# A Critical Long View Of Capital Markets And Institutions:

- 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.

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.

- 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. corporationsâ€™ increasing

utilization of long-term other liabilities

5

as a source of funds

6

, these mostly non-

interest bearing â€˜involuntaryâ€™ 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

firmsâ€™ 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

â€˜involuntary contributors.â€™ 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 â€˜involuntary contributorsâ€ finance.

General Motorsâ€™ projected difficulty with funding retiree liabilities is but one example of

the consequence of financing assets through â€˜involuntary contributorsâ€™ without creating

sufficient matching assets to pay the liabilities. Without explicit analysis of these

7

â€˜involuntaryâ€™ fund providers, it is impossible to determine how much of the realized

returns to equity and bond holders come at the expense of the â€˜involuntary contributors.â€™

2.3 Internal rate of return (IRR) calculation â€“ 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 firmâ€™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 firmâ€™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 firmsâ€™ 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 â€˜closed system fund

flowâ€™ 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

Jensenâ€™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 Jensenâ€™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 shareholdersâ€™ 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 firmâ€™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 providersâ€™ 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:

âˆ‘

=

+

+

+

=

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 FernÃ¡ndez (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, â€¦) through the

horizon given (10 years, 20 years, â€¦) 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 â€˜other liabilities.â€™

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, â€œeven

at the expense of long term performance.â€ 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 firmsâ€™ 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 â€“ 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 firmsâ€™ 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, â€œRobustness tests.â€ 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 â€˜other liabilitiesâ€™, 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 firmsâ€™ 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 â€˜axiomsâ€™ or â€˜givensâ€™ 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

â€˜other liabilitiesâ€™ 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.

FernÃ¡ndez, 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

âˆ‘

=

+

+

+

=

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 firmsâ€™

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

âˆ‘

=

+

+

+

=

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, â€¦ 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 firmsâ€™ 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 â€“ 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 â€“ 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

Ïƒ

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.

Ïƒ

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

âˆ‘

=

+

+

+

=

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.

â€¢ 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).

â€¢ 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.

â€¢ 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.

â€¢ If interest is missing, it is approximated by multiplying the book value of long-term

debt (data9) times the yield on Moodyâ€™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 â€œliabilities otherâ€ is from long term â€œliabilities â€“

otherâ€. 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

â€¢ Warranties

34

â€¢ Post employment benefits

â€¢ Unpaid losses under self-insurance programs

â€¢ 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

â€œThis 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

itemsâ€ â€¦ such as â€œremediation actions to clean up hazardous wastes as required by

federal and state lawsâ€.

Example # 5: Dow Chemical (Financial year 2000)

Pension and other post retirement benefitsâ€”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. Shareholdersâ€™ 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 yearâ€™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 firmâ€™s market value of equity â€“ 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 firmsâ€™ equity â€“ 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 â€“ 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 â€“ these are firms with an increase in their â€œScaled change in other

liabilitiesâ€ and (b) Decreasing other liabilities â€“ these are firms with a decrease in their â€œScaled change in other liabilitiesâ€. The correlation, i.e. rho, between

the IRR (5 year or 10 year) and â€œScaled change in other liabilitiesâ€ 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 â€œScaled change in other liabilities.â€ 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 â€“ 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 â€“ 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 â€“ 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 â€“ 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

Ïƒ

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.

Ïƒ

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 â€“ 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 firmâ€™s market value is an unbiased estimate of the firmâ€™s future cash flows (i.e. the

firmâ€™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

- Web Chapter Financial
3 MB

- Transparency In Financial
15.67 kB

- Financial Markets And
5.92 MB