ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Measuring Banking Intermediation Services

This paper reviews the methodology for estimating output from banking intermediation services in India. The existing practice is compared with the guidelines provided in the United Nations System of National Accounts 1993. The paper identifies certain unsettled issues and problem areas in measuring banking intermediation services.

Measuring Banking Intermediation Services

Issues and Challenges for India

This paper reviews the methodology for estimating output from banking intermediation services in India. The existing practice is compared with the guidelines provided in the United Nations System of National Accounts 1993. The paper identifies certain unsettled issues and problem areas in measuring banking intermediation services.


I Introduction

he system of national accounts (SNA) 1993 published by the United Nations (UN) brought about major changes in the methodology for compiling national accounts statistics (NAS) contained in UN SNA 1968. A significant modification came in the context of financial intermediation services. There are four pertinent issues: first, measuring the output generated from intermediation services at current prices; second, measuring the same output at constant prices (i e, at the prices of the base year); third, allocating the output so measured (both current and constant price estimates) among the user sectors/ sub-sectors of the economy; and fourth, constructing price and output/quantity indices of intermediation services. While the first three issues are generally talked about in the context of compiling NAS for a country, the fourth is also important for constant price estimates of NAS and measurement of inflation.

A major part of the revenue of the banking sector comes from the financial intermediation services rendered by banks to their customers. These intermediation services include a component which is not charged explicitly and needs to be measured indirectly. The SNA 1993 covers these activities under “financial intermediation services indirectly measured (FISIM)”.1 The SNA 1993 defines financial intermediation as “a productive activity in which an institution unit incurs liabilities on its own account for the purpose of acquiring financial assets by engaging in financial transactions on the market”. Financial intermediaries obtain funds by incurring liabilities on their own account either by taking deposits or by issuing bills, bonds or other securities. On this principle, financial intermediaries cover not only banks but also several other financial institutions, such as mutual funds and insurance companies.

The other component of output from the banking sector consists of the direct/auxiliary services the sector provides to its customers. Indeed, the earning from these services has been rapidly increasing in the recent past. Direct services are explicitly charged for; thus, measuring them involves no major conceptual problem.

In India, the Central Statistical Organisation (CSO) has implemented the SNA 1993 recommendations to the extent possible for compilation of the country’s NAS. A number of revisions with respect to coverage and methodology were carried out first for the 1993-94 series of NAS and further refinements/improvements have been made in the latest NAS series with base year 1999-2000, released in February 2006 [CSO 1999, 2006].2 But there are still many unsettled issues which call for fresh debate and research for improving the quality of NAS further.

The estimates of output from the banking sector according to the latest NAS series (i e, 1999-2000 series) have been revised to a substantially lower level now – from Rs 1,07,527 crore according to the 1993-94 series to Rs 94,114 crore in the 1999-2000 series. Also, the corresponding figure for 2000-01 has been adjusted further lower at Rs 93,014 crore according to the new series. The estimated gross domestic product (GDP) for two sub-sectors of banking, viz, the banking department of RBI and non-bank financial companies, have been highly volatile, at times negative, during the period 1999-2000 to 2004-05. Though the methodology for estimating value added has been revised with respect to the banking sector/sub-sectors, high volatility and substantial fall for the years 1999-2000 and 2000-01 in banking sector output according to the new series raise several concerns on methodo logical aspects. The banking sector plays a pivotal role in promoting growth and development of an economy and a fall in output from such an important sector results in structural change, which has major policy implications. Thus, there is a need to understand the limitations in the existing practices, identify unsettled issues and fine-tune the methodology for improving the quality and stability of NAS further.

In this paper we focus on the measurement of output and prices of banking intermediation services. Towards this objective, the paper reviews the existing practices of measuring output from the banking sector in India and identifies certain issues and problem areas, which would have to be addressed in future research and debate. Wherever possible, some suggestions are also made. The organisation of the rest of the paper is as follows: In Section II, a broad outline of the literature on measures for banking intermediation services is presented. While Section III discusses the treatment given to output from the banking sector in the latest NAS series in India, Section IV presents our reflections on the latest NAS with respect to the sector. Some general issues with regard to measures of banking intermediation services are discussed in Section V. Finally, Section VI concludes the paper.

II Measures of Banking Intermediation Services in the Literature

The output from the banking sector consists of (1) those services which are charged explicitly (direct/auxiliary services), and (2) FISIM.3 Thus, the total output from the banking sector

can be expressed as:

Total output (Banking Sector)

= FISIM + Explicit Charges (on Direct/Auxiliary Services)

= ΣFISIMg + Σ (Auxiliary Services/Explicit Charges)g ...(1)

g g where ‘g’ is indexing over different sub-sectors of the banking sector.

The calculation of output from auxiliary services is straightforward: sum of the explicit charges (fees, commissions, etc) on such services. Thus, there is no conceptual problem in measuring this component of banking output, though at times possible data gaps may cause a little computational difficulty. In the case of intermediation services, however, there are certain grave concerns even at the conceptual level. It is now recognised that financial intermediation services have to be measured indirectly, thus the literature on FISIM has evolved. FISIM are defined as “the total property income receivable by financial intermediaries minus their total interest payable, excluding the value of any property income receivable from the investment of their own funds, as such income does not arise from financial intermediation” [United Nations 1993, pp 14041]. From this perception FISIM is sometimes expressed, in simple form, as

FISIM = R – R = (Σ Lj rj – Σ Dk ik) ...(2)


j k where

R = interest flow receivable on loans/assets (excluding those

rfrom own funds) as part of financial intermediation Rp = interest flow payable on deposits/liabilities L = volume of loans/assets r = rate of interest on loans/assets D = volume of deposits/liabilities i = rate of interest on deposits/liabilities j and k index various loan/asset categories and deposit/liabilities categories, respectively. The expression for FISIM in Equation (2) is an illustration to show how output from financial intermediation services would be expressed. In practice, the words “interest flow receivable/ payable” are interpreted depending upon the typical nature of a sub-sector and financial intermediation. For example, the CSO considers Unit Trust of India (UTI) as a financial intermediary and measures its output (as per the 1999-2000 series) as the interest and dividends receivable minus interest and dividends paid. Unlike banks, UTI invests funds of unitholders/investors. Thus, for UTI, interest payment to depositors is negligible and it generally pays dividends to unitholders/investors and earns dividends on its investment in the process of its inter-mediation services.

User Costs and Prices of Intermediation Services

Equation (2) just gives a broad understanding of total FISIM for a financial intermediary (FI)/sub-sector – it gives neither any information about the service/product wise price-quantity break-up of the total FISIM nor an idea of how the total FISIM is consumed by the user sectors/sub-sectors. The derivation of such detailed estimates/break-ups is difficult on several counts:

(1) these services are charged implicitly, thus prices of these services are not readily/directly available; (2) much of these services are sold and priced in a bundle and it is conceptually difficult to derive the service wise break-up of price and quantity; and (3) till now no consensus has emerged on what components of intermediation services constitute the banking sector’s output bundle4 [Srimany and Bhattacharya 1998, CSO 1989]. In practice, the task is accomplished by means of what is called “reference rate method”, which actually has its root in the “user cost approach” [Fixler 1993, Hancock 1985]. The user cost of a financial product can be obtained from its holding cost less the reference rate.5 The price of the product is the negative of its user cost. In this literature, a financial product with negative user cost is identified as financial output and one with positive user cost is a financial input. But extreme difficulty in estimation of the reference rate has been a major hurdle in implementing the user cost approach.

In the process of financial intermediation, FIs incur liabilities on their own account by way of receiving funds (say, deposits) from others and produce loans from these funds. They do not explicitly charge their customers (borrowers and depositors) for the intermediation services they render. The interest charged on loans is higher than what would have been the pure cost of borrowing funds (i e, the reference rate) and the interest rate paid on deposit funds is less than the pure borrowing cost. Thus, intermediation services are charged implicitly – rates charged/paid on loans/deposits have at least two components: the reference rate and the charge made on intermediation services attached to each monetary unit transacted. Under certain simplifying assumptions, it is thus quite reasonable that Interest rate on loan = reference rate + price for inter

mediation services attached to each

monetary unit of loan Interest rate on deposit = reference rate – price for inter mediation

services attached to each monetary

unit of deposit …(3) In other words, Price of services on loan = interest rate on loan – reference rate

price of services on deposit = reference

rate – interest rate on deposit …(4)

In her pioneering paper, Hancock (1985) derived more general expressions for prices of financial intermediation services under more general settings (considering in addition, direct/ explicit charges, such as fees and commissions, and discount factor) [also see Fixler and Zieschang 1992 and Fixler 1993]. However, the simple forms of prices as in Equations (3) and

(4) are extensively used in the empirical literature.

Estimating FISIM and Allocating It to User/Consumer Sectors

The explicit charges, if any, made on intermediation services are relatively less than the expenditure incurred, hence the usual procedure of estimating value added in the banking sector would produce very small or negative value added.6 This is a paradox considering the importance and contribution of the banking sector to an economy. The proponents of NAS were well aware of this problem and as a remedy proposed imputing an additional component to the gross output of the sector, which consisted of “free services” that the sector was assumed to be providing to its customers. The imputed income, equivalent to the interest and dividend receipts of banks net of interest paid to depositors, was considered to be the income earned by the banks through financial intermediation [CSO 1989, OECD 1998].

The SNA 1968 retained the convention of imputation but made certain modifications, such as: (1) defining imputed service charge more precisely by noting that “the property income [banks] receive as a result of investing own funds should not be taken into calculating imputed service charge”, thus restricting the imputed service charge as being generated through loans produced using only other people’s money; and

(2) the imputed services charge was treated as intermediate consumption of a “nominal industry” [OECD 1998]. The process, however, suffers from a limitation in that the imputed service charge had no impact on GDP as no part of the imputed charge entered into final consumption expenditure, exports or imports. The SNA 1993 rectifies the problem by allocating the output from intermediation services (i e, FISIM) to the consuming sectors (enterprises, households, government, non-profit institutions serving households (NPISHs), and the rest of the world). The FISIM consumed by households, government and the rest of the world constitutes part of final demand and to that extent contributes to total GDP. To maintain the consistency of a country’s NAS, accuracy in allocating FISIM to consuming sectors/sub-sectors of the economy is thus extremely important.

Following OECD (1998), an example of FISIM calculation in respect of loans and deposits would be as below. FISIM on loans = average balance on loans * (effective interest

rate on loans – reference rate) ...(5) FISIM on deposits = average balance on deposits * (reference rate

– effective interest rate on deposits) …(6) where effective interest rate on loans/deposits has been derived as the interest received/paid by the bank on loans/deposits divided by corresponding average money balance.

From Equations (5) and (6), it is clear that the FISIM estimates could be positive or negative. The negative FISIM on loans arises if the reference rate exceeds the effective interest rate on loans. On the other side negative FISIM on deposits occurs when the reference rate is lower than the effective rate on deposits. Interestingly, it is clear to see that the expressions within the first parenthesis in Equations (5) and (6) actually are simplified forms of prices (ignoring direct payments, such as, fees and commissions, and also discount factor) given by Fixler (1993) and Hancock (1985). The user cost approach assumes that service/output is attached to each unit of money being transacted through financial intermediation, so it derives the value of output from financial intermediation (i e, FISIM) through expressions as in Equations (5) and (6), i e, the price for services attached to each unit of outstanding loan/ deposit product multiplied by average money balance in the product.

For a commercial bank providing intermediation services through loans and deposits, total FISIM may be estimated by the sum of the FISIM on loans and FISIM on deposits. In other words, total FISIM here can be expressed as Total FISIM (loans and deposits)

= FISIM on loans + FISIM on deposits

= L × (rL– r*) + D × (r* – rD)

= (RL – RD) + (D – L) r* …(7)

where L = outstanding amount of loans rL = rate of interest earned on loans D = outstanding amount of deposits rD = rate of interest paid on deposits RL = interest income from loans (excluding that from own funds) RD = interest flow payable to deposits

The estimated FISIM from the banking sector needs to be allocated to the user sectors. Given suitable data, the “reference rate” based approach has the ability to apportion the total FISIM from the banking sector over different user/consumer sectors.

Price Indices for Financial Intermediation Services

We have already seen how prices of intermediation services can be estimated through the reference rate based approach. So, knowing the prices of such services in the base year and a reference period, it would be possible to construct a price index of such services. The weighting diagram would be derived from the contribution of the products/services to the FISIM – the weighting diagram may be static pertaining to the base period or it would vary over time if updated for different time points/ periods.

Under the reference rate approach, one can estimate constant price FISIM directly as a sum of the values obtained by multiplying current period quantity with the base period’s price as obtained through the reference rate method for all services.

It may be noted that the estimation/allocation of FISIM and construction of indices for such services are two related but distinct tasks. In the case of FISIM, the emphasis is more on making complete coverage of services but the indices can be constructed based on a representative sample/basket of services. The total estimated FISIM from the banking sector has to be allocated to user sectors for the consistency of NAS statistics. The indices, however, do not require focusing on such allocation. Both the tasks, however, require an estimate of the reference rate, and are thus related.

III Measuring Output from Banking Sector in India

The banking sector broadly refers to banks and related financial institutions. In India this sector covers the following subsectors [Barman 1995, CSO 1989, 1999, 2006]: 1 the banking department of the Reserve Bank of India

(RBI); 2 commercial banks; 3 public non-banking financial corporations; 4 organised non-banking financial companies engaged in trad

ing in shares, investment holdings, loan finance and like

activities; 5 unorganised non-banking financial undertakings such as

professional moneylenders and pawnbrokers; 6 post office savings banks including operations concerning

cumulative time deposits and national saving certificates;

and 7 cooperative credit societies.

The coverage and methodology for compiling NAS in India with respect to the banking sector have undergone significant changes over different NAS series. This section focuses on the methodology used in the latest NAS series (i e, the series with base year 1999-2000) for measuring output (FISIM + auxiliary services) from the banking sector. The topic is covered under three heads7, viz, (1) methodology to estimate gross value added (GVA) at current prices; (2) corresponding estimates at constant prices; and (3) allocation of estimated FISIM to various user sectors and sub-sectors.

Gross Value Added at Current Prices

Banking department of the RBI: Like in the previous two series (i e, the 1980-81 and 1993-94 series), the RBI accounts have been divided into those pertaining to the issue department and those pertaining to the banking department. While the activities of the issue department are classified under government administration, the banking department is included under the corporate financial sector.

The output of the issue department is measured on cost basis, i e, sum of intermediate consumption, compensation to the employees and consumption of fixed capital (CFC) as is done for public administration. Output so measured is deducted from the total income of the RBI and treated as government final consumption expenditure.

The GDP of the banking department of the RBI is obtained as actual income net of output of the issue department plus imputed income (interest and discounts received minus interest paid by the RBI) minus intermediate consumption, and is included in the banking sector. Since the 1993-94 series, value added in the form of operating surplus of the issue department of the RBI, which was being somewhat ignored in the 1980-81 series, was added to the GDP of the banking sector [CSO 1999]. The same practice continues in the 1999-2000 series also. Commercial banks: There has been no change in the new series with respect to this sub-sector [CSO 2006], implying that the methodology followed in the 1993-94 series has been continued for the 1999-2000 series. But CSO (1999) says that the 1993-94 series, in effect, adopted the methodology followed in the 1980-81 series. Thus, the methodology adopted in the latest series is the same as that of the 1980-81 series, which measures GVA from financial intermediation as the interest and dividend receipts net of interest paid to depositors. While the data on income, expenditure and appropriation of commercial banks are collected from the RBI (Statistical Tables Relating to Banks in India), data on rent are obtained directly from the respective banks. Non-banking financial corporations and companies (NBFCs) including UTI: The estimates of public NBFCs are made based on their respective annual reports. In the case of organised private NBFCs, the RBI’s annual study for sample organised private companies is used. In order to obtain population estimates, sample estimates in the RBI’s study are blown up using the ratio of total paid-up capital of all such companies as available from the department of company affairs (DCA) to the paid-up capital of sample companies in the RBI’s study. The data gap with respect to unorganised NBFCs, however, is severe and on the basis of its internal study and discussions with subject experts, the CSO considers that the total output of all such unorganised companies would be equal to one-third of the output from organised private NBFCs.

The methodology of calculating GVA has been changed substantially in respect of (1) mutual funds (MFs) like the UTI and (2) non-government non-banking financial companies (NGNBFCs). In the 1993-94 series, UTI was considered a financial intermediary, and the GVA of UTI was computed like that of banks and other financial intermediaries. But unlike banks, MFs/UTI do not work on deposits of other sectors. Instead, they employ the funds raised by collective investments made by their unit/shareholders, who receive dividends, not interest payments. Thus, treating UTI as a bank and so subtracting only interest payments (which is insignificant for UTI) from interest/dividend earnings was resulting in in appropriate GVA from UTI. In the 1999-2000 series, this problem has been solved and the imputed value of services of UTI is now calculated as total income on account of dividend, interest, profit on sale/ redemption of investment minus interest, dividend paid to unitholders, and undistributed profit.

In the 1993-94 series, the current price GVA of NGNBFCs was being estimated based on the RBI annual study on “performance of non-government financial and investment companies”. This annual study presents data for three consecutive years for a common set of companies but the sample size changes each year. Thus, for a particular year, three different estimates might be available from studies of three consecutive years. At times, the results for a particular year vary considerably and have a bearing on the estimation of GVA from this sub-sector. The 1999-2000 series makes an attempt to derive relatively more stable estimates by pooling the results of a particular year appearing in all available RBI annual studies. For quick estimates, however, the method adopted in the 1999-2000 series is the same as in the 1993-94 series as no RBI study is available for the reference year. Post office savings banks: The GVA for this sub-sector comprises compensation of employees and rent. In India, however, due to data limitations GVA from this sub-sector is computed as a proportion of management expenses, where the proportion is assumed to be the same as observed with respect to commercial banks. Cooperative credit societies: The factor incomes of cooperative credit societies are available from ‘Statistical Statements Relating to Cooperative Movement in India’, published by the National Bank for Agriculture and Rural Development (NABARD). These data pertain to profits after tax but GDP compilation requires corresponding data on gross income/profits before tax. In order to obtain gross profit (before tax), detailed information on profit and loss accounts is obtained directly from select cooperative societies.

Output at Constant Prices

In order to estimate output at constant prices, the output from different activities needs to be valued at the prices that prevailed in the base year of the NAS series. These estimates are done separately for each sub-sector. If quantity and price of a service were known, corresponding output would be obtained simply by multiplication of price and quantity. But defining quantity of banking services and estimating unit prices for financial intermediation services is extremely difficult. Therefore, a simple practice across countries has been to carry forward the base year estimates using certain indicators measuring the volume of activity in the corresponding sub-sector. In India too, a similar practice is followed for estimating constant price output from the banking sector. The volume indicator if measured in monetary terms is divided by the wholesale price index (all commodities) to obtain the quantity index. Strikingly, CSO documents [CSO 1989, 1999, 2006] do not provide the details of volume indicators used for different sub-sectors of the banking sector.

In the 1999-2000 series, changes have been made in deriving the volume indicator in respect of certain sub-sectors of the banking sector. Notably, a significant departure is made from the earlier practice of estimating constant price GVA from the banking department of the RBI. In both 1980-81 and 1993-94 series, corresponding constant price values were being estimated by carrying forward the base year estimates with the volume index prepared by deflating aggregate deposits and credit of all scheduled commercial banks by the WPI. This practice resulted in a certain anomaly between current and constant prices estimates – while the GVA for the banking department of the RBI had been showing up as negative in recent years, corresponding constant price estimates were positive. In order to reduce such possible inconsistency, the 1999-2000 series deflates the current price GVA from the RBI’s banking department with the help of the implicit price deflator of commercial banks.

Allocating FISIM

Having discussed the principles and methodology followed for measuring output from the banking sector in India, we now turn our attention towards the allocation of FISIM to the users of such services. This is important for the purpose of maintaining consistency of NAS, particularly to avoid double/multiple counting of the same output. FISIM is allocated as (1) intermediate consumption by enterprises/businesses; (2) final consumption by households and government; and (3) exports to the rest of the world. Thus, FISIM attributed to businesses does not affect the GDP at the national level, while that attributed to households, the government and the rest of the world, which is part of final demand, is part of GDP.

In the 1980-81 series of NAS in India, the imputed services of the banking sector were partly treated as intermediate consumption of industries and partly as final consumption of the government and households. The proportions of imputed services allocated to various sectors was being determined based on the basic data on loans and deposits relating to enterprises and consumers [CSO 1989].8 In the 1993-94 series, the CSO made a refinement in allocating FISIM to user sectors/subsectors by implementing the UN SNA 1993 guidelines to the maximum possible extent [CSO 1999]. Unlike the 1980-81 series, where the FISIM was allocated only to a few organised segments of user industries and consumers, the 1993-94 series allocated FISIM to all user industries. In the latest NAS series (with base year 1999-2000), the CSO has made a few more changes for allocating FISIM from a few sub-sectors, viz, UTI and NBFCs [CSO 2006]. Surprisingly, however, the CSO has not reported the details of the methodology followed in both the 1993-94 and 1999-2000 NAS series to allocate FISIM from different sub-sectors (under the banking sector) to their users. Based on a discussion with experts,9 it is presumed that the output from the RBI is entirely allocated to the government. If so, then the entire output from the RBI is presumed to meet final demand (government consumption), which means increase of national GDP by the same amount.

IV Measures of Banking Services in India: Unsettled Issues and Possible Solutions

The CSO has been making constant and credible efforts at updating estimates of NAS for the country. There have been a lot of changes in methodology and coverage even in the latest three NAS series (namely, the 1980-81, 1993-94, and 1999-2000 series). There might be further scope for improving measurement practices and hence improving the quality of NAS. It is with this spirit that we would like to present some of our reflections on the 1999-2000 series of NAS with regard to the banking sector. We also raise certain relevant points/issues, deliberate upon those and make some suggestions for improvement of the data quality further.

To begin with, an issue that comes naturally is the treatment given to output from the RBI. Whether the RBI as a whole or only its banking department should form part of the banking sector is a debatable issue. For compiling NAS for India, CSO considers the GVA of the issue department of the RBI as part of the government on the grounds that the function of the issue department is akin to a function of the government.

The SNA 1993 guidelines say, “If the monetary authority type functions like issuing of currency or the maintenance of a country’s international reserve is carried out by any agency or agencies of general government, which remain financially integrated with central government and are directly controlled and managed by government itself, such agency’s accounts can be included with that of general government and not otherwise.” Obviously, the CSO’s current practice of bifurcating the RBI has been under debate for long, a point raised earlier by Subba Rao and Nag (1995). The modification published in the January 1996 issue of News and Notes of the UNSD suggests flexibility in SNA 1993 guidelines for treatment of accounts of a central bank, thereby keeping the issue open for further debate and refinement. The important issue here, therefore, is what should be the appropriate treatment of output for the central bank (RBI) for the purpose of NAS? Should a part of RBI’s total output (to the extent of output from the issue department minus operating surplus of the issue department) still be considered as part of the government’s output? Or should the total output of RBI be added to the GVA of the banking sector?

Second, the volatile and occasional negative or small GDP from a central bank/the RBI is an issue of major concern (Tables 1 and 2). Not only India, but several other countries have also faced a similar measurement problem. According to the SNA 1993, the implicit part of the output of central banks should be calculated similarly to that of other financial intermediaries: the difference between the interest receivable less interest payable. But a point then is, how much of a central bank’s output really comes from intermediation services. There is a suggestion now from some forum that where the SNA 1993 treatment leads to inappropriate results consistently, a second best approach would be to measure the output at cost as for other non-market producers [OECD 1998, UN 2006]. Further, OECD (1998) proposes that commercial operations, if any, performed by central banks should be separated and measured using the reference rate approach. Some researchers/experts feel that under no circumstances should a central bank be considered part of the general government sector, regardless of how its output is measured [UN 2006]. Thus, the approach for estimating output from the RBI has to be chosen carefully. The main function of the RBI is to formulate and implement monetary policy, issue and replace banknotes, manage the public debt, supervise the banking sector for stability and provide other banking services. Clearly, the RBI is involved in certain financial intermediation services. But a part of the RBI’s output comes from other services. So, if we apply the FISIM concept, the RBI’s output would possibly remain underestimated. Following the suggestion for central banks in OECD (1998), it might be necessary to measure output from intermediation/commercial activities of the RBI using the reference rate based approach and output from other services based on the cost approach.

The use of the implicit price deflator observed for commercial banks to convert the RBI’s current price output to the corresponding constant price measure might not be very accurate, though this is definitely an improvement over the practices followed in the past. But then construction of an appropriate deflator for the RBI will not be easy as long as the treatment of intermediary and direct service components remain unsettled. At the same time, the presumably current practice of allocating the RBI’s entire output to the government needs further examination. The RBI renders services not only to the government but banks and like institutions also. So, a part of FISIM from RBI should be allocated to those sectors/sub-sectors, which would form part of intermediate consumption.

Third, there is an urgent need to improve upon the current practice of estimating output from the unorganised financial sector, such as unorganised NBFCs. The “one-third rule” currently in use to derive the contribution of unorganised NBFCs from that of organised NBFCs, needs to be thoroughly re-examined. This is a figure which was derived long ago and deserves to be revised/updated in the light of the fast changing scenario of the Indian economy in past two decades, particularly during the post-reform period. So long as we have no mechanism to capture data on the activities of the unorganised sector directly, it might be useful to carry out focused surveys periodically to update such parameters.

Fourth, allocation of FISIM from the banking sector to various user sectors/sub-sectors is an extremely challenging task. Particularly, how to allocate FISIM rendered by mutual funds, NBFCs, etc, has to be addressed properly. The coverage of the subject in available documents is too sketchy [CSO 1989, 1999, 2006].

Fifth, the SNA 1993 recommendations do not consider services rendered by lending from owned funds under FISIM. Although lending from own funds by a financial corporation cannot be precisely defined as FISIM, borrowers of these funds still receive a service from the financial corporation. For example, suppose a financial intermediary earns an interest of Rs 10 from lending Rs 100 from its own funds. A portion of this interest can be considered a service, and, therefore, the output of a financial intermediary.

Sixth, there has been a lot of debate recently around the world on how to measure banking intermediation services at constant prices. Conventionally current price estimates are converted to corresponding constant price estimates by using some quantity/ volume indicators. It would be interesting to construct an implicit price deflator or a price index for such services. But intermediation services are produced and sold in a bundle, and individual services under the bundle are not explicitly charged for. It is extremely difficult to compute the break-up of total output into price and quantity. Nowadays the “user cost” approach has been a potential tool for measuring price/output indices of intermediation services. But a major practical hurdle in implementing this approach lies in its dependence on a rate variable, called “reference rate” or “opportunity cost of money”. The “reference rate” is one “that represents the pure cost of borrowing funds – that is, a rate from which the risk premium has been eliminated to the greatest extent possible and which does not include any intermediation services” [Ahn 1998]. Clearly, the reference rate is a purely theoretical construct, one not observable in the market directly, and thus needs to be estimated from the available data. But a consensus has not yet emerged on how to estimate a suitable reference rate, though empirical literature has proposed several potential approaches in recent years.

V Measuring Financial Intermediation Services: Some General Issues

As has been seen, two major challenges faced by NAS experts are (1) how to allocate FISIM to user sectors; and (2) how to obtain estimates of output at constant prices. In the past, several

Table 1: GDP from Banking Sector (Current Prices) – 1999-2000 Series (Rs crore)

Sub-sector 1999-2000 2000-01 2001-02 2002-03 2003-04 2004-05
Banks 40485 45966 55811 69222 83427 85035
Banking department of RBI 10875 6847 9360 8054 -412 -3156
Post office saving banks 745 887 1016 1072 1203 1271
NBFCs 34088 29475 29931 27682 32097 35412
Cooperative credit societies 7733 8631 10046 11319 12210 13171
EPFO 188 208 222 234 249 265
Total banking sector 94114 93014 106186 117583 128774 131998
Source: Central Statistical Organisation (CSO).
Table 2: GDP from Banking Sector (Constant Prices) – 1999-2000 Series
(Rs crore)
Sub-sector 1999-2000 2000-01 2001-02 2002-03 2003-04 2004-05
Banks 40485 48639 52835 61423 68772 79736
Banking department of RBI 10875 6799 8893 7147 -340 -2959
Post office saving banks 745 855 939 952 1030 1046
NBFCs 34088 30044 29115 27393 27849 29202
Cooperative credit societies 7733 7584 8003 8912 9353 9788
EPFO 188 201 205 208 213 218
Total banking sector 94114 92122 99990 106035 106877 117031
Source: Central Statistical Organisation (CSO).
Economic and Political Weekly September 15, 2007 3759

ad hoc techniques have been employed to solve the problem. But experience says that the application of such techniques may at times result in some inconsistencies. For instance, the application of the deflated credits and deposits of all scheduled commercial banks used for deriving the RBI’s constant price output resulted in an anomaly and the practice is now abandoned in the new NAS series.

The “user cost” approach is useful in this context. This approach can provide the break-up of product-wise output into quantity and price. Thus, knowing the estimates of current year’s quantity of services and the unit price of the base year, it is easy to obtain the output at constant prices. Additionally, the price-quantity break-up of product-wise output would be useful to construct both price and quantity indices for banking intermediation services.

On Estimating Reference Rate

As stated earlier, a major hurdle in implementing the “user cost” approach in practice relates to the estimation of the reference rate itself. Estimates of user cost and FISIM are sensitive to the choice of reference rate and the literature is still not very clear about how to go about estimating the reference rate.

We consider here alternative proxies available for the reference rate (Table 3) and discuss the issues associated with such approaches. First, a solution to the problem would be to consider a suitable market/benchmark rate, such as bank rate, repo rate, or treasury bill/bond rates, as a proxy for the reference rate. The SNA 1993 suggests the interest rate on short-term loans between banks (inter-bank rate) and Fixler (1993), in the context of US banks, considers the average of rates on treasury bills and notes as a proxy for the reference rate. These rates relate to assets with virtually no risk and negligible transaction costs. Under this approach, the problem boils down to the selection of the best representative rate from several available market rates. If one considers the reference rate as a risk free rate, then yields on treasury bills/bonds, repo rate, etc, may appear quite a reasonable proxy. In a country with developed and well integrated financial markets, other benchmark rates, such as the bank rate, may also serve the purpose reasonably well. But this approach ignores banking technology completely and determines a proxy for banks’ reference rate exogenously. In practice, it is also seen that the use of these “benchmark rates” as measures of the “reference interest rate” may lead to highly volatile or even negative FISIM [OECD 1998, Banco de Espana 2005, Samanta 2003, Barman and Samanta 2004].

Second, another potential choice for the reference rate has been the average of the interest rates charged/paid on different assets and liabilities in the bank’s balance sheet. (As reported by Arpaia and Scafuri (1997), this was a practice followed by Statistics Canada. Also see Collins (1993) and OECD (1998).) The choice of reference rate here does not fully disregard banking activities. In their recent studies, Samanta (2003) and Barman and Samanta (2004) have experimented with two types of proxy for reference rate; select benchmark rates and the weighted average of interest earned/paid on assets/deposits, weights being proportional to the respective outstanding amounts. Interestingly, they found that the choice of the second category of reference rate results in relatively more stable and robust estimates of FISIM as well as price/quantity indices. However, this approach is also subject to criticism.10

The third approach refers the estimation of the reference rate endogenously from banks’ production technology. Although empirical literature has a general preference for the first two approaches, primarily due to simplicity and analytical convenience, the only exception found by the authors is Fixler and Zieschang (1992), who applied their theory of extracting the reference rate as an endogenous variable under banking technology. But while applying this approach, one has to be careful in modelling banking technology (including the choice of input and output sets), for the extracted reference rate would depend heavily on the postulated model for banking technology, and thus may be model dependent.

There are certain other issues which need to be addressed irrespective of what approach for reference rate is adopted. For example, the introduction/recognition of term structure would complicate the issue of reference rate substantially. If we view reference rate as a “risk free rate”, then the issue that arises naturally is that there is no reason to believe risk free rate (hence reference rate) to be constant across all maturity periods. The issue has already drawn attention from some quarters (for instance Banco de Espana (2005) experimented with two reference rates: one for short-term maturities and the other for long-term).

It is well documented in the literature and also observed in the Indian context that the zero coupon yield curves on central government bonds need not necessarily be flat. Thus it is very

Table 3: Select Articles/Discussions on User Cost/Reference Rate Based Approach

Article/Note/Discussion Subject of Study Reference Rates Used/Considered Remarks

Ahn (1998) FISIM in Korea

Arpaia and FISIM in Italy Scafuri (1997)

Barman and Samanta Price index of

(2004) banking services (also partially FISIM) in India Simple average of interest rate on loans and deposits

Inter-bank rate Average of inter-bank rate and the rate on long-term securities issues by the state Weighted average of lending and borrowing rates

Bank rate Average call money rate Implicit yield on 91-day T bills Implicit yield on 364-day T bills Weighted average of rates on assets/loans and deposits/ liabilities Paper presents estimate of FISIM (provisional in nature) and its allocation in 1995 FISIM was allocated based on the “financial assets and liabilities approach”.

Provides empirical results for Italy Final and collective consumption of FISIM estimated using the third proxy of reference rate is smaller than estimates using the other two proxies Also impact on GDP is lower when the reference rate is chosen as the weighted average of lending and borrowing rates

Price index is sensitive to choice of reference rate Selection of benchmark rates (i e, the first four alternatives) results in highly volatile estimates of index Selection of the last proxy for reference rate leads to relatively more stable estimates Interestingly, price deflator based on current and constant price FISIMs is not sensitive to selection of reference rate


Economic and Political Weekly September 15, 2007

Table 3: Select Articles/Discussions on User Cost/Reference Rate Based Approach (Contd)

Article/Note/Discussion Subject of Study Reference Rates Used/Considered Remarks
Banco de Espana (2005) FISIM in Spain Internal Reference Rate Presents a trial exercise for seven years: 1995-2001
Interest charged on inter-bank loans divided by balance of these loans The internal reference rate described by the first alternative was selected based on which FISIM was allocated between
Weighted average of the interest rate on inter-bank loans resident users
and securities (other than select categories of shares) Two reference rates: one for short-term loans and deposits; Import and export of FISIM was estimated based on the external reference rate
and the other for long-term loans and deposits
Different weighted averages of (1) rate on loans and deposits; or (2) those rates obtained under different
methods above
External Reference Rate
Weighted average of inter-bank loans and deposits bet
ween resident and non-resident financial institutions
Begg et al (1996) FISIM for France For France The fact that the estimates of FISIM using reference rate based
and UK Weighted average of rates on (intermediated) bond assets and liabilities method for two countries with similar economic structures are reasonably close in magnitude is encouraging
Money market rate There are problems in extending the coverage of FISIM,
Benchmark government bond yield For UK particularly central banks’ note issue In order to allocate FISIM consistent tables of financial balance
three-month inter-bank sterling rate, except for sheets and interest flows for all financial intermediaries have
(1) borrowing and lending in other currencies; and (2) building societies (the source of mortgage lending) to be constructed
For foreign currency, relevant foreign currency inter-bank
rates were used
For building societies, weighted average of rates on
building society loans and deposits
Eggleston (2002) Price index for Central bank lending rate The note discusses broad framework (no empirical results)
banking in US Inter-bank lending rate Weighted average of the interest rates on all banks’
securities holdings
Fixler (1993) Price and quantity 90-day T bill rate Provides detailed theory
indices for banking Presents some empirical results
Fixler and Zieschang Theory for price Modelled using the distance function and user cost Gives detailed theory
(1992) and quantity indices of banking intermediation approach 90-day treasury Required rate to cover the interest cost of liabilities Presents empirical estimates of price and quantity indices of banking output in US for the years 1984-1988 Indices using model based estimates of reference rate
services. Also empirical results for banks in US Rate of return on asset UNSO opportunity cost (opportunity cost) are not practically different from those obtained by using other benchmark rates as proxy for reference rate
Hancock (1985) Theory on user Interest rate either paid on deposits or received on Proposes a theory on user costs, production in financial
costs and others loans firms. Gives empirical results.
OECD (1998) FISIM (discussion in the context of LIBOR Average of deposit and borrowing rates This note discusses issues and problem areas in dealing with FISIM
OECD countries) Presents illustrative example on how to calculate FISIM
based on loans and deposits Makes the observation that use of inter-bank rate as reference
rate may lead to negative FISIM for some sectors in some
countries – suggests using average of deposit and borrowing rates to reduce/eliminate the problem
Palmer (2002) Price index for banking and related Repo rate LIBOR (1 month and 3 months) Reports that the data analysis and anecdotal evidence from discussions with banks suggested that the repo rate was the
services to Swap most widely used as a reference rate to determine pricing
customers in UK
Samanta (2003) Price and quantity indices for banking intermediation Weighted average of the rates on loans/assets and deposits/liabilities Bank rate, call money rate, etc Estimates of indices are sensitive to choice of reference rate Weighted average of rates on loans/assets and deposits/
services in India liabilities helps reduce volatility in the estimates of indices
Srimany and Bhattacharya (1998) A review of the measures (both Implicit rate for the 91-day treasury bills sold in auction This paper provides a rich literature review on the subject and some issues for Indian context
FISIM and indices)
for financial servi
ces with special
reference to
banking in India
Statistics Finland (2005) FISIM in Finland Internal Reference Rate Average interest rate of producers of FISIM Average of rates on loans and deposits between credit institutions External Reference Rate The note provides the methodological revisions to national accounts in 2005 No empirical results of FISIM are presented, though illustrative example for estimation and allocation of FISIM are shown
Average of interest rate of loans and deposits between domestic and foreign credit institutions

clear that there would be different risk free rates (hence, reference rates) for different maturities. As the components of asset portfolios of banks generally have relatively longer maturity (as compared to deposit/liability portfolios) and wider/variable maturity patterns, it is likely that the constancy assumption of the reference rate would induce an element of measurement bias in estimation of FISIM from loan/asset portfolios.

Single or Multiple Reference Rate Structure?

Even if we resort to some of the strategies mentioned above for estimating the reference rate, another question is, should we use the same reference rate for all segments of the banking sector? For example, would the reference rate for estimating FISIM from commercial banks be identical with that for the central bank (RBI)? The answer to this question is not obvious. It may be necessary to have multiple reference rate series, each being applicable to a specific segment of the banking sector. Recently, Banco de Espana (2005) and Statistics Finland (2005) have proposed using two different types of reference rates for the purpose of allocating FISIM: the internal reference rate for the domestic sector and the external reference rate for the rest of the world (import and export of FISIM). The issue is not well discussed in the literature and deserves in-depth investigation.

Suitability of User Cost Approach

While estimating FISIM by the reference rate based approach, it would be necessary to examine the suitability of implementing the underlying method/concept. Putting it differently, we need to know when this approach works well and when not. As is well documented in the literature, a necessary precondition of meaningful estimates from the “user cost approach” is a deregulated environment in which banks behave like profit maximising firms, facing interest rates fully determined by market forces. Thus, the estimates provided by the user cost approach would be crucially related to the profitability of the banking sector. If the risk of default is high, banks might not be willing to disburse more credit as the amount disbursed might turn into a non-performing asset (NPA). If NPAs of banks increase, the effective returns from these assets would decrease. In such a situation, banks might tend to deploy a substantial portion of their funds in approved securities. Thus, if the profitability of the banking sector decreases, returns from advances would become closer to the return from the benchmark/reference rates and for some periods, might be less than these rates. This would lead to negative and at times highly volatile output. Thus, in the strict sense the scope of the user cost approach is limited in the context of many developing countries, where fixed/ administered interest rate mechanisms are in existence or NPAs are a major issue. However, considering that in India the administered interest rate regime is now dismantled, except for a very small segment, and NPAs have also reduced greatly, this may not be viewed as a serious limitation.

VI Concluding Remarks

Measuring output from financial intermediation services is extremely difficult. For the purpose of NAS of a country, allocation of output from such activities to their users is another challenging problem. There has also been a growing concern in recent years over how to construct price and quantity/output indices for banking intermediation services. The SNA 1993 provides a broad framework on the first two aspects, and the topic along with indices for intermediation services is dealt with in Hancock (1985), Fixler and Zieschang (1992) and Fixler (1993), among others. In India, SNA 1993 guidelines are implemented to the maximum possible extent in the latest two NAS series (i e, 1993-94 and 1999-2000 series). But there are still many unsettled issues which deserve careful treatment. Constructing price and output indices of banking intermediation services is also the need of the hour. This paper raises some important issues related to these aspects. We believe in-depth research is needed to address these vital issues.




[The authors are thankful to S L Shetty for his constant encouragement and suggestions on the work. The authors are also thankful to the participants in the seminar for their feedback and suggestions. The views expressed in the paper are purely personal and do not represent the opinions of the organisation with which they are associated.]

1 The SNA 1993 defines imputed bank service charges in the same way as the SNA 1968 but gives them a more precise description, namely, FISIM. The word “imputed” was dropped in order to emphasise that just like the bank services directly charged for, the intermediation(free) services are really being produced and consumed; the only difference being that they have to be measured by indirect methods [OECD 1998].

2 In post-independence India, the first series of NAS corresponds to the base year 1948-49 (released by the CSO in 1956). The methodology has undergone revisions from time to time, coverage has improved over time and the base year has been shifted periodically in the light of the latest economic development and growth. Accordingly, different NAS series published earlier by the CSO had pertained to the base years 1960-61 (shifted from 1948-49 to 1960-61 in August 1967; 1970-71 (in January 1978), 1980-81 (in February 1988), 1993-94 (in February 1999) and the latest series with base year 1999-2000 has been releasedin January 2006 [CSO 2006].

3 Although the banking sector provides valuable services and is considered an important and integral segment of the financial system of an economy, debate continues on what the sector exactly produces and how to measure the corresponding output. For the sake of brevity, discussions on the literature and relevant debates are skipped here.An interesting review of the methodology for measuring output from financial intermediation services with certain illustrative examples is available in Srimany and Bhattacharya (1998).

4 This issue refers to, for instance, the debate on input-output status of deposits.

5 In order to avoid possible confusion, and thus to enhance clarity in presentation, we refer to the “reference rate based approach” synonymously to the “user cost approach”.

6 If banks are treated like any other production enterprise, their income in the production account would be limited to the charges made on customers, which would mean that the banks have negative/small operating surplus and most likely negative/small value added [CSO 1989].

7 This section borrows heavily from three publications by CSO, viz, CSO (1989, 1999, 2006).

8 A similar approach is also followed by many other countries. For instance, in France and the UK, the basis of allocation of FISIM by industry/sector (household, company, government and foreign) is the sector-wise information on assets and liabilities [Begg et al 1996].

9 Participants in the seminar titled ‘Growing Size of the Services Sector in the Indian Economy: A Database Seminar’, March 29, 2006, organised jointly by Indira Gandhi Institute of Development Research (IGIDR) and EPW Research Foundation (EPWRF) at IGIDR, Mumbai.

10 Arpaia and Scafuri (1997) argue that in this case distribution of FISIM between final and intermediate consumption would be more stable but this would be at the cost of losing the link with the market rate.


Ahn, Kil-Hyo (1998): ‘The Estimation and Allocation of FISIM in Korea, 1995’, Joint OECD/ESCAP Meeting on National Accounts, Nangkok, May 4-8.

Arpaia, Alfonso and Emilia Scafuri (1997): ‘The Impact of FISIM on GDP: An Empirical Evaluation on Italian Data, STD/NA(97)27’, http://www.

Banco de Espana (2005): ‘Registering Financial Intermediation Services on the National Accounts as of 2005’, be/notes.

Barman, R B (1995): ‘Impact of Liberalisation on Collection of Data for Estimation of National Income and Related Aggregates: Banking and Financial Institutions’, The Journal of Income and Wealth, Vol 17, No 1, January, pp 26-30.

Barman, R B and G P Samanta (2004): ‘Banking Services Price Index: An Exploratory Analysis for India’, IFC Bulletin, No 19, Irving Fisher Committee on Central Bank Statistics, International Statistical Institute, November, pp 110-24.

Begg, Iain, Jacques Bournay and Martin Weale (1996): ‘Financial Intermediation Services Indirectly Measured: Estimates for France and the UK Based on the Approach Adopted in the 1993 SNA’, Review of Income and Wealth, Series 42, No 4, December, pp 453-72.

CSO (1989): National Accounts Statistics – Sources and Methods, Central Statistical Organisation, Department of Statistics, Ministry of Planning, Government of India.

  • (1999): Brochure on New Series on National Accounts Statistics (Base Year 1993-94), Central Statistical Organisation, Department of Statistics and Programme Implementation, Ministry of Planning and Programme Implementation, Government of India, May.
  • (2006): New Series of National Accounts Statistics (Base Year 1999-2000), Central Statistical Organisation, Ministry of Planning and Programme Implementation, Government of India, February.
  • Eggleston, Deanna (2002): ‘US Producer Price Index for Banking’,

    17th Voorburg Group Meeting, Nantes, France, September.

    Fixler, Dennis J (1993): ‘Measuring Financial Service Output and Prices in Commercial Banking’, Applied Economics, Vol 25, pp 983-93.

    Fixler, Dennis J and Kimberly D Zieschang (1992): ‘User Costs, Shadow Prices, and the Real Output of Banks’ in Z Griliches (ed), Output Measure ment in the Service Sector, University of Chicago Press, pp 219-43.

    Hancock, D (1985): ‘The Financial Firm: Production with Monetary and Non-Monetary Goods’, Journal of Political Economy, Vol 93, pp 859-80.

    OECD (1998): ‘FISIM’, Joint OECD/ESCAP Meeting on National Accounts, Organisation for Economic Cooperation and Development, Bangkok, May 4-8.

    Palmer, Nick (2002): Challenges in the Development of a Price Index for Banking Services in the UK, Mini Presentations on Producer Price Indices, Voorburg.

    Samanta, G P and K Bhattacharya (2000): ‘Prices of Financial Intermediation Services: An Index Based on Spread’, International Journal of Development Banking, Vol 18, No 1, pp 71-76.

    Samanta, G P (2003): ‘User Cost Approach for Production Index of Banking Services: The Case of the Indian Economy’ in Das, Tarun, Rajaram Dasgupta, Rohit Kumar Parmar and Asish Saha (eds), Preparation of an Index of Services Production – Seminar Papers and Proceedings, National Institute of Bank Management (NIBM), India, January 3-4, 2002 (Chapter 5, pp 127-48).

    Srimany, A K and K Bhattacharya (1998): ‘Measures for Financial Services: A Review with Special Reference to Banking in India’, Reserve Bank of India Occasional Papers, Vol 19, No 1, March, pp 1-38.

    Statistics Finland (2005): Methodological Revisions to National Accounts in 2005,

    Subba Rao, K G K and A K Nag (1995): ‘Production Account in the Revised System of National Accounts (SNA) 1993 with Reference to the Financial Sector’, The Journal of Income and Wealth, Vol 17, No 2, July, pp 119-23.

    United Nations (1993): System of National Accounts, 1993, Washington DC.

    Call for Papers for a Conference on

    “Growth and Macroeconomic Issues and Challenges in India”

    Institute of Economic Growth, Delhi February 11-12, 2008

    With the increased globalisation of the Indian economy large amounts of foreign capital has come in. This poses both challenges and opportunities for financial, monetary and real economic functioning and policy making in this country. It is also expected to change the structure and behaviour of many economic relations. To discuss these impacts and its policy implications in the broader framework of open-economy macroeconomics, the Institute of Economic Growth, as part of its Golden Jubilee Celebrations, proposes to organize a two day conference.

    Research papers (of 7000-10,000 words) are invited on the following six sub-themes for discussions and deliberations:

  • 1. Issues relating to growth and macroeconomic behaviour
  • 2. Changing roles of fiscal and monetary policies
  • 3. Search for new channels of policy transmission
  • 4. Exchange rate regimes and capital account convertibility
  • 5. Impact of opening up of the economy on trade, investment and labour markets.
  • 6. Issues relating to the level of foreign exchange reserves and external shocks.
  • Papers are expected to focus on the Indian Economy. Please send full papers to Dr N R Bhanumurthy, conference organizer, by 15th December 2007 through e-mail ( All the papers would be refereed and the authors of accepted papers would be informed by the end of December 2007. The Institute has limited travel grants and will reimburse domestic travel to and fro Delhi for one author per paper selected for the conference.

    Kanchan Chopra Delhi Director

    Dear Reader,

    To continue reading, become a subscriber.

    Explore our attractive subscription offers.

    Click here

    Back to Top