a significant p-value < 0.05 for these two models. In many cases, the association between the LCR and TE is found to be significant. Thus, similar results are revealed, ensuring robust results.

5. Discussion of Results

5.1. Hypothesis Validation and Comparison with Existing Studies

The study has two primary objectives. First, to analyze how the LCR affects the banks’ technical efficiency. Second, to examine how this affiliation interacts with other bank characteristics like T&D, investment in ICT, and ownership composition (promoter vs. institutional holdings).

The study’s results indicate that the LCR positively influences banks’ technical efficiency while considering the CRS approach. Hence, evidence is supportive for the first hypothesis (H1). The findings are consistent with those of Alam (2012) and Bitar et al. (2015), who suggest that bank efficiency tends to improve with increased liquidity. It is observed that banks make prudent investment decisions and build a more stable and reliable revenue-generating asset pool when operating under high liquidity requirements. This leads to enhanced efficiency as banks become more effective in their role as financial intermediaries through better capital mobilization and allocation.

When studying the association of liquidity with efficiency via VRS, the authors find the existence of an inverted U-shaped relationship. This suggests that initially, as liquidity increases, the efficiency of banks improves; however, beyond a point, as more and more liquidity is added, the efficiency starts taking a hit. Thus, hypothesis H2 has supportive evidence. The explanation for this can be found in the existing literature, which highlights that eventually, the higher costs associated with each additional unit of high-quality liquid assets start outweighing their relatively low return (Le et al. 2020; Bordeleau and Graham 2010; Tran et al. 2016), resulting in a decline in efficiency.

Turning to understand how the affiliation between the LCR and TE interacts with the ownership structure of banks, the authors find that the LCR promotes efficiency under high promoter holdings. However, the study shows opposing results for institutional investors, where the LCR tends to be detrimental to TE of banks as the stake of institutional investors increases. The outcome is consistent with the formulated hypothesis that ownership structure plays a considerable role in altering the relationship between the LCR and TE of banks. Hence, both hypotheses H3 and H4 cannot be refused.

While very scarce research is available that investigates the association between ownership and bank efficiency in light of liquidity, we have studies that evidence the impact of promoter vs. institutional holdings on a bank’s key efficiency parameters. Iannotta et al. (2007) and Barry et al. (2011) emphasize that high promoter holdings reduce the default and insolvency risk of banks. Therefore, a high promoter holding works in conjunction with the LCR to enhance the stability and resiliency of Indian banks by reducing liquidity risk. This explains how the TE of banks with high promoter stakes becomes better through evolved risk management practices adopted by banks while conforming with the new directives on liquidity.

On the other hand, banks with a higher representation of institutional investors are understood to indulge in higher risks (Barry et al. 2011). Such banks are tempted to take on more risk as the cushion of liquidity increases with the implementation of the LCR. However, excessive risk assumed by banks can lead to higher levels of non-performing assets, consequently impacting bank efficiency.

Interestingly, the results demonstrate that T&D is not significant in explaining the relationship between liquidity and efficiency, and thus our hypothesis (H5) that T&D positively impacts the relationship between liquidity and bank efficiency is rejected.

Regarding technological advancement, the results demonstrate that the LCR lowers the efficiency of banks as technology enhances in Indian banks. This is contrary to our expectations that investment in ICT would positively impact the relationship between liquidity and bank efficiency. Hence, hypothesis H6 is rejected. Such findings can assist policymakers and regulators in assessing both sides of the equation while making policies/changes.

5.2. Contribution and Implications

The current study has a significant contribution to the existing knowledge relating to bank studies. The findings deliver novel and reliable evidence on the empirical connectivity of bank liquidity and TE in Indian banking.

The first implication of this study is towards the regulators and policymakers. The liquidity conundrum needs monitoring as, after the initial period of boosting efficiency, the further infusion of liquidity leads to efficiency decline. It is essential to ascertain the potential factors attributing to the turning point in the relationship between liquidity and efficiency in order to factor in these when further policy re-alignment is undertaken.

The ownership structure is the subject of the second implication. In light of the study’s findings, the authors recommend a diverse ownership structure with the ideal distribution of promoters and institutional investors. Leveraging the promoters’ active risk management techniques and the essential experience of institutional investors would make it easier for banks to realize a more stable and beneficial effect of the liquidity regime on the efficiency of banks.

The third implication of this study relates to the fundamental principle of corporate governance reflected via T&D. One of the primary objectives of any banking regulation is to boost the resiliency of the banking sector by improving its overall corporate governance practices (Alexander 2006). This implies that corporate governance and regulations together have a critical influence on the efficiency of banks. However, the study results indicate that the same does not hold true for Indian banks. Therefore, it is imperative to understand the plausible factors that hamper this association in India and work towards stabilizing it to ensure meaningful regulation impacts.

The fourth and last implication of the investigation concerns technical advancement. Since technological advancement is an absolute necessity given the ever-evolving nature of the banking sector, it is crucial to investigate how the mandate for adequate liquidity and fintech development can be aligned to generate desired bank efficiency.

6. Conclusions

Although several countries have embraced liquidity regulations, empirical evidence of the impact of the LCR on bank efficiency is scarce. Provided that the national regulators in India vehemently promote the implementation of liquidity regulations, focalizing this issue is long overdue. By examining the influence of the LCR on bank technical efficiency and how this connection fluctuates under the moderating impact of bank characteristics from an emerging economy perspective, this study intends to abridge this identified gap in the literature on liquidity restrictions and bank efficiency.

Using the panel data analysis, this study inspects the impact of the LCR on the technical efficiency of Indian banks from 2010 to 2019. When studying the CRS_TE, we find that the LCR positively impacts banks’ efficiency. On the other hand, for VRS_TE, findings show that initially, as liquidity increases, the efficiency of banks improves; however, after reaching its peak, the efficiency starts deteriorating as more liquidity is introduced into the system. On ownership structure, the results suggest that if the banks have higher promoter holdings, the LCR will enhance the bank’s efficiency. On the contrary, if the banks have a high concentration of institutional investors, the efficiency of banks tends to suffer as the liquidity increases. However, T&D is not significant in explaining the relationship between liquidity and efficiency. As far as technological advancement is concerned, the results demonstrate that the LCR lowers the efficiency of banks as technology enhances in Indian banks.

Given the statistical significance of this study, the authors believe that the results would help better understand the nature of liquidity-impacting variables and contribute significantly to the existing corpus of studies. From a policy perspective, this research paves the way to further assist in refining the current policies and contribute to formalizing the next set of policies. For Indian banks and the economy as a whole, the liquidity policymaking direction needs to be tailored given the unique economic and geographic conditions. This study contributes in this regard by bringing to the fore explicit factors that can be considered during the policymaking process.

The study also opens up doors of discussion to the other attributes that might have a similar relationship with liquidity. The authors recognize that owing to the investigation’s internal nature, external market factors that might impact the liquidity structures of respective banks are not thoroughly studied. This is a limitation of the study that requires more investigation. In terms of future scope, the combination of internal factors (non-performing assets, leverage ratio) along with market factors (peer liquidity, prevailing market conditions) can be probed to evaluate the holistic effect of liquidity on the efficiency of banks. The authors also observe the shortcomings of this study in assessing the qualitative and non-modelable risk factors that are an inherent part of a bank’s risk ecosystem. Further research could be undertaken to cover these crucial factors.