What do we know about
executive compensation at
privately held firms?
Rebel A. Cole
DePaul University - Chicago
Hamid Mehran
Federal Reserve Bank of New York
Discussant: Massimiliano Stacchini
Banca d’Italia
contribution of the paper
The paper makes a significant contribution to the
literature on executive pay at privately held
firms…
…. which has been strongly limited by lack of data.
‘Published studies on compensation in privately
held firms are essentially nonexistent because
the data generally has not been accessible’.
Kee, et al (1999)
outline
sampling
multivariate analysis
- firm heterogeneity and pay-size elasticity
- return to education
other comments
conclusions
sampling
Analysis on privately held firms is based on
data from 2 samples of firms conducted in
1993 and 2003 (SSBF).
The Surveys rely on a stratified random
sampling design.
sampling
…Consistency between population and
sample is provided for by sampling
weights…
Some groups are over sampled (e.g., employee
groups of size 20 and above), and the surveys
contain weights to ensure that sample statistics
represent the population.
(Herrants et al. 2009)
sampling
‘Stylised facts’ apparently do not control for the
structure of the sampling design
(i.e. sampling weights are omitted)
..even if the risk of sampling error
seems to be non negligible…
(the sampling fraction is around 0.002)
The total population of S-Corp and C-corp consists of about
2,400,000 firms in 1993
(2,800,000 firms in 2003)
and is represented by
4,356 firms in 1993
(4,240 firms in 2003)
The ‘final’ C&M sample consists of:
1,630 firms for 1993 (1,668 firms for 2003)
(obtained by excluding firms who refused to response, firms whose the primary
owner is not the day-to day manager and public firms)
sampling
Suggestion:
- Take into account sampling weights.
item nonresponse
Item nonresponse is significant in SSBF.
‘The 2003 Survey of Small Business
Finances (SSBF) screening interview had
significant unit nonresponse and therefore
some type of nonresponse adjustment
was deemed necessary’
(Federal Reserve Board Finance and
Discussion Series Paper, 2007).
item nonresponse
The authors say that some firms were
excluded as they refused to divulge their
amount of CEO compensation.
Suggestion:
- The number of nonresponding firms
should be indicated and compared with
sampling size
- Item nonresponse should be controlled for
in case
item nonresponse
E.g.
Are non responding firms more likely to
combine high compensations with lowperformances than their peers?
(correlating characteristics of nonresponding
firms to the Ceo pay question, with those of
responding firms, that exhibit this pattern)
multivariate analysis
The authors find a much higher pay-size
elasticity at private companies than for
public firms…
…but also a drop of the pay-size elasticity
for privately held firms from 1993 to 2003.
multivariate analysis
…which is supposed ‘to be driven by the
growing familiarity with the use of the 0.30
benchmark among accountants of
privately held firms’...
multivariate analysis
Do heterogeneous pay-size elasticities exist
within privately held firms?
If differences exist but are not modelled,
a variation in the firms’ composition could
translate in an apparent variation of the
estimated (single coefficient for) pay-size
elasticity.
multivariate analysis
‘The relatively uniformity of the elasticity of
executive pay with respect to scale across
firms, industries, countries and periods of
time is puzzling
because the technology which sustains
control and scale should vary across these
disparate units of comparison’.
Rosen (1990)
multivariate analysis
Zhou (2000) models variations of the pay-size
elasticity for firms having different size.
(he finds a higher elasticity for larger firms).
multivariate analysis
Kostiuk (1990)
- models firm heterogeneity by interacting
‘pay-size elasticity’ with ‘capital intensity’
(capital–labor ratio)
(pay-size elasticity is lower at firms having
higher capital-labor ratios).
multivariate analysis
Size distribution: Firm heterogeneity within
the period.
total employment
(number of
individuals)
1st Quartile
4th Quartile
2003
0-4
54 - max
1993
0-6
62.5 - max
(the survey may include firms having less than 500 employees)
multivariate analysis
Size distribution: Firm heterogeneity across
periods
median
∆% (2003 vs 1993)
ASSETS ($)
-22%
580,000
450,000
Total employment
(numb.of ind)
-32%
22
15
SALES ($)
-7%
1993
1,500,000
2003
1,400,000
In 2003 the median size of firms was lower
than in 1993.
multivariate analysis
Suggestion:
Interact pay-sale elasticity with
- Proxies for firms’ complexity
- Proxies for firms’ dimension
E.g. ‘Large’ (=1 for firms having sales larger
than the sample median), Zhou (2000)
multivariate analysis
Legal type: heterogeneity across periods
C-corporations account for:
- 60% of sampling firms in 1993
- 30% of sampling firms in 2003
The relative importance of C-corporation (vs Scorporation) get reversed across periods.
multivariate analysis
Suggestion:
Interact pay-sale elasticity with ‘legal type’
return to education
Return to education may be expected to vary
with ‘family ownership’.
…Coeteris paribus, return to education for a
manager of a relatively large private firm
inherited by the family may differ from that for
a manager starting a new enterprise.
Suggestion:
Interact the ‘Graduate’ and ‘College’ degree
dummies with
- family ownership,
- age of firms, - size of firms.
other comments
Different specifications between the 1993 and
2003 models prevent the comparison of results,
‘all being equal’:
- ‘firm’s D&B Credit Score’ is included only in the
2003 model
Suggestion:
Presenting a model for 2003 by removing ‘firm’s
D&B Credit Score’
(which apparently correlates with RoA)
other comments
Regressions for pay-size elasticity, including the entire
list of determinants, are available for sale as a measure
of size.
In addition, ‘pay-assets’, ‘pay-employment’ elasticities
are even scrutinized by C&M in the bivariate section…
…and by other authors.
Suggestion:
Regressions for pay-size elasticity, including the entire
list of controls, should be presented even for ‘payassets’ and ‘pay-employment’.
other comments
The interaction term:
Ownership*C-corporation
is discussed in the text…
(The relation Ceo pay vs ownership is stronger at
C-corporations)
…. but apparently is not included in the set of
regressions.
other comments
As a corollary…
Interactions between pay-size elasticity and or
size (or complexity) may signal the extent of
the ‘measurement error’…
The Survey’s question on executive pay is:
‘Which is the total amount of officers’ compensation? ‘,
(not CEO pay)
other comments
Exercises conducted by C&M seem to indicate
that the measurement error is low.
The measurement error is expected to increase
with firms’ size.
If interactions between pay-size elasticity and
size, after controlling for the other determinants,
prove to be insignificant…
…the measurement error would be implicitly
negligible.
conclusions
The paper is definitely original, full of
information, and very enjoyable to read.
The empirical analysis is comprehensive and
shed light on important up till now obscure
issues.
conclusions
Compositions effects may have an impact on
the results.
Among privately held firms, heterogeneity in
size and complexity should be properly dealt
with.
Causal effects, controlling for endogeneity,
have still to be scrutinized.
Thanks for your attention.
Trhoughout the regression
Firm heterogeneity and
measurement error
Chrinsman’s (2007)
(Privately held family firms, US)
Number of employees (mean) = 32
Number of family managers (mean) =
2.5
Cole & Mehran’s sample
Number of employees (third quartile) = 62 (1993) ,
53 (2003)
Eisemberg (1998)
(Closely held firms, Finland)
Total assets (median) = $ 800,000
Board size (median) =
3
Cole & Mehran’s sample
Total assets (median) = $ 575,000 (1993)
Total assets (third quartile) =
$ 2,315,000 (1993)
to tal assets
Pay-size elasticity, by quartiles of the
size distribution
0.8
0.7
0.6
0.5
0.4
93
0.3
03
blu:
1993
brown: 2003
0.2
assets
0.1
0
f ir st
second
t hir d
f our t h
emplo yement
1
0.9
0.8
0.7
0.6
0.5
93
0.4
03
0.3
0.2
employment
0.1
0
f ir st
second
t hir d
f our t h
sales
0.9
0.8
0.7
0.6
0.5
93
0.4
03
0.3
0.2
sales
0.1
0
f ir st
second
t hir d
f our t h