University of South Carolina University of South Carolina
Scholar Commons Scholar Commons
Faculty Publications Finance Department
10-1997
The Coexistence of Multiple Distribution Systems for Financial The Coexistence of Multiple Distribution Systems for Financial
Services: The Case of Property-Liability Insurance Services: The Case of Property-Liability Insurance
Allen N. Berger
University of South Carolina - Columbia
, aberger@moore.sc.edu
J. David Cummins
Mary A. Weiss
Follow this and additional works at: https://scholarcommons.sc.edu/<n_facpub
Part of the Finance and Financial Management Commons, and the Insurance Commons
Publication Info Publication Info
Published in
Journal of Business
, Volume 70, Issue 4, 1997, pages 515-546.
http://www.jstor.org/action/showPublication?journalCode=jbusiness
© 1997 The University of Chicago Press
This Article is brought to you by the Finance Department at Scholar Commons. It has been accepted for inclusion
in Faculty Publications by an authorized administrator of Scholar Commons. For more information, please contact
digres@mailbox.sc.edu.
Allen
N.
Berger
Board
of
Governors
of
the Federal Reserve System
and Wharton
Fmanczal Instztutions Center
J. David Cummins
Umverslty
of
Pennsylvania
Mary
A.
Weiss
Temple Umverslty
The
Coexistence
of
Multiple
Distribution
Systems
for
Financial
Services:
The
Case
of
Property-Liability
Insurance*
I.
Introduction
Economic theory predicts that in long-run com-
petitive equilibrium the price
of
a good or service
will equal the minimum average costs associated
with the most efficient production
technology-
firms that have inefficient technologies and
higher average costs will not survive. The coexis-
tence over long periods
of
time
of
alternative
technologies performing the same function thus
poses an interesting economic puzzle.
Prominent
examples are alternative distribution systems for
the same or sImilar financial service, such
as
full-
service and discount brokers for performing
securities trading; automatic teller machines and
human tellers for distributing cash; banks, sav-
ings and loans, and credit unions for delivering
* The opmlOns expressed do not necessanly reflect those
of
the Board
of
Governors or its staff The authors thank the
anonymous referee, the editor (Doug DIamond), DWight Jaffee,
and the participants m the NatIOnal Bureau
of
EconomIc Re-
search Property-Llablltty Insurance
Semmar for helpful com-
ments. The authors acknowledge the capable research assIs-
tance
of
Jalal Akhavein and Joe Scalise
(Journal
of
Busmess, 1997, vol 70, no 4)
© 1997 by The Umverslty
of
ChIcago All nghts reserved
0021-9398/9717004-0003$02 50
515
Copyright © 2001. All Rights Reserved.
Property-liabIlIty insur-
ance
IS
distnbuted
through a direct-writer
system, where agents
represent one lllsurer,
and an independent-
agency system, where
agents represent sev-
eral insurers. Indepen-
dent-agency insurers
have
hIgher costs than
direct writers. The
market -impe rfections
hypothesis
attributes
the coexistence
of
the
two types
of
insurers
to impediments to com-
petition,
whtle the
product-quality hypoth-
esis holds that indepen-
dent-agency insurers
provide hlgher-quahty
servIces. We measure
cost efficiency and
profit effiCiency for
property-liabIlIty in-
surers and find strong
support for the prod-
uct-qualIty
hypotheSIS,
ImplYlllg that indepen-
dent-agency lllsurers
produce higher-quality
outputs and are com-
pensated by higher rev-
enues.
516
Journal
of Business
depository servIces; and banks and capital markets for providing loans
to busmesses (Fama 1985).
This article focuses on a partIcularly interesting case
of
financial-
services distribution, property-liability insurance. Property-liability in-
surance is distributed by two different types
of
firms: direct-writing
insurers that distribute insurance through exclusive agents who repre-
sent only one insurer, and independent-agency insurers that distribute
their product through independent agents who represent multiple insur-
ers. These systems have long interested researchers because they have
coexisted in insurance markets for many decades, even though indepen-
dent-agency insurers are known to have higher costs (e.g., Joskow
1973; Cummins and VanDerhei 1979;
Pauly, Kunreuther, and Klein-
dorfer 1986; Kim, Mayers, and SmIth 1996). The purpose
of
this article
is
to
analyze the reasons for the long-term coexistence
of
the direct-
writing and independent-agency dIstribution systems.
Two primary hypotheses have been advanced to explain the coexis-
tence
of
the two systems. According to the market-impeifections hy-
pothesis, independent-agency insurers survIve while providing essen-
tially the same services
as
direct-wnting insurers because
of
market
imperfections, such
as
pnce
regulation (Joskow 1973; Cummins and
VanDerhei 1979; Weiss 1990), slow diffusion
of
information in insur-
ance markets (Berger, Kleindorfer, and Kunreuther 1989), or search costs
that permit inefficient firms to survive alongside efficient firms (Dahlby
and West 1986). Under the market-impeifections hypothesis, effiCIent
firms are expected to earn supernormal risk-adjusted profits, while mef-
ficient firms will earn risk-adjusted profits closer to normal levels.
In contrast, according to the product-quality hypothesis, the higher
costs
of
independent -agency insurers represent expenses associated
with producing higher product quality
or
greater service intensity, such
as providmg additional customer assistance with claims settlement, of-
fering a greater variety
of
product choices, or reducing policyholder
search costs
(Pauly et
aI.
1986; Kim et al. 1996; Regan and Tennyson
1996). This hypothesis predicts normal risk-adjusted profits for both
direct-writing and independent-agency firms.
The product-quality hypothesis Imphes that firms are sorted into
product-quality or service-intensity market niches,
WIth
customers who
prefer higher quality paying more for the product. The higher prices
receIved by the higher-quality providers cover their extra production
costs, allowing these firms to survive in equilibrium. This rationale
is broad enough to encompass agency-theoretic explanations for the
eXIstence
of
alternative technologies (e.g., Mayers and Smith 1981;
Kim et al. 1996). For example, principal-agent problems, such as
company/buyer incentive conflicts, may be more important
to
some
buyers
or
for some product vanants, leading to the survival
of
dIstribu-
tIOn
systems that deal efficiently with
thIS
type
of
incentive conflict.
Thus, independent-agency insurers may survive because they more ef-
Copyright © 2001. All Rights Reserved.
Multiple Distribution Systems
517
fectively discipline insurers into paying legitimate claims promptly and
fairly. Independent agents can credibly threaten to shift business to an
alternate insurer because their contracts with insurers convey owner-
ship
of
the policyholder list to the agent (i.e., the company cannot ap-
proach policyholders directly), whereas exclusive agents usually
do
not
have this ownership right.
Because product quality in insurance is essentially unobserved, re-
searchers have been unable to reach a consensus on whether the market-
imperfections hypothesis or the product-quality hypothesis
is
more
consistent with the observed cost data. This lack
of
consensus leaves
open the interesting economic question
of
whether the market works
well in minimizing product-distribution costs and leaves unresolved
the issue
of
whether marketing costs in property-liability insurance are
excessive and perhaps should receive regulatory attention.
I
This article proposes a new methodology for distinguishing between
the two hypotheses.
Usmg frontier efficiency methods, we estimate
both cost and profit efficiency for direct-writing and independent-
agency insurers. Measuring cost efficiency enables us to determine
whether the cost efficiency difference between direct-writing and inde-
pendent-agency insurers found by prior researchers persists under our
methodology. Measuring profit efficiency helps to identify unobserved
product-quality differences because customers should be willing to pay
extra for higher quality. Thus, our approach allows for the possibllity
that one group may provide higher-quality service on average and be
rewarded with higher average revenues that are reflected in profit effi-
ciency. That is, the profit-efficiency approach allows for the possibllity
that some firms may incur additional costs providing superior service
and be compensated for these costs through higher revenues.
A key statistic in our analysis will be the proportion
of
the difference
in measured cost efficlency between the firms employing the two distri-
bution systems that remains when we estimate profit efficiency.
If
most
of
the measured cost-efficiency difference remains as a profit-efficiency
difference, then the market-imperfections hypothesis would be sup-
ported. In this event, the profit efficiency, which includes both cost
efficiency and revenue efficiency, would reinforce the efficiency differ-
ence between the two groups. In contrast, if most
of
the measured cost-
efficiency difference is eliminated when the more encompassing profit
efficiency is measured, then the product-quality hypothesis would be
supported. This event would be consistent with the difference in service
quality being reflected in higher revenues.
By way
of
preview, we find data on 472 insurers over the period
I Regulators
In
several states, including CalifornIa, Flonda, and Massachusetts, have
argued that the hIgh costs
of
automobIle Insurance are partly attnbutable to Insurer ineffi-
cIency
In
marketing, admlntstratlOn, and claims settlement and that such
ineffiCIenCIes
should be
dISCIplined
through pnce regulatton
Copyright © 2001. All Rights Reserved.
518
Journal of Business
1981-90 to be fairly consistent with the product-quality hypothesis.
We measure independent-agency insurers as less cost efficient on aver-
age than direct writers, but most
ofthis
measured cost-efficiency differ-
ence does
not translate into a profit-efficiency difference. Indeed, after
conditioning on firm characteristics, such as size and business mix, the
profit-efficiency difference between the two groups
of
firms
is
quite
small and not statistically significant, even though a large, significant
cost-efficiency difference is still present.
The article is organized
as
follows. Section II summarizes some
of
the problems encountered in the extant empirical literature and dis-
cusses in an intuitive manner how our methodology addresses these
difficulties.
Section III gives the details
of
our methodology and model
specification.
Section IV discusses the measurement
of
inputs, outputs,
and prices in property-liabihty insurance, and
Section V describes the
data set.
Section VI presents
"simple"
tests based on the average effi-
ciency differences between direct-writing and independent-agency in-
surers, and
Section VII provides" sophisticated" tests, which condition
on other firm characteristics that may affect efficiency.
Section VIII
concludes.
II. Methodological Difficulties in the Extant Literature
Three major methodological problems have been encountered in the
literature on insurance distribution systems. First, product quality is
essentially unobserved.
If
some firms incur additional costs in provid-
ing a hIgher-quality product to consumers, such
as
extra assistance with
claims settlement or greater product variety, this may be incorrectly
identified as cost inefficiency unless proper controls for product quality
are used. Ex ante, we might expect better service from independent
agents because they can offer customers choices among the products
of
many insurance companies, perhaps better tailoring the insurance
product to the needs
of
the individual customer. In addition, indepen-
dent agents may be more likely to act
as
advocates for customers in
claims-settlement disagreements than exclusive agents since indepen-
dent agents are not tied to the individual insurer and can threaten to
steer business elsewhere
if
settlements are unsatIsfactory (see Kim et
al.
1996). Unfortunately, control variables for insurance product quality
are generally lacking in the data sets available to researchers.
2
In
this article, we estimate profit efficiency, which incorporates both
cost and revenue efficiency and should net out most
of
the unobserved
differences in product quality. In an efficiently functioning output mar-
2 Although
pnor
research has consistently shown mdependent-agency msurers to have
higher costs than direct wnters,
It
IS
not
obvIOUS
a
pnon
that
thiS
should be the case For
example, mdependent-agency msurers might benefit by shanng their agents' fixed costs
With
other msurers, Yleldmg lower costs than direct wnters However, any such gams may
be
diSSipated m praclice because
of
the difficulty of dealmg
With
multiple sets
of
fOflllS,
procedures, and computer systems by mdependent agents.
Copyright © 2001. All Rights Reserved.