`Services so Expensive?
`
`Thomas Philippon, New York University
`
`Abstract.
`Despite its fast computers and credit derivatives, the current financial system does not seem
`better at transferring funds from savers to borrowers than the financial system of 1910.
`
`"/would rather see Finance less proud and Industry more content."
`Winston Churchill, 1925
`
`The role of the finance industry is to produce, trade and settle financial contracts
`that can be used to pool funds, share risks, transfer resources, produce information
`and provide incentives.
`
`Financial intermediaries are compensated for providing these services. Total
`compensation of financial intermediaries (profits, wages, salary and bonuses) as a
`fraction of GOP is at an all-time high, around 9% of GOP.
`
`What does society get in return? Or, in other words, what does the finance industry
`produce? I measure the output of the finance industry by looking at all issuances of
`bonds, loans, stocks (IPOs, SEOs), as well as liquidity services to firms and
`households. Measured output of the financial sector is indeed higher than it has
`been in much of the past. But, unlike the income earned by the sector, it is not
`unprecedentedly high.
`
`Historically, the unit cost of intermediation has been somewhere between 1.3% and
`2.3% of assets. However, this unit cost has been trending upward since 1970 and is
`now significantly higher than in the past. In other words, the finance industry of
`1900 was just. as able as the finance industry of 2010 to produce loans, bonds and
`stocks, and it was certainly doing it more cheaply. This is counter-intuitive, to say
`the least. How is it possible for today's finance industry not to be significantly more
`efficient than the finance industry of John Pierpont Morgan?
`
`What happened? Why did we get the bloated finance industry of today instead of the
`lean and efficient Wal-Mart? Finance has obviously benefited from the IT revolution
`and this has certainly lowered the cost of retail finance. Yet, even accounting for all
`the financial assets created in the US, the cost of intermediation appears to have
`
`VERSATA EXHIBIT 2054
`SAP v. VERSATA
`CASE CBM2012-00001
`
`
`
`increased. So why is the non-financial sector transferring so much income to the
`financial sector?
`
`One simple answer is that technological improvements in finance have mostly been
`used to increase secondary market activities, i.e., trading. Trading activities are
`many times larger than at any time in previous history. Trading costs have
`decreased, but I find no evidence that increased liquidity has led to better (i.e., more
`informative) prices or to more insurance.
`
`Measuring the Cost of Financial Intermediation
`The sum of all profits and wages paid to financial intermediaries represents the cost
`of financial intermediation.
`
`There are various ways to define the size of the financial sector. Conceptually, the
`measure is:
`
`Cost = Income of Finance Industry / Total Income
`
`
`
`The three most important issues are:
`1. Definition of “Finance.” For the most part, financial activities are classified
`consistently over time (but sub-sectors within finance are not). The main
`issue is with real estate. The value added of the “real estate” industry
`includes rents and imputed rents for homeowners. Whenever possible, I
`exclude real estate. In my notations, all variables indexed with “fin” include
`finance and insurance and exclude real estate.
`2. Definition of “Income.” The best conceptual measure is Value Added. In this
`case, “Cost” is GDP of the finance industry over the GDP of the US economy.
`However, this is only acceptable if we can exclude real estate, or at least
`imputed rents. When this is not possible, a good alternative is to use the
`compensation of employees. In this case, “Cost” is the compensation of
`employees in finance over the total compensation of employees in the US. For
`the post-war period, the two measures display the same trends, even though
`annual changes can differ. This simply means that, in the long run, the labor
`share of the finance industry is the same as the labor share of the rest of the
`economy. In the short run, of course, profit rates can vary.
`3. Definition of “Total Income.” During peacetime and without structural
`change, it would make sense to simply use GDP. WWI and WWII take
`resources away from the normal production of goods and services. Financial
`intermediation should be compared to the non-war related GDP. To do so, I
`construct a measure of GDP excluding defense spending. This adjustment
`makes the series more stationary.
`
`
`
`
`
`Figure 1: GDP Share of Finance Industry
`
`
`
` I
`
` measure this cost from 1870 to 2010, as a share of GDP, and find large historical
`variations, shown in Figure 1 (with the various data sources, see Philippon 2011 for
`details).
`The first important point to notice is that the measures are qualitatively and
`quantitatively consistent. It is thus possible to create one “extended” series simply
`by appending the older data to the newer ones.
`
`The cost of intermediation grows from 2% to 6% from 1870 to 1930. It shrinks to
`less than 4% in 1950, grows slowly to 5% in 1980, and then increases rapidly to
`almost 9% in 2010.
`
`This pattern is not driven by globalization or by structural changes in the economy.
`The pattern remains the same if finance is measured as a share of services, and if net
`financial exports are excluded (see Philippon, 2011).
`
`The second key point is that finance was smaller in 1980 than in 1925. Given the
`outstanding real growth over this period, it means that finance size is not simply
`driven by economic development.
`
`Measuring the Output of Financial Intermediation.
`Next comes the issue of measuring the output of the financial sector. Following
`Merton (1995) and Levine (2005), one can propose the following four categories of
`financial services or functions:
` Provide means of payment (ease the exchange of goods and services)
` Produce information about investment opportunities
`
`
`
` Monitor investments and exert corporate governance
` Provide markets for insurance (diversification, risk management, liquidity)
`
`
`These services are the output of the finance industry and its source of
`economic value. To the extent that this higher total cost is met with
`proportionally more output, the greater compensation of the sector should
`not be surprising.
`
`These services are provided to both households and firms, and facilitate the creation
`of financial assets. The most important contracts involve the credit markets. I
`measure the production of credit separately for households, farms, non-financial
`corporate firms, financial firms, and the government.
`
` I
`
` show in Philippon (2011) that a simple benchmark can be constructed using the
`workhorse of modern macroeconomics, i.e., neo-classical growth model. This
`benchmark is a weighted average of the financial assets created by the financial
`sector for the real economy.
`
`The most important trends in recent years are the increase in household debt, and
`in financial firms’ debt. Figure 2 shows the outstanding bonds issued by Farms,
`Households and Non Profit Organizations, and the Government. Household debt
`exceeds 100% of GDP for the first time in history (see Figure 2), while financial debt
`exceeds non-financial corporate debt for the first time. Surprisingly, the non-
`financial corporate credit market is smaller today than it was at its peak of the late
`1920s.
`
`Figure 2: Debt over GDP (selected sectors)
`
`
`For the corporate sector, we need to look at bonds and stocks, and for stocks, we
`want to distinguish seasoned offerings and IPOs. We also need to look at the
`
`
`
`
`
`liquidity benefits of deposits and money market funds. When we put all the pieces
`together, we obtain a series for output for the finance industry.
`
` I
`
` then aggregate all types of non-financial credit, stock issuance, and liquidity
`services from deposits and money market funds.
`
`Figure 3: Financial Intermediation Output
`
`
`
` I
`
` construct two series of output in Figure 3: One using the flows (gross issuances
`over GDP) and one using the levels (debt over GDP). Note that both are relevant in
`theory. Screening models apply to the flow of new issuances, while monitoring
`models apply to the stocks. Trading applies to both.
`
`The two series are displayed in Figure 3. The production of financial services
`increases steadily until WWI, and rapidly after 1919 until 1929. It collapses during
`the great depression and WWII. It increases steadily until 1975 and more randomly
`afterwards. The flow and level measures share the same long term trends, but there
`are clear differences at medium frequencies. The flow variable is more stationary
`before WWI, suggesting a steady buildup of financial assets. The flow variable
`collapses much faster during the great depression and the great recession. The level
`variable peaks in 1933 because of deflation and the need to deal with rising default
`rates.
`
`The Decreasing Efficiency of Intermediation in the U.S.
`
` I
`
` can then estimate the cost of financial intermediation, defined as the value added
`share divided by output series. The cost of intermediation in the US (expressed as a
`share of outstanding assets) is between 1.3% and 2.3%.
`
`
`
`
`However, the cost of intermediation per dollar of assets created has increased over
`the past 130 years, and especially since the 1970s. In other words, according to this
`measure, the finance industry that sustained the expansion of railroads, steel and
`chemical industries, and the electricity and automobile revolutions was more
`efficient than the current finance industry.
`
`
`
`Figure 4: Financial Intermediation Unit Cost
`
`
`
`
`This is counter-intuitive. If anything, the technological development of the past 40
`years (IT in particular) should have disproportionately increased efficiency in the
`finance industry. How is it possible for today's finance industry not to be
`significantly more efficient than the finance industry of John Pierpont Morgan?
`
`It is important to understand that using the GDP share of finance to measure the
`costs of financial intermediation captures all fees and spreads but it ignores hidden
`costs of systemic risk. The neo-classical benchmark also assumes that all agents
`borrow rationally ex-ante (of course, that does not rule out the fact that they might
`end up with too much debt ex-post, but that means that they understand the risks
`involved and choose to borrow). We can debate this assumption, but the point I
`want to emphasize is that this provides an upper bound on financial efficiency. If
`anything, adding excessive risk taking and over-borrowing would decrease the risk-
`adjusted efficiency.1
`
`
`1 For an insightful discussion, see Haldane and Madouros (2011), Popov and Smets (2011), and
`Arcand, Berkes, and Panizza (2011).
`
`
`
`Information Technologies: Where is Wal-Mart When We Need
`It?
`
`An obvious driving force in financial intermediation is information technology. I
`often hear the argument that improvement in IT explains the increase in the share
`of finance. This argument, however, is either incomplete or misleading. One reason
`it is incomplete is simply that IT cannot explain the evolution of the GDP share of
`finance before 1970.
`
`What makes the IT argument misleading is that it is far from clear why IT should
`increase the share of finance. The neo-classical growth model predicts that, in most
`cases, technological improvement should lower the share of GDP spent on financial
`intermediation. In particular, this prediction is unambiguous for most retail finance.
`Essentially, the physical transaction costs of buying and holding financial assets
`must have decreased because of IT. This effect must have lowered the amount spent
`on intermediation.
`
`An apt analogy is with retail and wholesale trade, since these are also
`intermediation services.2 As Blanchard (2003) explains in his discussion of Basu et
`al. (2003), “fully one-third of the increase in TFP growth from the first to the second
`half of the 1990s in the United States came from the retail trade sector. For this
`reason, the general merchandising segment, which represents 20% of sales in the
`sector, was one of the sectors examined in a McKinsey study (McKinsey Global
`Institute, 2001) aimed at understanding the factors behind U.S. TFP growth in the
`1990s.”
`
`Figure 5(a,b) shows the evolution of GDP shares and IT investment in wholesale
`trade, retail trade.
`
`
`
`
`2 For instance, one can compare retail finance and retail trade. Households go to grocery stores not
`because they derive utility from doing so, but rather to have access to groceries. Similarly,
`households use financial intermediaries to gain access to the financial products that they need.
`
`
`
`Figure 5 IT Investment and GDP Shares of Retail and Wholesale Trade
`
`
`
`
`
`Figure 6 shows this evolution for finance. The contrast is striking. Based on what we
`see in wholesale and retail trade, IT should have made finance smaller, not larger.
`
`Figure 6: IT and GDP Share in Finance
`
`
`
`
`
`Trading
`
`What happened? Why did we get the bloated finance industry of today instead of the
`lean and efficient Wal-Mart? Finance has obviously benefited from the IT revolution
`and this has certainly lowered the cost of retail finance. Yet, even accounting for all
`the financial assets created in the US, the cost of intermediation appears to have
`increased. So why is the non-financial sector transferring so much income to the
`financial sector?
`
`
`
`
`One simple answer is that technological improvements in finance have mostly been
`used to increase secondary market activities, i.e., trading. Figure 7 shows trading in
`the stock market. Foreign exchange trading volumes are more than 200 times larger
`today than in 1977. Finally, trading accounts for a large fraction of revenues for the
`largest banks.
`
`Trading, of course, is neither a good nor a bad thing. It all depends on its impact on
`the real economy. The output measures developed above, however, only capture the
`production of financial assets (equity, bonds, money, etc.).3 Two important functions
`of financial markets are not captured: the production of price information, and the
`provision of insurance.
`
`Figure 7: Equity Trading Volume over GDP
`
`
`
`
`
`It is then important to ask the following question: If improvements in financial
`intermediation lead to more informative prices or better risk sharing, where would
`these improvements be seen in equilibrium?
`
`Informativeness of Prices
`
`In a model where managers learn from prices, better prices should lead to better
`capital allocation and higher productivity.
`
`Are prices more informative about future income streams? Preliminary evidence in
`Bai, Philippon and Savov (2012) suggests no. They use a large panel of stock price
`data and ask whether a stock’s price (relative to it’s assets) contains more
`
`
`3 Note that the impact on average user costs is already taken into account. If trading lowers
`borrowing costs, firms can borrow more and invest more. This would be captured by the previous
`measures.
`
`
`
`information about future earnings growth today than it has in the past. The answer
`from this statistical analysis is no.
`
`
`Risk Sharing
`
`Another benefit of financial intermediation is risk sharing. Risk sharing can affect
`firms and households.
`
`At the firm level, risk sharing is commonly called risk management. Better risk
`management would, in equilibrium, mostly translate into lower cost of fund, more
`issuances and more investment. This first effect would be captured by our measures
`of debt and equity issuances. Better risk management could also increase TFP if high
`productivity projects are also riskier. I am not aware, however, of any evidence
`suggesting improvement in risk management. The most obvious index, that of
`precautionary savings by businesses, suggests even the opposite: corporate cash
`holdings have increased over the past 30 years. There is also no direct evidence of
`credit derivatives leading to better risk management, and it is commonly believed
`that hedging represents a small fraction of all trades in the CDS market.
`
`At the household level, better risk sharing should lead to less consumption risk.
`Income inequality has increased dramatically in the US over the past 30 years. If
`financial markets have improved risk sharing, however, one would expect
`consumption inequality to have increased by less than income inequality. This is a
`controversial issue, but Aguiar and Bils (2011) find that consumption inequality has
`closely tracked income inequality over the period 1980-2007. It seems difficult to
`argue that the vast sums of money spent on intermediation are justified by better
`risk sharing among households. It is also unclear that any of the main financial
`innovations of the past 20 years have improved risk sharing opportunities.
`
`One area where there is evidence of improved consumption smoothing is in the
`housing market. Gerardi et al. (2010) find that the purchase price of a household’s
`home predicts its future income. The link is stronger after 1985, which coincides
`with important innovations in the mortgage market. The increase in the relationship
`is more pronounced for households more likely to be credit constrained. My model,
`however, captures this type of smoothing and is reflected as increased mortgage
`borrowing. Therefore, it does not bias my estimates.
`
`
`Financial Derivatives
`
`Derivatives markets have grown enormously. As of June 2011, the notional amount
`of outstanding OTC contracts is $700 Trillions (BIS, December 2011), with interest
`rates contracts (mostly swaps) accounting for $550T, and CDS for 32T. Measured at
`gross market values the numbers are 20T, 13T, and 1.3T, respectively.
`
`
`
`
`These numbers are certainly impressive, but the relevant question is: do they bear
`any connection to measures of market and economic efficiency? The short answer
`is: no.
`
`Most people are struck when they hear the market for financial derivatives is 700
`Trillion. These numbers are sometimes used to justify the costs of financial services.
`This is misleading. Derivatives are just the plumbing of the financial markets. End-
`users do not care about the plumbing, only about the quality of the service.
`
`Another analogy comes to mind. Most people would probably be struck if they heard
`that each Airbus A380 contains 40,300 connectors and 100,000 wires with a total
`length of 330 miles (530km). The wires in a single airplane would be enough to go
`from Philadelphia to Boston, or from Paris to London or Frankfurt. Should we
`congratulate Airbus when it manages to increase the length of its wires? Or should
`we only care about the safety, comfort, speed and fuel economy of the plane?
`
`Similarly, suppose you were told that the “complexity and interconnectedness”
`inside your computer had increased. Would you pay for the pleasure to have a
`complex and interconnected computer, or would you only care about its speed,
`design and battery life?
`
`To understand better the relevance (or lack thereof) of financial derivatives,
`consider the following example. Corporation A needs a long-term fixed-interest
`loan. Making the loan would expose the lender(s) to duration risk and to credit risk.
`How these risks are allocated, however, depends on the internal organization of the
`finance industry. Consider two polar cases. Suppose first that the loan is made and
`retained by bank B. Bank B must be compensated for bearing duration and credit
`risks. For instance, B must monitor its credit exposure and maintain a buffer of
`equity against credit risk. B must also monitor and hedge its interest rate risk. These
`activities are costly and the costs are passed through to the borrowers through
`spreads and fees.
`
`Assume now that B can transfer credit risk to fund C using a CDS. Fund C now bears
`the credit risk, while B retains the duration risk. B and C must be compensated
`accordingly. The key point is that B and C together hold exactly the same risk as B in
`the earlier example. Absent other frictions, the two examples are exactly equivalent
`in terms of economic efficiency. Comparing the two polar cases, one can see that the
`size of the CDS market bears no connection to any measure of efficiency.
`
`Let us now extend the example. In terms of economic theory, derivatives can add
`real value in one of two ways: (i) risk sharing; (ii) price discovery. Risk sharing
`among intermediaries would not create a bias in my measurements, however. To
`see why let us go back to the simple example. Suppose there are frictions that
`rationalize why B and C should be separate entities, and why they gain from trading
`with each other (i.e., B has a comparative advantage at managing duration risk, and
`C at managing credit risk).
`
`
`
`
`improve risk sharing among
` Then the existence of CDS contracts can
`intermediaries, lower the risk premia, and lead to a decrease in the borrowing costs
`of A. With free entry, the total income going to intermediaries {B+C} would
`decrease. The unit cost measure developed earlier would correctly capture these
`effects: either borrowing costs would go down, or borrowing volumes would go up,
`or both. In all cases, my approach would register an increase in efficiency.
`
`Therefore, the only bias from derivative contracts must come from better risk
`sharing or price discovery among non-financial borrowers. The correct way to
`measure the value added of derivatives is to directly measure the informativeness of
`prices, or the welfare gains from risk sharing among non-financial firms and
`households. As explained earlier, however, I am not aware of any evidence
`suggesting better risk sharing or better prices.
`
`
`
`
`
`
`
`Conclusion
`
`The finance industry of 1900 was just as able as the finance industry of 2000 to
`produce bonds and stocks, and it was certainly doing it more cheaply. But the recent
`levels of trading activities are at least three times larger than at any time in previous
`history. Trading costs have decreased (Hasbrouck (2009)), but the costs of active
`fund management are large. French (2008) estimates that investors spend 0.67% of
`asset value trying (in vain, by definition) to beat the market.
`
`In the absence of evidence that increased trading led to either better prices or better
`risk sharing, we would have to conclude that the finance industry's share of GDP is
`about 2 percentage points higher than it needs to be and this would represent an
`annual misallocation of resources of about $280 billions for the U.S. alone.
`
`
`
`
`
`References
`
`Aguiar, M., and M. Bils (2011): “Has Consumption Inequality Mirrored Income Inequality?”
`
`Working Paper, University of Rochester.
`
`Arcand, J.-L., E. Berkes, U. Panizza (2011), "Too much Finance?", VoxEU.org, 7 April.
`
`Bai, J., T. Philippon, and A. Savov (2011): “Have Financial Prices Become More Informative?”
`
`Basu, S., J. G. Fernald, N. Oulton, and S. Srinivasan (2003): “The Case of the Missing
`
`Productivity Growth,” in NBER Macroeconomics Annual, ed. by M. Gertler, and K. Rogoff,
`
`vol. 18, pp. 9–63.
`
`Blanchard, O. (2003): “Comment on Basu et al.,” in NBER Macroeconomics Annual
`
`Bolton, P., T. Santos, and J. Scheinkman (2011): “Cream Skimming in Financial Markets,”
`
`Working Paper, Columbia University
`
`French, K. R. (2008): “Presidential Address: The Cost of Active Investing,” The Journal of
`
`Finance, 63(4), 1537–1573.
`
`Gerardi, K. S., H. Rosen, and P. Willen (2010): “The Impact of Deregulation and Financial
`
`Innovation on Consumers: The Case of the Mortgage Market,” Journal of Finance, 65(1),
`
`333–360.
`
`Haldane, A.G. and V. Madouros (2011), "What is the Contribution of the Financial Sector,
`
`VoxEU.org, 22 November.
`
`Hasbrouck, J. (2009): “Trading Costs and Returns for U.S. Equities: Estimating Effective
`
`Costs from Daily Data,” Journal of Finance, 64(3), 1445–1477.
`
`Levine, R. (2005): “Finance and Growth: Theory and Evidence,” in Handbook of Economic
`
`Growth, ed. by P. Aghion, and S. Durlauf, vol. 1A, pp. 865–934. Elsevier, Amsterdam.
`
`McKinsey Global Institute (2001), "US productivity growth 1995-2000. Understanding the
`
`Contribution of Information Technology Relative to other Factors. Washington, D.C.
`
`Merton, R. C. (1995): “A Functional Perspective of Financial Intermediation,” Financial
`
`Management, 24, 23–41.
`
`Philippon, T. (2011) “Has the U.S. Finance Industry Become Less Efficient?” mimeo NYU.
`
`Popov, A. and F. Smets (2011), “On the tradeoff between growth and stability: The role of
`
`financial markets”, VoxEU.org, 3 November.