The Evolution of the US Financial Industry from 1860 to
2007: Theory and Evidence.∗
New York University, NBER, CEPR
The share of ﬁnance in U.S. GDP displays large historical variations. I argue, using
evidence and theory, that corporate ﬁnance is a key factor behind these evolutions.
Corporate demand for intermediation depends crucially on the relative investment op-
portunities of ﬁrms with low cash ﬂows (young ﬁrms) and ﬁrms with high cash ﬂows
(incumbents). A simple general equilibrium model is developed in order to separate
demand and supply factors in the market for ﬁnancial intermediation. The demand
parameters accord well with historical evidence on the importance of entrants during
technological revolutions. The supply parameters suggest ﬁnancial regress in the 1930s
and progress in the 1990s. The model accounts for much of the variation in the income
share of the ﬁnancial sector from 1860 to 2001. Only the period 2002-2007 appears
Keywords: Financial development, corporate ﬁnance, ﬁnancial intermediation, func-
tional analysis, moral hazard, structural change.
JEL: E2, O16, G2, G3
∗I thank seminar participants at NYU, Duke, Toulouse, the Society for Economic Dynamics and the
NBER for their comments.
†Stern School of Business, 44 West Fourth Street, Suite 9-190, New York, NY 10012-1126. Tel: +1 212
998 0490, email@example.com.
Financial institutions provide services to households and corporations. The ﬁnancial
sector’s share of aggregate income reveals the value that the rest of the economy attaches
to these services. Historical data from the United States shows surprisingly large variations
in the economic importance of Finance. It was high in the 1920s, but, after a continuous
collapse in the 1930s and 40s, it was down to only 2.5% of GDP in 1947. It recovered slowly
until the late 1970s, and then grew more quickly to reach almost 8% of GDP in 2006.
The interactions between Finance and the rest of the economy are complex. Economic
growth in the 1960s was outstanding, but seemed to require little ﬁnancial intermediation.
Finance grew quickly in the 1980s while the economy stagnated, and the pattern changed
again in the 1990s. These evolutions could oﬀer valuable insights into the process of eco-
nomic growth, but they have received little attention so far.
The ﬁnance industry, unlike other sectors, would not exist in an Arrow-Debreu economy.
If markets were complete, ﬁnancing and insuring would be trivial. How then should we
interpret the dramatic growth of this industry over the past 60 years? Does this imply that
the U.S. economy has drifted away from the Arrow-Debreu benchmark? Is information
becoming more asymmetric, or enforcement more diﬃcult? Why is the U.S. devoting a
growing share of its human capital to the provision of ﬁnancial services? Which economic
forces pin down the equilibrium size of the ﬁnancial sector? This paper attempts to answer
these questions by bringing together new evidence and theory.
The paper focuses on the role of corporate ﬁnance, because the evidence suggests that
other explanations for the evolution of the ﬁnancial sector are either incorrect or incomplete.
Contrary to common wisdom, there is neither theoretical nor empirical support for the idea
that total factor productivity (TFP) growth in the non ﬁnancial sector has a direct inﬂuence
on the size of the ﬁnancial sector. Neither ﬁnancial globalization, nor increased trading of
securities, nor the development of the mutual funds industry can account for the increasing
share of ﬁnance in GDP. Moreover, the usual explanations given for the increasing share of
the service sector in the economy do not apply to the ﬁnance industry.
To further motivate the focus on corporate ﬁnance, I present new evidence on the evo-
lution of the cross-sectional distribution of cash ﬂows and investment expenditures. I ﬁnd
that ﬁrms with low cash ﬂows account for a growing share of total capital expenditures.
In the 1950s, most corporate investment was done by incumbents with high cash ﬂows.
In 2000, half of total investment was done by (young) ﬁrms whose cash ﬂows covered less
than a third of their capital expenditures. What we observe, however, is the distribution
of cash ﬂows and realized investments, and not the distribution of investment opportuni-
ties. This evolution might reﬂect either increasing demand for ﬁnancial intermediation, or
improvements in ﬁnancial intermediation (supply shocks), or both.
An equilibrium model of intermediation is therefore needed to interpret the evidence.
Greenwood and Jovanovic (1990) and Bencivenga and Smith (1991) are important references
in the literature. These models, however, are not suﬃcient to address the issues raised by
the empirical evidence. Models with exogenous ﬁxed costs of trading do not provide an
interpretation for the link between investment and cash ﬂows, which is the cornerstone
of corporate ﬁnance. In addition, models based on trading costs are unlikely to explain
the growth of the ﬁnance industry since, as Freixas and Rochet (1997) argue, “the progress
experienced recently in telecommunications and computers implies that FIs would be bound
to disappear if another, more fundamental, form of transaction costs were not present.” In
the model of Bencivenga and Smith (1991), it is possible to study the balance sheet of
ﬁnancial intermediaries, but not their value added since the modelled intermediaries do not
The model economy is populated by overlapping generations of agents who choose to
work in the ﬁnancial or in the non-ﬁnancial sector. Following Diamond (1984) and Holm-
ström and Tirole (1997), among others, I focus on corporate ﬁnance and monitoring in the
presence of moral hazard. Agents in the non-ﬁnancial sector have diﬀerent productivities
and receive diﬀerent investment opportunities. Moral hazard limits borrowing, and agents
with good opportunities but low cash ﬂows cannot always invest. Agents in the ﬁnancial
sector help alleviate these borrowing constraints. In equilibrium, the market for corporate
ﬁnancial services must clear, and young agents must be indiﬀerent between careers in the
ﬁnancial and non-ﬁnancial sectors.
The model clariﬁes the economic forces that pin down the size of the ﬁnancial sector.
First of all, the Finance share of income is constant along a balanced growth path: the
level of productivity in the non ﬁnancial sector does not inﬂuence the size of the ﬁnancial
1 To be empirically useful, the model must also allow for both direct and intermediated ﬁnance. Bencivenga
and Smith (1991) assume that all savings are intermediated.
sector. Under some conditions, this also applies to the rate of productivity growth, and this
is consistent with the historical evidence.
On the supply side, eﬃciency gains in ﬁnance always reduce credit rationing, but have
ambiguous eﬀects on the relative size of Finance. When the ﬁnancial sector is ineﬃcient,
eﬃciency gains increase its income share, but when the sector is already quite eﬃcient,
further gains reduce its income share. As a result, the size of the ﬁnancial sector is not a
straightforward measure of market incompleteness, although it does contain useful informa-
On the demand side, there is an important distinction between moral hazard at the
micro level, and the relevant measure of moral hazard at the macro level. At the macro level,
the joint distribution of current productivity and investment opportunities is a fundamental
determinant of the aggregate demand for ﬁnancial services. The intuition is that, when ﬁrms
with investment opportunities also receive high cash ﬂows, there is little macro demand for
intermediation, even if moral hazard is potentially severe at the micro level. Finance is
more needed when young ﬁrms or young industries have better investment opportunities
than large established ﬁrms or industries. This prediction receives strong empirical support.
To go deeper into the analysis of the ﬁnancial sector, the technological parameters of the
model are estimated by matching empirical moments. The model is then used to study the
historical data. The model quantiﬁes the eﬃciency of the ﬁnancial sector in each period,
and recovers the distribution of productivities and investment opportunities.
The model-implied structural parameters shed light on the evolution of the US economy
over the past 150 years. The 1890s and 1920s are a time a rapid entry and investment
by ﬁrms with large ﬁnancial needs. These needs decrease during the two World Wars. A
structural shift happens during World War 2. After the War, large established ﬁrms with
high cash ﬂows appear to have the best investment projects. As a result, the demand
for ﬁnancial intermediation is small. Starting in the 1970s, investment opportunities shift
away from large proﬁtable ﬁrms towards young ﬁrms with low current cash ﬂows, and the
demand for intermediation increases. These predictions of the model are consistent with the
historical evidence on General Purpose Technologies, the role of Electriﬁcation in the 1920s
and Information Technology starting in the 1970s (David (1990), Jovanovic and Rousseau
On the supply side of ﬁnancial services, the 1930s represent a ﬁnancial disaster. The
eﬃciency of the ﬁnancial sector collapses and a large number of ﬁrms become credit con-
strained. Eﬃciency in ﬁnance recovers after the War but remains below its level of the
1920s for three decades. In the 1980s the ﬁnancial sector is not eﬃcient enough to meet
the demands of the corporate sector and credit constraints rise in the 1980s. In the 1990s,
ﬁnancial eﬃciency improves and leads to an investment boom.
The model can further be tested by considering its predictions for the income share of
(corporate) ﬁnance and for the size of the corporate credit market. The income share of
ﬁnance is nothing more that a broad-based measure of the fees, spreads, and other income
paid by the non-ﬁnancial sector to the ﬁnancial sector. The model correctly predicts much
of the variations in the income share of ﬁnance over more than 150 years. The model also
accounts reasonably well for the evolution of the credit market.
This paper is a ﬁrst attempt to understand, qualitatively and quantitatively, the evo-
lution of the Finance industry in the US economy. In doing so, the paper makes several
empirical and theoretical contributions. The paper develops a new model of the equilibrium
size of the Finance industry that incorporates moral hazard, ﬁrm heterogeneity, monitoring
and career choices. The paper also presents new evidence for the U.S., at the industry level,
at the ﬁrm level, and at the individual level (regarding the tasks performed by employees
in the ﬁnancial sector).
The paper is related to the literature on ﬁnancial intermediation, reviewed in Gorton and
Winton (2003), and to the literature on ﬁnancial development: King and Levine (1993), Do
and Levchenko (2007) and Greenwood, Sanchez, and Wang (2007) among others. Do and
Levchenko (2007), in particular, ﬁnd that ﬁnancial development responds to the demand
for external ﬁnance generated by international trade. In this paper, I estimate how the
ﬁnancial sector responds to the demand for external ﬁnance generated by changes in the
distribution of investment opportunities.
The remaining of the paper is organized as follows. Section 1 presents the evidence and
discusses alternate explanations. Section 2 presents the model and Section 3 characterizes
the equilibrium allocations. Section 4 estimates the parameters of the model. Section 5
tests the prediction of the model using historical data. Section 6 concludes.
This section describes the evolution of the U.S. ﬁnancial sector. Its goal is to present the
facts that the theory must explain, but also to clarify a number of widespread misconceptions
regarding the ﬁnancial sector and its evolution.
Value added and compensation shares
Figure 1 displays the share of the Finance and Insurance industry in the GDP of the
United States, estimated from 1850 to 2007. For the period 1947-2007, I use value added
measures from the Annual Industry Accounts of the United States, published by the Bureau
of Economic Analysis (BEA). For 1929-1947, I use the share of employee compensation
because value added measures are either unavailable or unreliable. I extend the series using
data from Kuznets (1941) and Martin (1939) for the period 1900-1929. For 1850-1900, I use
the Historical Statistics of the United States2, and Census data. The data are described in
Philippon and Reshef (2007).
The growth of the U.S. ﬁnancial sector is the story of three waves and two crashes (the
2008 crash is not yet visible in the data). The ﬁnancial industry was around 1.5% of GDP
in the mid-19th century. The ﬁrst large increase between 1880 to 1900 corresponds to the
ﬁnancing of railroads and early heavy industries.
The second big increase between 1918 and 1933 corresponds to the ﬁnancing of the Elec-
tricity revolution, as well as automobile and pharmaceutical companies. General Electric
did its IPO in 1913, General Motors in 1920 and Procter & Gamble in 1932. Key discover-
ies of the 1920s and 1930s, such as insulin and penicillin, became mass-manufactured and
After a continuous collapse in the 1930s and 40s, the GDP share of ﬁnance and insurance
industries was down to only 2.5% of GDP in 1947. It recovered after the war and was
mostly stable at around 4% until the late 1970s. The third large increase, from 1980 to
2001, corresponds to the ﬁnancing of the IT revolution. The ﬁnancial industry accounted
for 8.3% of GDP in 2006.
Figure 2 shows the evolution of the shares of various subsectors (starting in 1977 because
of data limitations). In the National Income and Product Accounts (NIPA), the ﬁnance in-
2 Carter, Gartner, Haines, Olmstead, Sutch, and Wright (2006).
dustry is split into 4 categories: (i) Credit intermediation; (ii) Investment banking, venture
capital, brokerage, and portfolio management; (iii) Insurance and reinsurance; (iv) Pension
funds, mutual funds (open- and closed-end), and trusts. Figure 2 shows that credit inter-
mediation is the dominant activity, and, together with investment banking and brokerage,
the fastest growing. Funds and trusts account for a negligible share of value added. This
brings up two important issues: the distinction between assets and value added, and the
classiﬁcation of the various activities in a theoretical model.
Value added versus assets under management
Figure 3 shows the allocations of value added and assets within the ﬁnancial sector in 2005.
The data on value added is from the NIPA, as described above. The data on assets is from
the Flow of Funds Accounts. To create Figure 3, I have mapped the Flow of Funds into the
NIPA classiﬁcation. Figure 3 makes it clear that there is no simple relationship between
value added and assets under management. As a result, the common wisdom that the rise
of the pension and mutual funds industry is the main factor behind the evolution of ﬁnance
severely misses the point. In fact, from a theoretical perspective, funds and trusts resemble
the Arrow-Debreu benchmark: they control a lot of assets without using much economic
To understand the growth of the ﬁnancial sector, one should not focus on mutual funds,
but rather on credit intermediaries, investment banks and private equity. From a theo-
retical perspective, it means that models based purely on trading costs are unlikely to be
useful. Similarly, it is probably more important to explain value added than assets under
From a theoretical perspective, industry classiﬁcations are useful only to the extent that
they can be mapped into economic functions, as advocated by Merton (1995). To do so, one
must look at the tasks performed by employees of the ﬁnancial sector. Figure 4 presents
estimates of the share of ﬁnance activity that is presumably related to corporate ﬁnance
and credit intermediation. The data is from Philippon and Reshef (2007) and the primary
source is the Current Population Survey. The estimates are based on the compensation of
employees, i.e., on employment weighted by relative wages.3
The baseline share is constructed by removing the jobs that are not related to corpo-
rate ﬁnance or to credit intermediation. The excluded categories are: insurance specialists;
traders of stocks, bonds, commodities and other assets; personal ﬁnancial advisors; janitors,
private security and miscellaneous employees. The baseline share is one minus the compen-
sation share of the excluded categories. Figure 4 shows that the baseline share is relatively
stable, around 75%. This measure might overestimate corporate ﬁnance services, however,
because it includes all clerical and administrative jobs. Some of these workers keep track
of loans to businesses, but some also provide services to households. I therefore compute
a second estimate that excludes all clerks and administrative workers. This adjusted share
increases over time because of the decrease in clerical employment. Based on Figure 4, I
will assume that approximately 60% of ﬁnance activity is related to corporate ﬁnance, and
that this share has not declined over time.
Globalization does not account for the growth of the U.S. ﬁnancial sector, for two reasons.
First, the U.S., unlike the U.K., is not a large exporter of ﬁnancial services. According
to IMF statistics, in 2004, the U.K. ﬁnancial services trade balance was +$37.4 billions
while the U.S. balance was -$2.3 billions: the U.S. was actually a net importer. In 2005,
the U.K. balance was +$34.9 billions, and the U.S. balance was +$1.1 billions.4 Second,
ﬁnancial globalization is a relatively recent phenomenon (see Obstfeld and Taylor (2002),
and Bekaert, Harvey, and Lumsdaine (2002)), while Figure 1 shows that the growth of the
ﬁnancial sector has been continuous since the end of World War II.
Finance versus other industries
Is ﬁnance diﬀerent from other service industries? Yes. The other fast growing service
industry is health care, but it does not share the U-shaped evolution of Finance from 1927
to 2006. In theory, Finance is clearly diﬀerent because it would not exist in a Arrow-Debreu
economy, while almost all other services would. This fact has important implications that
3 Weighting is important because wages vary with occupations. For instance, traders earn more than
4 There is, of course, much trade within the ﬁnancial sector, notably between the U.S. and the U.K., but
it would be misleading to argue that growth in value added of the U.S. ﬁnance and insurance industry is
due to large net exports.
are not well understood. In particular, it renders much of the literature on structural
change irrelevant for understanding of the evolution of the ﬁnancial sector. The traditional
explanations for the growth of the service sector focus on the elasticity of substitution
between goods and services, and on rapid productivity growth in manufacturing (Baumol
(1967)). These ideas cannot be applied to the ﬁnancial sector, for at least two reasons.
First, ﬁnancial services do not enter directly into the utility function of agents; they enter
the budget constraint. As a result, there is no elasticity of substitution to speak of.5
The second reason why the usual models of structural change do not apply to Finance is
that the economic value added of one unit of ﬁnancial services grows with the productivity
of the agents who beneﬁt from these services.6 As a result, the equilibrium size of the
ﬁnancial sector is independent of productivity in the non ﬁnancial sector. In an earlier
version of this paper, I showed that productivity growth is indeed uncorrelated with the
size of the ﬁnancial sector. The 1960s was the period of fastest productivity growth but it
was also the period where the ﬁnancial sector remained essentially unchanged. The growth
of ﬁnance was strong in the 1980s, a time of low productivity growth. In the 1990s, however,
productivity and ﬁnance grew together. Productivity growth by itself does not account for
the evolution of the ﬁnancial sector.
Investment share of low cash ﬁrms
The evidence just discussed suggests that neither the development of mutual and pension
funds, nor globalization, nor the traditional explanations for the rise of the service sector
can account for the evolution of the Finance industry. Some periods of economic growth
appear to be more intensive in ﬁnance than others, but, until now, we simply do not know
5 One can of couse imagine a reduced form model with ‘ﬁnancial services in the utility function,’ just as
some monetary models use ‘money in the utility function’ as a shortcut. But these reduced forms model
cannot be used to study the evolution of the ﬁnancial sector for the same reasons one would not use ‘money
in the utility function’ to study the structural determinants of transaction costs and money demand.
6 When the utility function depends on real quantities of goods and services, it is conceptually (if not
always empirically) straightforward to deﬁne the relative productivity of various sectors. This is not the
case with ﬁnancial services. To take a simple example, suppose that screening and monitoring services by
one banker allow one entrepreneur to obtain ﬁnancing. What is the productivity of the banker? The answer
is that it is proportional to the productivity of the entrepreneur. In this simple example, the eﬃciency of
ﬁnance is deﬁned in terms of the number of entrepreneurs that a banker can screen, monitor and advise.
This is the most natural way to think about corporate ﬁnance, and it implies that, for a given eﬃciency of
screening and monitoring, productivity in ﬁnance grows with productivity in the non-ﬁnancial sector. This
property holds in the particular model developed below.
Figures 2 and 4 highlight the role of credit intermediation and corporate ﬁnance. What
is needed now is to look at the non ﬁnancial sector in order to understand how ﬁnancial
services are used. Corporate ﬁnance is fundamentally related to investment and internal
funds. Firms with low cash ﬂows and high capital expenditures must raise external ﬁnance,
and they require ﬁnancial services to do so. Empirically, I compute the share of investment
done by ﬁrms whose cash ﬂows are less than a fraction α of their capital expenditures:
I use all ﬁrms in the industrial Compustat ﬁles with non missing values for income and
capital expenditures, excluding ﬁnance, insurance and real estate. In equation (1), i is
the ﬁrm identiﬁer, income is income before extraordinary items (Data #18), and capex is
capital expenditures (Data #128). To avoid issues with the timing of income and investment
— because of time-to-build or accounting rules — capexit and incomeit are the sum of capital
expenditures and income in year t − 2, t − 1 and t.
Figure 5 displays the shares of investment accounted for by low cash ﬁrms, deﬁned
using three diﬀerent cutoﬀ values: α = 0.33, 0.25 and 0.15. All three measures show strong
upward trends. The investment share of low cash ﬁrms and the size of the corporate credit
market before 1940, presented in Table 1, are indirect estimates whose construction will be
Let me discuss brieﬂy the issue of sample section and composition eﬀects. First of all,
the coverage of Compustat in terms of capital expenditures and employment has remained
constant since 1975.7 Since the fastest growth in s33 happens after 1975, increased coverage
cannot explain the trend. Moreover, the increasing investment share of low cash ﬁrms
mostly reﬂects the fact that ﬁrms have been going public at a younger age over the past
50 years. Far from being a statistical sample bias, this is the consequence of ﬁnancial
and technological innovations, and it is in fact evidence in support of the model presented
7 The ratio of employees of all Compustat ﬁrms to Non Farm Payrolls is around 40%. The ratio of capital
expenditures to non residential ﬁxed investment is around 80%. These ratios are fairly constant from 1975
8 See Davis, Haltiwanger, Jarmin, and Miranda (2006) and Fink, Fink, Grullon, and Weston (2005) for