Indranil Sen Gupta, Indranil Bhattacharyya, Satyananda Sahoo and Siddhartha Sanyal*

The management of liquidity poses a major challenge to the conduct of monetary policy in an environment of financial liberalisation. Recent research has attempted to assess liquidity conditions in the market for bank reserves in terms of central bank balance sheet flows. This paper constructs the concepts of autonomous liquidity (AL) and discretionary liquidity (DL) in the Indian context and finds that there is a systemic response in the Reserve Bank's discretionary operations to offset 'autonomous' shocks to the market for bank reserves.

Introduction

The contemporary conduct of monetary policy has assumed a market orientation the world over, following financial liberalisation. A major challenge for any monetary authority today is to tune liquidity conditions in the financial markets consistent with the overall macroeconomic objectives, although monetary policy operating procedures vary according to the relative efficacy of the respective transmission channels. The key issue before central bank liquidity management is then to assess the demand in the market for bank reserves and initiate policy action in consonance with a targeted quantum or price of liquidity – or in many instances, a combination of both.

India, like several emerging market economies, embarked on a programme of financial liberalisation in the 1990s. This necessitated changes in the monetary policy framework to accommodate the resultant shifts in emphasis on monetary policy transmission channels and therefore, the information content of policy indicators (RBI, 1998). The contemporaneous evolution of inter-linked money, government securities and foreign exchange markets improved the efficiency of monetary management, but also posed challenges arising out of the possibilities of rapid contagion (Jalan, 2000). The need to ensure orderly conditions in the financial markets has intensified the quest for an effective liquidity assessment framework.

Recent research has attempted to assess liquidity by partitioning bank balances with the central bank in terms of central bank balance sheet flows emanating from discretionary policy liquidity operations and other "autonomous" factors (Borio, 1998; ECB 2001; Schaechter et al, 2001). This paper examines their applicability in the Indian context. Section I introduces the conceptual framework. Section II attempts to construct such measures from the Reserve Bank balance sheet. Section III reviews the recent Indian experience. Finally, Section IV sets out some concluding observations.

Section I

The Framework

The market for bank reserves evolves largely through the dynamic interaction between the central bank and the banks, which are the principal financial intermediaries in an economy and the most important participants in financial markets. Autonomous liquidity (AL) aggregates the primary liquidity available to banks, stemming from regular central banking functions as the currency issuing authority and banker to banks and the government. In a stylised central bank balance sheet, this could comprise the claims on the Government and the rest of the world (RoW) netted for leakages from the banking system, such as currency. From Table 1,

AL = A1 + A3 – L1 – L3 …(1)

Discretionary liquidity (DL), the balance primary money flowing to the banking system, arises out of the central bank’s money market operations and captures the reaction of the monetary authority to autonomous changes in market liquidity,

DL = A2 …(2)

Table 1 : A Stylised Central Bank Balance Sheet
	Component	Source
	1	2
L1.	Currency	A1. Credit to Government
L2.	Bank Reserves (=R)	A2. Credit to Banks
L3.	Net Other Liabilities	A3. Net Foreign Assets
	Total Liabilities	Total Assets

so that the supply of reserves (R^s) works out to

R^s = AL + DL …(3)

Central banks can - and often do - predict AL and the demand for bank reserves (R^d) (conventionally decomposed into required reserves (RR) and the demand for excess reserves (ER^d), inclusive of settlement balances, etc.). The net liquidity (NL), prior to central bank liquidity operations, could then be estimated, ex ante, as

NL = R^d (= RR + ER^d) - AL ...(4)

If the central bank decides to maintain the existing liquidity conditions, it could bridge NL with DL (Bindseil, 2001). Alternately, interest rates would change to clear the market for bank reserves. For example, interest rates would harden (soften) if the central bank chose to ‘short’ (over-supply) the market. The realised liquidity in the market for bank reserves is simply the balances banks maintain with the central bank, in an ex post sense,

R = RR + ER = AL + DL = L2 …(5)

Employing the format of the central bank balance sheet presented in Table 1, suppose the Government draws Rs.100 as credit from the central bank to pay salaries to its employees who hold half the income in cash (Table 2). For the banking system, this results in a supply of funds of the order of Rs.50, since the currency component is a direct claim of the public on the Reserve Bank and therefore, does not impact bank liquidity (Table 3).

Table 2 : Autonomous and Discretionary Liquidity in a Central Bank Balance Sheet
	Component Flow	Rupees		Source Flow	Rupees
	1	2		3	4
L1.	Currency	50		A1. Credit to Government	100
L2.	Bank Reserves (=R)	0		A2. Credit to Banks	-50
L3.	Net Other Liabilities			A3. Net Foreign Assets
	Total Liabilities	50		Total Assets	50
Memo Item
D AL = A1+A3-L1-L3 = 100-50=50				D NL = D Rd (=0, by assumption) -
D DL = A2 = -50				D AL= 0-50 =(-) 50
D R = L2 = DAL + DDL = 50-50 = 0

Thus, AL amounts to Rs.50¹. Assuming that the deposit is exempt from reserve requirements (RR) (without loss of generality) and that there is no change in the demand for settlement balances (i.e., implicitly assuming that there is no mismatch between deposit mobilisation and credit offtake), there is, thus, an excess supply of bank reserves of Rs. 50. If the central bank contracts credit to banks, i.e., DL, by the entire Rs. 50, there would be no change in liquidity (and hence interest rates) in the market for bank reserves. Note that in case of an incremental credit offtake of say, Rs. 50, which is fully held in cash, the AL generated by the Government’s salary disbursement would be fully absorbed and therefore, the maintenance of existing liquidity conditions would not warrant central bank action. Secondly, in case of capital flows, of say Rs. 50, AL increases to Rs.100 and necessitates a withdrawal of Rs. 50 through DL to keep liquidity conditions intact. This discretionary action of the central bank to suck out liquidity can even take the form of imposing stricter reserve requirements.

Table 3 : Balance Sheet of the Banking System
	Source Flow	Rupees		Use Flow		Rupees
	1	2			3	4
1.	Deposits	50		1.	Bank Reserves
2.	Credit from Central Bank	-50		2.	Bank Credit
	Total	0		Total		0

The supply of bank reserves can be, alternatively, decomposed into borrowed reserves (BR), - essentially standing facilities available at the central bank rate - and non-borrowed reserves (NBR). The critical point of difference is that while the AL-DL classification bifurcates bank reserves on the basis of the central bank’s control over its balance sheet in general, the BR-NBR categorisation splits it in terms of the commercial banks’ ability to access primary money on own account (Table 4).

Table 4 : Analysis of Supply of Bank Reserves
Central Bank Action	Exogenous to liquidity		Accounted in
	operations of
	Central bank	Banks	AL	DL	BR	NBR
1	2	3	4	5	6	7
Monetisation of deficit	Ö	Ö	Ö			Ö
Foreign exchange operations	Ö	Ö	Ö			Ö
Cash issuance	Ö	Ö	Ö			Ö
Open market operations	Ö	Ö		Ö		Ö
	(Offtake)	(Amount/price)
Standing facilities	Ö	Ö		Ö	Ö
	(Utilisation)	(limit & price)

The AL-DL framework is intuitively more appealing from the practitioners’ perspective in case of central banks which i) combine internal debt and monetary management, ii) perform development functions, and iii) operate in economies in which cash remains a major medium of transaction. The critical importance of the AL-DL construct lies in the fact that it is operable in case of both quantity and interest rate operating targets as well as the intermediate regimes (Chart 1).

In case of quantum targets, central banks could supply a pre-determined level of reserves (R^*), so that the interest rate i^* would clear the market, given the reserves demand curve R₀^D. In case of a shift of reserves demand from R₀^D to R₁^D, interest rates would have to rise along R₀^S to A₁ for market clearing, if the central bank sticks to the same quantum of reserves. In case of interest rate targets, central banks would adjust the liquidity at A₂ along the desired trajectory say, R₁^S, so that the interest rate is maintained at i^*, and the supply of liquidity accommodates the shift in reserves demand. In many economies, especially the emerging markets, both quantity and rate transmission channels usually operate side by side, and so central bank policy spans both the quantum as well as price of the supply of bank reserves along R₂^S, inducing a shift in the market equilibrium from A to A₃. Only in the extreme cases, need the central bank give up one target for the other.

Recognising that discretionary operations often encompass changes in the policy rates, the initial framework can be extended to a simple model of the market for bank reserves in which the central bank uses the entire array of instruments – quantum and rate - at its disposal for stabilising the price of liquidity. This is set out in Appendix I. In this scenario, the market interest rates depend on the autonomous liquidity generated and the discretionary operations of the central bank, including changes in policy rates. Given that i) an increase (decrease) in the quantum of discretionary liquidity softens (hardens) the market interest rate and that ii) an increase (decrease) in the price of discretionary liquidity similarly raises (lowers) the market interest rate, it is possible for the central bank to maintain orderly conditions in the financial markets through a policy of countervailing changes in the price and quantum of liquidity. This could, for example, include calibrated counterbalancing of i) changes in the reserve requirements and the policy rate, ii) open market operations (OMO) and the policy rate and iii) reserve requirements and open market operations. The common "sufficient" condition, requiring that the market interest rate sensitivity of discretionary operations should be greater than that of autonomous factors, is derived in the Appendix I. This underscores the potency of the central bank’s discretionary use of policy instruments in fine-tuning liquidity. Central banks with a credible reputation would, thus, be able to influence market conditions with the signalling effect (which is consistent from the market perspective) reinforcing the liquidity impact of policy measures.

Cross-Country Experiences

Most central banks adjust market liquidity pro-actively in pursuance of either a quantum target or a price target, or a mix of both (Borio, 2001; Table 5). The most common operating procedure appears to be a two-step process of i) working out the net demand for bank reserves (through a forecast of AL), and then ii) undertaking discretionary operations, as determined by the estimated liquidity effect to maintain interest rates consistent with macroeconomic objectives (Ugolini 2002; and Appendix 2). The European Central Bank (ECB), for example, manages liquidity conditions through a policy mix of OMO (including repos), marginal refinance facilities (both deposit and lending facilities) and changes in the policy rates based on estimates of the autonomous liquidity² of the banking system. The US Federal Reserve targets the inter-bank federal funds rate essentially through OMO guided by its estimates of the demand of bank reserves and the technical factors³ affecting bank reserves. The Bank of Japan (BoJ) targets the outstanding balance of the current accounts with it through money market operations which are based on its projections of the autonomous sources of bank reserves⁴. Operating procedures of monetary policy in emerging market economies are also coalescing into similar strategies of liquidity management (Kamin et al, 1998). A number of central banks, including the BoJ and the ECB, publish their forecasts of autonomous factors.

Table 5 : Operating Procedures Of Liquidity Management
Country	Operating Target	AL forecast	Frequency of	Key Instruments of Discretionary Liquidity
		horizon, if any	Market Operations	Quantum				Rate	Others
				CRR	OMO	Repo	Standing
							Facilities
Brazil	Overnight inter-	1 month	Daily	Ö	Ö	Ö			Financial assis-
	bank rate								tance for liquidity
ECB		1 month	Once a week		Ö	Ö	Ö	Policy rate
			plus once addi-
			tionally a month,
			on a regular basis
Indonesia	Monetary base	1 week	Daily	Ö	Ö	Ö		Discount	Moral
	and REER							rate	suasion
Japan	Bank reserves	1 day	More than one		Ö	Ö	Ö
			per day
Malaysia	Intervention rates,	1 day	Daily	Ö	Ö	Ö			Selective credit
	inter-bank rates								and moral suasion
Mexico		1 day	Daily		Ö	Ö
South Africa	Repurchase rate	1 to 6 months	Daily	Ö	Ö	Ö	Ö	Repo rate	Foreign cur-
									rency swaps
UK		1 day to 13	Daily		Ö	Ö		Repo rate
		weeks
USA	Federal Funds Rate	2 weeks	Typically one per day		Ö	Ö	Ö	Discount rate

Section II

The Indian Context

The autonomous (DAL) and discretionary (DDL) liquidity flow measures adapt the standard literature with the caveat that the policy measures are "discretionary" and not "rule bound" (RBI 1999; RBI 2000a,b; RBI 2001a; RBI 2002a,b,c).

The DAL, in the Indian case, could be defined as the sum of the following:

1. the Reserve Bank’s primary monetisation of the fiscal deficit, through a) ways and means advances (WMA), netted for the Government balances, b) net primary subscriptions to Treasury Bills, dated securities and non-marketable securities (such as ad hoc T-Bills funded into non-transferable special securities without any maturity, etc.) and c) holdings of rupee coins, (i.e., the change in the net RBI credit to the Government, adjusted for secondary market operations);
2. incremental claims on banks (other than credit to commercial banks);
3. incremental claims on the commercial sector (other than credit to primary dealers (PDs), typically driven by development objectives;
4. incremental net foreign assets;
5. less incremental liabilities (other than scheduled commercial bank and government balances with the RBI), which constitute a leakage from the banking system, comprising a) cash, governed by demand for transactions balances which, in turn, depends on the level of economic activity and on seasonal factors, such as harvests and festivals, b) balances maintained by co-operative banks, financial institutions and foreign central banks, and c) net non-monetary liabilities, mainly constituting the Reserve Bank’s claims on itself.

The DDL, in quantum terms, could be the sum of i) the Reserve Bank’s secondary market operations in the government securities market in the form of OMO (including repo), incremental credit to ii) commercial banks and iii) PDs at a pre-determined interest rate netted for iv) changes in reserve requirements.

The construction of DAL and DDL measures is judgemental and at best time-specific, especially in view of the on-going transition from direct to indirect instruments⁵ of monetary control (Appendix III). For instance, the Reserve Bank often accepts private placement and devolvement of Treasury Bills/government securities auctions during tight liquidity and offloads them to the market when conditions ease. As the Reserve Bank is entrusted with the dual responsibility of internal debt and monetary management, the first leg is classified as DAL and the second leg as DDL. Secondly, the critical difficulty in case of policy instruments such as OMO and standing facilities is that while the Reserve Bank often determines either the price or the potential quantum and, at times, both, the response really depends on banks and PDs. In this case, the precise utilisation levels, are viewed as a function of the "enabling" conditions, in both price and quantity terms, set by the Reserve Bank and are hence classified as DDL. Finally, while there usually exists an inverse relationship between bank reserves and short-term interest rates (i.e., the liquidity effect), it is necessary to adjust bank balances with the Reserve Bank for changes in required reserves in a regime of frequent CRR changes. While the impact of CRR changes is actually dynamic and best captured by constructing a series for adjusted bank reserves, the first round release of resources on account of CRR changes is taken in DDL as a first approximation in absense of a time series on required reserves (RBI, 2001a).

Section III

The Indian Experience

The Reserve Bank faces the monetary policy dilemma of funding both the Government and the commercial sector at reasonable cost without stoking inflationary pressures and at the same time maintaining exchange rate stability (RBI, 1998a). The Reserve Bank announced a multiple indicator approach in April 1998 to accord itself the necessary flexibility for drawing policy perspectives in the face of financial liberalisation. The array of indirect instruments of monetary control has been simultaneously expanded to ensure orderly conditions in the money and foreign exchange markets.

Table 6 : AL and DL - Select Operational Cases
Autono-	AL	Likely Liquidity Operations					Adjusted	Monetary Conditions
Mous		CRR	OMO	Standing	DL	Bank	Bank	Interest	Exchange
Factors			(inc.	Facilities		Rate	Reserves	Rates	Rates
			repos)						(Re/US$)
1	2	3	4	5	6	7	8	9	10
Cash	¯		­	­	­		«	«	«
demand
Government	­		¯		¯		«	«	«
demand
Capital	¯	­	¯@	­#	¯#	­	¯	­	¯
outflows
On reversal,	«	¯	­$	¯*	­*	¯	­	¯	«
@ Especially higher cost repos.			# At higher cost.		$ At higher prices.		* At lower cost.

The liquidity management is now carried out by OMO (including repo operations) supplemented by direct interest rate signals through changes in the policy rates such as the bank rate and the Liquidity Adjustment Facility (LAF) rates, besides the traditional reserve requirements and standing facilities (Vasudevan, 1998; Reddy, 2001, 2002; Table 6). It is in this milieu of multiple objectives, indicators and instruments that partitioning bank reserves on the basis of liquidity management provides a useful tool for analysing central bank operations.

There are four stylised facts. First, DDL, more or less, offsets DAL⁶ (Chart 2 and Table 7). Secondly, the Reserve Bank has operated through different instruments⁷ - quantum and rate – at different points of time to ensure orderly conditions in the money markets.

Table 7 : Measures of Volatility
Variable		Monthly Flow		Co-efficient
		Average (Rs. crore)		of Variation
1		2		3
AL		2,502		2.5
Currency in circulation		2,010		2.2
Monetisation of the Centre’s Fiscal Deficit		2,223		3.1
Net foreign assets of the RBI		3,507		1.1
DL		-1,510		-3.8
Memo Item:
Variable/Co-efficient of Variation	1999-2000		2000-01		2001-02
Adjusted bank reserves (Ra)	0.06		0.07		0.07
Call rate	0.12		0.19		0.05
Exchange rate (Re/US $)	0.01		0.02		0.01

Thirdly, there has been a gradual reduction in the volatility in the inter-bank call rates, without any appreciable change in the variability of bank reserves, adjusted for first-round CRR changes (Chart 3). Finally, DAL (and consequently DDL) is heavily influenced by seasonalities in cash demand and government payments.

The Anecdotal Evidence

The recent monetary experience is best analysed in terms of phases of easy and tight monetary conditions (Chart 4 and Table 8). The fiscal year 1999-2000 provides a good starting point since the present operating procedure of liquidity management was, by and large, in place. The foreign exchange market saw excess demand conditions by early 1999-2000 emanating from a mix of domestic uncertainties, border tensions and bulk crude oil imports, depleting the foreign exchange reserves. This gap was funded by refinance drawals by commercial banks and PDs at the Bank Rate, which pushed up inter-bank call rates above central bank lending rate. The Reserve Bank continued its policy of private placements/ devolvements combined with subsequent OMO to deflect the pressures of Government borrowing. As a result, while DAL amounted to Rs. 14,228 crore, DL declined by Rs. 8,354 crore during June-October 1999 to constrain average monthly change in bank reserves (adjusted for CRR changes) (DR^a) to Rs.1,175 crore. During this period, the discretionary operations were essentially quantum based, in the absence of any change in the policy rates.

Capital flows revived after November, adding to the foreign exchange reserves. Given seasonal cash demand, AL increased by Rs. 12,603 crore during November 1999 to April 2000.

Table 8 : AL and DL – Phase-wise Analysis
Variable	April-	June-	November	May-	November
	May	October	1999-	November	4, 2000-
	1999	1999	April	3, 2000	March
			2000		2002
1	2	3	4	5	6
1. AL (1+2+3+4-5-6-7-8)	9,184	14,228	12,603	7,804	44,103
1.1 Net RBI credit to Government	16,528	15,789	12,890	21,378	15,198
(other than OMO)
1.1.1 Primary subscription to	16,000	11,000	0	30,149	26,679
dated GoI securities
1.2 RBI's claims on banks	-1,064	1,060	-224	383	337
(other than credit to commercial banks)
1.3 RBI credit to commercial sector	-36	-267	94	-367	-1,121
(other than PDs)
1.4 RBI’s net foreign assets (NFA)	9,250	599	23,171	-4,581	97,811
1.4.1 net of revaluation	9,047	-2,893	27,554	-10,545	89,760
1.5 RBI’s net non-monetary liabilities	1,678	4,757	6,329	2,554	28,735
1.6 Notes in circulation	14,894	-2,119	16,440	6,421	37,063
1.7 "Other" deposits with the RBI	-1,138	30	98	-262	-151
1.8 Bankers’ deposits with the RBI	61	285	462	296	2,475
(other than commercial banks)
2. DL (1+2+3+4)	-4,234	-8,354	-4,722	-1,040	-37,613
2.1 Open Market Operations	-11,301	-12,385	-12,749	-5,815	-47,717
2.1.1 Repo operations#	702	256	-1,984	0	-4,022
2.2 RBI’s credit to commercial banks	2,066	2,382	-2,274	820	-2,272
2.3 RBI’s credit to primary dealers	1,751	1,649	-3,155	4,155	-4,224
2.4 Release of resources through changes in	3,250	0	13,456	-200	16,600
CRR on NDTL of commercial banks
3. Bank Reserves (adjusted for CRR) (Ra)	4,950	5,875	7,881	6,763	6,491
(1+2 = 3.1+2.4)
3.1 Commercial bank deposits with the RBI	1,700	5,875	-5,575	6,963	-10,109
Memo item
Average Inter-bank Call Rate	8.4	9.3	8.4	9.8	7.7
# : Pertains to the LAF since June 2000.

The Reserve Bank eased monetary conditions, reducing the CRR in November 1999 and April 2000 and cutting the Bank Rate and the fixed repo rate effective April 1, 2000, to ease average call rates to 8.4 per cent. The surplus liquidity enabled banks and PDs to redeem their drawals from the Reserve Bank and revived interest in gilts, reducing DL by Rs. 4,722 crore. The average monthly DR^a amounted to Rs. 1,313 crore.

The foreign exchange market again saw excess demand pressures due to an increase in the oil import bill and the drying up of capital inflows, draining the Reserve Bank’s foreign exchange reserves. The Reserve Bank initially conducted high-cost reverse repos under the newly-introduced LAF to bridge the liquidity gap and at the same time, stabilise the foreign exchange market. As the rupee continued to depreciate, the Reserve Bank raised the Bank Rate and the CRR and halved the refinance facilities available to banks on July 21, 2000. The Reserve Bank accepted private placements/devolvements of government paper (Rs. 20,151 crore during end-July-November 3) and simultaneously conducted aggressive repo operations (averaging Rs. 9,267 crore during August-October 2000) at attractive interest rates (14.5 per cent on August 14). During May-November 3, 2000, DAL worked out to be Rs. 7,804 crore while DL declined by Rs. 1,040 crore, on a point to point basis, limiting average DR^a to about Rs. 1,000 crore. The adjustment of liquidity was reinforced by strong interest rate signals, especially in terms of sharp changes in LAF repo rates.

Capital flows revived in November 2000, initially with the proceeds of India Millennium Deposits (IMDs), followed by strong portfolio inflows, resulting in an accretion of Rs. 89,760 crore (adjusted for revaluation) to the Reserve Bank’s foreign currency assets between November 2000 and March 2002, enabling the Reserve Bank to ease monetary conditions pulling the average inter-bank call rates down to 7.7 per cent. During November 4, 2000-March 2002, DAL amounted to Rs. 44,103 crore while DL was tightened by Rs. 37,613 crore. As a result, average monthly DR^a amounted to about Rs. 400 crore.

An Empirical Exercise

We examine the dynamic inter-relationships among DAL, DDL and the changes in the call money rate (i_mr), through an unrestricted vector auto regression (VAR) model, over the period April 1996 - March 2002. The exercise has been undertaken in two steps. First, the monthly interaction between DAL and DDL has been examined along with the policy rate (i_pr), proxied by the Bank Rate, as exogenous, in Model 1. Second, the interaction between DR^a and changes in the call money rate (Di_mr) has been examined along with Di_pr and a dummy representing the South-East Asian crisis as exogenous in Models 2 (on monthly basis) and 3 (on weekly basis). We follow the standard VAR methodology of model estimation, block-causality, impulse response and variance decomposition.

The underlying variables were first examined for stationarity with the lag length chosen by the appropriate model selection, viz., Akaike Information (AIC) and Schwarz Bayesian (SBC) criteria. In the monthly data, all the series were found to be stationary at levels both in terms of DF and ADF tests (Table 9). In the weekly data, except for the call money rate, which is stationary at levels, all other series were found to be stationary at first-differences, i.e., I(1).

We first test the causality between DAL and DDL (Table 10). It will be noted that the x²-tests, which measure the statistical significance of lags of other variables in predicting the left-hand side variables, in addition to own lagged variables, are equivalent to Granger (1969) causality involving non-stationary variables in a VAR framework.

Table 9 : Unit Root Tests
Variable	Monthly Models I and II						Weekly Model III
	Without Trend			With Trend			Without Trend	With Trend
1	2			3			4	5
DAL	-3.56	(4)		-3.80	(4)
DDL	-3.52	(4)		-4.13	(4)
Ra	-5.59	(10)		-5.59	(10)		-0.91 (1)	-9.73 (1)
Imr	-5.25	(1)		-5.25	(1)		-8.87 (1)	-9.62 (1)
Ipr	-6.41	(2)		-6.37	(2)		-1.28 (1)	-2.88 (1)
Note: Based on Akaike Information Criterion (AIC). In case of the monthly models, the 95% critical value for the ADF statistic is –2.9023 for the regression without a trend and –3.4730 with a trend. In case of weekly data, the 95% critical value for the ADF statistic is –2.8710 for the regression without a trend and –3.4258 with a trend both at levels and first differences.

Table 10 : Test of Granger’s Block Causality in a VAR Framework
	Null Hypothesis		Test Statistic		Accept/	Inference
			x2		Reject Null
					Hypothesis
1			2		3	4
I.	DAL does not Granger cause	DDL		8.10*	Reject	DAL Granger
	DDL does not Granger cause	DAL		1.91	Accept	causes DDL
II.	DRa does not Granger cause	Dimr		5.42**	Reject	Bi-directional
	Dimr does not Granger cause	DRa		8.69*	Reject	causality
III.	DRa does not Granger cause	Dimr		16.6*	Reject	DRa Granger
	Dimr does not Granger cause	DRa		2.37	Accept	causes D imr
Note : Models 1-2 and 3 are estimated with 12 and 4 lags of the endogenous
	variables, respectively.
	* Significant at 1 per cent level.		** Significant at 5 per cent level.

As expected, DAL causes DDL, without evidence of reverse causation, which is consistent with previous research (RBI, 2001). Note that the computed D² test-statistic for the null hypothesis of no-causation running from DAL to DDL is highly significant at 1 per cent level.

We then turn to the causal relationship between market interest rates (i_mr) and adjusted bank reserves (R^a) to extend the existing research. We find that there exists bi-directional causality between DR^a and D i_mr in terms of monthly data in Model II. An empirical verification of the liquidity effect in the Indian context is subject to a number of limitations of the data. For example, while the liquidity numbers are point-to-point flows between the monthly last reporting Fridays, the call money rate is taken as the weighted average lending rate of the month. Thus, in as much DR^a affects Di_mr , intra-month Di_mr influences the monthly DR^a8. It is difficult to match the periodicity between the two series since i) time series data on daily bank reserves are not available and ii) it would be inappropriate to work with point call data since call rates used to fall to very low levels on reporting Fridays earlier. To reduce the gaps in the information content of the data, an exercise was also conducted on a weekly basis. In this case, a strong block Granger causality running from DR^a to Di_mr is evident, while at the same time, the causality running from Di_mr to DR^a has been rejected. The i_pr is taken as an exogenous variable and its presence in the model has been vindicated by a significant x² statistic.

We used the unrestricted VAR model for analysing the impulse responses and forecast error variance in each of the three cases. In each equation, iterations are made till 12 lag periods by imposing a one-standard deviation shock to each variable in a generalised impulse response approach. This has the advantage of circumventing the problem of the dependence of the orthogonalised impulse responses on the ordering of the variables in the VAR (Koop et al, 1996).

In Model 1, in terms of impulse response, the response of DDL to DAL is immediate and pronounced in the first and second months, although the tendency to react continues up to 5-6 months (Charts 5-8). On the other hand, in Model II, the response of Di_mr to DR^a is immediate and more pronounced with the impact lasting up to 8-9 months (Charts 9-12). In Model III, the response of Di_mr to DR^a is similarly immediate and strong with the effect persisting for 8-10 weeks (Charts 13-16).

In terms of variance decomposition, in the Model I, about one-third of the total variation in DDL is due to the innovations in DAL. Similarly, about half of the total variation in Di_mr is due to the innovations in DR^a in the monthly data. In terms of weekly data, almost the total variation in i_mr is due to the innovations in DR^a

Conclusion

The management of liquidity in financial markets has emerged as a key policy issue in central banking. Given the large size of the fiscal deficit, sudden switches in capital flows and the seasonal character of cash demand, the Reserve Bank steers short-term liquidity conditions by a policy mix of adjustments in market liquidity through changes in reserve requirements, standing facilities and open market operations, reinforced by interest rate signals via changes in policy rates. This paper simply formalises this natural partition of sources of primary liquidity into the emerging reserve concepts of autonomous and discretionary liquidity. Given the empirical result that changes in bank reserves impact short-term interest rates, especially in the short run, we believe that the AL-DL dissection provides a reasonably good analytical framework for mapping the interaction of the central bank and participant banks in the market for bank reserves. This is buttressed by the fact that DL is empirically found to capture the policy response to the autonomous factors in the AL.

Further research could essentially proceed in two directions. First, there is a need to forecast autonomous liquidity. Second, it may be apposite to explore the impact of combinations of policy instruments on the stability of liquidity conditions to evolve useful feedback rules for monetary policy formulation. Empirical work, beyond the present modest attempt, is constrained by the shifts in the relationship between the money market rates and the instruments of DL, such as, refinance and repo operations, as a result of frequent changes in regulations in respect of average reserve requirements, on the one hand, and the lack of an acceptable representative policy rate in view of the infrequent changes in the Bank Rate and the intermittent character of repo auctions, especially in the pre-LAF years, on the other. These limitations of quantification, natural in transition, are likely to be ironed out with the evolution of the LAF as a principal operating instrument of monetary policy and the recent CRR stipulation of a daily minimum maintenance of 80 per cent of required reserves for commercial banks.

Notes

1. Following the Reddy Working Group, we divide the economy into six sectors, viz., households (A), the Reserve Bank (B), the banking system in India (C), other financial corporations (D), general government (E) and the non-financial commercial sector (F), which interact within themselves and with the rest of the world (RoW) sector. It may be useful to work out the numerical example through the sectoral balance sheets of the economy, including the Government and the commercial sector, which comprises the other agents of the domestic economy. In case of the Government, the monetisation of the deficit to fund employees’ salaries expands the balance sheet by Rs.100, generating an AL of a like amount in the first instance (Tables 2 and 11).

Table 11 : Balance Sheet of the Government Sector
Income Flow	Rupees	Expenditure Flow	Rupees
1	2	3	4
1. Credit from the Central Bank	100	1. Salaries	100
Total	100	Total	100

The net impact on AL, however, depends on the public demand for cash, which is a leakage from the banking system. In this case, since the commercial sector splits the salary evenly into cash and bank deposits, net AL increases by Rs.50 (Tables 3 and 12). It is useful to mention that if there was an increase in the public’s demand for currency, irrespective of the Government’s salary disbursement, there would have been a drainage of liquidity from the banking system as deposits would have been drawn down and been substituted for by currency.

Table 12 : Balance Sheet of the Commercial Sector
Income Flow	Rupees		Expenditure Flow	Rupees
1	2	3		4
1. Salaries	100	1.	Currency with the Public	50
2. Bank Credit to Commercial Sector		2.	Deposits	50
Total	100		Total	100

2. Defined as the sum of the liquidity-injecting factors such as the purchases of net foreign assets by the Eurosystem, netted for liquidity-absorbing factors such as issuance of bank notes in circulation, government deposits with the Eurosystem and other factors (net) The ECB has published weekly forecasts of autonomous factors since June 2001, with a view to providing the counter-party public and private credit institutions in the Eurosystem a reliable basis for assessing its allotment decisions in variable tender auctions for central bank support.

3. Includes shifts in cash demand, size of treasury balances at Federal Reserve Banks and the volume of the Federal Reserve float.

4. Includes net issuance of bank notes and changes in treasury funds. The Bank of Japan releases projections of sources of changes in current account balances with it and market operations a day in advance.

5. For instance, the 14-day Treasury Bills (introduced June 1997) emerged as a key instrument for mopping up surplus liquidity generated by capital inflows during the first half of 1997-98 because they were more attractive than repos. In the latter half of the year, when external pressures on account of the South-Asian crisis warranted monetary tightening, the Reserve Bank raised the repo rates but left the 14-day T-Bill rates unchanged since the Government borrowing programme had already been completed (RBI, 1998a). The EPW Research Foundation adjusts the autonomous liquidity for Treasury Bill auctions in its monthly money market review in the Economic and Political Weekly.

6. The empirical test of the strategy of the AL-DL framework should ideally involve a verification of the relationship between the market liquidity gap (i.e., net liquidity in (4)) and discretionary operations. The computation of net liquidity, however, is difficult given the lack of data on required reserves (or net demand and time liabilities relevant for CRR) necessary for a forecast of the demand for bank reserves (R^d). While this is undoubtedly a limitation, the difference may not be as material given the secular trend in long-run NDTL growth and adjustment of bank reserves for changes in CRR.

7. For instance, the Reserve Bank has adjusted liquidity (i.e., rejected bids) 143 times and adjusted rates 20 times in the operation of the 1-day auctions under the LAF between June 2000-March 2002.

8. For instance, once the Reserve Bank tightened monetary policy on January 16, 1998, in the wake of the South-Asian crisis, inter-bank call rates shot up from 6 per cent on the reporting Friday of January 16, 1998 to 30 per cent per cent on January 17 and 70 per cent on January 24 before returning to 7 per cent on the next reporting Friday on January 30.

References

Bindseil, Ulrich (2001): "Central Bank Forecasts of Liquidity Factors: Quality, Publication and the Control of the Overnight Rate", ECB Working Paper, No. 70.

Borio, Claudio E.V. (1997): "The Implementation of Monetary Policy in Industrialised Countries", BIS Economic Papers, No. 47.

Borio, Claudio E.V. (2001): "A hundred ways to skin a cat: comparing monetary policy operating procedures in the United States, Japan and the euro area", BIS Policy Paper, No. 9.

European Central Bank (2001): The Monetary Policy of the European Central Bank.

Granger, C.W.J. (1969): "Investigating Causal Relations by Econometric Models and Cross-Spectoral Models", Econometrica, 37.

Jalan, Bimal (2000): "Finance and Development – Which Way Now?", RBI Bulletin, February.

Kamin, Steven, Philip Turner and Jozef Van’t Dack (1998): "The Transmission Mechanism of Monetary Policy in Emerging Market Economies: An Overview", BIS Policy Paper, No. 3.

Koop, G., M.H. Pesaran and S.M. Potter (1996): "Impulse Response Analysis in NonLinear Multivariate Models", Journal of Econometrics, 74.

Reddy, Y.V. (2001): "Developments in Monetary Policy and Financial Markets", RBI Bulletin, May.

Reddy, Y.V. (2002): "Parameters of Monetary Policy in India", RBI Bulletin, February.

Reserve Bank of India (1998): Report of the Working Group on Money Supply: Analytics and Methodology of Compilation (Chairman: Dr. Y.V. Reddy), June.

Reserve Bank of India (1998a): Annual Report.

Reserve Bank of India (1999): Annual Report.

Reserve Bank of India (2000a): Macroeconomic and Monetary Developments.

Reserve Bank of India (2000b): Annual Report.

Reserve Bank of India (2000c): Report on Currency and Finance.

Reserve Bank of India (2001a): Annual Report.

Reserve Bank of India (2001b): Report on Currency and Finance.

Reserve Bank of India (2002a): Monetary and Credit Policy Statement 2002-03. Reserve Bank of India (2002b): A Short-term Liquidity Forecasting Model for India. Reserve Bank of India (2002c): Annual Report.

Schaechter, Andrea (2001): "Implementation of Monetary Policy and the Central Bank’s Balance Sheet", IMF Working Paper, No. 149.

Schaechter, Andrea et al (2001): "Forecasting Liquidity", IMF Working Paper.

Ugolini, Peiro (2002): "Development of Money Markets and Co-ordination with Monetary Policy Operations", Workshop on Developing Government Bond Markets in Eastern Europe, Istanbul, April.

Van’t dack, Jozef (1999): "Implementing Monetary Policy in Emerging Market Economies: An Overview of Issues", BIS Policy Paper, No. 5.

Vasudevan, A. (1998): "Analytical Issues of Monetary Policy in Transition", RBI Bulletin, January.

Appendix 2 : An Operational Scheme of Liquidity Management

 

Appendix 3 : Classification of Autonomous and Discretionary Liquidity Flows in the Reserve Bank Balance Sheet
Account	Nature		Claim	Select Remarks
	of Flow		of/on
	AL	DL	Sector
1	2	3	4	5
Issue Department: Liabilities
Notes held in Banking Department	Ö		B
Notes in Circulation	Ö		Other
			Than B
Issue Department: Assets
Gold Coin and Bullion	Ö		Sovere-
			ign asset
(a) Held in India
(b) Held outside India
Foreign Securities	Ö		RoW
Rupee Coin	Ö		E
Government of India Rupee Securities	Ö		E	Secondary market transac-
				tions classified DL.
Internal Bills of Exchange and	Ö		A+D+F
other Commercial Paper
Banking Department: Liabilities
Capital paid up	Ö		E
Reserve Fund	Ö		B
National Industrial Credit (Long	Ö		B
Term Operations) Fund
National Housing Credit (Long	Ö		B
Term Operations) Fund
Deposits
(a) Government			E
(i) Central Government	Ö
(ii) State Governments	Ö
(b) Banks			C
(i) Scheduled Commercial Banks				Changes in required re-
				serves for scheduled com-
				mercial banks under Sec-
				tion 42(1) classified DL.
(ii) Scheduled State Co-operative Banks	Ö
(iii) Other Scheduled Co-operative Banks	Ö
(iv) Non-scheduled State	Ö		C
Co-operative Banks
(v) Other Banks	Ö
(c) Others	Ö		A+D+F
Bills Payable	Ö		A+D+F
Other Liabilities	Ö		B
Banking Department: Assets
Notes	Ö		B
Rupee Coin			E
Small Coin
Bills Purchased and Discounted
(a) Internal	Ö		A+D+F
(b) External	Ö		RoW
(c) Government of India Treasury Bills			E
Balances held Abroad	Ö		RoW
Investments
Investments in Subsidiaries/	Ö		C+D
Associate Institutions
Foreign Securities	Ö		RoW
Government Securities	Ö		E	Secondary market transa-
				ctions classified DL.
Loans and Advances to
(i) Central Government	Ö		E
(ii) State Governments	Ö		E
Loans and Advances to
(i) Scheduled Commercial Banks		Ö	C
(ii) Scheduled State Co-operative Banks	Ö
(iii) Other Scheduled Co-operative
Banks
			C
(iv) Non-Scheduled State Co-operative
Banks
(v) NABARD	Ö		D
(vi) Others			D	Liquidity support to
		Ö		primary dealers
				classified DL.
Loans, Advances and Investments
from National Industrial Credit
(Long Term Operations) Fund	Ö		D
(a) Loans and Advances to
(i) Industrial Development	Ö		D
Bank of India
(ii) Export Import Bank of India
(iii) Industrial Investment
Bank of India
(iv) Others
(b) Investments in bonds/
debentures issued by
(i) Industrial Development
Bank of India
(ii) Export Import Bank of India
(iii) Industrial Investment
Bank of India
(iv) Others
Loans, Advances and Investments
from National Housing Credit
(Long Term Operations) Fund
(a) Loans and Advances to
National Housing Bank
(b) Investments in bonds/debentures
issued by National Housing Bank
Other Assets	Ö		B	Includes gold.

---

* Indranil Sen Gupta is Assistant Adviser and Satyananda Sahoo and Siddhartha Sanyal are Research Officers in the Department of Economic Analysis and Policy while Indranil Bhattacharyya is Research Officer in the Monetary Policy Department. The views expressed are those of the authors and not of the institution to which they belong. The authors are grateful to R. Kannan, Madhusudhan Mohanty, Partha Ray, Muneesh Kapur, Dhritidyuti Bose, Jeevan Kumar Khundrakpam, Arghya Kusum Mitra, Kaushik Bhattacharya and Sarat Chandra Dhal for their comments, although the usual disclaimer applies. The authors thank Nitin Sahasrabudhe and Rajesh Salvi for computational support.