A Critical Long View Of Capital Markets And Institutions:

    Financial Markets and Institutions.. It is often taken for granted that: 1) capital markets and institutions allocate funds to firms where realized returns on real assets are highest; 2) the net gains to the. www.kelley.iu.edu.

  • Id: # 768c8ed
  • File Type: pdf
  • Author: www.kelley.iu.edu
  • File Size 330.75 kB
  • Read by User: 105 Times
  • Published: Thursday, April 3, 2014
  • index: Financial Markets And Institutions.

Rating

  • Read Online
A Critical Long View of Capital Markets and Institutions:
Realized Returns from Corporate Assets, 1950-2003
By
James S. Ang
BankAmerica Eminent Scholar
Department of Finance
Florida State University
Tallahassee, Florida 32306
Gregory L. Nagel
Assistant Professor
Department of Finance and Economics
Mississippi State University
Starkville, Mississippi 39762
Jun Yang
Assistant Professor
Department of Finance
Indiana University
Bloomington, Indiana 47405
We would like to acknowledge financial support from the PriceWaterhouseCoopers Global Competency
Centre Grant Programme. We are grateful to Phil Dybvig, Jia He, Chong-En Bai, two anonymous referees,
session participants at the 2006 Journal of Banking and Finance in Beijing, and seminar participants at
Tsinghua University, Shanghai Jiaotong University, and Shanghai University of Finance and Economics.
A Critical Long View of Capital Markets and Institutions:
Realized Returns from Corporate Assets, 1950-2003
Abstract
It is often taken for granted that: 1) capital markets and institutions allocate funds to
firms where realized returns on real assets are highest; 2) the net gains to the economy
from investments by corporations have improved in the last 30-50 years due to
innovations and better risk management techniques in the financial markets; and 3) the
agency cost-reducing role of markets and institutions ensures that real assets funded with
external funds would earn higher returns. However, corporate real assets are long lived,
and realized returns have to be tracked over a long period to verify these assertions. We
perform large-scale calculations of the realized returns on real assets to all firms available
in the Compustat database for periods of 10, 20, 30, 40, and 50 years. Our methodology
relies only on cash flow between the firms and all their fund providers. In particular, we
focus on capital markets, institutions and non interest bearing liability holders. It
circumvents the potential problem in using market expectations of future cash flows if
markets are inefficient over long periods as suggested by Shiller (1981). We found
several new and surprising results. Returns on real assets by corporations derived from
actual cash flow over long periods are, on the whole, lower than expected by the fund
providers. They suffer a long-term decline, and have been below the yields of 10 year
Treasury bonds since 1973. Real assets that received more external financing (from
capital markets and institutions) actually report even lower realized long-term returns.
These unexpected results may stimulate fresh debate on the roles and long-term
performance of capital markets and institutions.
2
A Critical Long View of Capital Markets and Institutions:
Realized Returns from Corporate Assets, 1950-2003
1. Introduction
We evaluate the long-term performance of capital markets and financial institutions
based on the actual returns on the funds they allocated to the corporate sector.
Specifically, we report the findings of a large-scale and long-term study of the realized
returns to U.S. corporations on the total funds employed, and on the funds supplied by the
capital market participants.
Well functioning capital markets and financial institutions are generally considered
to be the mechanisms that allocate the supply of funds (savings) to the demanders of
funds who yield the highest returns. This suggests that one should judge how well capital
markets and institutions perform their resource allocation role by examining the long-
term realized returns of the real assets they financed. Since the U.S. is reputed to have
the most developed capital markets and institutions, one would assume that its markets
and institutions, via price signals and monitoring, play a major role in deciding the
amount of funds allocated to corporations for investment in real assets. Thus, a study of
the long-term realized returns to U.S. corporations is a natural starting point to investigate
the long term allocational efficiency of the capital markets and institutions in general.
We examine three predictions for an economy with a system of well functioning
capital markets and institutions. First, U.S. corporations as a whole, using funds and
investment guidance (through market prices) provided by the capital markets, earn
adequate if not superior realized returns. Second, because allocational efficiency is
expected to increase due to visible improvements in the operations of the markets and
institutions (better risk management, information disclosure, corporate governance, and
regulations, etc.) in the last 30-50 years, funds allocated to the corporate sectors are
expected to earn increasingly higher realized returns over the period. Third, due to the
monitoring role of markets and institutions in reducing agency costs, corporate
investments financed with external funds are expected to yield higher realized returns
than those financed with internal funds, or free cash flows.
3
We measure realized returns from long-lived corporate investments by considering
only cash flows that are distributed to the capital suppliers and cash flows contributed by
capital suppliers to fund corporate real investments. These cash flows are then
summarized by the familiar internal rate of return (IRR) measure to enable comparison
among firms, over time, and against the yield of newly issued 10-year Treasury bonds.
The procedure yields true returns on investments, unlike calculations based on reported
earnings or market values of securities. Reported earnings have several known
shortcomings, as they are prone to manipulation by the management and there is no
guarantee that earnings not distributed as cash flows to capital providers would be able to
maintain their present value due to waste or poor reinvestments. Finally, we use book
value
1
of assets instead of market value to approximate the terminal value, because
calculation of returns on corporate real assets that involves market values of corporate
securities captures expectations, which may be biased if the market is not efficient
2
,
i
. In
effect, we are testing market efficiency; therefore, we cannot use market value. In
robustness studies we calculate IRRs using replacement values for the subset of firms that
has all of the required data. These IRRs are shifted downward by an average of 0.7
percent when compared to IRRs computed using book values (see Figure A4). Thus, this
and other robustness studies strengthen our conclusions.
The three principal findings are:
1. Realized internal rates of return on all assets utilized by U.S. corporations, as a whole,
are not only less than expected but are also consistently less than the 10-year Treasury
bond yield since 1973. We compute a 10-year IRR for all Compustat firms. We
1
. Ideally we would use replacement value to calculate IRR; however, the data necessary to calculate
replacement value is not available for many firms and is not available at all until about 1975, so we use the
book value of total assets instead. We observe that replacement value is primarily calculated by adjusting
the book value of fixed assets for inflation. In unreported results we show that fixed assets as a proportion
of total assets has declined through time. Thus the adjustment for inflation affects a smaller portion of
fixed assets as time goes on. The net results is that, since 1975, realized IRRs calculated using book value
are, on average, greater than that calculated from replacement value; we show this in Figure A4
2
One could be ensnared in a paradox when market values of corporate securities are included in the
calculation of realized or ex post returns. This is because if market value is a rational unbiased estimate of
future realizable cash flows, then q ratio, or market value to cost of investments, should be a sufficient
statistic to infer realized returns.
4
count only the actual inflow to and outflow from the firms, and use total assets
3
as the
terminal value. Each year we then compute the total asset weighted IRR for all firms
that survive for the subsequent ten years. This yearly cross-sectional weighted
average IRR shows a steady decline, from nearly 10 percent in the early 1950s to 4.8
percent in late 1980s and 1990s. Since 1973, this IRR has not been able to exceed the
newly issue 10-year Treasury bond yield
4
. This fact alone provides strong evidence
that many corporations made suboptimal use of the funds they have. We further
compute IRRs having horizons of 10, 20, 30, 40 and 50 years. The long horizon IRRs
should largely incorporate the cash flow consequence of growth options (exercised or
expired), and so these IRRs are expected to equal to or exceed the cost of capital if
firms make profitable investments. Surprisingly, as horizon lengthens, the total firm
IRRs are consistently lower than the short horizon IRRs, which are less than the risk-
free rate starting in 1973.
2. To cross check the observed decline in realized returns over the period, we calculate
the cross sectional median rate of return on assets (ROA) of S&P500 firms, every
year from 1950-2003 and find a steady decline in ROA in every decade. Median
ROAs in recent years are less than half of their values at the beginning of the period.
More importantly, the decline in median ROA is not attributable to decreasing risks,
as we also find the standard deviations of ROA actually increase steadily during this
period. Because the S&P 500, through its process of addition and deletion, has a
survival bias in favor of the strong, we calculate an aggregated economy wide ROA
for all publicly traded firms. We find the decline in ROA is even larger; ROA is
reduced by three-quarters, while standard deviation has steadily increased by a factor
of about eight since the early 1950s. Moreover, aggregate return on equity (ROE)
3
If a firm exits Compustat before the end of the database and is not a target firm, in the exit year we assign
it a terminal value equal to the market value of equity plus the book value of total liabilities. This situation
occurs for 10,039 of the 20,354 firms in our database. In cases of bankruptcy etc, this procedure biases the
IRR upward. For robustness, we also assign exiting firms a terminal value equal to the maximum of (a) the
book value of total assets or (b) the market value of equity plus the book value of total liabilities. By doing
so, the result should go against our finding of low realized returns. However, our conclusions do not
change.
4
For reference purposes we also compute the 10-year total firm IRRs by setting the terminal value equal to
the market value of equity plus the book value of total liabilities rather than setting it equal to the book
value of total assets. These IRRs do not exceed the 10-year Treasury bond rate starting in 1973 but do
attain equality with it starting in 1986.
5
declines by about one-third while the variability in ROE, economy wide and for
individual firms, increases even more than it does for ROA. The decline in net profit
margin since the early 1960s is largely attributable to an increase in selling, general,
and administrative costs. Over time, these costs have more than cancelled out large
improvements in (a) gross profit margin and (b) operating efficiency attributable to
improving inventory turnover. However, because of U.S. corporations’ increasing
utilization of long-term other liabilities
5
as a source of funds
6
, these mostly non-
interest bearing ‘involuntary’ capital contributors have been subsidizing equity
holders to make up for the low overall returns on total corporate assets.
3. As a group, companies that obtain the highest external financing as a percent of total
assets, from financial institutions (bank loans), and capital markets (bonds and stocks)
earn a lower IRR on corporate investments than those that mainly use internal funds.
The result holds for all levels of new investment, and the IRRs are particularly low
for the firms that had the highest investment and mainly financed externally.
The paper is structured as follows: Section 2 discusses the issues in measuring
realized returns and presents our procedure. Section 3 specifies the data and discusses the
main empirical results. Section 4 then contains a series of robustness tests and alternative
measures of return on real assets, followed by the summary of Section 5.
2. Methodology
2.1 Why measure realized returns?
To know whether funds are allocated to their best use, one needs to know the
realized, not expected cash flows. After all, ex ante, all investments are expected to yield
superior returns. Although corporate investments are inherently risky, a well functioning
system that allocates funds for corporate investments should produce adequate, if not
superior, realized returns in the aggregate of the economy. Unfortunately, corporate
investments have a long life and take 10, 20, or more years to fully realize all their cash
5
See the appendix for a description of long term other liabilities.
6
In 1950 1.7 percent ($1.4 billion) of total assets was financed by long term non interest bearing liabilities,
by 2003 it had increased to 9.8 percent ($2.3 trillion). See the Appendix, Figure A1, for details.
6
flows. Realized returns on invested capital, however, are not commonly tabulated and
studied.
2.2 Measurement Issues
There are three technical problems that have to be overcome in order to correctly
calculate realized returns: 1) Reported earnings or reported cash flows are both noisy,
due to accounting convention and other measurement errors, and probably biased, due to
firms’ attempts to manage earnings or even to manipulate accounting statements; 2) For a
finite period estimation, a terminal value has to be imputed, and 3) All capital
contributors need to be included when calculating realized returns. The first problem is
well known; we now elaborate on the other two problems.
Market value of securities is often a convenient choice for terminal value; however,
using market values assumes that it is an unbiased estimate of the present value of future
cash flows, yet there may be deviation due to market inefficiency. Further, using market
values to determine whether the market efficiently allocates resources involves an
inherent contradiction. If the market price at a later date is assumed to be an unbiased
estimate of the present value (at that point) of all subsequent cash flows, then the
calculation of realized returns is not necessary. This follows since the current market
price, by iterative expectations, should also be an unbiased estimation of all cash flows
henceforth. Then a comparison of market value to asset costs (or book value of the
security) is the only information needed to judge whether real assets earn adequate
returns.
Not only is the measurement of realized returns to capital suppliers vaguely
understood, but the measurement of capital supplied often excludes the set of
‘involuntary contributors.’ Examples are employees in the case of under-funded
pensions, the government in the case of deferred taxes, and suppliers and other liability
holders that are not paid yet. Analysis of firm performance is incomplete without
explicitly determining returns on the assets that ‘involuntary contributors” finance.
General Motors’ projected difficulty with funding retiree liabilities is but one example of
the consequence of financing assets through ‘involuntary contributors’ without creating
sufficient matching assets to pay the liabilities. Without explicit analysis of these
7
‘involuntary’ fund providers, it is impossible to determine how much of the realized
returns to equity and bond holders come at the expense of the ‘involuntary contributors.’
2.3 Internal rate of return (IRR) calculation – overall approach
We take a new approach to calculate realized returns to corporate investment. First,
we take the corporate sector as a black box, and count as inflows only new sources of
funds contributed by capital providers to the corporations, and count as outflow only
actual funds distributed to the capital providers. This way, reported but undistributed
earnings, regardless of the amount, stay in the black box and have no cash flow
consequence. This approach sidesteps the vagaries in accounting earnings, such as the
timing of recognition of revenues and costs.
IRR is calculated using the capital budgeting procedure for two different groups of
fund providers: (a) capital markets (bond and stock holders); (b) all fund providers to the
firm including bond holders, stock holders, as well as non-interest-bearing liability
holders. We call the former the capital market IRR, and the latter the total firm IRR.
We emphasize here that the total firm IRR incorporates funds provided by non-
interest bearing liability holders; they financed 3.3 trillion dollars (i.e., 13.9 percent) of
assets in 2003. Figure A1 in the Appendix shows that the percentage of assets financed
by long term other liabilities starting at about 1.7 percent ($1.4 billion) in 1950 and
steadily increased to 9.8 percent ($2.3 trillion) in 2003. Fama and French (1999) assume
non-interest bearing liabilities earn an implicit return, which shows up in a firm’s net
income. Since our approach uses actual cash flows rather than accounting estimates of
net income, we are able to explicitly determine the returns earned by non-interest bearing
liability holders.
The standard textbook approach calls for determining the initial investment made
by each fund provider and all the subsequent funds received by each, including the
terminal value. This means that the replacement cost should be used as the initial value
and the terminal value should be the discounted sum of all future cash flows.
2.4 Estimation of initial values and terminal values for realized IRR computation
8
2.4.1 Initial value estimation
The firm’s initial value includes all items on the balance sheet and, therefore, does
not include unreported items such as the value of some intangibles (ex. brand equity).
Replacement value of balance sheet assets is then the best estimate of the initial value.
However, the data to compute replacement values is not available till 1975 and even then
many firms do not have sufficient data. Therefore, we use book value in most of our
analyses to approximate replacement value. If possible, we compare those IRRs to ones
computed using replacement values.
2.4.2 Terminal value estimation
The objective in estimating terminal value is to determine the present value of all
future cash flows. Some researchers assume that market value is an unbiased estimate of
the terminal value; see for example, Fama and French (1999). However, the estimation
will be biased if the market is inefficient (See LeRoy and Porter (1981), Shiller (1981),
Shiller (2003)). Shiller argues that the deviation may last for long periods of time.
Moreover, our objective is to determine what investors actually realize, and we have
already pointed out there is an inherent contradiction if we use market values. Therefore,
we seek an alternative approach.
To reduce the need to rely on market value, first we compute the IRR on assets
employed, ignoring the value of future growth options. Thus, the appropriate terminal
value is the replacement value of assets. We will show that the IRR on assets declines
over time and declines as the horizon of computation goes beyond 10 years. In
particular, after 1973, the short horizon (10 year) IRR on assets is less than the risk free
rate (which is less than firms’ cost of capital). Therefore, the net present value (NPV) in
the corporate sector is negative. Observe that NPV must be greater than zero at some
point if the present value of future cash flows is to ever exceed the replacement value of
the firm. Thus, replacement value is likely to be an upper bound on the present value of
future cash flows (which does not equal the market value if the market is inefficient).
Admittedly, it is possible that the increase in IRR occurs beyond the end of our
database. We address this issue in two ways. First, we use horizons as long as 50 years.
Over such a long period, whatever is realizable from growth options or intangibles should
9
have largely been realized. Second, in unreported results (available upon request) we
focus on firms that survive. We find surviving firms have higher IRRs than firms that
cease to exist due to mergers, bankruptcy, etc.; in other words, the strong survive. For
cohorts of these surviving firms we observe that IRRs
7
consistently decline as the horizon
increases. It is well known that firms mature; so if the cohort of surviving
8
firms has not
realized an IRR above the risk free rate in 10 or 20 years (which is what we observe
starting in the early 1970s), there is no reason to think it ever will. Therefore, it is
reasonable to conclude that unfulfilled growth is unlikely be realized after the end date of
the database for surviving firms.
2.5 Fund flows for individual firms
We first calculate realized IRRs for individual firms. Then, we calculate value-
weighted realized IRRs for the corporate sector. Notice that results are similar if we
calculate the aggregate IRR directly.
Our procedure in calculating total firm IRR is based on the ‘closed system fund
flow’ approach, in which only funds that providers actually contribute and funds that
firms actually distribute to these providers are counted. Thus, inflows are fresh funds
from two sources: all fund providers or just capital markets. Fund outflows are fresh
funds received by the fund providers; the funds received depend upon the claims made by
the fund provider. We do not count non-distributed earnings. Nor do we consider the
type and amount of investment. That is subject to classification and estimation error
(whatever cash is not paid out has to be invested, however efficiently or inefficiently).
We also exclude conversion or call of convertibles as they have no cash consequence.
We apply the same principle in dealing with mergers, and more specifically firms
that are acquired. For the acquiring firm, an acquisition is like any other investment. If it
is financed by issuing debt or stock, these cash flows from fund providers are
7
Some portion of the declining realized IRR that we observe as horizon is lengthened is due to the effect of
Jensen’s inequality. We quantify this effect in a number of ways. In unreported results, each year we
compute both total asset weighted IRR and aggregate IRR for cohorts of surviving firms; our conclusions
are unchanged. Next, we observe that the annual aggregate return on assets declines steadily through time
as does the short horizon IRR for cohorts of surviving firms; therefore, the decline in IRR as horizon
lengthens is likely due to declining firm performance over time, not Jensen’s effect.
8
In unreported results (available upon request), we show that the performance of firms that do not survive
is on average lower than that of surviving firms, so survival bias free IRRs would be lower than the IRRs
for cohorts of survivors.
10
incorporated in the yearly cash flow analysis. Acquisitions involving securities (stocks
and bonds, etc.) are counted as an inflow to the acquirer; acquired assets are valued at the
market price of equity less the cash paid plus the book value of liabilities. This approach
is not affected by the accounting treatment of merger premiums. For target firms, the
terminal value of shareholder equity is estimated as the market value of equity as reported
by Security Data Corporation; liabilities (interest bearing debt) are valued at their book
value.
2.5.1 Itemized fund flows
The initial investment from capital market participants is the sum of interest bearing
debt and shareholders’ book equity. Yearly cash flows are then the sum of the following
four items less the market value of acquired firms: (a) dividends, (b) repurchase minus
new stock issues (Seasoned Equity Offering, SEO, and Initial Public Offering, SEO), (c)
cash flows to debt holders (i.e., repayment of interest bearing debt), and (d) interest paid.
Finally, at the end of the analysis horizon, shareholder book equity plus the book value of
debt becomes the terminal value received by debt holders and shareholders
9
.
We use the book value of total assets to measure the initial investment from all fund
providers. Yearly fund flows include all cash flows, not only from the capital markets
but also funds from other stakeholders retained by the corporations (e.g., suppliers,
employees, federal and state governments). Significant fund flows occur due to changes
in non-interest bearing liabilities
10
. These other liabilities finance assets of the firm
11
.
Therefore, the yearly fund flow for the firm is the cash flows to equity holders (i.e.,
dividends plus repurchase minus new stock issues) plus interest payment minus the
change in total liabilities minus acquired firm’s market value of equity plus cash paid for
9
If a firm exits the Compustat database and is not identified as a target by the Security Data Corporation
Mergers and Acquisitions database, its terminal value is set equal to the market value of equity plus the
book value of debt. This upward biases the IRR computation in the case of bankruptcies.
10
For instance, we find that long-term non-interest-bearing liabilities (data75) have steadily increased from
less than two percent of total assets in 1950 to 9.8 percent (i.e. $2.3 trillion dollars) of total assets in 2003.
11
Non-interest-bearing long term liabilities are commitments made by the firm. Examples are given in the
Appendix; these commitments were rare in the 1950s and 1960s, but have become a major liability since
then. Not paying for them in the current year frees that cash for use by the firm and so is a source of cash
to the firm. On the other hand changes in equity [excluding changes due to (a) stock purchases, (b) stock
issuances, or (c) dividends paid)] caused by reclassifying equity as a liability are not a commitment to pay
stockholders, so they are not counted as a fund outflow.
11
the firm and minus its book value of total liabilities. Finally, at the end of the analysis
horizon, the book value of total assets becomes the terminal value.
Table A1 (see Appendix) shows the yearly cash flows and variables used to
calculate them at individual firms for capital markets and for all fund providers.
2.6 Computation of IRR for individual firms
Our objective is to determine the internal rate of return (IRR) that managers actually
produce by making investments. We look at realized IRR from the fund providers’ point
of view, which is simply the return that equates all the funds contributed by the providers
to all the subsequent fund flows, including the terminal value of the company at the end
of the analysis horizon. Algebraically, IRR
12
is the discount rate that solves:
∑
=
+
+
+
=
T
1t
T
)
i
IRR1(
T,i
TV
t
)
i
IRR1(
t,i
CF
0,i
INV
(1)
where:
INV
i,0
is the initial investment shown in Table A1 for the fund provider at time 0;
TV
i,T
is the terminal value for the fund provider at the analysis horizon;
T is the number of years over which IRR
i
is computed, ranging from 10 to 50 years;
CF
i,t
is net cash flow from firm i at time t to the fund provider shown in Table A1.
The timing in the calculation corresponds to the period funds are actually paid or
received.
Thus, IRR is the rate of return that equates the discounted sum of net fund flows to
fund providers over the analysis horizon to the initial investment made by the fund
providers at time zero.
2.7 Summary of the closed system approach to calculate IRR calculation
The advantages of the closed system approach to measure returns to fund providers
are: (1) it includes all fund provider, such as non interest bearing liability holders, that
12
A common issue associated with IRR is that multiple solutions may be found. We take the root closest
to a 100 percent return. Thus our results are biased upward if multiple roots occur. We do check for
multiple roots greater than negative 100 percent and find it occurs very rarely, even though there are cash
flow sign changes. The reason is that large terminal values almost always drive any multiple roots to be
either imaginary or less than negative 100 percent. See Fernández (2004) for a broad discussion of issues
encounter in computing IRR for firms.
12
support the profitability of the firm; (2) it is insensitive to accounting manipulations, or
arbitrary designation (as regular or extraordinary incomes), and (3) it is not biased by
mergers.
3. Results and Analysis
3.1 Data
Our sample includes all U.S. firms (excluding firms in the insurance industry
13
) with
the required Compustat
14
data. Our sample period starts in 1950 and concludes in 2003.
Due to the calculation procedure used, firms must survive at least two years to be
included in the analysis. Since the analysis starts in 1950, certain data are not always
available; however, for all qualified firms we require that total assets (data6) be greater
than one million and total liabilities are not missing. Other missing values used in
computations are derived from other available data or approximated; see the Appendix
for the approximations.
In the Appendix, Table A1, we show the sample sizes used in the calculation of IRR
for all fund providers. The second column shows the number of firms used in the
computation. Using this information, IRRs of various horizons are computed each year
for analysis of trends. Returns are reported by year as IRR from adjoining years may
share overlapping data and are therefore correlated. The remaining columns in Table A1
show the number of firms that survived from the year given (1950, 1951, …) through the
horizon given (10 years, 20 years, …) for which an IRR could be computed. These IRRs
are used to compute both total asset weighted IRRs for surviving firms, and returns to
surviving firms as a function of investment and amount of external financing.
3.2 Realized returns for the total firm
We analyze total firms IRR (IRR for all fund providers). These fund providers are
shareholders, bond holders, banks, and non-interest-bearing liability holders. As shown
13
Insurance firms are identified by SIC codes between 6300 and 6499. These firms are excluded because
their business is to insure liabilities and therefore they are uniquely different than the rest of firms. Insured
liabilities are often recorded as long term ‘other liabilities.’
14
Merger data is obtained from Security Data Corporation, this database starts in 1977. Interest rate data is
obtained from the St. Louis Federal Reserve website.
13
in Figure 1, IRR for surviving firms declines as horizon lengthens. This consistent
pattern over the fifty year time period is a basis for concluding that a rational expectation
for the present value of future cash flows is the book value of assets.
Figure 1 also shows that total firm IRR is nearly monotonically declining as time
goes on (while stock prices increase)
15
. This quantifies the issue Porter (1998) points
out: that the U.S. financial system focuses on near-term stock price appreciation, “even
at the expense of long term performance.” In fact, Figure 1 shows that realized total firm
IRR declines to 6.45 percent by the early 1970s and continues its tendency to decline in
the 1980 and 1990s. This leads to the comparison of returns for the total firm versus
capital market providers provided below.
3.3 IRR for capital market participants
Figure 2 shows IRR for capital market participants (stock and bond holders). It also
shows an overall declining trend in IRR as horizon increases. The capital market IRRs
have declined from an average of 12 percent in the early 1950s to 8.6% in the 1990s.
Overall, the analysis of capital market IRR reveals no increase of IRR as horizon
lengthens which would be required to support market valuation greater than book value.
We next compare total firm IRR and IRR for capital markets to known benchmarks.
3.4 Comparison of IRRs to the yield of ten-year treasury bonds
We compare ten-year IRRs to the yield of newly issued ten-year Treasury bonds,
which we use for the risk-free rate. Figure 3 shows the striking fact that total firm IRR
for the corporate sector has not exceeded the risk-free rate since 1973. This is a departure
from the prediction, under any model with risk aversion, that firms with risky real assets
are expected to earn more than the risk-free rate on average. Further, total firm IRR is
4.8 percent in the 1990s while the IRR for capital market investors is nearly four percent
greater. The results imply a significant wealth transfer from other capital suppliers, the
no- interest-bearing liability holders in particular, to capital market participants.
15
The IRRs with horizons ending in 2001 and 2002 show a significant decline due to the worsened
economic conditions in those years and the IRRs ending in 2003 increase for the survivors.
14
Figure 3 also shows that on an asset-weighted basis during the high inflation years
around 1980, all fund providers realized IRRs that are at least four percentage points less
than the long-term risk-free rate, suggesting that the high cost of funds at the time was
not fully incorporated into firms’ capital budgeting decision
16
. Next we compare capital
market expectation to realized returns.
Figure 3 shows that in 1950 the capital market IRR significantly exceeded the
Treasury bond rate. Starting in 1980, the capital market IRR became either significantly
less than or equal to the Treasury bond yield. The weighted average cost of capital
(WACC), which is the required return by capital markets, should be higher than the
Treasury bond yield. Thus it is likely that returns to capital markets have not exceeded
the WACC since 1980
17
.
Overall, the results show that the total firm IRRs in the corporate sector have been
less than the risk free rate since 1973. However, IRRs to capital markets are nearly four
percentage points greater than total firm IRRs. This is accomplished by a wealth transfer
from non-interest bearing liability holders to share holders and bondholders. The wealth
transfer interpretation is supported by an analysis of cash flows to the fund providers. In
the Appendix, Figure A2 shows that cash flows to all fund providers have usually been
negative, and increasingly so in recent decades. The figure also shows that cash flows to
capital markets almost always exceeded that to all fund providers, especially in recent
decades. This implies that the long-term other liabilities holders (pension funds, etc.)
have experienced an increasingly negative cash flow (i.e. corporate liabilities are piling
up – See Figure A1).
Since our prior expectation, that fund providers earn at least their cost of capital, has
not been realized, we question whether capital markets have allocated capital to its best
use. This is the subject of the next section.
16
Unfortunately, firms also failed to earn realized returns that exceed the long run average Treasury bond
rate as well; thus, it is also not likely the firms went ahead with the investments with refunding at lower
interest rates in mind.
17
Interestingly, we observe no correlation between nominal IRR for any fund provider and the riskless rate.
Yet, if firms followed the criterion that investment returns should exceed the cost of capital, then
investments returns should also exceeds the inflation rate plus a premium. The lack of correlation may
indicate: (a) there is no such thing as inflation hedged investments; (b) that realized inflation rates do not
equal expected inflation rates; or (c) that corporations have little ability to change prices with inflation.
15
3.5 The role of capital market participants in the allocation of capital
Our objective in this section is to determine whether increased external funding (by
shareholders, bondholders and banks) yields greater realized returns on newly invested
capital than investments funded internally. We address this issue empirically because
there are two possible theoretical outcomes. On the one hand, there is a long tradition of
literature starting with Jensen and Meckling (1976) suggesting that external monitoring
reduces agency costs. Theoretically, this will in turn improve allocational efficiency of
capital. Many studies support this point of view. For instance, Berger, Ofek and
Yermack (1997) find large stockholders prefer to increase leverage and monitoring from
lenders. On the other hand, there is a trade-off associated with external financing. Not
only is external monitoring costly, but external fund providers may also shorten
managerial investment horizons, resulting in avoidance of profitable long term projects
(see Von Thadden (1995)). Finally, the availability of funds from external sources, such
as through an IPO or SEO, may enable management to over invest. Thus it is an
empirical question whether external financing actually leads to greater realized returns
than internal financing.
To address the issue, we sort firms into three groups each year based on the amount
of new investments made. Investment is measured by the sum of: (a) capital
expenditures; (b) research and development; and (c) advertising over the preceding two
years. Then each is scaled by total assets in the current year. Additionally, within each
of the three investment groups we further sort firms into three groups based on the
amount of their new external financing. We define external financing to be the net stock
and debt issued by the firm over the preceding two years, all scaled by total assets in the
current year. The sorting process creates nine groups of firms each year based on their
level of scaled investment and level of scaled external financing. For each year, we then
determine the median IRRs for 5, 10 and 15-year horizons. In addition, each year, we
measure the risk using the standard deviation of the 10-year IRRs. The average yearly
risk and the average yearly median IRRs (5 year, 10 and 15) are reported in the tables.
16
Selected results for firms in the top decile of market capitalization
18
are shown in
Table 1. Since 1975, among firms that invest the most, the 10-year capital market IRR
for internally financed firms is on average three percentage points higher than that for
externally financed firms. The yearly data (not reported) shows there was only a single
year (1990) in which firms with the greatest external financing realized a higher median
capital market IRR than that of the internally financed firms. Despite the dramatic
difference in average IRRs, the ex-post risks, as measured by the standard deviation of
the 10 year IRRs, are very similar for the two groups of firms.
In short, the results indicate that external financing does not lead to better allocation
of capital than internal financing, given similar level of risks. Indeed, external capital
markets might even distort firms’ efforts to efficiently allocate capital to projects. How
the distortion could occur is a subject of future study that may provide new insights into
how to improve the operation of financial markets.
4. In depth analysis
4.1 Robustness tests
Robustness tests are used to: (a) check whether our conclusions hold on non-
financial firms; (b) test for the influence of other liabilities on firm performance and
survival rates, and; (c) compare the 10 year total firm IRR of externally financed firms to
that of internally financed firms in the group of firms that invest the most. In addition, an
alternative measure of IRR is used to ensure that our conclusions do not hinge on the
method of computation. These tests are now discussed in order.
To start, we check whether our conclusions hold on non-financial firms because they
have experienced less regulation than financial firms have. To accomplish this, we rerun
the total firm IRR analyses for the sector of non-financial firms (SIC less than 6000 and
SIC greater than 6999). As shown in Figure 4, the 10 year total firm IRRs for the
corporate sector of non-financial firms decline over time and have never exceeded the
yield of the 10 year Treasury bonds since 1977.
18
We also have this data for all fund providers as a group (i.e total IRR); it is given in the Appendix, Table
A3, and discussed in the next section, “Robustness tests.” Conclusions are the same.
17
Next, we test for the influence of other liabilities on firm performance and survival
rates. This analysis is shown in the Appendix, Table A3. When it comes time to pay off
the accumulated ‘other liabilities’, firms have a very difficult time performing well or
even surviving.
Table A4 and Figure A3 in the Appendix compare the 10 year total firm IRR of
externally financed firms to that of internally financed firms for the group of firms that
invest the most (top investing firms). Since 1950, the median IRRs of the externally
financed group have never exceeded those of the internally financed group. Further, the
average risk for each group, as measured by the standard deviation of IRRs, is identical.
In unreported results, we analyze the IRRs of non-financial top investing firms. The
same trend is observed. These analyses reinforce the conjecture that unidentified
imperfections in the capital markets prevent efficient allocation of capital in the long
term.
Finally, an alternative measure of IRR is used to check whether our conclusions are
robust to the method of computation. This measure of IRR uses replacement value of
total assets instead of book value when computing total firm IRRs. To determine
replacement value, we use the Lewellen and Badrinath (1997) method
19
. Although their
estimate of replacement value is considered to be the best (see Erickson and Whited
(2001)), the data required limits our sample to post 1974. In the Appendix, Figure A4
shows that IRRs computed using replacement value follow the same trend but are slightly
lower than the IRRs computed using book value. Given the different methods and
samples used to compute realized corporate returns, it is reasonable to believe that our
conclusions are consistent and valid.
4.2 Cross-checks: ROA and common sized statement analysis
Since our overall objective is to understand how firms’ assets are being utilized over
the long term, we begin our cross-checks by examining corporate performance using
Return on Assets (ROA). Figure 5 shows that median ROA has decreased sharply since
1950 for both S&P 500 firms and all firms listed in the Compustat database.
19
When data is unavailable we use the Lee and Tompkins (1999) approximation.
18
Figure 5 also shows that the decline in the median ROA is not due to declining risk
since the standard deviation of ROA has steadily increased over time. The risk has
increased by a factor of about two for S&P 500 firms and by a factor of about eight for all
Compustat firms. We observe decreasing median ROAs (and IRRs) accompanied by
increasing risks. Next, we investigate the causes of the decreased ROA over time.
Figure 6 shows the results of breaking ROA, for the aggregate of all firms, into its
component parts by using the DuPont system analysis. The analysis shows that ROA is
declining due to a deteriorating profit margin (net income / sales) and decreasing
management efficiency (sales / total assets)
20
.
The common size income statement shown in Table 2 further analyzes the source of
the decline in ROA from 1964 through 2003. We compare the mid-sixties to the bull
market period in the late 1990s. Gross profit margins have increased by 4.4 percentage
points from the mid 1960s to the late 1990s, indicating significant pricing power relative
to costs. This may be the result of reduced cost of materials
21
and improved operating
efficiencies. In unreported analyses, we find operating efficiency increased as indicated
by the improvement in inventory turnover, which increased from six in the early 1960s to
ten in the late 1990s. A closer look at Table 2 indicates the improvement in the gross
profit margin is especially strong after the 1980s. On the other hand, the soaring sales,
general and administrative (SG&A) costs have more than cancelled out the improvement
in gross profit margins. Additionally, depreciation and net interest expenses have
steadily risen. Non-operating income has decreased and special item costs have
increased. The only bright spot for the firm (other than decreasing cost of goods sold,
COGS) is that taxes have declined sharply. Thus although operating efficiencies have
improved significantly over time, the results have not fallen through to the bottom line,
largely because SG&A expenses have increased. As a result, net income as percentage of
sales is declining over time.
20
The lack of persistence in earnings beyond chance expectations (see Chan, Karceski and Lakonishok,
2003) is supportive of the declining ROA and IRR reported here. In unreported results, we find that sales
growth for the corporate sector has experienced an overall decline since the 1950s. Additionally, there is no
evidence of a change in the pattern of declining sales growth since the late 1970s.
21
The popular press reports significant cost saving by outsourcing to China and other countries with low
labor costs.
19
Overall, the standard analysis using common measures of profitability indicates that
the performance of U.S. firms is declining while risks are increasing over time, despite
the significant improvements in operating efficiency and reductions in procurement costs.
5. Summary and Conclusions
We calculate realized IRRs for U.S. corporations from 1950 to 2003 measured over
several horizons: 10, 20, 30, 40 and 50 years. Our calculation overcomes the accounting
bias in earnings and the bias introduced by using market valuations. First, we measure
firms over sufficiently long periods (up to 50 years) to make their growth options a lot
less valuable: they are eventually either exercised or expired. Second, we use an upper-
bound value on the present value of all future cash flows in place of market valuations.
We calculate IRRs for all fund providers and for capital market participants (debt and
equity holders).
Our three new findings should provoke rethinking of certain ‘axioms’ or ‘givens’ in
contemporary conceptions about the roles of capital markets and institutions:
1. Despite technological advances and regulatory improvements, we find pervasive
evidence that U.S. corporations earn poor returns by any standard on the funds at
their disposal. In particular, the aggregate 10 year realized returns to all fund
providers have fallen below the yields of newly issued 10 year Treasury Bonds
since 1973. This result is further supported by the long-term decline in the
average ROA.
2. The decline in realized returns can not be attributed to declining risks as the
volatilities of realized returns have been increasing over time.
3. Contrary to the prediction of agency theory that capital markets and financial
institutions should improve the allocation efficiency of capital, we find that
corporate investments financed by external funds actually earn lower long-term
returns than those financed by internal funds. The common size financial
statement analysis further shows that agency costs, which are included in the
broader selling and general administrative expenses, may have actually increased
and contributed to the poor realized returns to fund suppliers.
20
Our findings question the ability of capital markets and institutions to allocate
resources efficiently for long-lived assets and hopefully direct the attention of researchers
to the long-term performance of financial markets and institutions. We believe future
research needs to address the following questions:
A. If U.S. corporations did not earn adequate returns for funds provided by the
capital markets and institutions, how could this fact stay unknown after all these
years? (We have shown that part of the reason is the increasing reliance on the
‘other liabilities’ and a significant wealth transfer from this group to lenders and
stockholders. This wealth transfer provided additional cash flows to firms to
discharge debt claims, pay dividends or repurchase shares.)
B. Since the duration of long-lived real assets spans over several generations of
capital market investors and loan officers, how can we improve the operations of
capital markets and institutions to enable their participants to take a long-term
view?
C. Why do market participants fail to see through the veil shielding poor long-term
corporate performance? Could it be due to the confidence created by generally
rising stock markets in the last 30-50 years? Regardless of the reason, we are
left with a major puzzle: what has sustained overall high and increasing stock
prices when the underlying real assets earn poor
22
and declining returns
accompanied by increasing risks?
22
Since realized total firm IRR has been less than the 10 year Treasury bond since 1973, realized net
present value on assets has been less than zero for the last three decades.
21
References
Berger, Philip G., Eli Ofek, and David L. Yermack, 1997, Managerial Entrenchment and
Capital Structure Decisions, The Journal of Finance 52, 1411-1438.
Chan, Louis K.C., Jason Karceski, and Josef Lakonishok, 2003, The Level and
Persistence of Growth Rates, Journal of Finance 58, 643-684.
Erickson, Timothy, and Toni M. Whited, 2001, On the Information Content of Different
Measures of Q, SSRN working paper.
Fama, Eugene F., and Kenneth R. French, 1999, The Corporate Cost of Capital and the
Return on Corporate Investment, Journal of Finance LIV, 1939-1967.
Fernández, Pablo, 2004, 80 Common Errors in Company Valuation, SSRN, Working
Paper.
Hall, Robert E., 2003, Corporate Earnings Track the Competitive Benchmark, SSRN,
Working Paper 1-31.
Jensen, Michael C., and William H. Meckling, 1976, Theory of the firm: Managerial
behavior, agency costs and ownership structure, Journal of Financial Economics
3, 305-360.
Lee, Darrell E., and James G. Tompkins, 1999, A Modified Version of the Lewellen and
Badrinath Measure of Tobin's Q, Financial Management 28, 20-31.
LeRoy, Stephen F., and Richard D. Porter, 1981, The Present-Value Relation: Tests
Based on Implied Variance Bounds, Econometrica 49, 97-113.
Lewellen, Wilbur G., and S.G. Badrinath, 1997, On the measurement of Tobin's q,
Journal of Financial Economics 44, 77-122.
Porter, Michael, 1998. On Competition (Harvard Business School Press).
Poterba, James M., 1998, The rate of return to corporate capital and factor shares: New
estimates using revised national income accounts and capital stock data,
Manuscript MIT.
Shiller, Robert J., 1981, Do Stock Prices Move Too Much to be Justified by Subsequent
Changes in Dividends?, American Economic Review 71, 421-436.
Shiller, Robert J., 2003, From Efficient Markets Theory to Behavorial Finance, Journal
of Economic Perspectives 17, 83-104.
22
Von Thadden, Ernst-Ludwig, 1995, Long-Term Contracts, Short-Tern Investment and
Monitoring, The Review of Economic Studies 62, 557-575.
West, Kenneth D., 1988, Dividend Innovations and Stock Price Volatility, Econometrica
56, 36-71.
23
Total firm IRR - for all surviving firms; Nominal values
Total asset weighted; 10, 20, 30, 40, 50 year horizon
0
0.02
0.04
0.06
0.08
0.1
1950 1960 1970 1980 1990 2000
IRR
TAw_IRR10
TAw_IRR20
TAw_IRR30
TAw_IRR40
TAw_IRR50
Figure 1: IRR for all fund providers (Total firm IRR) for surviving firms weighted by total assets
Total firm IRR for each firm is the return for firm i that solves
∑
=
+
+
+
=
T
1n
T
)
i
IRR1(
T,i
TV
n
)
i
IRR1(
n,i
CF
0,i
INV
.
For firm i, INV
i,0
is the book value of total assets at time 0. The variable n ranges from time equals 1 to
the horizon length (T=10, 20, 30, 40, and 50 years). The variable CF
i,n
is the net cash flow for firm i in
year n. For all fund providers the flow = common and preferred dividends + purchases of common and
preferred stock - the sales of common and preferred stock + interest - the market value of acquired firms’
total assets + cash paid to acquire a target - change in total liabilities = data21 + data19 + data108 - data115
+ data15 - the market value of all assets from acquired firms + cash paid to acquire a target - change in
data181. The variable TV
i,T
is the book value of total assets unless the firm was acquired (i.e is a target) or
exits the database in the last year of the horizon other than 2003 (the last full year of the database). Then,
for target firms, the terminal value is set to the total market value of the merger (SDC variable VALM) plus
the book value of total liabilities (data181). For firms that exit for other reasons in the horizon year, the
terminal value is set to the market value of stock plus the book value of total liabilities (This upward biases
our IRRs since bankruptcies etc are given an upward biased terminal value. We make this conservative
estimate for 10,039 of the 20,354 firms in our sample). The plot line labeled TAw_IRR10 is the total asset
weighted IRR for the corporate sector computed over a 10 year horizon. Plot lines for the other horizons
are similarly defined.
24
Capital market IRR for all surviving firms; Nominal values
Book equity + debt weighted; 10, 20, 30, 40, 50 year horizon
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
1950 1960 1970 1980 1990 2000
IRR
CMw_IRR10
CMw_IRR20
CMw_IRR30
CMw_IRR40
CMw_IRR50
Figure 2: Capital market IRR for all surviving firms
The plot shows the value of capital market IRRs for the corporate sector (i.e. all surviving and non
surviving USA Compustat firms excluding insurance firms having SIC codes between 6300 and 6499)
through horizons of 10, 30, and 50 years after the start year of the IRR computation. These IRRs are found
each year beginning in 1950 using the following procedure. IRRs are computed for all capital market
participants (as an aggregate) in Compustat through the horizon needed to compute the IRR. Each year,
the IRRs are found by solving
∑
=
+
+
+
=
T
1n
T
)
i
IRR1(
T,i
TV
n
)
i
IRR1(
n,i
CF
0,i
INV
. For the capital market, INV
i,0
is
the sum of the book value of owners equity + interest bearing debt for all firms in Compustat. n ranges
from 1 to the horizon length (T=10, … 50 years). The variable CF
i,n
is the net cash flow for firm i in year
n. For the capital market participants the flow, CF
i,n
= common and preferred dividends + purchases of
common and preferred stock - the sales of common and preferred stock + interest - the market value of
acquired firms’ total assets + cash paid to acquire a target - change in interest bearing debt = data21 +
data19 + data108 - data115 + data15 - the market value of all assets from acquired firms + cash paid for
acquired firms - change in (data34+data9). The variable TV
i,T
is the book value of owners equity plus
interest bearing debt unless the firm was acquired (i.e is a target) or exits the database in the last year of the
horizon other than 2003 (the last full year of the database). Then, for target firms, the terminal value is set
to the total market value of the merger (SDC variable VALM) plus the book value of interest bearing debt
(data34 + data9). For firms that exit for other reasons in the horizon year, the terminal value is set to the
market value of stock plus the book value of interest bearing debt (This upward biases our IRRs since
bankruptcies etc are given an upward biased terminal value. We make this conservative estimate for
10,039 of the 20,354 firms in our sample). The plot line labeled CMw_IRR10 is the book equity + debt
weighted IRR for the corporate sector computed over a 10 year horizon. Plot lines for the other horizons
are similarly defined.
25
26
10 year asset weighted IRR - for all surviving firms
Nominal values
0.1
0.15
0.2
1990
IRR
0
0.05
1950 1960 1970 1980
Cap_Mkt_Nominal
Total_Nominal
Tbond10
Figure 3: Nominal values of corporate sector 10 year IRRs for fund providers
Nominal value of corporate sector 10 year IRRs for capital markets (Cap_Mkt_Nominal) and for the total
firm (Total_Nominal) are defined in Figure 1 (see TAw_IRR10) and Figure 2 (see CMw_IRR10). The
plot line labeled Tbond10 is the yield on new issue 10 year Treasury bonds for the years shown.
Table 1: Firms in the top decile of market capitalization – Capital Market IRR versus external financing at increasing levels of investment (1975-2003)
The sample is all U.S. firms in Compustat (excluding insurance firms) between 1975 and 2003. Further, firms must be in the top decile of market value and have
the information necessary to compute capital market IRR. To be included in the computation of IRR, a firm must survive through the horizon length (5, 10, or 15
years). Firms in the top decile of market value (data25*data199) are found by ranking firms by market value of equity each year. The reported IRR is the
average of the median value of the nominal realized capital market IRRs found each year; the computational details for computing IRR are given in the
footnote
23
. For the IRR computation, the initial investment is the book value of owners equity + interest bearing debt. Terminal value is usually the book value
of owners equity plus interest bearing debt for firms that survive through the horizon. However, if the firm exits the database in the horizon year, then terminal
value is set equal to market value of equity and interest bearing debt. Ranking is first done each year by amount of investment then by amount of external
financing (both are scaled by total assets in the ranking year). Investment is the last two years of capital expenditure (data128), research and development
(data45), and advertising (data46). Unknown R&D and advertising are set to 0. External money is the a) net debt issued over a two year period (data34+data9 at
time t – their values at time t-2) plus (b) the net issuance of stocks over the last two years (i.e. sale of common and preferred stock less the purchase of common
and preferred stock over a two year period = data108 at time t -data115 at time t + data108 at time t -1 - data115 at time t-1. Book to market is the book value of
equity / market value of equity = data60/(data25*data199). The variable
σ
for 10 year IRR is the average of the yearly standard deviations
Ranks Average of the yearly median values
Investment
External
financing
Investment External money Total assets
Book-to-
market
IRR:
5 year horiz.
IRR:
10 year horiz.
IRR:
15 year horiz.
σ
for 10
year IRR
0 min 0 min 0.056 -0.050 3305 0.605 0.113 0.108 0.106 0.082
0 min 1 0.057 0.030 4661 0.695 0.105 0.102 0.101 0.059
0 min
2 max 0.055 0.174 4901
0.723
0.093 0.094 0.094 0.052
1 0 min 0.104 -0.059 2809 0.505 0.126 0.115 0.108 0.073
1 1 0.105 0.029 3469 0.579 0.112 0.105 0.100 0.068
1 2 max 0.104 0.150 3639 0.594 0.099 0.098 0.099 0.056
0 min
0.182
-0.053
2476
0.363
0.161
0.144
0.140
0.092 2 max
2 max 1 0.181 0.040 2648 0.391 0.130 0.126 0.120 0.076
2 max 2 max 0.197 0.191 1855 0.354 0.116 0.111 0.102 0.084
23
Capital market IRR for each firm is the return for firm i that solves
∑
=
+
+
+
=
T
1n
T
)
i
IRR1(
T,i
TV
n
)
i
IRR1(
n,i
CF
0,i
INV
. For capital market providers at firm i, INV
i,0
is
the book value of equity (data6-data181) + debt (data34+data9). n ranges from 1 to the horizon length (T=5, 10, 15 years). The variable CF
i,n
is the flow of funds
in the nth year of the horizon for the IRR computation at firm i, CF
i,n
is defined in table 2. TV
i,T
is the book value of book equity plus interest bearing debt unless
the firm was acquired or exits the database in the horizon year. Then, for acquisitions, the terminal value is set to the total market value of the merger (SDC
variable VALM) plus the book value of debt (data34+data9). For firms that exit for other reasons in the horizon year, the terminal value is set to the market
value of stock plus the book value of debt; this biases our IRRs upward. Note: We report the IRR closes to 100 percent in the rare case of multiple roots.
27
Non financial firms; Total firm IRR - for all surviving firms;
Nominal values; Total asset weighted
10 year horizon
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
1950 1960 1970 1980 1990 2000
IRR
TAw_IRR10
Tbond10
Figure 4: Total IRR for non financial firms
The plot for non financial firms shows the value of total firm IRRs for all surviving
24
firms excluding financial firms having SIC codes between 6000 and 6999.
The plot line labeled TAw_IRR10 is based on total firm IRRs computed over a horizon of 10 years; the computation is described in Figure 1.
24
A selection bias is introduced before 1978 because Compustat did not have full coverage of all listed firms until that date.
28
Economy wide
median ROA
0
5
10
1950 1970 1990
Percent
Economy wide
standard deviation of ROA
0
40
80
120
1950 1970 1990
Percent
S&P 500 ROA (median and standard
deviation)
0
5
10
15
1950 1970 1990
Percent
mROA
sROA
Figure 5: Yearly median ROA and standard deviation of ROA for individual firms
The economy wide sample in the first two figures includes all U.S. firms in Compustat (excluding insurance firms with SIC codes between 6300 and 6499) that
report their net income (data172) and have total assets (data6) greater than $1 million. Return on Assets (ROA) for firm i = net income
i
/ total assets
i
=
data172/data6. Economy wide standard deviation of ROA is the standard deviation of ROA for all firms in the sample each year. In the last figure, the sample
is all S&P 500 firms, where mROA is the median value of ROA for S&P 500 firms each year, and sROA is the standard deviation of ROA for S&P500 firms in
the sample each year.
29
Aggregate ROA
0
5
10
1950 1970 1990
Percent
Aggregate net income / sales
0
4
8
1950 1970 1990
Percent
Aggregate sales / total assets
0
50
100
1950 1970 1990
Percent
Figure 6: Yearly analysis of firm performance
The sample is all U.S. firms in Compustat (excluding insurance firms; i.e. SIC codes 6300 to 6499) that have total assets (data6) greater than $1 million and have
reported sales (data12), net income (data172) and total liabilities (data181). Each year the values of each variable are summed together. In year t, Aggregate
ROA is the return on assets for all firms combined = Sum of Net Income for all firms / Sum of total assets for all firms = sum(data172)/sum(data6). Aggregate
net income / sales = sum(data172) / sum(data12). Aggregate sales / total assets = sum(data12) / sum(data6).
30
Table 2: Common size income statement for USA firms - The mid 60s versus the late 90s
The sample is all firms in Compustat (excluding insurance firms; i.e. SIC codes 6300 to 6499) that have all of the variables used to form the common size income
sheet. Each year the values of each variable are summed together then divided by the sum of sales for all firms in the sample. The total value of each variable is
reported as a percentage of the total value of sales in each year. The variables reported are (A) sales (data12), (B) cost of goods sold (COGS is data41), (C)
Gross margin = data12-data41, (D) Selling general and administrative costs (SG&A=data189), (E) Operating income before depreciation (data13), (F) Operating
income after depreciation (data178), (G) Interest expense (data15), (H) Non operating income (data61), (I) Special items (data17), (J) Pre tax income (Data170),
(K) Taxes (data16), (L) Minority interest (data49), and (M) Net income before extraordinary items (data237)
C=A-B E=A-B-D
J=F-
G+H+I
M=J-
K-L
year
Number
of firms
Sales
(A)
COGS
(B)
Gross
margin
(C)
SG&A
(D)
Operating
Income
before
depreciation
(E)
Operating
Income after
depreciation
(F)
Interest
(G)
Non
operating
Income
(H)
Special
items
(I)
Pre tax
income
(J)
Tax
(K)
Minority
Interest
(L)
Net
income
(M)
1964 1599 100.0 69.9 30.1 14.9 15.2 10.8 0.8 1.0 0.0 11.1 4.8 0.1 6.2
1965 1748 100.0 69.7 30.3 15.0 15.3 11.0 0.8 1.0 0.0 11.2 4.8 0.1 6.4
1966 1883 100.0 70.0 30.0 15.0 15.0 10.9 0.9 1.0 0.0 11.0 4.7 0.1 6.2
1967 2038 100.0 70.2 29.8 15.5 14.2 10.0 1.0 0.9 0.0 9.9 4.1 0.1 5.7
1968 2546 100.0 70.0 30.0 15.6 14.5 10.4 1.2 1.0 0.0 10.2 4.6 0.1 5.6
1969 2671 100.0 70.5 29.5 15.0 14.5 10.1 1.3 1.2 0.0 9.9 4.6 0.1 5.3
1970 2720 100.0 71.3 28.7 15.7 13.0 8.4 1.6 1.2 0.0 8.0 3.5 0.1 4.5
1971 2859 100.0 71.1 28.9 15.3 13.6 9.2 1.5 1.1 0.0 8.8 4.0 0.1 4.7
1972 2965 100.0 70.9 29.1 15.2 13.9 9.7 1.4 1.1 0.0 9.3 4.3 0.1 4.9
1973 3204 100.0 71.1 28.9 14.6 14.3 10.3 1.5 1.1 0.0 9.9 4.6 0.1 5.3
1974 4102 100.0 72.6 27.4 13.7 13.7 10.1 1.8 1.1 -0.1 9.4 4.7 0.1 4.6
1975 4046 100.0 72.6 27.4 14.2 13.2 9.4 1.8 1.0 0.0 8.6 4.3 0.1 4.3
1976 4069 100.0 72.6 27.4 13.9 13.4 9.8 1.6 1.1 0.0 9.3 4.4 0.1 4.8
1977 4039 100.0 72.3 27.7 14.0 13.7 10.0 1.5 1.0 -0.1 9.5 4.5 0.1 4.9
1978 3925 100.0 72.3 27.7 14.0 13.7 10.0 1.6 1.0 -0.1 9.3 4.3 0.1 4.9
1979 3831 100.0 73.8 26.2 13.4 12.7 9.4 1.6 1.2 -0.1 9.0 4.1 0.1 4.8
1980 3822 100.0 74.3 25.7 14.0 11.7 8.1 1.8 1.4 0.1 7.7 3.5 0.1 4.2
1981 3823 100.0 74.5 25.5 14.0 11.5 7.8 2.2 1.8 0.1 7.5 3.2 0.1 4.2
1982 3931 100.0 74.1 25.9 15.0 10.9 6.6 2.5 1.6 -0.2 5.6 2.5 0.1 3.1
1983 4049 100.0 71.0 29.0 17.1 11.9 7.8 2.2 1.4 -0.3 6.7 3.1 0.1 3.6
1984 4083 100.0 70.1 29.9 17.5 12.4 8.4 2.2 1.4 -0.2 7.4 3.1 0.0 4.3
1985 4198 100.0 70.7 29.3 17.3 12.0 7.7 2.2 1.4 -0.8 6.2 2.9 0.1 3.2
1986 4325 100.0 69.0 31.0 19.3 11.7 7.2 2.5 1.4 -0.7 5.5 2.7 0.0 2.8
31
Table 2 continued: Common size income statement for USA firms - The mid 60s versus the late 90s
C=A-B E=A-B-D
J=F-
G+H+I
M=J-
K-L
year
Number
of firms
Sales
(A)
COGS
(B)
Gross
margin
(C)
SG&A
(D)
Operating
income
before
depreciation
(E)
Operating
income after
depreciation
(F)
Interest
(G)
Non
operating
Income
(H)
Special
items
(I)
Pre tax
income
(J)
Taxes
(K)
Minority
Interest
(L)
Net
income
(M)
1987 4290 100.0 68.3 31.7 19.1 12.6 8.2 2.6 1.4 -0.3 6.9 3.0 0.1 3.8
1988 4122 100.0 68.0 32.0 19.4 12.6 8.5 3.0 1.2 -0.1 6.7 2.5 0.1 4.0
1989 4016 100.0 67.2 32.8 19.7 13.2 8.8 3.3 1.4 -0.6 6.5 2.6 0.1 3.8
1990 4016 100.0 67.2 32.8 20.3 12.5 8.4 3.2 0.9 -0.6 5.6 2.4 0.1 3.1
1991 4137 100.0 67.6 32.4 20.6 11.8 7.7 3.1 0.9 -1.2 4.4 2.0 0.1 2.3
1992 4491 100.0 68.0 32.0 19.9 12.1 7.9 2.5 0.7 -0.8 5.4 2.2 0.1 3.1
1993 5608 100.0 65.4 34.6 20.9 13.7 9.6 2.9 -0.3 -1.3 6.0 2.4 0.1 3.5
1994 5897 100.0 65.9 34.1 20.0 14.0 9.9 3.0 -0.1 -0.5 7.5 2.8 0.1 4.6
1995 6333 100.0 64.8 35.2 20.5 14.6 10.5 3.0 0.0 -1.3 7.1 2.7 0.1 4.2
1996 6141 100.0 65.2 34.8 20.5 14.4 10.3 2.9 0.0 -0.9 7.4 2.9 0.1 4.4
1997 5937 100.0 66.2 33.8 19.3 14.5 10.1 3.1 0.2 -1.2 7.0 2.7 0.1 4.1
1998 5801 100.0 65.0 35.0 20.2 14.8 10.3 3.5 0.0 -1.8 6.5 2.8 0.1 3.6
1999 5728 100.0 65.7 34.3 19.5 14.8 9.9 3.8 0.1 -0.2 7.6 3.1 0.1 4.3
2000 5454 100.0 65.8 34.2 19.9 14.3 9.1 3.6 0.2 -1.4 5.7 3.1 0.0 2.6
2001 5165 100.0 64.8 35.2 20.8 14.4 8.1 3.8 -0.4 -5.1 0.5 2.2 0.1 -1.8
2002 5052 100.0 65.1 34.9 20.0 14.9 10.0 3.5 -0.4 -3.2 4.5 2.7 0.1 1.7
2003 4441 100.0 63.1 36.9 20.5 16.5 11.4 3.5 -0.2 -0.8 8.3 2.8 0.1 5.3
Average (64-68): 100 70.0 30.0 15.2 14.8 10.6 0.9 1.0 0.0 10.7 4.6 0.1 6.0
Average (96-00): 100.0 65.6 34.4 19.9 14.6 9.9 3.4 0.1 -1.1 6.8 2.9 0.1 3.8
Difference between the average in the 1990s to the average in the 1960s:
0 -4.4 4.4 4.7 -0.3 -0.7 2.5 -0.9 -1.1 -3.8 -1.7 0.0 -2.2
32
Appendix
A.1 Approximations used in the calculation of realized IRR
The Compustat back dataset sometimes misses values used in the calculation of
realized IRRs. The approximations follow.
• If the book value of equity is missing we approximate it with the value of total assets
(Compustat item data6) less the value of total liabilities (data181).
• If dividends for either common or preferred shareholders are missing we set the value
to zero. These data are missing is the early years of the Compustat database. Setting
the values to zero downward biases Internal Rates of Return (IRRs) computed in the
1950s and 1960s. However, increasing the IRRs in this time period would only
strengthen our conclusions. Internal rates of return are highest in the 1950s and
1960s even with a downward bias.
• If sale of common stock and preferred stock is missing, its value is set to zero. If
purchase of common and preferred stock is missing it is also set to zero. These
approximations are used because Compustat does not provide this data before 1971.
The biases caused by this missing data are largely offsetting and are not likely to
affect conclusions, as the values of these variables tend to be very small by
comparison to other terms in the computation of IRR.
• If interest is missing, it is approximated by multiplying the book value of long-term
debt (data9) times the yield on Moody’s seasoned corporate bonds rated Baa (these
yields are obtained from the St. Louis Federal Reserve web site) and adding to that
the book value of short-term debt times the Baa rate minus 1.5 percentage points.
For computation of investments made by a corporation, we use capital expenditure
(data128), research and development (data46) and advertising (data45). We require that
capital expenditure exist. If R&D or advertising are missing we set them to 0; this occurs
before 1971 because Compustat does not provide that data before this date. However, the
bias is small because R&D and advertising were relatively small in proportion to capital
expenditures before 1971.
A.2 Examples of long term other liabilities
The dominant source of increase in “liabilities other” is from long term “liabilities –
other”. In 2001 the value of this item for all firms is $1.0 trillion. Examples of items
included in long term other liabilities are shown below for selected firms (in millions of
dollars).
Example #1: Ford Motor (Financial year 1999)
Automotive sector
:
Post retirement benefits other than pensions $15,458
Dealer and customer allowances and claims 7,271
Employee benefit plans 4,525
Unfunded pension obligation 1,189
Financial Services sector:
Other liabilities and deferred income 6,775
Example # 2: International Business Machine (Financial year 2000)
Non pension post retirement $7,128
Deferred income 1,266
Restructuring actions 854
Executive compensation accruals 769
Post-employment/pre-retirement liability 585
Environmental accruals 226
Other 497
Example # 3: General Motors (Financial year 2000)
Automotive operations
Post retirement benefits other than pensions 34,306
Pensions 3,480
Other liabilities and deferred income taxes 15,768
• Warranties
34
• Post employment benefits
• Unpaid losses under self-insurance programs
• Other
Financing and Insurance Operations
Other liabilities and deferred income taxes 12,922
Example # 4: General Electric (Financial year 2001)
All other liabilities (see below) $32,921
“This caption includes noncurrent compensation and benefit accruals at year-end 2002
and 2001 of $8,826 million and $8,745 million, respectively. Also included are amounts
for deferred income, interest on tax liabilities, product warranties and a variety of sundry
items” … such as “remediation actions to clean up hazardous wastes as required by
federal and state laws”.
Example # 5: Dow Chemical (Financial year 2000)
Pension and other post retirement benefits—non current: $1,746
Other non current obligations: $2,178
Description of other non current obligations: Silicone breast implant litigation, Superfund
environmental cleanup, other.
Note that according to COMPUSTAT the long term other liability classification does not
include:
1. Capital leases;
2. Deferred taxes (when reported separately);
3. Investment tax credit;
4. Long-term debt;
5. Minority interest;
6. Shareholders’ equity;
7. Unearned deferred compensation related to redeemable preferred stock, and;
8. Insurance liabilities, reserves and annuity benefits
35
A.4 Supporting information
Table A1: Individual firm yearly fund flow variables for capital markets, and the total firm
Book value of equity is total assets minus total liabilities (data6-data181). There may be hidden liabilities
like a pending lawsuit that could reduce the value of equity; we would count these items but have no way at
present to identify them (Enron is an example). Common dividends is data21; preferred dividends is
data19. If a dividend value is missing its value is set to 0. Sale of common and preferred stock (data108) is
sometimes combined with other figures or missing. We then compute it from the change in Capital Surplus
(data210); if this information is missing we set data108 to zero. Similarly there are times that Purchase of
common and preferred stock (data115) is combined with other figures or missing. We then compute it
from the change in Treasury Stock (data88); if data88 is missing we set data115 to zero. Interest expense
(data15) may also be combined with another figure or missing. In that case we let interest expense equal
that year’s Baa rate times the book value of long term debt plus the Baa rate minus 1.5 percentage points
times the book value of short term debt. Interest bearing debt is short term debt plus long term debt
(data34+data9). Total liabilities is data181, this includes all interest bearing debt plus non interest bearing
liabilities. Total assets is data6. Book value of equity equals total assets minus total liabilities (data6-
data181). Value of merger equity is reported by Security Data Corporation (variable name VALM).
Cash outflow from an individual firm Cash inflow to an individual firm
Panel A: Capital markets
Initial investment = book value of equity + interest
bearing debt
Purchase of own common and preferred stock Sale of common and preferred stock
Reduction in interest bearing debt Increase in interest bearing debt
Common and preferred dividends Acquired firm’s market value of equity – cash paid
for the firm + book value of its interest bearing debt
Interest expense
Terminal value = book value of equity + interest
bearing debt. For all firms (including target firms)
that exit COMPUSTAT in the horizon year,
terminal value = market value of the equity +
interest bearing debt.
Panel B: Total firm
Initial investment = book value of total assets
Reduction in total liabilities Increase in total liabilities
Purchase of common and preferred stock Sale of common and preferred stock
Common and preferred dividends Market value of acquired firms’ equity – cash paid
+ book value of total liabilities
Interest expense
Terminal value = book value of total assets. For all
firms (including target firms) that exit
COMPUSTAT in the horizon year, terminal value =
market value of the equity + book value of total
liabilities
36
Table A2: Samples sizes
The column labeled Corporate sector firms gives the number of firms used each year in the computation
of total firm IRR for the corporate sector. Details of the corporate sector IRR computation for the total firm
are given in Figure 1. The number of firms that survive and have sufficient information to compute their
total firm IRRs for a 10 year horizon is shown in the column labeled 10 year IRR. Similarly, the number
of firms that survive and have sufficient information for computation of their total firm IRR for horizons up
to 50 years are shown in the remaining columns. Details of the computation of total firm IRR for
individual firms are given in Figure 1.
Year
Corporate
sector firms
10 year IRR 20 year IRR 30 year IRR 40 year IRR 50 year IRR
1950 557 426 424 355 231 163
1951 658 434 431 347 234 159
1952 666 440 438 338 236 155
1953 674 446 445 332 231 146
1954 690 464 462 328 237
1955 708 482 473 323 238
1956 726 499 480 312 243
1957 746 514 484 310 241
1958 807 537 490 306 234
1959 865 565 493 311 227
1960 1427 1013 819 489 320
1961 1564 1281 1000 645 415
1962 1766 1522 1118 736 451
1963 1984 1766 1265 826 499
1964 2124 1857 1268 856
1965 2274 1950 1269 854
1966 2423 2024 1261 876
1967 2587 2087 1265 872
1968 3145 2287 1354 911
1969 3334 2332 1380 901
1970 3399 2334 1389 857
1971 3584 2408 1432 863
1972 3707 2434 1478 869
1973 4028 2599 1543 881
1974 5031 2996 1768
1975 5179 2887 1707
1976 5212 2758 1670
1977 5221 2684 1639
1978 5157 2602 1565
1979 5094 2549 1512
1980 5186 2593 1454
37
Table A2 continued:
Year
Corporate
sector firms
10 year IRR 20 year IRR 30 year IRR 40 year IRR 50 year IRR
1981 5329 2657 1459
1982 5562 2806 1491
1983 5831 2927 1498
1984 5849 3015
1985 6085 3158
1986 6372 3323
1987 6409 3291
1988 6328 3206
1989 6216 3102
1990 6295 3032
1991 6435 2991
1992 6781 3045
1993 7757 3231
1994 8051
1995 8557
1996 8396
1997 8102
1998 7968
1999 7857
2000 7541
2001 7046
2002 6896
2003 6447
38
Sum of other liabities for all firms
0
500000
1000000
1500000
2000000
2500000
3000000
3500000
1950 1970 1990
$ millions
Long term OL
Total OL
Figure A1: Aggregate other liabilities
The sample is all surviving and non surviving Compustat firms excluding insurance firms having SIC codes
between 6300 and 6499. Further, to be in the sample, the firm must report total assets (data6) greater than
0. Current other liabilities is data72. Long term OL (other liabilities) is sum of long term other liabilities
= sum(data75). Total OL is sum of long term other liabilities plus the sum of current other liabilities =
sum(data75) + sum(data72). If either data72 or data75 is missing, their value is set to 0. Before 1963,
current other liabilities (data72) is not reported.
39
Figure A2: Aggregate cash flows for fund providers
The sample is all Compustat firms (excluding insurance firms, i.e. SIC codes between 6300 and 6499) with
total assets greater than 0. CFs to all fund providers is the yearly fund flow to all providers; it is defined
in Figure 1. Net to capital markets = fund flow to capital markets less the fund flow to all providers.
Yearly fund flow to capital markets is defined in Figure 2.
40
Aggregate cash flows
-700.00
-200.00
300.00
800.00
1950 1970 1990
Dollars (Billions
)
-1200.00
CFs to all fund providers
Net to capital markets
Table A3: Capital markets: The effect of changing other liabilities on IRR and firm survival rate
The sample is all Compustat firms (excluding insurance firms, i.e. SIC codes between 6300 and 6499) in existence between 1950 and 1993 that have the
information needed to do the required IRR computations. Panels A and C are for all firms in the period; Panels B and D are for S&P 500 firms in existence
between 1984 and 1993. S&P 500 firms are selected if the firm is in the S&P 500 list in the base year of the IRR computation. Firms are ranked each year into
five groups based on scaled change in other liabilities. Scaled change in other liabilities (%) is the difference between (a) other liabilities at time 10 and (b)
other liabilities at time 0 all scaled by total assets at time 0 = (Other liabilities at t10 – other liabilities at t0) / (total assets at time 0). The variable T0 is the base
year of the IRR analysis where investment is determined. The variable T10 is the terminal year of the 10 year IRR computation. 10 year IRR is computed for
capital markets as is described in Table 2. 5 year IRRs used in panels C and D are computed in a similar fashion, but over a horizon of 5 years. Other liabilities
is total liabilities less the book value of interest bearing debt = data181-data34-data9. Total assets is data6.
In Panels C and D firms are categorized into two groups: (a) Increasing other liabilities – these are firms with an increase in their “Scaled change in other
liabilities” and (b) Decreasing other liabilities – these are firms with a decrease in their “Scaled change in other liabilities”. The correlation, i.e. rho, between
the IRR (5 year or 10 year) and “Scaled change in other liabilities” is found for each of the two samples. Spearmans rank correlation test is used to determine
the significance of the correlations. dOL is an abbreviation for “Scaled change in other liabilities.” Survival rate is the fraction of firms in existence at five
years that survive through at least ten years. The difference in proportions test is used to determine if the survival rates are significantly different.
Change in other liabilities versus IRR The correlation of change in other liabilities to IRR and
survival rate versus change in other liabilities
Medians
Rank on
scaled change
in other
liabilities
Sample
size
Scaled change
in other
liabilities (%)
Other
liabilities
at t0
Other
liabilities at
t10
Total assets
at t0
10 year
IRR (%)
Increasing
other
liabilities
Decreasing
other
liabilities
Panel A: All firms (1950-1993) Panel C: All firms – correlations and survival rates
0 (low) 17,765 -2.4 7.3 5.7 40 5.9 10 year rho: IRR to dOL 0.3* 0.08*
1 18,014 16.7 11.6 23.1 66 8.8 10 year sample size 78,713 10787
2 18,027 37.4 18.3 53.1 93 10.3 5 year rho: IRR to dOL 0.2* 0.10*
3 17,982 71.0 17.4 75.8 83 11.8 5 year sample size 115,073 25756
4 (high) 17,822 190.8 7.8 88.5 35 12.9 Survival rate 0.68 0.42^
Panel B: S&P 500 firms (1984-1993) Panel D: S&P 500 firms – correlations and survival rates
0 (low) 721 12 1727 2,359 6,118 9.4 10 year rho: IRR to dOL 0.1* -0.02
1 729 34 1029 2,250 3,451 10.5 10 year sample size 3,464 116
2 725 62 685 2,125 2,337 12.1 5 year rho: IRR to dOL 0.1* 0.10**
3 711 122 692 3,565 2,197 11.0 5 year sample size 5,304 398
4 (high) 694 430 137 3,257 480 11.3 Survival rate 0.65 0.29^
* p-value < 0.0001, ** p-value < 0.05; ^ p-value < 0.0001 for Ho: Survival rate for firms with increasing other liabilities = survival rate for firms with decreasing
other liabilities.
Table A4: Firms in the top decile of market value – Total IRR versus external financing at increasing levels of investment (1975-2003)
The sample is all U.S. firms in Compustat (excluding insurance firms) between 1975 and 2003. Further, firms must be in the top decile of market value and have
the information necessary to compute total firm IRR. To be included in the computation of IRR, a firm must survive through the horizon length (5, 10, or 15
years). Firms in the top decile of market value (data25*data199) are found by ranking firms by market value each year. The reported IRR is the average of the
median value of the nominal realized total firm IRRs found each year; the computational details for computing IRR are given in the Figure 1. For the IRR
computation, the initial investment is the book value of total assets. Terminal value is usually the book value of total assets for firms that survive through the
horizon. The exception occurs when the firm exits the database in the horizon year, then terminal value is set equal to market value of equity plus book value of
total liabilities Ranking is first done each year by amount of investment then by amount of external financing (both are scaled by total assets in the ranking
year). Investment is the last two years of (a) capital expenditure (data128), (b) research and development (data45), and c) advertising (data46). If R&D or
advertising are unknown they are set to 0. External money is the a) net debt issued over a two year period (data34+data9 at time t – their values at time t-2) plus
(b) the net issuance of stocks over the last two years (i.e. sale of common and preferred stock less the purchase of common and preferred stock over a two year
period = data108 at time t -data115 at time t + data108 at time t -1 - data115 at time t-1. Book to market is the book value of equity / market value of equity =
data60/(data25*data199). The variable
σ
for 10 year IRR is the average of the yearly standard deviations.
Ranks Average of the yearly median values
Investment
External
financing
Investment External money Total assets
Book-to-
market
IRR:
5 year horiz.
IRR:
10 year horiz.
IRR:
15 year horiz.
σ
for 10
year IRR
0 min 0 min 0.028 -0.137 53 0.984 0.069 0.067 0.067 0.093
0 min 1 0.031 0.011 82 0.907 0.067 0.069 0.069 0.302
0 min
2 max 0.028 0.238 109
0.736
0.064 0.063 0.065 0.096
1 0 min 0.087 -0.076 105 0.771 0.077 0.072 0.069 0.107
1 1 0.089 0.030 166 0.764 0.074 0.072 0.072 0.060
1 2 max 0.092 0.221 98 0.650 0.062 0.064 0.067 0.081
0 min
0.185
-0.032
60
0.572
0.079
0.075
0.075
0.125 2 max
2 max 1 0.193 0.113 81 0.562 0.072 0.071 0.071 0.085
2 max 2 max 0.252 0.511 27 0.355 0.022 0.049 0.053 0.125
42
Total firm median 10 year IRR; Nominal values
Top third of all
investing firms
Investment with internal vs external money
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
1950 1960 1970 1980 1990
10 year IRR
internal
external
Figure A3: All firms in the top rank of investment and top decile of market value – Median yearly total firm IRR for internal and external financing
This figure plots the yearly median 10 year IRR for firms with the highest rank of investment level (rank = 2) for high (rank = 2) and low (rank=0) levels of
external financing. Internal means firms ranked 0 for external financing; external means firms ranked 2 for external financing. The sample, details of IRR
computation, and ranking procedure are given in Table A4.
43
Asset weighted total firm IRR
assets values at book versus at replacement value
0
0.02
0.04
0.06
0.08
0.1
0.12
1975 1980 1985 1990 1995
IRR
RVTAw_IRR10
TAw_IRR10
Figure A4: Total firm IRR computed using the replacement value of total assets versus the book value of total assets
25
The sample is all firms with sufficient data to calculate the 10 year horizon IRR using both replacement value of total assets and the book value of total assets
(data6). Replacement value is computed using procedures given by Lewellen and Badrinath (1997) and Lee and Tompkins (1999). Since replacement value
requires data that large manufacturing firms usually have, the sample of firms is reduced by about two-thirds. For example, in 1980, total firm IRR for a 10 year
horizon was calculated for 2,593 firms using the book value of total assets. In the same year there were only 952 firms with sufficient replacement value data to
calculate the total firm IRR for a 10 year horizon. In 1980, this figure shows IRR computed both ways for just the 952 firms. This figure shows the 10 year
horizon IRR calculated using book value (TAw_IRR10) for initial and terminal value versus using replacement value (RVTAw_IRR10). The calculations are
done on the same set of firms from 1975 through 1990. The calculation procedure for TAw_IRR10 is described in Figure 1. The procedure for RVTAw_IRR10
is the same except replacement value of total assets is used in place of the book value of total assets.
25
In unreported results we show that fixed assets are declining as a proportion of total assets through time.
44
Endnotes
i
Fama and French (1999) (FF99) is the previous key investigation of long-term return on U.S. corporate investment. They conclude that the return on corporate investment is
significantly greater than the weighted average cost of capital (WACC). We take issues with their study for a number of reasons.
First, the FF99 conclusion is methodologically predetermined. Return on investment is set to the discount rate that equates corporate cash flows to book value of equity plus
interest bearing debt. WACC is set, by FF99, to be the discount rate that equates corporate cash flows to initial market value. The cash flows over time in these two mathematical
problems are negligibly different. Since investors will not initially buy stocks unless they believe they will receive a return greater than the WACC, FF99 must find that that U.S.
corporations only invest in positive net present value (NPV) projects regardless of corporate cash flows.
Second, FF99 do not explicitly include assets financed by non-interest bearing liabilities for two reasons. One is that they assume these assets earn an implicit return (equal to
the cost of capital) that affects firm earnings; however, their method does not enable one to determine whether there is a wealth transfer from non-interest bearing liability holders
to capital market participants. The second reason FF99 give is the claim that non-interest bearing liabilities are mainly due to an increase in short term trade receivables. However,
long term other liabilities such as benefits promised to employees (See the appendix for extensive examples) have increased significantly. These long term type of management
commitments are now backed by approximately 9.8 percent ($2.3 trillion) of all corporate assets compared to less than 1.7 percent ($1.4 billion) before 1960.
Third, the FF99 study follows the efficient market theory by making the assumption that a firm’s market value is an unbiased estimate of the firm’s future cash flows (i.e. the
firm’s terminal value). Shiller (1981), LeRoy and Porter (1981), and West (1988) have long argued that stocks are misvalued by the market. Shiller (2003) uses standard analysis
of over 120 years of data to show that in the last 50 years U.S. stocks likely were significantly overvalued relative to the present value of dividends for all but a brief stretch of
about seven years in the late 70s and early 80s. Return computations are sensitive to the assumed terminal value of the firm. If broad misvaluations occur for extended periods as
the literature suggests, then the FF99 reported returns are biased significantly upward, especially over the time period of the FF99 study; however, FF99 do not explore the
sensitivity of their results to extended periods of market misvaluation.
Fourth, FF99 compute firm cash flows by using accounting earnings and many values from both the asset and liability sides of the balance sheet. In the process FF99 make
many assumptions. The net cash flow that investors actually receive is never found. The values that FF99 use are subject to manipulations as is evident from recent revelations
about accounting irregularities in large U.S. corporations.
Finally, FF99 only investigate aggregate corporate returns. A study of individual firm returns is required to enable analysis of the characteristics of corporate capital allocation
and the associated returns.
Technically our results differ from FF99 mainly because (a) we explicitly find the return earned on assets financed by non interest bearing liability holders, and (b) we avoid the
use of, possibly biased, market valuation as the best estimate of the present value of future cash flows (PVFCFs) and instead derive an upper bound on the PVFCFs.
45