What Explains Excess Liquidity of Banks? Empirical Evidence from India

Abstract

We study excess liquidity in the banking system using data for India during 2005–2020. We apply Autoregressive Distributed Lag model and panel regressions to identify the factors determining excess liquidity at both aggregate and bank levels. We find that required reserves, private sector credit, and government securities held by banks have negative, positive, and negative effects on excess liquidity, respectively. Other factors such as exchange rate and inter-bank call rate have varying effects at the two levels. Our results suggest that banks can chalk out mechanisms to optimize their liquidity management and avoid the cost of excess liquidity.

JEL Classifications: C23, E50, E58, G00, G21

Keywords

Excess liquidity required reserves exchange rate autoregressive distributed lag panel regression

1. Introduction

Excess liquidity is the difference between actual reserves maintained by banks and the central bank and statutory reserves (Omer et al., 2015). High level of excess liquidity can cause unfavorable effects in an economy, such as price instability and ineffectiveness of monetary policy transmission. For instance, Nissanke and Aryeetey (1998) and Agénor and El Aynaoui (2010) argue that effectiveness of monetary policy is hampered by the prevalence of excess liquidity in developing economies. Several other studies have also pointed out that excess liquidity works as an impediment for the intended monetary policy (Agénor et al., 2004; Khemraj, 2007; Saxegaard, 2006). These studies used aggregated banking data to examine the demand and supply components of excess liquidity. Being a developing economy, India is no exception to the potential problems of excess liquidity. In this article, we examine excess liquidity of Indian banks and identify factors which influence it. This is the first study to analyze the determinants of excess liquidity, both at the bank level and the aggregate level using bank balance sheets and macroeconomic data, respectively

We chose to limit our study to India for the following reasons. First, India is an important emerging economy that is also the world’s sixth largest and the fastest growing major economy. Second, analyzing a panel of countries at bank level would be difficult due to very different macroeconomic and institutional conditions. Studying one country has the advantage of relative homogeneity in the sample that can lead to generalizable results. While previous studies have examined the determinants of excess liquidity either at the aggregate level or at bank level, our contribution is the study of Indian banking system using both aggregate- and bank-level data and a comparison of these with policy implications.

There are mainly two ways to explain the high level of excess liquidity in banks, that is, structural and cyclical (Omer et al., 2015; Saxegaard, 2006). Structural explanations range from the lack of financial development and high-risk aversion, causing banks to hold large reserves (Agénor et al., 2004; Mohanty et al., 2006; Wyplosz, 2003). Cyclical explanations focus on the long-term effects of macroeconomic instability due to factors such as inflation or policy changes in the capital account of the country (Agénor & El Aynaoui, 2010; Zhang & Pang, 2008). Excess liquidity with banks could have an important implication for corporate governance as it might lead to a dilution of prudential norms in loan disbursal, which puts investors’ interest at risk. This primarily happens due to the pressure to utilize excess liquidity for some productive use rather than holding on it. Kaminsky and Reinhart (1998) find that a high level of excess liquidity leads to an increase in bank credit, which makes banks vulnerable to exogenous shocks and financial instability. Similarly, Darvas and Pichler (2018) find that excess liquidity with banks might incentivize banks to engage in risky lending, which further creates financial stability risks. On the other hand, Odonkor et al. (2016) find that banks in Ghana prefer high excess reserves when they perceive the market to be risky.

Excess liquidity held by banks has important implications for the pass-through of monetary policy signals. Excess liquidity works as a buffer, which is freely available to banks, that can be used for lending in the event of liquidity constraint caused by a revision of the monetary policy. Saxegaard (2006) decomposes excess liquidity into voluntary and involuntary components and shows that the latter has implications for the monetary transmission mechanism as it is an undesired liquidity imposed from outside. Other studies such as Agénor and El Aynaoui (2010) and Nguyen and Boateng (2015) also find that banks with large excess reserves are less vulnerable to monetary policy shocks.

Most of the existing literature examines the impact of monetary policy on lending by banks with liquidity deficit, which is a norm in the banking system as liquidity deficit is desired for effective monetary policy transmission (https://core.ac.uk/download/pdf/6961926.pdf). However, the literature does not analyze in detail situations of surplus liquidity, which is a characteristic of developing countries such as India. Surplus liquidity may hinder or facilitate the effectiveness of monetary policy in India.¹ However, there are no studies that estimate the extent of excess liquidity in India, its (voluntary and involuntary) components and determinants. In this article, we examine the prevalence of and factors that influence excess liquidity in the Indian banking system. A novelty of this article is that we first examine the factors which determine excess liquidity in the banking system as a whole with aggregate banking data. Second, we examine the determinants of excess liquidity at bank level with banks’ balance sheet data. Thus, our article. contributes to the literature by examining the determinants of excess liquidity at both macro (aggregate) as well as micro (bank) levels, which enhances the scope and policy relevance of the results. Our study provides insights for an effective transmission of monetary policy and for minimizing the opportunity cost of excess liquidity held by banks.

We find numerous bank-level as well as macroeconomic factors that determine excess liquidity in India. For instance, required reserves have a positive and significant effect on excess liquidity at the bank level as well as at the aggregate level. At the aggregate level, excess liquidity reacts negatively to the exchange rate and positively to the interbank call rate. However, at the bank level, exchange rate has a positive effect while the call rate has a negative effect on excess liquidity. This difference in the exchange rate’s effect on excess liquidity is because of heterogeneity in the response of different banks. At an individual bank level, unrestrained use of forex is not allowed, forcing some banks to hedge for currency depreciation. This leads them to increase their liquidity holding in the face of currency depreciation. At the aggregate level, the decline in excess liquidity may be driven by large banks who may not worry about depreciation, either due to ample forex reserves available with the RBI, or because they already have hedged positions. Instead, their response may be to switch their portfolio from liquid assets to supporting export credit. This response may get reflected in the aggregate level analysis. In the case of call rate, the negative effect of weighted average call money rate (WACR) at bank level could be because of some banks that draw their reserves to turn lenders in the interbank market. However, large banks may be enjoying a surge in deposits due to higher interest rates, which end up as higher reserves. This behavior may dominate in the aggregate level analysis.

The remainder of this article is structured as follows. In Section 2, we present a brief overview of the literature on excess liquidity Section 3 explains the data and reports the methodology. Section 4 presents the analysis of excess liquidity using aggregate data while in Section 5 we discuss the determinants of excess liquidity at bank level. Finally, Section 6 concludes the study.

2. Literature Review

High level of excess liquidity has been explored from both structural and cyclical points of view (Omer et al., 2015; Saxegaard, 2006). In the case of structural view, there are two reasons for the presence of excess reserves. First, low financial development may compel banks to hold large reserves due to unreliable payment systems, high costs of processing information, high costs of monitoring borrowers, etc. Second, high-risk aversion by banks can lead to high-risk premia and low credit demands, which then leads to excess reserves held by banks. These factors support our choice of India as an empirical setting since it has low levels of financial development with high-risk aversion exhibited by state-owned banks. Empirical evidence on the structural view can be found in Wyplosz (2003) for Euro area, Agénor et al. (2004) for Thailand, and Mohanty et al. (2006) for a panel of countries, who argue that accumulation of excess liquidity is due to weak credit demand. From the cyclical point of view, an increase in inflation leads to uncertainty about the value of collateral pledged by the borrowers; consequently, banks may either increase the premium or ration the credit. In both cases, excess liquidity increases significantly (Agénor & El Aynaoui, 2010).

While the above-mentioned sources of excess liquidity are endogenous in nature, there are also exogenous sources such as policy changes. A large amount of capital flows facilitated by a liberalized capital account and intermediated by the banking system also contributes to rise in excess liquidity in developing countries. Asymmetric opening (that is, lifting of restriction on capital movement for non-residents while controlling foreign exchange operation by residents) of capital accounts by developing countries accompanied by privatization of large state-owned enterprises have led to large capital inflows (Khemraj, 2007). India is one such country where the capital account is partially convertible with liberal inflows but there are restrictions on outward investments by residents. In both managed float and pegged foreign exchange rate regimes, restricted inflows of capital (with sterilization of capital flows through exchange rate regulation) act as a deterrent to mopping up the liquidity in the economy.² In the absence of these checks, the monetary base blows up, which results in surplus liquidity in the economy. During 2006–2008, countries such as India, China, and South Korea increased their required reserve ratios so that excess liquidity could be absorbed and thereby addressed the destabilization in macroeconomic scenario (Agénor & El Aynaoui, 2010).

There are a few other studies as well which examine the sources of excess liquidity in developing economies (Hasanovic & Latic, 2017 for Bosnia and Herzegovina; Pontes & Sol Murta, 2012, for Cape Verde; Nwakanma & Sc, 2014; Ukeje & Amanze, 2015, for Nigeria). Such developing economies suffer from structural deficiencies in the banking system, which strengthens the argument that developing countries are more prone to excess liquidity issues (Saxegaard, 2006). However, it should be noted that the case of India has been under‑researched so far.

The consequences of excess reserves encourage banks to lend more and take more risks (Acharya & Naqvi, 2012; Nguyen & Boateng, 2015). Darvas and Pichler (2018) examined excess liquidity and bank lending risk in the Euro area and found that excess liquidity might create financial instability because banks can indulge in risky lending.

Excess liquidity has implications for monetary policy, which is also a dimension we include in our analysis. An optimal amount of liquidity with banks is necessary for monetary policy to bring intended changes in the macroeconomic variables in an economy. Both excess liquidity and liquidity deficit can jeopardize the intended impact of monetary policy. Voluntary and involuntary excess liquidity have different implications for monetary transmission (Saxegaard, 2006). The former relates to liquidity management (excess reserves provide insurance against possible increase in reserve requirements and signal liquidity strength to customers, etc.), which does not have major implications for monetary policy transmission while the latter is undesired liquidity that is externally imposed and has implications for the monetary transmission mechanism.

Agénor and El Aynaoui (2010) show in a theoretical model that excess liquidity weakens the impact of monetary policy. An empirical support for this comes from Nguyen and Boateng (2015), who show that Chinese banks with larger involuntary excess reserves are less vulnerable to monetary policy shocks. On the other hand, Demiralp et al. (2021) study banks in the Euro area to find that excess liquidity and reliance on retail deposit funding may increase banks’ responsiveness to a negative interest rate policy. The above studies show that while there is a theoretical support for the role of excess liquidity in impeding effective transmission of monetary policy, the evidence is mixed. We add to the evidence by offering the first comparative analysis of aggregate data with bank-level data for India, for which excess liquidity has not been analyzed before.

3. Data

We use data from the Reserve Bank of India (RBI) website. We have two sets of variables: one for aggregate-level estimation and another for bank-level estimation. We use fortnightly data for all aggregate variables and annual data for bank-level variables, both spanning the period 2005–2020. Excess liquidity is measured as the difference between actual reserves maintained by banks with the central monetary authority (RBI) and required statutory reserves mandated to be maintained with RBI as a precautionary need (Omer et al., 2015). Statutory reserves in India are cash reserve ratio (CRR) and statutory liquidity ratio (SLR). CRR is a portion of net demand and time liabilities (NDTL) of banks to be mandatorily kept with the RBI in the form of liquid assets (4% of NDTL as of ending in March 2021). SLR is a percentage of NDTL kept by banks in the form of liquid cash, government securities, government-approved other securities, or gold (18.25% of NDTL as of ending in March 2021). We take the ratio of excess liquidity to total deposits as the dependent variable.

The explanatory variables are mainly based on two papers: Saxegaard (2006) and Omer et al. (2015). The list of variables along with their descriptive statistics and their descriptions are presented in Table A1. We first explain the variables used in the aggregate‑level estimation. We include required reserves (measured as the ratio of required reserves to total deposits) as it has a direct negative effect on excess liquidity. The central bank’s discount rate is proxied by the repo rate, which can have a negative effect on excess liquidity. Exchange rate captures the exchange rate risk (Pontes & Sol Murta, 2012). Interbank liquidity risk is proxied by volatility in WACR as it makes banks more cautious about managing their liquidity holdings. Hence it is expected to have a positive effect on excess liquidity. Private sector credit captures the non-food credit (Pontes & Sol Murta, 2012; Wyplosz, 2003) as, in India, food credit is determined by the government’s procurement policy. Index of Industrial Production (IIP) is a common measure for production activities in India. The higher the IIP, the more the excess liquidity, because increase in economic activities increases money demand. Hence, banks are expected to maintain large liquidity holdings. able A1. Description of Variables.

Volatility in RBI credit to the government increases the volatility of current deposits with banks, indicating better management of precautionary liquidity holdings by banks. RBI credit to the government results in new money creation and this increases the banks’ deposits, which further increases the excess liquidity holdings with banks. Commercial banks’ credit to the government may lead to drawing down of excess liquidity with banks but the effect may be transient because it is just a matter of time when banks are replenished with almost the same amount of money once the government spends it (Omer et al., 2015).

Government securities indicate government borrowings from sources other than the central bank and commercial banks. We also use the ratio of demand deposits to total deposits as a larger proportion of demand deposits to total deposits obliges banks to maintain higher level of liquidity. Figure 1 shows the trends in required reserves and excess liquidity in the Indian banking system over the period of 2005–2020, where an overall declining trend in required reserves can be observed, while excess liquidity declined till 2008–2009 and since then has shown an upward movement. Figure 1 shows the prevalence of excess liquidity in Indian banking as ranging between 3.7% and 17.1%.

Figure 1.

Source: The authors.

For the bank-level estimation, we have again followed Saxegaard (2006) and Omer et al. (2015) along with a few other papers in choosing the variables. While many of the variables are common to the aggregate-level analysis, we also use the following additional variables: WACR captures the banks’ short-term borrowing cost. Banks may have to hold more excess liquidity if WACR goes up. Cash-deposit ratio is included because fluctuations in demand for cash as a percentage of deposits have an effect on excess liquidity. Higher the ratio, higher the excess liquidity. Banks need to know it precisely for their day-to-day transactions. We use the ratio of internal debt to GDP to observe the crowding out of liquidity from commercial banks. Larger domestic debt of the government reduces the liquidity of banks. Hence, we expect a negative relationship between excess liquidity and internal debt of the government. The ratio of demand to saving deposits captures the banks’ need for more cash or liquid assets if demand deposits are more in their liabilities, so an unexpected increase in withdrawal from current account deposits can be honored (Moussa, 2015). We include output gap to capture the demand for cash because there would be less demand for cash when there is a cyclical gap in the economy. The ratio of total advances by banks to GDP (Hasanovic & Latic, 2017) measures the magnitude of lending by banks in the economy and it is expected to show that more advances would result in less excess liquidity with the banks.

In some cases, empirical considerations led us to choose slightly different but comparable variables for aggregate and bank-level analysis. For instance, we have taken volatility in WACR for the aggregate-level analysis while for the bank-level analysis, we have taken WACR. This is because, for aggregate analysis, the frequency of data is fortnightly, and therefore it is feasible to take a 5-period (fortnight) moving average. However, in the case of bank-level analysis, the data is annual, which makes it impractical to compute a moving average. For a similar reason, the volatility in RBI credit to the government is not included in the bank-level analysis.

Instead of taking private sector credit in bank-level analysis, we have taken the ratio of total advances to GDP because of availability of the data with the same frequency. RBI does not provide data on private sector credit at bank level on its website. For a similar reason, we have included RBI advances to commercial banks instead of banks’ credit to government in the bank-level analysis. It may be noted that the former has an opposite effect on excess liquidity compared to the latter because RBI advances to commercial banks is an asset for the central bank but liability for the commercial banks. To capture the business cycle, we have used IIP at aggregate level since it is available at a monthly frequency, and it took us only one level of interpolation to make it fortnightly. In our bank-level analysis, we have used the output gap estimated by the usual Hodrick–Prescott filter based on annual GDP.

We use the ratio of demand deposits to total deposits for the aggregate-level analysis instead of the ratio of demand deposit to savings deposit used at bank level because at the aggregate level, RBI does not provide separate data for savings bank deposits. To capture monetary policy, we have used the repo rate (a proxy for discount rate used by Omer et al., 2015) in the aggregate level analysis. However, we do not use it in the bank-level analysis because of the presence of WACR in the model, which is the first leg of monetary policy transmission.

4. Methodology

4.1. Unit Root Test

Before estimating the relationship among time-series variables we test for their stationarity using Augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) unit root tests by estimating the following equation (Pesaran, 2015).

Δ y_{t} = μ_{0} + μ_{1 τ} + p y_{t - 1} + \sum_{p = 1}^{k - 1} y_{p} Δ y_{t - p} + ε_{t}

(1)

where y_t is the series to be tested, μ₀ and μ₁ are parameters while τ is the deterministic trends. ρ and y stand for coefficients of unit root and lagged dependent variable. ε stands for the error term. The null hypothesis for the unit root tests is that the series is non-stationary. When the null hypothesis is rejected, we assume the series to be suitable for analysis at levels, implying a short-term analysis. If the null hypothesis is not rejected, then a long-term analysis is considered appropriate. Later, for the bank-level analysis, we use panel unit root tests to assess the stationarity of the variables.

4.2. Estimation of Excess Liquidity: Aggregate Level

Next, we follow the approach of Saxegaard (2006) and Omer et al. (2015) to study the determinants of excess liquidity in the banking system and its components (voluntary and involuntary excess liquidity). In the first step, we examine the relationship between excess liquidity and its determinants by applying an autoregressive distributed lag (ARDL) model:

E L_{t} = μ + \sum_{j = 1}^{m} \sum_{k = 1}^{n} ψ_{k}^{​} x_{t - k}^{j} + \sum_{k = 1}^{n} α_{k} E L_{t -}^{}_{k}^{} + ς_{t}

(2)

In Equation (2) x_t is the set of regressors, $ψ_{k}^{j}$ are coefficients for any jth regressor at lag k, and αk represents the stickiness in dependent variable at lag k. We apply ARDL for the following reasons. First, since we have a mix of variables that are stationary and non-stationary, ARDL analysis is suitable as it does not impose a restriction of all variables to be integrated of the same order. Second, applying ARDL allows a large number of explanatory variables to be used (Omer et al., 2015). Third, for small samples, ARDL is suitable for examining long-term relationships, which is not the case with other multivariate time series techniques. Finally, ARDL gives unbiased estimates even if some of the independent variables are endogenous (Odhiambo, 2009).

In the second step, we decompose excess liquidity into its voluntary and involuntary components using the following equations:

E L_{}^{s}_{t}^{} = a^{s} \hat{c} + {\hat{α}}_{2} (L) X_{t}^{1}

(3)

E L_{t}^{d} = (l - a^{s}) \hat{c} + {\hat{a}}_{3} (L) X_{t}^{2} + v_{t}

(4)

We use Equations (3) and (4) to separately estimate the voluntary and involuntary excess liquidity, respectively. In the above equations, a^s and (1–a^s) are the intercepts of the voluntary and involuntary components, respectively, which are not distinguishable. However, as we are interested in long-term relationships and long-term coefficients, estimating the separate intercept values is not necessary. Similarly, voluntary and involuntary parts of the lagged dependent variable are also not separable (Omer et al., 2015). From the list of the determinants of overall excess liquidity, we consider determinants of voluntary excess liquidity to be the ratio of required reserves to total deposits, repo rate, volatility in WACR, ratio of demand deposits to total deposits, and volatility in RBI credit to the government. On the other hand, private sector credit, exchange rate, IIP, RBI credit to the government, banks’ credit to the government, and government-dated securities are the determinants of involuntary excess liquidity.

4.3. Estimation of Excess Liquidity: Bank Level

Additionally, we estimate the determinants of excess liquidity at bank level. There are few studies on the determinants of excess liquidity at the bank level and none for India. An exception is Hasanovic and Latic (2017) for Bosnia and Herzegovina. However, they used fewer variables to explain excess liquidity, but in order to be as close as possible with our aggregate level analysis, we have incorporated more variables in our model:

\begin{array}{l} E x c e s s L i q u i d i t y_{i t} = β_{0} + β_{1} Re q u i r e d Re s e r v e s_{i t} + β_{2} D i s c o u n t R a t e_{t} + \\ β_{3} C a s h - D e p o s i t R a t i o_{i t} + β_{4} I n t e r n a l D e b t_{t} + \\ β_{} 5 R B I A d v C o m_{i t} + β_{6} R B I A d v G o v t_{t} + \\ β_{7} E x c h a n g e R a t e_{t} + β_{8} D D S B R a t i o_{i t} + \\ β_{9} O u t p u t G a p_{t} + β_{10} T o t a l A d v G D P_{i t} + \\ β_{11} G o v t S e c u r i t i e s_{t} + ε_{i t} \end{array}

(5)

where RBIAdvCom is RBI advances to Commercial Banks as a ratio of GDP, RBIAdvGovt stands for RBI advances to the government as a ratio of GDP, TotalAdvGDP represents total advances by banks as a ratio of GDP and DDSB Ratio is the ratio of demand to saving deposits.

We estimate the above equation using standard panel regression and dynamic panel regression methods (GMM), adding a lag of the dependent variable excess liquidity. We use panel regression and GMM method because of the following reasons. First, they improve the efficiency of estimation compared to pooled analysis of panel data and are more informative. Second, they control for the effect of omitted variables because of the presence of intertemporal and individual entity’s dimensions. Third, they provide the micro foundations of the aggregate data analysis presented in the first part of this article. Finally, the GMM method addresses the endogeneity problem by using internal instruments (Jacob & Lukose, 2018).

5. Results

5.1. Unit Root Test: Aggregate Level

Results from the unit root tests (not reported to save space) confirmed that all variables except excess liquidity, volatility in WACR, private sector credit, IIP, volatility in RBI credit to the government, and RBI credit to the government are stationarity at first difference (while considering ADF test with drift only). This indicates that there is a possibility of estimating the long-term relationship among the variables. Hence, we proceed with the estimation applying ARDL.

5.2. Long-term Determinants of Excess Liquidity: Aggregate Level³

In Table 1, based on the ARDL model, we report the long-term determinants of excess liquidity. We observe that required reserves has a negative and significant effect on excess liquidity implying that as required reserves (comprising of CRR and SLR) increase, the holdings of excess liquidity of banks decreases. The effect of repo rate (discount rate) on excess liquidity is positive and significant, which means that as discount rate increases, banks increase their liquidity holdings to avoid liquidity deficit. The negative coefficient of exchange rate shows that depreciation of Indian rupee forces banks to decrease their liquidity holdings. Excess liquidity responds positively to the volatility in overnight call money rate (WACR). The coefficient of private sector credit is negative and significant. It shows that as more credit is disbursed to private sector, less excess liquidity is left with banks. IIP has a negative but statistically insignificant impact on excess liquidity. In the aggregate-level analysis, IIP shows a negative effect on excess liquidity, which suggests that an expansion in industrial production leads to more investment demand for money, resulting in less excess liquidity with banks. During an expansionary period (strong credit demand), banks are drained of excess liquidity while during a contractionary period, banks are deluged with excess liquidity (weak credit demand).

Table 1.
Long-term Determinants of Excess Liquidity

Coefficient Standard Error

Required reserves –0.930* 0.167

Repo rate 0.733* 0.184

Exchange rate –4.980* 2.703

Volatility in WACR 2.262 0.988

Private sector credit –18.135* 1.768

IIP –0.016 1.506

Volatility in RBI credit to the government –2.699 3.261

RBI credit to the government –0.069 0.097

Banks’ credit to the government 18.189*** 2.425

Government-dated securities 0.407* 0.207

Ratio of demand deposits to total deposits 0.548* 0.127

Intercept 54.850 22.440

Critical values for I (1) Boundary F-statistics

1% 3.61

5% 3.04

10% 2.77

	Coefficient	Standard Error
Required reserves	–0.930***	0.167
Repo rate	0.733***	0.184
Exchange rate	–4.980*	2.703
Volatility in WACR	2.262**	0.988
Private sector credit	–18.135***	1.768
IIP	–0.016	1.506
Volatility in RBI credit to the government	–2.699	3.261
RBI credit to the government	–0.069	0.097
Banks’ credit to the government	18.189***	2.425
Government-dated securities	0.407*	0.207
Ratio of demand deposits to total deposits	0.548***	0.127
Intercept	54.850**	22.440
Critical values for I (1) Boundary	F-statistics
1%	3.61
5%	3.04
10%	2.77

Source: The authors.

Notes: Dependent variable: Ratio of excess reserves to total deposits (%). Excess Reserves have been calculated following Omer et al. (2015). *, ** and *** indicate significance at 10%, 5% and 1%, respectively.

The coefficients of both RBI credit to the government and volatility in RBI credit to the government are negative but not significant, which alludes to the lack of role of government borrowings from RBI in determining excess liquidity in the banking system. This finding refutes the argument Ganley (2002) made, who said that monetizing government budget deficit is one of the main causes of excess liquidity. Banks’ credit to the government and government-dated securities have positive and significant effect on excess liquidity. It means that relying on these sources of finance by government causes an increase in excess liquidity. The effect of the ratio of demand deposits to total deposits is positive and significant, implying that a larger proportion of demand deposits warrants larger excess liquidity holdings.

5.3. Short-term Determinants of Excess Liquidity: Aggregate Level

Table 2 reports the estimation of short-term determinants of excess liquidity from the ARDL model. It shows the response of excess liquidity to the lagged values of required reserves, exchange rate, volatility in WACR, private sector credit, and banks’ credit to the government.

Table 2.
Short-run Determinants of Excess Liquidity

Contemporaneous Response Lag (1) Lag (2) Lag (3) Lag (4)

Excess liquidity

0.558*
(0.054) 0.176*
(0.060) 0111
(0.053)

Required reserves
–0.369*
(0.068) 0.139
(0.096) –0.066
(0.096) 0.152
(0.069)

Repo rate
0.114*
(0.027)

Exchange rate 2.833
(1.104) –1.104
(1.583) –2.502*
(1.141)

Volatility in WACR 0.279**
(0.101) –0.187*
(0.114) –0.056
(0.113) 0155
(0.112) 0.160
(0.101)

Private sector credit 18.595*
(1.973) 6.844
(2.675) 2.677
(2.614) 6.261***
(2.201)

IIP
–0.002
(0.234)

Volatility in RBI credit to the government 1.675
(1.385) –2.093*
(1.165)

RBI credit to the government –0.011
(0.014)

Banks’ credit to the government 15.653*
(1.178) –1.206
(1.457) –10.502*
(1.561) –2.855
(1.487) 1.730
(1.096)

Government-dated securities 0.063
(0.031)

Ratio of demand deposits to total deposits 0.085*
(0.024)

Intercept
8.504*
(3.125)

Diagnostic tests Statistics P-values

Adjusted R-squared 0.981

F-statistics 678.215 (0.000)

Normality test 742.486 (0.000)

Ramsey reset test 6.776 (0.010)

	Contemporaneous Response	Lag (1)	Lag (2)	Lag (3)	Lag (4)
Excess liquidity		0.558*** (0.054)	0.176*** (0.060)	0111** (0.053)
Required reserves	–0.369*** (0.068)	0.139 (0.096)	–0.066 (0.096)	0.152** (0.069)
Repo rate	0.114*** (0.027)
Exchange rate	2.833** (1.104)	–1.104 (1.583)	–2.502*** (1.141)
Volatility in WACR	0.279** (0.101)	–0.187* (0.114)	–0.056 (0.113)	0155 (0.112)	0.160 (0.101)
Private sector credit	18.595*** (1.973)	6.844** (2.675)	2.677 (2.614)	6.261*** (2.201)
IIP	–0.002 (0.234)
Volatility in RBI credit to the government	1.675 (1.385)	–2.093* (1.165)
RBI credit to the government	–0.011 (0.014)
Banks’ credit to the government	15.653*** (1.178)	–1.206 (1.457)	–10.502*** (1.561)	–2.855** (1.487)	1.730 (1.096)
Government-dated securities	0.063** (0.031)
Ratio of demand deposits to total deposits	0.085*** (0.024)
Intercept	8.504*** (3.125)
Diagnostic tests	Statistics	P-values
Adjusted R-squared	0.981
F-statistics	678.215	(0.000)
Normality test	742.486	(0.000)
Ramsey reset test	6.776	(0.010)

Source: The authors.

Notes: Dependent variable: Ratio of excess reserves to total deposits (%). Standard errors in parenthesis. *, ** and *** indicate significance at 10%, 5% and 1%, respectively.

The effects of various explanatory variables with lags may be responsible for structural persistence in excess liquidity. For instance, required reserves’ effect on excess liquidity is up to three lags and exchange rate has an impact of up to two lags. Similarly, private sector credit has effect on excess liquidity up to three lags.

5.4. Decomposing Excess Liquidity: Aggregate Level

Following the procedure from Saxegaard (2006) and Omer et al. (2015), we use the estimates of Equations (3) and (4), available from the long-term coefficients in Table 1, to decompose total excess liquidity into voluntary and involuntary parts. The outcome is shown in Figure 2.

Figure 2.

Source: The authors.

We observe that there was a spike in voluntary excess liquidity during the year of financial crisis of 2008. After that, there was a decline and since then there has been a steady rise in the voluntary excess liquidity. This rise in the voluntary excess liquidity could be because of the financial crisis of 2008 and implementation of Basel-III norms, which introduced the liquidity coverage ratio (LCR). As a part of post-global financial crisis of 2008 reforms, the Basel Committee on Banking Supervision introduced LCR under which banks are required to keep high-quality liquid assets. Unlike voluntary excess liquidity, involuntary excess liquidity shows a steady decline, especially after 2012–2013. This decline may be because of banks’ offsetting of increase in voluntary excess liquidity, which would have been caused by the stricter banking prudential norms initiated by RBI.

6. Determinants of Excess Liquidity: Bank Level

Analyzing bank-level excess liquidity is important as there are different factors that determine how much liquidity/excess liquidity is maintained by individual banks, which may be different from the aggregate-level factors. Considering the heterogeneous characteristics of banks in India, we study the determinants of excess liquidity holdings at bank level. In an economy, banks are of different sizes, capitalization, and profit, which could have bearings on the maintenance of liquidity among them. For instance, less profitable banks may not find it feasible to keep a large amount of liquidity or excess liquidity due to their limited capacity to withstand opportunity costs such as the return forgone on idle liquid assets. Such effects due to heterogeneity of banks cannot be studied in the aggregate analysis and require a bank-level estimation.

6.1. Panel Unit Root Test: Bank Level

First, we apply four tests for panel unit roots: Im, Pesaran, and Shin (IPS), Levin, Lin, and Chu (LLC), Fisher Phillips–Perron (Fisher PP), and Fisher Augmented Dickey–Fuller (Fisher ADF). Fisher ADF and Fisher PP tests are based on combining the p-values of the underlying ADF and PP statistics, while for LLC and IPS tests, the null hypothesis (non-stationarity) is based on zero value of the p parameter. Following Pesaran (2015), we estimate the equation given below.

Δ y_{i}^{}_{t}^{} = α_{1} + δ_{i t} + p_{i} Y_{i}^{} {_{,}^{}}_{t - 1}^{} + \sum_{l = 1}^{p i} φ_{i l} Δ y_{i}^{} {_{,}^{}}_{t}^{} {_{-}^{}}_{1}^{} + ε_{i t}

(6)

where y_it is the variable value for panel member i in period t, ϵ_it is assumed to be independent and identically distributed IID (0, σ²_ε).

The results from the panel unit root (not reported to save space) revealed that all variables are stationary at level in a majority of the tests, satisfying the condition to proceed for bank-level estimation applying panel regression analysis.

6.2. Determinants of Excess Liquidity: Bank Level

Table 3 reports the factors that influence the excess liquidity holdings at bank level in India. Indian banks are broadly categorized into two groups: public sector banks and private sector banks. As the ownership and policies of these banks are different, the determinants of excess liquidity may not be the same for both. Therefore, we studied the two groups separately as sub-samples. We report the estimation of the determinants of excess liquidity for the bank groups in separate columns in Table 3 for standard panel as well as for GMM regressions. For all banks, we observe from the fixed effects regression estimates (as recommended by the Hausman test) that there are various factors which have a statistically significant effect on excess liquidity holdings with banks.

Table 3.
Determinants of Excess Liquidity: Panel Regression and GMM Regression

All Banks
Public Banks
Private Banks

Fixed Two-step Fixed Two-step Random Two-step

Effects GMM Effects^† GMM Effects† GMM

Excess liquidity_t–1 0.081* 0.279* (0.034)

(0.003) (0.048) (0.010)

Required reserves –0.912* –1.521* –2.848* –1.737* 0.552 0.123

(0.285) (0.077) (0.250) (0.276) (0.524) (0.358)

WACR –0.797* –0.199* –0.622* –0.243* –0.640 –0.390

(0.299) (0.036) (0.219) (0.090) (0.587) (0.165)

Cash deposit ratio 0.329 3.774* 0.550 –1.764 0.884 5.356*

(1.578) (0.531) (1.359) (3.085) (2.766) (1.988)

Ratio of internal debt to GDP –0.985* –0.840* –1.243 –0.784* –0.524 0.138

(0.219) (0.047) (0.171) (0.109) (0.423) (0.198)

RBI advances to commercial banks 1.097 0.382* 0.415 0.217 2.582 0.223

(0.547) (0.066) (0.385) (0.207) (1.107) (0.181)

RBI advances to government –0.377 –0.851 6.061 –1.564 –5.511 –0.985

(4.139) (0.397) (2.901) (1.532) (8.226) (1.167)

Exchange rate 29.715 7.381* 17.054* 8.300* 28.728 13.234*

(12.684) (1.208) (9.160) (3.162) (24.924) (4.230)

Ratio of demand to saving deposits 0.010* 0.005* 0.173* 0.055* 0.007 0.016

Deposit (0.003) (0.002) (0.020) (0.014) (0.005) (0.010)

Output gap –9.322 –5.283 –48.555 15.240 46.296 42.361

(33.527) (5.658) (23.294) (7.788) (67.893) (22.291)

Ratio of total advances to –1.026 –3.156* –0.378 –1.210 –1.395 –11.255*

GDP (0.502) (0.517) (0.309) (1.580) (1.380) (4.326)

Government securities held by banks 3.870* 8.190* 6.632* 8.038* 3.260 3.459

(1.169) (0.657) (1.644) (1.837) (1.432) (1.524)

Intercept –82.429* –32.344*** –20.457 –40.957* 120.069 –75.888***

(47.696) (3.906) (35.505) (25.348) (90.759) (21.494)

R-squared 0.032 0.056 0.075

No. of observations 730 626 399 343 331 283

No. of instruments 117 117 117

Sargan tes 44.618 21.421 10.739

(1.000) (1.000) (1.000)

	All Banks	Public Banks	Private Banks
Effects	GMM	Effects^†	GMM	Effects†	GMM
Excess liquidity_t–1		0.081***		0.279***		(0.034)**
		(0.003)		(0.048)		(0.010)
Required reserves	–0.912***	–1.521***	–2.848***	–1.737***	0.552	0.123
	(0.285)	(0.077)	(0.250)	(0.276)	(0.524)	(0.358)
WACR	–0.797***	–0.199***	–0.622***	–0.243***	–0.640	–0.390**
	(0.299)	(0.036)	(0.219)	(0.090)	(0.587)	(0.165)
Cash deposit ratio	0.329	3.774***	0.550	–1.764	0.884	5.356***
	(1.578)	(0.531)	(1.359)	(3.085)	(2.766)	(1.988)
Ratio of internal debt to GDP	–0.985***	–0.840***	–1.243	–0.784***	–0.524	0.138
(0.219)	(0.047)	(0.171)	(0.109)	(0.423)	(0.198)
RBI advances to commercial banks	1.097**	0.382***	0.415	0.217	2.582**	0.223
(0.547)	(0.066)	(0.385)	(0.207)	(1.107)	(0.181)
RBI advances to government	–0.377	–0.851**	6.061**	–1.564	–5.511	–0.985
(4.139)	(0.397)	(2.901)	(1.532)	(8.226)	(1.167)
Exchange rate	29.715**	7.381***	17.054*	8.300***	28.728	13.234***
(12.684)	(1.208)	(9.160)	(3.162)	(24.924)	(4.230)
Ratio of demand to saving deposits	0.010***	0.005***	0.173***	0.055***	0.007	0.016
Deposit	(0.003)	(0.002)	(0.020)	(0.014)	(0.005)	(0.010)
Output gap	–9.322	–5.283	–48.555**	15.240**	46.296	42.361**
	(33.527)	(5.658)	(23.294)	(7.788)	(67.893)	(22.291)
Ratio of total advances to	–1.026**	–3.156***	–0.378	–1.210	–1.395	–11.255***
GDP	(0.502)	(0.517)	(0.309)	(1.580)	(1.380)	(4.326)
Government securities held by banks	3.870***	8.190***	6.632***	8.038***	3.260**	3.459**
(1.169)	(0.657)	(1.644)	(1.837)	(1.432)	(1.524)
Intercept	–82.429*	–32.344***	–20.457	–40.957*	120.069	–75.888***
	(47.696)	(3.906)	(35.505)	(25.348)	(90.759)	(21.494)
R-squared	0.032		0.056		0.075
No. of observations	730	626	399	343	331	283
No. of instruments		117		117		117
Sargan tes		44.618		21.421		10.739
		(1.000)		(1.000)		(1.000)

Source: The authors.

Notes: Dependent variable: Ratio of excess reserves to total deposits (%). Standard errors in parentheses.*, ** and *** indicate significance at 10%, 5% and 1%, respectively. For panel regression, we report only appropriate model suggested by Hausman Test. ^††p -value in parentheses.

Required reserves have a negative and significant impact on excess liquidity holdings with banks. It means that as the RBI increases the required reserves (CRR and SLR) the excess liquidity drains out from banks to the RBI or into government securities. It corroborates the theoretical prediction of Agénor et al. (2004) that increase in required reserves reduces the excess liquidity and is consistent with the empirical findings of Omer et al. (2015) and Saxegaard (2006) based on their macroeconomic analysis. Excess liquidity responds negatively to overnight call money rate (WACR), which means that monetary policy tightening is associated with lower excess liquidity of banks. The reaction of excess liquidity to the government debt is negative. The RBI advances to commercial banks have a positive effect on excess liquidity holdings with banks, implying that all money that is borrowed by banks from the RBI is not used for lending purposes but may end up as reserves. Exchange rate has a positive impact on excess liquidity. As the Indian rupee depreciates, banks tend to hold more excess liquidity. This implies that banks increase their excess liquidity holdings to hedge for the falling value of foreign currency assets. This is in contrast with Omer et al. (2015), who found a negative effect of foreign exchange on excess liquidity in Pakistan where foreign currency deposits have a larger share in banks’ deposits. This may be because of less foreign currency deposits in Indian banks but higher foreign currency assets, which necessitates banks to hedge for depreciation of Indian rupee.

The ratio of demand deposits to saving deposits has a positive impact on excess liquidity holdings with banks. More the demand deposits with banks, more the requirements for liquid assets to avoid dishonoring withdrawal from saving and current accounts due to a sudden surge in withdrawals (Saxegaard, 2006). Excess liquidity reacts negatively to the ratio of total advances to GDP. When banks lend more, excess liquidity is poised to go down as money is being disbursed from the deposit accounts of banks to the loan accounts of borrowers.Response of excess liquidity to government securities held by banks is positive and significant. Cash-deposit ratio, RBI advances to government, and output gap have, respectively, a positive, a negative, and a negative effect on excess liquidity. However, these responses are not statistically significant. Response of excess liquidity to overnight call money rate (WACR) and exchange rate is negative and positive, respectively, which is in contrast to the findings from the aggregate-level analysis. A possible reason for this could be a cross‑ sectional variation in response at bank level, as we observe that excess liquidity of private sector banks does not respond to these two variables. The results for GMM regression show similar coefficients to panel regressions except for cash-deposit ratio and RBI advances to the government. The former has a positive and significant effect while the latter has a negative a significant impact on excess liquidity (which is in contrast with panel regression, where coefficients had the same sign but were statistically insignificant).

For public sector banks (in panel regression), we find almost similar results as observed in the full sample discussed earlier but for private sector banks only two of the variables have significant coefficients. In particular, we highlight the following differences in the results for the sub-samples: RBI advances to commercial banks have a positive effect on excess liquidity in the case of public sector banks but the coefficient is not statistically significant. It means that borrowings from the RBI do not result in excess liquidity for public sector banks, possibly because of their capital stressed nature due to high non-performing assets. However, for private sector banks, the borrowings from the RBI seem to result in excess liquidity as they are not otherwise capital stressed.

RBI advances to the government has a positive and significant effect on excess liquidity of public sector banks only. When the government borrows from the central bank (through sale of government securities, which are purchased by the RBI in the secondary market), it creates new deposits since borrowed money is credited to the government’s account with RBI. Whenever the government spends out of this account, it is more likely to end up as deposits with public sector banks (through various government agencies who hold accounts in these banks) resulting in excess liquidity. The output gap has a negative and statistically significant effect on excess liquidity of public sector banks due to credit demand as a result of improvement in economic activity. An output gap implies a difference between an economy’s potential output and actual output. The results for bank-level analysis show a negative response of excess liquidity to output gap, suggesting that a booming economy with a lower output gap (one that needs more funds for investments) leads to low levels of excess liquidity,while in the case of a recession, there exists a larger output gap (where investment demand is lower), leading to high levels of excess liquidity.

Excess liquidity responds negatively to the ratio of total advances to GDP, but the response is not statistically significant. The coefficient of government securities held by banks is positive and statistically significant for both public and private sector banks. It means that banks’ investments in these securities are in excess of the required levels and thereby affect the excess liquidity.

For GMM regression, we find almost similar results for public sector banks except RBI advances to government and output gap, where the former becomes negative and insignificant while latter changes to positive and significant (as compared to panel regression). For private sector banks, there is a considerable improvement in the results compared to panel regression. In addition to the variables with significant coefficients in panel regression analysis, we observe the following significant coefficients in case of GMM.WACR and ratio of total advances to GDP are negative and have a statistically significant effect on excess liquidity. Cash-deposit ratio, exchange rate, and output gap have a positive and significant impact on excess liquidity.

7. Conclusions

We have investigated the factors that influence excess liquidity in the Indian banking system at aggregate as well as bank levels. Overall, we identify various bank-level and macroeconomic factors that determine excess liquidity in India. Our results suggest that RBI must design monetary policies in a way to discount excess liquidity so that the amount of adequate liquidity is determined and the cost of having excess liquidity is eliminated. At the aggregate level, a push for private sector credit by banks can be helpful in sorting their problem of excess liquidity. Less investment in government-dated securities and better management of required reserves position (as the repo rate has a positive effect on excess liquidity) are important so that reliance on interbank market is reduced. Besides, understanding the distinction between the persistence of voluntary and involuntary excess liquidity needs to be considered by policymakers as the latter is responsible for the weakening of monetary policy (Saxegaard, 2006). Hence, the focus of policy measures should be to stabilize the cyclical fluctuation in the economy.

Our findings have the following implications for banks and policymakers. At the bank level, internal debt and total advances by banks can be important for tackling the prevalence of excess liquidity as they negatively impact it. Targeting required reserves to tackle the issue of excess liquidity may not be a useful policy measure. Required reserves only transfer the excess liquidity with banks or banking system to the RBI, which is either unproductive or less productive than other potential uses of the liquidity. Besides, reducing RBI advances to commercial banks could act as a deterrence to accumulation of excess liquidity with banks, which is a recurring phenomenon of recent origin. The positive effect of government securities on excess liquidity implies that banks are resorting to less productive use of their funds as they typically hold more than the mandated level of government securities. The regulator must incentivize banks to invest in more productive assets. Our results suggest that managing the funds required for honoring demand deposits is crucial because it leads to maintaining excess liquidity by banks.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Appendix A

Notes

References

Acharya

, & Naqvi

(2012). The seeds of a crisis: A theory of bank liquidity and risk taking over the business cycle. Journal of Financial Economics, 106(2), 349–366. 10.1016/j.jfineco.2012.05.014

Agénor

P. R.

, & El Aynaoui

K. E.

(2010). Excess liquidity, bank pricing rules, and monetary policy. Journal of Banking and Finance, 34(5), 923–933. 10.1016/j.jbankfin.2009.10.003

Agénor

P. R.

, Aizenman

, & Hoffmaister

A. W.

(2004). The credit crunch in East Asia: What can bank excess liquid assets tell us? Journal of International Money and Finance, 23(1), 27–49. 10.1016/j.jimonfin.2003.08.008

Darvas

, & Pichler

(2018). Excess liquidity and bank lending risks in the Euro area [Study for the Committee on Economic and Monetary Affairs of the European Parliament, Luxembourg] (No. 2018/16). https://www.europarl.europa.eu/RegData/etudes/STUD/2018/626069/IPOL_STU(2018)626069_EN.pdf

Demiralp

, Eisenschmidt

, & Vlassopoulos

(2021). Negative interest rates, excess liquidity and retail deposits: Banks’ reaction to unconventional monetary policy in the Euro area. European Economic Review, 136. https://www.ecb.europa.eu/pub/pdf/scpwps/ecb.wp2283~2ccc074964.en.pdf

Ganley

(2002). Surplus liquidity: Implications for central banks. Centre for Central Banking Studies, Bank of England.

Hasanovic

, & Latic

(2017). The determinants of excess liquidity in the banking sector of Bosnia and Herzegovina [No. 11–2,017]. Graduate Institute of International and Development Studies Working Paper. https://www.econstor.eu/bitstream/10419/184714/1/HEIDWP11-2017.pdf

Jacob

, & Lukose

P. J. J.

(2018). Institutional ownership and dividend payout in emerging markets: Evidence from India. Journal of Emerging Market Finance, 17(Suppl 1), S54–S82. 10.1177/0972652717751538

Kaminsky

G. L.

, & Reinhart

C. M.

(1998). Financial crises in Asia and Latin America: Then and now. American Economic Review, 88(2), 444–448.

10.

Khemraj

(2007). Monetary policy and excess liquidity: The case of Guyana. Social and Economic Studies, 56(3), 101–127.

11.

Mohanty

M. S.

, Schnabel

, & Garcia-Luna

(2006). Banks and aggregate credit: What is new? https://www.bis.org/publ/bppdf/bispap28b.pdf

12.

Moussa

M. A. B.

(2015). The determinants of bank liquidity: Case of Tunisia. International Journal of Economics and Financial Issues, 5(1), 249–259.

13.

Nguyen

V. H. T.

, & Boateng

(2015). An analysis of involuntary excess reserves, monetary policy and risk-taking behaviour of Chinese banks. International Review of Financial Analysis, 37, 63–72. 10.1016/j.irfa.2014.11.013

14.

Nissanke

, & Aryeetey

(1998). Financial integration and development: Liberalization and reforms in sub-Saharan Africa. Routledge.

15.

Nwakanma

P. C.

, & Sc

I. M. M.

(2014). Determinants of excess liquidity: The Nigerian experience. International Business Research, 7(6), 174. 10.5539/ibr.v7n6p174

16.

Odhiambo

N. M.

(2009). Energy consumption and economic growth nexus in Tanzania: An ARDL bounds testing approach. Energy Policy, 37(2), 617–622. 10.1016/j.enpol.2008.09.077

17.

Odonkor

T. A.

, Osei

B. A.

, & Sjö

(2016). Risk-taking, ownership and excess reserves in the Ghanaian Banking system. Journal of Emerging Market Finance, 15(2), 147–168. 10.1177/0972652716645890

18.

Omer

, De Haan

, & Scholtens

(2015). An empirical analysis of excess interbank liquidity: A case study of Pakistan. Applied Economics, 47(44), 4754–4776. 10.1080/00036846.2015.1034842

19.

Pesaran

M. H.

(2015). Time series and panel data econometrics. Oxford University Press.

20.

Pontes

, & Sol Murta

(2012). The determinants of the bank’s excess liquidity and the credit crisis: The case of Cape Verde. Faculdade de Economia da Universidade de Coimbra.

21.

Saxegaard

(2006). Excess liquidity and the effectiveness of monetary policy: Evidence from Sub-Saharan Africa (no. 6–115). International Monetary Fund.

22.

Ukeje

S. A.

, & Amanze

(2015). Sources and impact of excess liquidity on monetary policy in Nigeria. Economic and Financial Review, 53(2), 4.

23.

Wyplosz

(2003). Briefing notes to the Committee for Economic and Monetary Affairs of the European Parliament. Europe, 15, 20.

24.

Zhang

, & Pang

(2008). Excess liquidity and inflation dynamics in China: 1997–2007. China and World Economy, 16(4), 1–15. https://onlinelibrary-wiley-com-443.web.bisu.edu.cn/doi/10.1111/j.1749-124X.2008.00118.x

What Explains Excess Liquidity of Banks? Empirical Evidence from India

Abstract

Keywords

1. Introduction

2. Literature Review

3. Data

4.1. Unit Root Test

5.1. Unit Root Test: Aggregate Level

5.2. Long-term Determinants of Excess Liquidity: Aggregate Level 3

6.1. Panel Unit Root Test: Bank Level

Footnotes

Declaration of Conflicting Interests

Funding

Appendix A

Notes

References

5.2. Long-term Determinants of Excess Liquidity: Aggregate Level³