Are Estimates of the Economic Contribution of Financial Services Reliable?

Brian Sturgess - June 2017


Key Points
  • The methods used to estimate the contribution of financial services to national income are seriously flawed. Banking sector output in the UK was estimated to have increased in 2008 while the financial services sector was collapsing.
  • The relative contribution of service activities in GDP is not easy to measure, but there are many problems in measuring financial services in general and the output of banks in particular.
  • National income accounting standards, used to estimate the output of financial intermediation companies such as banks, rely on flawed indirect measurements based on interest rate spreads. Furthermore, many services are provided at no charge so price indexes cannot be meaningfully created.
  • The main method used, Financial Intermediation Services Indirectly Measured (FISIM), is arbitrary and fails to measure the quality of banking assets and risk. Over the period 2003–7, one study found that aggregate risk-adjusted output would have been only 60% of officially estimated output across the Euro area.



Introduction

During the period between 1997 and 2007 the financial services sector’s output in the UK grew at over 6% per annum according to official statistics. Oddly, measured growth continued into 2008 while survey data showed collapsing bank lending as a result of the severe market disruption that began in the middle of 2008. This anomaly places a large question mark over the reliability of measurements of the economic contribution of financial services to the economy.

Measuring the size of the sector’s contribution is extremely important. The impact of the United Kingdom’s decision to leave the European Union on the size, growth and future prosperity of the country’s financial services sector has attracted a great deal of comment. There have been fears that the pre-eminent position of the City of London as a global financial services hub will be diminished as financial firms move, or are lured away, to alternative European cities such as Paris, Frankfurt or Dublin. Anxiety that this might happen seems justified given the importance of banking, insurance and other financial services in providing employment and in contributing to economic activity and the prosperity of the UK. In 2016, it is estimated that financial and insurance services contributed £124.2 billion in gross valued added (GVA), or 7.2% of the UK’s total GVA. The sector also generated a trade surplus of over £60 billion for the country that year and in the financial year 2015–16, the banking sector alone contributed £24.4 billion to the Treasury in corporation tax, income tax and through the bank levy.[1]

In sharp contrast, it is often argued that the financial services sector in the UK is too large in relation to the rest of the UK economy. Furthermore, the inward liquid capital flows that it stimulates produce less desirable consequences such as an overvalued exchange rate, to the detriment of manufacturing and to the economic development of the country outside its south-eastern corner.[2] There are over 1 million jobs in the financial and insurance sector, 3.1% of all UK jobs, but 60.4% of the total financial and insurance sector GVA in 2015 was generated in London and the South-East.[3]  A related concern, which has been intensified by the 2008 financial crisis, is the fear that the financial sector in some countries may grow too quickly, becoming too large in size for regulators to ensure macro-economic stability. This begs the question of whether or not there is an optimal size for the financial services sector in any economy, developed or emerging.

However, to assess the validity of all of these concerns requires accurate measurement of the contribution of financial services to the national economy. National income statisticians have found measurement of the services sector more problematic than estimating GVA in manufacturing and agriculture, but establishing the value of the output of financial services is particularly challenging. The significance of financial services in an economy can be measured by estimates of financial depth – such as the ratio of credit or banking assets to GDP. These measures suffer from many problems, but do allow crude international comparisons to be made. Alternatively, the significance of the sector can be measured by the direct stand-alone contribution of the financial services sector to GDP through its GVA. This may in turn under- or overestimate the sector’s contribution to the economy since a functional financial services sector may have positive or negative external spillover effects. Relative ease of access to credit may help other industrial sectors to grow or may lead to dysfunctional excessive borrowing.

To estimate the sector’s contribution through the GVA method it is necessary to be able to measure the output of financial firms accurately, but all evaluations face practical and theoretical challenges. That is a problem for all statisticians involved in measuring output in finance and not one specific to the United Kingdom. Financial services output data in the United Kingdom are compiled in accordance with international best practice, but great care is often needed when interpreting them. The United Nations-recommended Standard System of National Accounts (SNA) only tackled the question of measuring financial sector output in SNA 1993, with an indirect method recommended to measure activities where fees are not charged, which has been refined in SNA 2008, but these methods are still very much work in progress.

In this paper the second section considers the application of standard methods and definitions to estimate the size and growth of financial sector output in the United Kingdom. The history of the sector’s real GVA growth is outlined and the estimated contribution of subsectors – banking, insurance, pensions, financial and other auxiliary activities – is enumerated. Measuring real output in each of the subsectors faces different theoretical and practical issues, but this paper focuses on measuring the contribution of financial intermediation – central banking, banking, building societies, etc. – for two reasons. Using SNA methods, this subsector is estimated to be the largest component of the entire financial services sector in the UK, accounting for around 57% of sector GVA and 4.4% of GDP.[4] Secondly, measuring output in this subsector is fraught with problems because many services provided are not charged for directly and a great deal of the income earned is related to interest rate spreads, asset values and risk-bearing activities. The third section analyses in depth the problems involved in measuring the output of the activities of banks and evaluates critically the evolution and effectiveness of the methods proposed by national income-accounting specialists to solve these issues. Many challenges remain, but the lack of consistent data and the slow and partial adoption of existing, imperfect, recommended standards across countries means that international comparisons based on the relative size of the financial services sector are meaningless. The fourth section concludes.


The size of the financial service sector

In order to measure the output and value of the financial services sector, first the sector must be defined. This section outlines the main activities within the financial services sector and reviews its historical growth and significance in GDP in the UK. For the purposes of national income accounting, the ‘financial services’ sector is defined to be the set of firms that are classified under Division J (‘Financial Intermediation’) in the Standard Industrial Classification (SIC) 2003 system.[5] The five main industry groups comprising the sector, along with their SIC numbers, a description of activities, their weights in GDP and a summary of some of the methods used to estimate output, are shown in Table 1. 





Official measurements of the growth of real GVA in the financial services sector using the above definition demonstrate interesting historical patterns of growth. Figure 1 compares rates of growth in the financial services sector with rates of growth in GDP for selected periods, roughly over the last century. These data can be contrasted with data supplied in a Bank of England paper by Burgess (2011), which surveys the period from 1856 onwards. The author notes that the financial sector deepened considerably over the period up until 1913, just before the First World War, with the sector growing at an average annual rate of 7.6%, more than three times the average annual growth rate in real GDP of 2.0%.





Between 1914 and 1970 financial services grew at a real annual average rate of 1.5%, which can be compared to an average rate of growth in real GDP of 1.9%. This period witnessed two world wars, the slump of the 1920s in the UK, the global depression of the 1930s and tighter regulations on the provision of credit, which were not relaxed until 1971. The sector again grew faster than GDP over the period 1971–96 with an average 0.5% per annum advantage. This expansion continued and financial services grew at rates well above GDP growth from 1997 onwards until 2007, averaging 6.1% per annum, more than twice the growth of 3.0% per annum in real GDP. Surprisingly, measured financial services output also grew strongly during 2008, by 5.0% per annum in real terms, in contrast with indicative surveys of financial sector output, which fell back at the onset of market disruption in mid-2007. There was a sharp reduction in lending growth and the provision of some financial services to the rest of the economy around that time, but financial markets themselves were very active. In 2008, the sharp tightening of credit conditions helped produce a contraction in GDP of –0.1%. Thereafter, despite this anomaly, measured financial services output reflected the problems faced within the banking sector and output fell more than GDP during the period 2009–10 and the sector has failed to recover, as the rest of the economy did, from 2011 until the end of 2016. Figure 2 shows real year-on-year changes in financial services sector GVA and the share of the sector in total UK GVA.





Within the subsectors of the financial services sector listed in Table 1, monetary financial institutions – banks and building societies – account for around 55% of value added, while other forms of financial intermediation accounted for a further 9%. These two categories make up the bulk of the sector and both rely heavily on an artificial construct devised by national income-accounting statisticians to measure nominal output indirectly known as Financial Intermediation Services Indirectly Measured (FISIM), as shown in the fifth column of Table 1. This method is recommended as best practice but, as will be discussed in the third section of this paper, it is far from perfect and is a cause of a number of the anomalies recorded, such as the perverse behaviour of the sector in relation to GDP in 2008.

Banking measurement: Problems evaluating banking output The early years of national income accounting, responding to the fiscal needs of war-time governments in the 1940s, concentrated on measuring contra flows of goods and services and money, with the latter playing the role of unit of account only. The role of the financial sector in overall economic activity was relatively neglected until the late 1970s when effort was put into the measure of the money supply through money aggregates because of the shift in the focus of macro-economic policy from employment to combatting inflation.

The contribution of any sector to the economy is generally measured by its GVA, which is the value of the gross output generated by the sector’s activities minus the value of intermediate consumption or the inputs used in production. But the output of the financial sector, like many service sector activities, is very different from manufacturing and it has been mostly measured indirectly, throwing up many theoretical and practical difficulties. Most sectors charge a price per unit or a fee for the goods and services supplied and pay an explicit cost for the inputs consumed in the process of production.This facilitates the direct measurement of GVA. However, only a part of the income of financial services companies is received directly in fees.

This paper will concentrate on problems in measuring the banking sector’s contribution to GDP because, in the UK at least, the industry group, Monetary Intermediation, includes central banking, banking and building societies, for two reasons. This industry is the largest part of the financial services sector, accounting for 57% of the total financial services sector and around 4.4% of GDP in 2016 and it is the area where the most difficulties arise and where the utmost caution is necessary when using official data.

The Office for National Statistics measures the real value of banking output by direct and indirect methods. Direct methods estimate the nominal value of fees generated, which are then deflated using a price index. Investment banks such as Goldman Sachs do charge fees to clients for advice in corporate transactions and for underwriting the issue of bonds or equities. In commercial banking some fees are also earned from overdrafts, international transactions or loan applications. In the UK, as shown in Table 1, real volume indicators of these services are created by deflating revenue data using the adjusted earnings (AWE) series for the financial services industry, excluding bonuses and adjusted for changes in productivity.Direct methods can be criticised in many ways, but the indirect methods used to measure other banking activities are close to fantasy.


Indirect measures
According to Haldane (2010) financial intermediaries ‘provide services to consumers, businesses, governments and the rest of the world for which explicit charges are not made’, including:

• accepting, managing and transferring deposits 
• providing flexible payment mechanisms such as debit cards 
• making loans or other investments and
• offering financial advice or other business services.

The task of attempting to measure the volume of such services provided directly and deflating the estimated nominal value with an appropriate price index is almost impossible. Instead, indirect measures of banking output concentrate on non-fee income as the reward for providing a large range of services to customers. The largest part of the earnings from banking in most countries, 40% of gross output in the UK,[6] is derived from a very different activity where income is earned not by producing an output, but by charging for interest on loans, which are banking sector assets, an excess over received deposits, which are banking sector liabilities. In economic terms banking sector total income is derived from the net rate of return on total assets employed, both tangible and intangible. The resulting net receipts of interest rate spreads are used to offset banking expenses and provide an operating surplus. Earning income from these spreads avoids the need to charge customers individually for services provided. From a national income-accounting perspective the banking sector’s contribution to gross output can then be estimated in theory by multiplying the return on net sector assets employed, after depreciation, by total assets employed. The latter can be sourced from balance sheet data, although in measuring banking assets assessing the relative quality of balance sheet assets is not straightforward, even when these are equal in nominal value.

The activity of financial services in general, and of banks in particular, has long been a challenging area for those who develop international standards. Guidance on measuring GDP in order to provide internationally comparable datasets is provided by the United Nations System of National Accounts. The issues involved in measuring the economic contribution of financial services activities that generate income through interest rate margins were originally addressed in the 1993 SNA update and refined in SNA 2008. The rules adopted in the SNA 1993 applied to European Union Member States through the European System of National and Regional Accounts (ESA95), with which Member States are legally required to comply, and were amended by ESA 2010.

The concept of Financial Intermediation Services Indirectly Measured (FISIM) was introduced in the SNA 1993 in order to estimate the value added by banking activities through financial intermediation. FISIM nominal output generated by banks should then be allocated between the users of the services for which no explicit charges are made. FISIM employed by national income statisticians is based on an imputed figure, to measure the ‘spread’ between a risk-free or ‘reference’ interest rate and a market lending rate, which is then multiplied by the stock of outstanding balances. The SNA 1993 guidelines defined the ‘reference rate’ as representing ‘the pure cost of borrowing funds – that is a rate from which the risk premium has been eliminated to the greatest extent possible, and that does not include any intermediation services’. According to the SNA 1993 the reference rate could be approximated by an inter-bank lending rate such as LIBOR or by using a central bank lending rate.

In effect the SNA 1993 states that any observed interest rate can be decomposed into three elements:[7]

 • pure cost of borrowing (=reference rate)
 • risk premium and
 • intermediation service. 


The value of the intermediation service (FISIM) is the difference between the first two elements. The impact of applying FISIM to estimate the nominal value of the services offered by the banking sector can be illustrated by the added value generated by banks through loans and the value of the services provided to depositors; a simple example is provided by Burgess (2011).

If a bank provides a loan of £10,000 charged at an annual rate of 9% and the Bank of England’s Bank Rate is 4% then the imputed service charge on the loan (the price of the service provided by the bank for lending money and accepting risk) over a year would be:

 Loan FISIM = (9% – 4%) × £10,000 = £500

Looking now at a bank deposit of £10,000 paid into the bank by a customer and carrying a deposit rate of interest of 5%, the imputed service charge (for the services provided by the bank) over a year would be:

 Deposit FISIM = (6% – 5%) × £10,000 = £100 

In this example total FISIM, or the implicit price of all banking services generated, would be £600. For each loan or deposit the total nominal value of output is simply the implicit price (FISIM) multiplied by volume, which is taken as the size, or monetary value, of the banking asset or liability.


Despite this simple example, in practice this method of calculating and allocating FISIM requires detailed sector data on stocks and interests for loans and deposits. It is almost impossible to calculate FISIM for all loans and deposits so an average rate and a total stock of loans and deposits are calculated for the sectors of the economy that banks deal with. The data are collected from various sources, but the degree of detail differs over time. The data to calculate FISIM output generated by the UK banks and building societies, namely monetary financial institutions (UK MFIs), are obtained from different sources for each of two time periods: From 1999 detailed data have been collected by the Bank of England from its own specially designed inquiries and, before 1999, detailed stocks data are sourced from the Bank of England and interest data are derived by ONS from the effective interest rates used elsewhere in the National Accounts.

The SNA 1993 did acknowledge that. although FISIM was conceptually satisfactory, there are practical problems in allocating between different users, even where the required data are available. SNA 1993 allowed two approaches which, with data deficiencies and different practices by national statistics offices (NSOs), restricts the international comparability of estimates of banking output even across areas such as the European Union.

Approach 1: Allocation of FISIM into a ‘nominal’ sector. The SNA 1993 permits a simplified approach, where (by convention) FISIM output is not allocated between users but is treated as absorbed by the intermediate consumption of a ‘nominal sector’. In consequence, the estimate of FISIM is not allocated into user sectors or industries. In this approach, GDP is not affected by the size of the FISIM output. ESA95, as originally published, did not require FISIM allocation to be introduced in National Accounts, because EU countries had concerns about the availability of source data and the reliability of the methodology. This approach is currently used in the UK.

Approach 2: Allocation of FISIM into user sectors. The recommended approach involves a full allocation of the use of FISIM across relevant sectors and industries. The purpose of allocating FISIM by sectors and industries is to identify the purchase of these services explicitly and to classify this as intermediate consumption, final consumption expenditure or exports according to which sector incurs the expenditure.

The FISIM measure of the implicit price of services is based on interest spreads, which are subject to monetary policy, the state of the economy and to competition within the banking sector. To prevent this having an impact on the price used for valuation purposes a further assumption is made when the nominal value of financial intermediation is deflated to estimate the real value of output. To introduce a constant price measure the interest rate spread is fixed at a historic base-year value and then the nominal values of loans and deposits, used as indicators of volume, are adjusted using the GDP deflator (excluding FISIM). For example, the real value of the £10,000 loan would then be:

 Real Loan FISIM = (Spread over Bank Rate in base year)*£10,000*GDP deflator in base year GDP deflator in current year

Despite the glaring deficiencies of the FISIM methodology the only improvements made since the introduction of the SNA 93 have been amendments to the reference rates employed in calculating spreads.


Measuring asset quality and risk

The interest rate spread used in FISIM measures the perceived risk that banks are taking at the margin so using the total stock of loans does not differentiate between assets of different quality and default risk. How much should the contribution of the banking sector to the economy by measured by the role of banks in bearing risk by aggregating and managing assets of differing qualities? The actual transfer of funds between borrowers and lenders is only a part of the process by which banks engage in financial intermediation. Banks screen and monitor potential and actual borrowers, acquiring valuable information that could aid in the allocation of capital and the management of risk. But this is not accounted for in the FISIM measure of banking services and real output. According to Burgess (2011): 

Banks with better risk management techniques should be regarded as providing higher quality services, and therefore as generating higher output when they provide finance. But this activity is almost impossible to measure ex ante. Conversely, the impact of poor decision making may only become apparent years later, and cannot easily be reflected in estimates of output when a loan is first made.[8]


In contrast, Haldane (2010) confirms that, although banks do take on risks on behalf of companies and individuals, it is not clear that bearing risk itself is a productive activity. Investment is a risky activity, but it is a fundamental feature of capital markets and not confined specifically to banks.

The SNA 1993 guidelines using risk-free policy rates as a reference rate assume that compensation for bearing risk is a major part of measured nominal output, but this can lead to odd results. If there is a stability crisis that raises the level of liquidity risk and the expected rate of corporate and individual defaults rises, banks will raise rates charged to cover expected losses. The FISIM method would record this as increased income and would indicate an increase in nominal output, unless the value of loans and deposits are not reduced in compensation for the fall in asset quality. For example, in 2008 interest rate spreads widened dramatically as the Great Financial Crisis, which had started in 2007, impacted heavily on the banking sector. Official statistics, reflecting these spreads, estimated real growth of UK financial services of 5.0% in 2008 compared with a decline of –0.1% in real GDP. The banking sector was close to collapse, but it still showed robust growth. (See Figure 2.) For most sectors signs of troubled economic times ahead would normally be detected by a decline in output and the sector’s contribution to aggregate value added. The problems with the measurement of financial services mean that at present the estimates are merely filling a gap in the measurement of GDP and have very little use for any other purpose, particularly as a tool for monitoring macro-prudential stability.

To be useful FISIM needs to take the cost of bearing risk into account separately from the nominal cost of providing banking services. For example, if a bank lends £100 million to an AAA-rated company it bears risk and is compensated for this by the difference between the reference rate and the rate charged and this is attributed as output within the national accounts framework. In contrast, if households hold newly issued bonds worth £100 million in the same company there is no output attributed despite the similarity in the two activities from a risk-bearing perspective. There are a number of ways to strip out the compensation for bearing risk from the cost of providing services. One solution is to move away from using the risk-free rate as the reference rate within the FISIM framework. It has been argued that the FISIM calculations should employ reference rates that match the credit risk and maturity profiles of loans and deposits. For example, if a bank lends £10 million for one year to a company at a loan rate of 8% and the risk-free rate is 5%, the loan FISIM would estimate bank output at £300,000. However, if the bank assesses the company as AAA-rated and the current rate on AAA-rated credits for one year was 6%, the risk-adjusted estimate of bank output on the loan would be only £200,000, a substantial reduction. This method of adjusting for risk could have a large impact on the size of FISIM output estimates. One study calculated that over the period 2003–7 aggregate risk-adjusted FISIM would be only 60% of estimated FISIM across the Euro area.


Conclusion

It is clear from the above discussion that the anomaly of 2008, showing banking output rising while the financial sector was in a state of deep crisis, is not a mere statistical wrinkle that can be ironed out by greater refinements of current methods of measuring the output of the financial services sector in general and of the banking industry in particular. The heavy reliance on the FISIM methodology introduced in the SNA 93 means that international comparisons of financial sector output and the relative contribution to different economies are almost useless. Changing the reference rate used to make adjustments for risk shows the unreliability of current methods. Until a better means of measuring banking output is derived any conclusions drawn by researchers from official data should be treated with extreme caution.


 
Reference
Burgess, S. (2011) Measuring financial sector output and its contribution to UK GDP. Bank of England Quarterly Bulletin Q3, pp. 234–45.

Haldane, A. (2010) The contribution of the financial sector – miracle or mirage? Speech by Mr Andrew Haldane, Executive Director, Financial Stability, of the Bank of England, at the Future of Finance conference, London, 14 July 2010, Bank of International Settlements. Retrieved from www.bis.org/review/r100716g.pdf. Accessed 19 May 2017.

Tyler, G. (2017) The financial sector’s contribution to the UK economy. Patliamentary research briefing, 4 April. Retrieved from http://researchbriefings.parliament.uk/ResearchBriefing/Summary/SN06193. Accessed 19 May 2017.

[1] Tyler (2017).
[2] Tyler (2017).
[3] A further 6.9% was generated in Scotland. See Tyler (2017).
[4] The weights are based on the share of value added in the base year 2006. See Burgess (2011).
[5]Under this system, firms are classified into industries according to the type of goods and services they are mainly involved in producing, not by their ultimate ownership.
[6] Burgess (2011).
[7] Paragraph 6.128 of SNA 93.
[8] Burgess (2011), p. 238.