Conglomerate Investment, Skewness, and the CEO Long‐Shot Bias

• Published date: 01 April 2016
• Author: OLIVER SPALT,CHRISTOPH SCHNEIDER
• DOI: http://doi.org/10.1111/jofi.12379
• Date: 01 April 2016

THE JOURNAL OF FINANCE •VOL. LXXI, NO. 2 •APRIL 2016

Conglomerate Investment, Skewness,

and the CEO Long-Shot Bias

CHRISTOPH SCHNEIDER and OLIVER SPALT∗

ABSTRACT

Do behavioral biases of executives matter for corporate investment decisions? Using

segment-level capital allocation in multisegment firms (“conglomerates”) as a labora-

tory, we show that capital expenditure is increasing in the expected skewness of seg-

ment returns. Conglomerates invest more in high-skewness segments than matched

stand-alone firms, and trade at a discount, which indicates overinvestment that is

detrimental to shareholder wealth. Using geographical variation in gambling norms,

we find that the skewness-investment relation is particularly pronounced when CEOs

are likely to find long shots attractive. Our findings suggest that CEOs allocate capital

with a long-shot bias.

MAKING INVESTMENT DECISIONS that maximize shareholder value is the central

task of corporate managers, and every MBA curriculum features state-of-the-

art valuation tools prominently. Nevertheless, making investment decisions in

the real world is difficult because even the best valuation tools rely to a consid-

erable extent on assumptions that are subjective.1Consistent with substantial

residual uncertainty around true project value, about half of the CEOs sur-

veyed in Graham, Harvey, and Puri (2015) mention “gut feel” as an important

or very important factor in their investment and capital allocation decisions.

As there is by now overwhelming evidence suggesting that intuitive reason-

ing in financial matters frequently leads to biased and therefore suboptimal

∗Christoph Schneider is at the University of Mannheim. Oliver Spalt is at Tilburg Univer-

sity. We would like to thank Malcolm Baker, Martijn Cremers, Joost Driessen, Sebastian Ebert,

Markus Glaser,Cam Harvey (the Editor), Byoung-Hyoun Hwang, Paul Karehnke, Han Kim, Ernst

Maug, Zacharias Sautner, David Thesmar,two anonymous referees, our colleagues at Mannheim

and Tilburg, and participants at seminars and conferences for comments and helpful discussions.

We are responsible for all remaining errors and omissions. Christoph Schneider gratefully ac-

knowledges financial support of the collaborative research center TR/SFB 15 “Governance and the

Efficiency of Economic Systems” at the University of Mannheim. We have read the Journal of

Finance’s disclosure policy and have no conflicts of interest to disclose.

1As a simple example, suppose that the cash flow of a project is $1 next year. The appropriate

discount rate is 5%. Assuming a perpetual growth rate of 2% leads to a project value of $33.3.

Using an equally defensible growth rate of 3% instead yields a value estimate of $50.0, which is

50.0% higher. Even with substantial resources spent on gathering relevant information, most val-

uation models necessarily retain a significant subjective component, and two equally sophisticated

individuals can obtain substantially different valuation results for most investment projects.

DOI: 10.1111/jofi.12379

635

636 The Journal of Finance R

decisions, a natural—and potentially very important—question is whether bi-

ases distort optimal capital allocation in firms.

A central difficulty in addressing this question is that researchers do not

usually observe the characteristics of individual projects considered by corpo-

rate decision makers. In this paper, we propose looking at segment-level capital

allocation in multisegment firms (“conglomerates”) as an identification strat-

egy.2Throughout our study, segments are defined as all operations of a firm in

the same Fama-French 48 (FF48) industry (Fama and French (1997)). Since

firms are required to disclose segment-level information, we can “look inside”

conglomerates and study how biases affect capital allocation across segments.

A particular advantage of this approach is that prior research suggests that

there may be a link between capital budgeting and executive biases. CEOs are

central to the capital allocation decision as they have “total and unconditional

control rights” and can “unilaterally decide” what to do with a segment’s physi-

cal assets (Stein (2003)). Further, almost 40% of U.S. CEOs say that they make

capital allocation decisions with very little or no input from others according

to the survey in Graham, Harvey, and Puri (2015). Hence, looking at capital

allocation in conglomerates allows us to obtain a large sample of economically

important investment decisions made by individuals who self-report a tendency

to rely on gut feel.

Our paper provides new evidence indicating that managerial biases can lead

to distorted capital budgets that do not maximize shareholder wealth. We fo-

cus on the implications of a powerful behavioral phenomenon that we label the

“CEO long-shot bias,” which is the tendency of CEOs to systematically over-

value projects with high perceived upside potential (as proxied by expected

skewness in our empirical tests). One prominent potential source of this phe-

nomenon is prospect theory’s probability weighting feature, but, as we discuss

in greater detail below, there may also be other drivers. We posit that the spe-

cial authority of conglomerate CEOs in capital allocation decisions, together

with the fact that assumptions in valuation models are partly subjective, al-

low the CEO long-shot bias to affect capital budgeting. CEOs subject to such

bias destroy shareholder wealth by investing too much in segments with high

perceived upside potential.

To fix ideas, consider a simple example of a hypothetical conglomerate with

two segments, A and B. The CEO oversees a fixed investment budget of I=5for

new projects that she can allocate to either segment A or segment B. Segment

A proposes the following project:

[(2,0.4); (8,0.6)],

which generates a present value of cash flows, discounted at the appropriate

risk-adjusted rate, before investment of 2 in the low state, which occurs with a

2The idea of using segment data to open up the black box of capital budgeting has a long

tradition in the literature, going back at least to Shin and Stulz (1998).

Conglomerate Investment 637

probability of 0.4, and of 8 in the high state, with a probability of 0.6. Segment

B proposes a project with a more skewed payoff distribution:

[(2,0.9); (30,0.1)].

This project yields 2 in the low state, with a probability of 0.9. However, with

a probability of 0.1, project B will be a major success, yielding a value before

investment of 30. Based on these numbers, because the expected net present

value (NPV) of project A is 0.6 and the expected NPV of project B is −0.2, a

rational CEO should allocate the investment budget to project A. But if the

CEO’s long-shot bias is strong enough, she may invest in project B and in turn

destroy shareholder wealth.

Below, we begin by showing that the above example is consistent with actual

investment patterns in the Compustat universe of U.S. conglomerates from

1990 to 2009: capital expenditure is significantly higher for segments with

projects that have higher expected skewness. This effect is particularly pro-

nounced for smaller segments, as predicted by the CEO long-shot hypothesis.

The positive relation between expected skewness and investment is robust to a

battery of standard controls and additional checks, including firm and segment

fixed effects.

When we match conglomerate segments to comparable stand-alone firms, we

find that investment in conglomerates is significantly higher when skewness

is high, even though we control for potentially greater debt capacity and other

factors known to affect investment. In fact, among stand-alones, there is no

skewness-investment relation after we control for industry-specific effects. This

evidence, together with several additional tests, indicates that high skewness

is not simply proxying for good investment opportunities.

We next look at value implications, using the method of Berger and Ofek

(1995) and controlling for endogeneity of the diversification decision using fixed

effects, instrumental variables, and selection models, as in Campa and Kedia

(2002). We find that conglomerate firms with skewed segments are valued at a

significant discount by the market. As in our simple motivating example above,

this suggests that conglomerates overinvest in segments with high expected

skewness and that this investment behavior is detrimental to shareholder

wealth.

The empirical patterns above are potentially consistent with a number of

channels other than the CEO long-shot bias. First, the project with more

skewed returns (project B) might be harder to value and therefore more prone

to idiosyncratic valuation errors. Second, agency theory suggests that CEOs

may strategically exploit subjectivity in valuations to tilt capital budgets to-

ward an allocation that maximizes private benefits (e.g., Rajan, Servaes, and

Zingales (2000), Scharfstein and Stein (2000)). Our analysis shows that random

valuation mistakes or agency problems cannot easily explain the skewness-

investment relation we document.

We propose, and provide evidence for, a third channel: managers subject to

the long-shot bias choose project B because it offers larger upside potential and