One could argue that the behavior of unsophisticated and agency-prone investors generates

a positive externality: there are surely more resources spent gathering information and
monitoring managers than there would be in a world in which investors refused to overpay
for active asset management. Absent investors’ willingness to overpay, equilibrium securities
prices could have less than the socially effi cient amount of information, and corporate
managers would be subject to insuf effi cient amount of information, and corporate
managers would be subject to insuf f i cient monitoring. For example, venture capital
funding of start-up fi rms, which f i cient monitoring. For example, venture capital funding of
start-up fi rms, which has arguably brought signifi cant positive externalities, would have
been less robust if investors in venture capital funds had required adequate compensation
for the risks they were taking. While one could make this sort of argument, it is not entirely
convincing. One important reason is that not all information collection performed by active
asset managers is socially valuable. For example, a hedge fund may be willing to pay $20,000
to form a more accurate prediction of a company’s earnings to be released in the next week.
To the extent that this information allows the hedge fund to profi t at the expense of other
less-informed market participants, the fund earns an profi t at the expense of other less-
informed market participants, the fund earns an excess return. Hirshleifer (1971) calls
information of this type “foreknowledge,” but explains that it has no social value. More
specifi cally, the $20,000 expenditure but explains that it has no social value. More specifi
cally, the $20,000 expenditure should be regarded as a social loss because getting this
information into prices one week earlier is unlikely to lead to a more effi cient allocation of
real resources. Modern fi nancial markets are rife with examples of such socially wasteful
invest ments. For example, consider the costs of “co-location hosting services,” which enable
electronic orders to arrive milliseconds faster because of their geographical proximity to
trading centers. These investments lend support to Paul Samuelson’s view, originally cited in
Shiller (2001, p. 243), that modern fi nancial markets display view, originally cited in Shiller
(2001, p. 243), that modern fi nancial markets display “considerable micro effi ciency”—
perhaps facilitated by active asset management— while at the same time retaining large
“macro ineffi ciency.” We fi nd it noteworthy while at the same time retaining large “macro
ineffi ciency.” We fi nd it noteworthy that over the last 15 years, despite increased resources
devoted to asset manage ment, there have been two large and socially costly valuation
errors: the Internet bubble at the end of the 1990s and the overvaluation of mortgage-
backed securities during the 2000s. Another reason to question the social benefi ts of
information production by active managers is the evidence that they cater to the
preferences of unsophisti cated investors. For example, mutual fund managers channel
investor fl ows into the cated investors. For example, mutual fund managers channel
investor fl ows into the sorts of securities that investors want to own (say, Internet stocks at
certain times, high-yield bonds at other times, and so on) rather than allocating capital to its
best use (Frazzini and Lamont 2008). Gennaioli, Shleifer, and Vishny (2012) suggest that
investment managers cater to unsophisticated investors’ preferences to earn their trust.11
Thus, we think there is good reason to question whether the marginal 11 Also, Scharfstein
and Stein (1990) and Froot, Scharfstein, and Stein (1992) show that reputational concerns
can lead active asset managers to herd in their investment decisions. Thus, the ineffi ciency
in active asset management does not depend on there being unsophisticated investors.
Robin Greenwood and David Scharfstein 17 dollar of active management makes securities
prices more informative. Indeed, Bai, Philippon, and Savov (2012) present evidence
suggesting that securities prices have not become more informative since the 1960s. Finally,
when investors overpay for active management, it creates rents in the sector. These rents
lure talented individuals away from potentially more productive sectors (Baumol 1990;
Murphy, Shleifer, and Vishny 1991).12 Indeed, during the period of rapid growth in asset
management, fi nance attracted more talent, at least period of rapid growth in asset
management, fi nance attracted more talent, at least as measured by the number of
students entering fi nance from elite universities. The as measured by the number of
students entering fi nance from elite universities. The cost of this reallocation of talent
depends, in large measure, on the industries that top students would have otherwise
entered and the marginal value of additional talent entering fi nance. If, for example,
students shifted into fi nance from science and engineering, where rents are low and
marginal productivity potentially higher, then the talent reallocation is costly to society. By
contrast, the social costs are much lower if the marginal entrant into fi nance would have
otherwise sought a career in other rent-seeking sectors, such as parts of legal services. In a
recent study of MIT undergraduates, Shu (2013) shows that fi nance attracts the best
students, as MIT undergraduates, Shu (2013) shows that fi nance attracts the best students,
as measured by their characteristics at the time of admission.13 The Growth of Credit
Intermediation Components of Growth As illustrated in Figure 1, the credit intermediation
industry (as defi ned by the As illustrated in Figure 1, the credit intermediation industry (as
defi ned by the BEA) grew on a value-added basis from 2.6 percent of GDP in 1980 to 3.4
percent in 2007, having peaked at 4.1 percent of GDP in 2003. The growth of credit
intermedi ation accounted for roughly one-quarter of the growth in the fi nancial sector,
which ation accounted for roughly one-quarter of the growth in the fi nancial sector, which is
less than the contribution of the securities industry to fi nancial sector growth and about
equal to that of the insurance industry. As with the securities industry, we examine in more
detail the activities that drove the growth of credit intermediation. Again due to data
limitations, we look at the output of these activities rather than their value-added. Table 2,
using data from the Bureau of Economic Analysis and Economic Census, breaks out credit
intermediation into its main components: traditional banking (lending and deposit taking)
and transactional services related to credit card accounts, deposit accounts, ATM usage, and
loan origination. The distinction between these broad categories is admittedly imprecise. 12
Murphy, Shleifer, and Vishny (1991) argue that talent fl ows to large markets, where there
are weakly diminishing returns, and talent is measurable and contractible. These are all
features of asset management. 13 However, Shu (2013) also shows that the students who go
into fi nance are not the best ones at the time of graduation. The best students at
graduation go to graduate school in science and engineering. Thus, it is possible that the lure
of fi nance induces the best MIT students at the time of admission to invest less in
coursework and focus more on preparing themselves for a career in fi nance. 18 Journal of
Economic Perspectives The output from transactional services is simply measured as the
fees collected The output from transactional services is simply measured as the fees
collected for these services. Measuring the output from traditional banking, which is divided
for these services. Measuring the output from traditional banking, which is divided into
lending and deposit-taking, is more complex. The output from lending is into lending and
deposit-taking, is more complex. The output from lending is imputed as the difference
between the interest earned on bank loans (that is, loans imputed as the difference between
the interest earned on bank loans (that is, loans on bank balance sheets including
commercial, consumer, and real-estate loans) on bank balance sheets including commercial,
consumer, and real-estate loans) and the interest that would have been earned, had the
funds been invested in and the interest that would have been earned, had the funds been
invested in Treasury and Agency securities (those guaranteed by government agencies such
Treasury and Agency securities (those guaranteed by government agencies such as the
Federal Housing Administration or government-sponsored enterprises such as the Federal
Housing Administration or government-sponsored enterprises such as Fannie Mae). These
calculations use the average interest rate earned on banks’ as Fannie Mae). These
calculations use the average interest rate earned on banks’ holdings of these securities: that
is, holdings of these securities: that is, Lending Output == Bank Loans ×× ( (Interest Rate on
Loans – – Interest Rate on Treasury and Agency Securities). This is meant to capture ). This is
meant to capture Table 2 Value Added and Output from Credit Intermediation Firms,
Selected Years $ billions % of GDP Industry outputs, by activity 1997 2002 2007 1997 2002
2007 Traditional banking (imputed output) 179.1 253.9 328.9 2.15% 2.39% 2.34% Lending
76.8 99.2 102.2 0.92% 1.32% 1.34% Deposit-taking 102.3 79.9 76.9 1.23% 1.07% 1.00%
Transactional services (fees) 186.1 328.0 487.7 2.25% 3.08% 3.47% Deposits and cash
management 24.7 57.5 78.4 0.30% 0.54% 0.56% Credit card accounts 23.8 23.7 29.6 0.29%
0.22% 0.21% Other products supporting fi nancial services 17.8 55.0 76.3 0.21% 0.52%
0.54% Loan origination, nonresidential 14.0 20.2 27.9 0.17% 0.19% 0.20% Loan origination,
consumer residential 11.3 76.8 62.3 0.14% 0.72% 0.44% ATM and electronic transactions 3.0
6.2 8.6 0.04% 0.06% 0.06% Other 91.5 88.6 204.6 1.10% 0.83% 1.46% Total credit outputs
365.2 582 816.6 4.38% 5.47% 5.82% Bank revenues from activities other than credit
intermediation 67.3 109 130.3 0.81% 1.02% 0.93% Total inputs 180.8 239.9 455.2 2.17%
2.25% 3.24% 3.8 15.2 14.9 0.05% 0.14% 0.11% Revenues collected by nonbanks for credit
related activities Value added by credit intermediation fi rms 247.9 436 476.9 2.97% 4.10%
3.40% Value added for all credit intermediation activities 211.2 374.7 415.1 2.53% 3.52%
2.96% Source: Bureau of Economic Analysis, Economic Census of the United States, and
authors’ estimates. Note: Firms engaged in credit intermediation are mostly banks, but also
include credit unions and other savings and lending institutions. The Growth of Finance 19
the ongoing services provided by banks in managing and monitoring loans on their balance
sheets, as well as the value of identifying the loans in the fi rst place. their balance sheets, as
well as the value of identifying the loans in the fi rst place. However, this basic measure
could overstate or understate the value of these services. It overstates the value to the
extent that it also includes the credit risk and maturity premium that banks (or any other
investors) earn by holding risky long-term loans (Ashcraft and Steindel 2008). The measure
could understate the value of these services to the extent that the fees associated with loan
origination are included in our transactional services category. The imputed output from
deposit-taking is measured as the quantity of deposits multiplied by the difference between
the rate earned on Treasury and Agency securities and the rate paid on those deposits; that
is, Deposit Services Output = Deposits × (Treasury Interest Rate – Average Interest Rate Paid
to Depositors). Depositors presumably accept yields below those of US Treasuries and
equivalent government guaranteed securities because they use deposits for transactional
purposes. Table 2 shows that the output from traditional banking as a percentage of GDP
was roughly the same in 2007 as it was in 1997. However, substantial growth occurred in
transactional services, which in turn were largely refl ected in fees associated with in
transactional services, which in turn were largely refl ected in fees associated with deposits,
residential loan origination, and the catchall category of “other products supporting fi
nancial services.” In 2002, in particular, residential loan origination supporting fi nancial
services.” In 2002, in particular, residential loan origination fees spiked as part of the largest
mortgage-refi nancing wave in US history. These fees totaled $76.8 billion—0.7 percent of
GDP, or 2.7 percent of the $2.85 trillion of residential mortgages issued in that year. As in the
previous section, we form our own estimates of the sector’s outputs going back to 1980.
Here, we follow the methodology of the Bureau of Economic Analysis and use data from the
Call Reports, which all regulated fi nancial institu Analysis and use data from the Call
Reports, which all regulated fi nancial institu tions must submit to the Federal Deposit
Insurance Corporation at the end of each quarter. As a consistency check, we verify that we
can replicate the total output numbers in the years in which the Economic Census is carried
out (that is, every f i ve years starting in 1982).14 As can be seen from Figure 4, imputed
output from lending as a share of GDP has fl uctuated around its mean of 1.2 percent of
GDP. Much of the variation comes from changes in the ratio of bank loans to GDP, which fell
from about 60 percent at the end of the 1980s to under 50 percent at the end of 1990s.
During the housing boom in the 2000 –2006 period, bank loans rose back to about 60
percent of GDP. 14 Output from lending and deposit-taking is calculated using data from
Federal Reserve’s Call Reports, and from the Historical Statistics on Banking of the Federal
Deposit Insurance Corporation. Fees on mortgage loans are imputed from BEA benchmark
year estimates using annual mortgage origination totals. Fees on credit card accounts are
imputed combining Flow of Funds data on total credit card debt outstanding with
Government Accountability Offi ce data on average credit card fees. Data on service charges
on deposit accounts are from FDIC’s Historical Statistics on Banking. 20 Journal of Economic
Perspectives Figure 4 Credit Intermediation Output 1980–2007 8% 7% 6% 2006 2004 2002
2000 1998 1996 1994 1992 1990 1988 1986 1984 1982 1980 5% 4% 3.% Output as a percent
of GDP 2% 1% 0% Imputed output loans Imputed output deposits Other Noninterest income
deposit accounts Mortgage originations Other fees on residential loans Credit card fees
Securitizations on bank balance sheets All other securitization Uncounted output from
securitization Mortgage origination and “other” mostly consumer related fees Traditional
bank-based credit intermediation Source: Call Reports, Flow of Funds Accounts of the United
States, Bureau of Economic Analysis, and authors’ estimates. Note: For imputed output, we
follow the BEA’s methodology. Figure 4 also shows that output from deposit-taking has
generally been falling over time. Some of the decline stems from reductions in spreads
between securities and deposits, but the main source of the decline is a reduction in
deposits relative to GDP, from its peak of about 70 percent at the beginning of the 1980s to
under 50 percent in the early 2000s. This decline mostly refl ects a shift of saving into money
market funds, bond funds, and the stock market. While traditional banking has declined
slightly as a share of GDP, Figure 4 illustrates that essentially all of the growth in the credit
intermediation industry has come from transactional services, largely refl ected in fees
associated with consumer and mortgage credit. A sizable share of the refl ected in fees
associated with consumer and mortgage credit. A sizable share of the fees can be traced to
the refi nancing of existing mortgages. Mortgage origination, in turn, is highly dependent on
the path of nominal interest rates, which were falling for most of the period we study here
and led to extraordinarily high levels of refi nancing most of the period we study here and
led to extraordinarily high levels of refi nancing for a number of years during the period