Income measures, ownership, capacity ratios and the dividend decision of the non-life insurance industry : some empirical evidence

UNIVERSITY OF

IL; LIBRARY

AT Ul CAMPAIGN

3KSTACKS

Digitized by the Internet Archive

in 2011 with funding from

University of Illinois Urbana-Champaign

http://www.archive.org/details/incomemeasuresow699leec

Faculty Working Papers

College of Commerce and Business Administration

University of Illinois at U r b a n a - C h a m p a i g n

f

FACULTY WORKING PAPERS
College of Commerce and Business Administration
University of Illinois at Urbana -Champaign
August 14, 1980

INCOME MEASURES, OWNERSHIP, CAPACITY RATIOS AND THE
DIVIDEND DECISION OF THE NON-LIFE INSURANCE INDUSTRY:
SOME EMPIRICAL EVIDENCE

Cheng F. Lee, Professor, Department of Finance
Stephen W. Forbes, Life Office Management
Association

#699

Summary

Dividend payout ratio, dividend yield for non-life insurance industry
are studied in detail. A dividend behavior model is developed for the
non-life insurance industry. It is found that income measures, ownership
and capacity ratio are important factors to be considered in doing the
above-mentioned empirical studies.

Acknowledgment

The authors gratefully acknowledge the financial support of the S.S.
Heubner Foundation for Insurance Education of the University of Pennsylvania.
We also acknowledge the assistance of Mr. Dongsae Cho, and the editorial
comments of J. David Cummins.

The dividend policy of the firm is important for several reasons. An
understanding of the factors influencing dividend payments contributes to
the theory of corporate savings. Dividends may also influence the price
per share of common stock, thus dividend behavior is of interest because
it affects the maximization of shareholder wealth. In addition, dividend
policy also plays a direct role in the firm's financing and investment
decision.

While the factors influencing the dividend policies of industrial
firms have been studied in some detail by Lintner [12], Brittain [2],
Fama and Babiak [5], Dhrymes and Kurz [4], and others, theories of divi-
dend behavior have not been as extensively developed and explored for
financial firms. The purpose of this research is to study the dividend
behavior of one type of financial intermediary, the non-life insurance
company, to test whether existing dividend behavioral theories are ap-
plicable to the non-life insurance firms.

The argument is based upon the fact that the financial management
principles of financial institutions are not necessarily identical to
those of industrial firms. Specifically, the non-life insurance company
deals with (1) different income measures which affect reported earnings
and retained earnings, (2) is subject to a unique borrowing-lending rate
relationship, and (3) has an asset portfolio comprised primarily of se-
curities of industrial firms. In addition, most insurance stocks are
traded over the counter instead of NYSE. If the dividend practices of
these firms depart from those anticipated by the theories used to explain
the dividend behavior of industrial firms, this is of interest to the
understanding of the financial behavior of such firms. On the other hand,

-2-

dividend payment patterns which follow the theoretical anticipations will
tend to strengthen them.

Brittain said that econometric modeling is an exercise in persuasion
[3], The purpose of this study is not to persuade the reader as to the
validity of a particular financial theory involving dividend behavior.
Rather, it has the less ambitious but useful objective of ascertaining
whether non-life insurance companies follow widely accepted financial
models found to be successful in describing dividend behavior.

Certain problems are somewhat unique to the non-life insurance in-
dustry. They include the following: (1) There are several income mea-
sures that may be used as the basis for the dividend decision; statutory
or generally accepted accounting principle earnings; including unrealized
capital gains and losses or excluding them, and (2) the capital and sur-
plus position of the non-life insurer, as a measure of financial capacity,
may influence the dividend decision; there are also other matters par-
ticular to the study of insurance companies which include: (1) some
non-life insurers pay dividends to policyholders as well as shareholders,
(2) a widespread parent-subsidiary relationship found in the non-life
insurance industry, and (3) the total dividend freeze of 1971 and partial
freeze of 1972 in the United States. These factors may complicate any
study of non-life insurance company dividend behavior.

Two econometric models will be used in this study. First, the model
developed by Lintner [12] and Fama and Babiak [5] which defines current
dividends as a function of current after-tax profits and the preceding
year's dividends will be used to do the empirical study. The rationale
underlying this model is that the ability to pay, as measured by corporate

-3-

earnings, should have a great influence on dividend payments, and the
prior dividend should influence the current dividend to the extent that
dividend stability is viewed as desirable. Secondly, the seemingly un-
related regression technique [SUR] developed by Zellner [15] will be
used in order to take care of the possible simultaneous relationship
over time.

The next section of this paper presents the earnings payout ratios
and dividend yields for 61 non-life insurers during 1955-1975; the divi-
dend decision behavior for non-life insurers is also explored. In the
second section, the dividend behavior models used in research involving
industrial firms are specified and modified in order to describe the
dividend behavior of non-life insurance companies. The possible impli-
cations arising from the empirical results are justified in accordance
with the nature of insurance accounting procedures and financial opera-
tions. In the third section, the SUR technique is used in order to in-
vestigate the possible simultaneous relationship among the factors pre-
sented in the model described in the second section. Aggregate dividend
behavior is also examined in this section. The final section of the paper
summarizes the implications of the empirical results for financial theory
as it relates to the non-life insurance industry.

I. THE DIVIDEND YIELD AND EARNINGS PAYOUT
RATIO AND DIVIDEND DECISION BEHAVIOR

Information regarding the dividend yields and the earnings payout
ratios for non-life insurers is of interest to both investors and finan-
cial managers. Theoretically, the magnitude of the earnings payout ratio
for a firm is jointly determined by its investment opportunities and its
shareholders' preferences. This decision is complicated in the situation

-4-

of a non-life insurer by the payment of policyholder dividends. The
nature of policyholder dividends are not necessarily identical to those
of equity holder dividends. This is due to the fact policyholders are
not necessarily owners of non-life insurance companies.

If a firm has an investment opportunity with a return exceeding its
cost of capital, and the internal sources of funds are cheaper than the
external sources, then a financial manager will generally reduce his firm's
earnings payout ratio.

In relation to external financing, stock non-life insurers are limited
to the following alternatives: (1) mergers and acquisitions involving
other insurance companies, (2) new stock issues, (3) contributions from
a parent insurer or holding company, and/or (4) borrowing of funds for gen-
eral business purposes (the latter by Section 76, New York Insurance Code,
amendment effective September 1, 1969).

In the situation of a merger, one of the insurers loses its identity
and the surviving company absorbs all of its assets, liabilities, and
legal rights. The shareholders of the merged insurer usually retain a
financial interest in the new firm consistent with their interest in the
acquired firm. The acquisition may involve either the use of cash or a
tax-free exchange of shares.

A stock non-life insurer can also acquire a subsidiary insurer by
gaining ownership of more than 50 percent of its voting stock using either
cash or an exchange of securities. As a practical matter, Forbes [7]
found that non-life insurers usually use exchanges of stock in acquiring
subsidiaries because of the attractiveness of this approach from tax and
liquidity standpoints. Forbes also found that new stock issues are a

-5-

relatively unimportant form of non-life insurance external financing,
comprising slightly more than 5 percent of the total financing volume
during the 1955-66 period studied [7].

The advantages of external financing involving contributions from
a parent insurer or holding company are its simplicity and the lack of
substantial transactions costs. The borrowing of funds for general
business purposes is a relatively new external financing option in the
non-life insurance industry which has not been explored at length in
the financial literature except for Nye [13].

Forbes found in another study [8] that new money flowing into the
non-life insurance industry played a minor role in the industry's capital
and surplus growth during 1956-70. Given this behavior, the conservation
of capital and surplus would appear to be a primary non-life insurance
company objective in the typical situation. Inasmuch as dividend policy
provides one of the important mechanisms for controlling the level of
retained earnings, one would expect dividend policies of non-life insur-
ance companies to be geared to the insurer's capital and surplus require-
ments. Empirical results reported later in this study suggest a direct
relationship between dividend policies and capital and surplus adjust-
ments in this industry.

Haugen and Kroncke [10] have argued that policyholder funds also
represent a source of external financing to the non-life insurance in-
dustry. Other things remaining equal, an increase in the ratio of an
insurer's unearned premium and loss ar.d loss adjustment expense reserves
to its capital and surplus will affect the risk/return relationships in-
volving its shareholders. However, this interesting problem is not
studied in this paper.

-6-

Data associated with four different income measures and the divi-
dends for 61 firms (see Appendix A) during 1955-75 are used in order to
analyze the earnings payout behavior for the non-life insurance industry.
The stratified random sampling technique is used to select the sample
insurers. Three strata represent three alternative ownerships (see the
discussion in this section and Appendix A). The insurers were selected
at random from the population of firms in Best's Property Liability (for-
merly Best's Fire and Casualty) Insurance Reports having complete series
of 1955-1975 financial data. The stock price data were taken from the

Bank Quotation Record. The four different methods of calculating the net

2
income of a non-life insurer involve the following:

(A) earnings without the amortization of underwriting expenses
and without unrealized capital gains and losses

(B) earnings without the amortization of underwriting expenses
and with unrealized capital gains and losses

(C) earnings involving the amortization of underwriting ex-
penses without unrealized capital gains and losses

(D) earnings involving the amortization of underwriting ex-
penses with unrealized capital gains and losses

Under the accounting procedures used to measure (A) and (B), the
first year acquisition costs for insurance policies are written off imme-
diately against earnings without proper allocation to the periods in
which the associated premiums are earned. This method is required under
statutory accounting. It is also the method of accounting used in fed-
eral income tax calculations.

In the situation of an insurer with an expanding premium volume,
this accounting technique usually understates profits (overstates losses)
and understates capital and surplus. This is viewed by regulators as

-7-

desirable because the resulting excess valuation in the unearned premium
reserve (UPR) may provide additional surplus if the insurer encounters
financial difficulty (there will be no excess valuation of course if
all of the UPR is required for the payment of losses and loss adjustment
expenses). The lack of underwriting expense amortization under statutory
accounting also tends to make insurers more cautious in obtaining new
business since there are large surplus reductions if premium expansion
is too rapid.

Income measures (C) and (D) involve the proper amortization of
underwriting expenses. This is accomplished by adding to earnings the
after-tax prepaid expense involving the increase in the unearned premium
for the period. The increase in the UPR is multiplied by (l-x)E , where
t is the marginal federal income tax rate for the year and E equals the
ratio of underwriting expenses to net written premiums.

The other adjustment in the paper involves the inclusion of un-
realized capital gains and losses in income measures (B) and (D). This
"flow-through" approach to measuring earnings eliminates the potential
for earnings manipulation through the selective taking of realized capital
gains or losses (usually involving the taking of realized capital gains
in order to improve poor results). The primary argument against including
unrealized capital gains and losses in earnings is the "realization
principle". Under this principle, it is argued that only realized income
and loss items should be included in the income statement [1].

The American Institute of Certified Public Accountanzs has not
taken a clear position on the treatment of unrealized capital gains and
losses in insurance company earnings. Most insurers take such a gain or

-8-

loss as a direct credit or charge, respectively, to surplus rather than
a "flow-through" to earnings. The "flow- thro ugh" approach to earnings
measurement does not give an insurer an incentive to take realized capi-
tal gains in order to disguise bad performance. Thus, on an a priori
basis, one would expect earnings measures (B) and (D) to be more closely
associated with dividend decisions than measures (A) and (C), other
things remaining equal.

As a negative net income for an individual insurer in a particular
year is not unusual, a time aggregate earnings payout ratio is calculated
for each firm over the 21 year period. The resulting 61 payout ratios
for the 21 years are listed in Table 1.

During 1955-75, the ownership arrangement for these insurers can be
classified as (i) majority of common stock never held by a single entity
during the period, (ii) majority of common stock acquired by a single

entity sometime during the period, and (iii) majority of common stock

3
held by a single entity throughout the period (see Appendix A). One

would expect that the earnings payout policies of insurers having widely
held common stock under ownership arrangement (i) would most closely
follow (or fit) the models found successful in describing dividend be-
havior in other industries. This is because the shareholders of these
insurers have the same objectives as other investors in widely held
equities. On the other hand, the dividend policies of wholly owned sub-
sidiaries under ownership arrangement (iii) would be determined by the
managerial discretions of the parent insurers or holding companies.
Dividends may be declared in these situations in order to reduce per-
ceived excess surpluses in the subsidiaries. Sometimes subsidiaries are

-9-

acquired or holding companies are formed for the express purpose of trans-
ferring surplus from relatively unprofitable insurance operations to other
activities. The dividend policies of the insurers in group (ii) would be
expected to vary depending upon the individual circumstances surrounding
each acquisition.

The analysis of variance technique is used to test whether significant
differences arise among the earnings payout ratios for these three groups.
It is found that the average earnings payout ratios are significantly dif-
ferent among the three groups with a 5 percent level of significance if
the (A) or the (B) net income definition is used. Furthermore when either
the (A) or (B) income definition is used, the average earnings payout ratio
is higher than 50 percent. This figure is close to the earnings payout
ratio of the electric utility industry as indicated in Lee [11]. Empirical
studies related to this issue and the possible implications of a high earn-
ings payout ratio on the cost of equity capital will be done in separate
research.

As only a portion of the firms listed in Appendix A had actively
traded common stock during 1955-75, the dividend yield results are based
upon a subset of these insurers. The annual average shareholder dividend
yields calculated for these insurers are listed in Table 2. The analysis
of variance technique is used to test whether the dividend yield for the
non-life insurance companies changed over the 25 year period studied.

It is found from Table 2 that the shareholder dividend yields among
the years are significantly different at a one percent level of signifi-
cance. The average dividend yield for the 21 years studied is 3.38 per-
cent. This fluctuation is related to business cycles and economic

-10-

conditions. The average dividend yields for 1974 and 1975 are 5.57 and
6.03 percent, respectively. These figures are as high as the time de-
posit interest rates in these respective years. This information sug-
gests that non-life insurance companies' stock dividend yields are
similar to those found for low risk investments.

The results of this section give both investors and decision makers
some cross-sectional and time series information about the earnings pay-
out ratios and dividend yields of non-life insurers.

II. DIVIDEND FORECASTING MODEL FOR THE
NON-LIFE INSURANCE INDUSTRY

Lintner [12] has used the partial adjustment assumption in order to

derive a dividend forecasting model for an industrial firm. This model

takes the following form:

AD, = y(D * - D„ .) (1)

where,

and,

t ' v t t-1'

Dt* = 6 Et (2)

D = actual total cash dividend payment in period t

*

D = desired total cash dividend payment in period t

D .. = actual total dividend payment in period t-1
E = total earnings in period t

Y = partial adjustment coefficient
6 = target earnings payout ratio

-11-

Af ter substituting (2) into (1) , we can formulate the following al-
ternative time series regression models in order to describe an individual
firm's dividend behavior over time:

Dt = Ao + *l"t + Vt-1 + Elt (3)

Dt = BlEt + B2Dt-l + £2t (4)

where A. = B- = 3y and A„ = B„ = 1-y; A is the intercept; and both

e.. and e„ are disturbance terms for the regressions. To accommodate

the two special circumstances we have encountered in this study (full and

partial dividend freezes and capital and surplus capacity considerations),

equation (3) is modified as

CS
Dt = ao + alXt + a2Et + a3Dt-l + a4 AlT + £t (5)

where,

= 0 for 1955-70

Xt

=1 for 1971-75

as dummy variables

and, CS

77— = capacity ratio

t

where,

CS = capital and surplus at the end of the period =

capital stock, plus paid in surplus, plus retained
earnings

AA = admitted assets at the end of the period

The change in a non-life insurer's capital and surplus is explained

by the following:

ACS =I+U-D+F+M

-12-

where,

I = net income or loss after taxes [the sum of the statutory
underwriting gain (loss) , net realized capital gain (loss) ,
interest, dividends, and rents, reduced by net loss from
agents' premium balances charged off, and adjusted for the
federal and foreign income tax liability (rebate)]

U = unrealized capital gain (loss)

D = dividends declared to shareholders and/or policyholders

F = external financing

M = miscellaneous adjustments (the sum of the change in the
excess of bodily injury liability and compensation statu-
tory and voluntary reserve over the case basis and loss
expense reserve, change in nonadmitted assets, change in
liability for unauthorized reinsurance, change in foreign
exchange adjustment, and net remittances to or from the
home office). Nonadmitted assets include furniture and
office equipment, unpaid balances over 90 days late, and
other items considered to be lacking in liquidity under
statutory accounting.

From the above model, it can be seen that an insurer needs to
retain capital and surplus in order to absorb (1) net losses from
operations (defined by I), (2) unrealized capital losses, and (3) mis-
cellaneous adjustments (defined by M) . Generally, items (1) and (2)
will cause the greatest surplus fluctuations in a given accounting
period. Net losses from operations may be compounded by the upward
adjustment of inadequate loss and loss adjustment expense reserves
arising from claims incurred in prior years. These adjustments can
be especially large during periods of rapid inflation.

The model indicates that the adjustment of dividends is one of
the most realistic alternatives in attempting to conserve capital and
surplus in the typical situation. This is because the raising of exter-
nal financing through new equity issues involves a transactions cost,

-13-

uncertain proceeds (especially in volatile stock markets) , and is time
consuming as well as troublesome to management. Empirical data indicate
that new equity issues have not been an important form of external
financing in the non-life insurance industry [7]. External financing
involving mergers and acquisitions is not motivated by the need to con-
serve surplus as it does not affect the CS /AA ratio in most situations.
Contributions of surplus from parent insurers or holding companies, or
the borrowing of funds, have not been common external financing practices.
Thus, dividend policy is the only decision variable left.

It is also relevant to note the CS /AA ratios fluctuate widely
from period to period for a given insurer because of the impact of un-
realized capital gains and losses involving common stock portfolios and
fluctuating underwriting results (see Forbes [8]). Thus a non-life
insurer does not have time to consider external financing as a means of
stabilizing capital and surplus in the usual situation. The adjustment
of dividends is a more direct and immediate method of correcting capital
and surplus deficiencies. As an alternative hypothesis, it might be
argued that rapidly changing capital and surplus positions would make a
non-life insurer more cautious in its earnings payout policies. These
issues are explored by including the CS /AA ratio in equation (5).

In Table 2, we have calculated the average capacity ratios for 78 in-
surers for each of the years studied. Observation of the Table will in-
dicate that the average capacity ratio for the non-life insurance industry
fluctuates over time. This fluctuation may result from changes in the
value of insurance company equity portfolios and/or variations in under-
writing results. In addition, the coefficients of variation associated

-14-

with the average capacity ratios in Table 3a suggest wide variations
in capacity ratios among the insurers for each year.

Table 3a presents a frequency distribution of the temporal coeffi-
cients of variation for the capacity ratios of the 78 insurers. It should
be noted that 73 percent of the coefficients in the Table fall within a
range of .1 to .3.

Table 3b shows the percentage distribution of the average 1955-75
capacity ratios for the 78 insurers. Approximately 64 percent of the
insurers had average capacity ratios within the range of .3 to .5.

The empirical results based upon this specification for 61 non-life
insurers during 1955-75 are reported in Table 4. From the t-values asso-
ciated with the regression coefficients of the dummy variable (a.. ) , it is
found that 15 out of 61 firms appeared to change their dividend payment
behavior because of the dividend freeze. Similarly, from the t-values
associated with the regression coefficients of the capacity variable (a,),
only 17 of the 61 firms had an a, coefficient significantly different
from zero. This implies that most of the insurers' dividend decisions were
not affected by a change in the capacity variable. This may be due to the
fact that a change in retained earnings is only one of two alternatives
for adjusting the capacity ratio. In general, a non-life insurance company
can also issue new equity in order to raise its capacity ratio. It should
also be noted that four alternative earnings definitions are used to fit
equation (5) . These empirical results in terms of income measure (A) are
reported in Table 4. The overall results are relatively independent of
the different income definitions used.

-15-

One of the main purposes of equation (5) is to forecast a firm's

future dividend payment. The adjusted coefficient of determination

—2
(R ) provides an indication of a regression equation's statistical

—2
fit to historical data. Based upon Table 4, it is found that the R

—2
ranges from .0209 to .9845. The frequency distribution of R for these

61 firms indicated that more than 90 percent of these firms' dividend

—2
behaviors can be described by equation (5). Furthermore, the R are

classified according to the ownership arrangements (i) , (ii) , and (iii)

defined above. From the analysis of variance results indicated in Table

—2
5 it is found that significant differences exist for the R among these

three groups. The implication arising from these results is that the

ownership arrangement has an impact upon the dividend payment behavior

of a non-life insurer. This follows the expectations discussed earlier

—2
in the paper. Other possible explanations for the low R include the

presence of negative earnings in some of the years studied for a partic-
ular insurer and policyowner dividends. The percentages of policyowner
dividends to total dividends are listed in Table 6.

The estimated partial adjustment coefficient y and the percentage of
optimal dividend related to current earnings are of interest to both
investors and financial managers in the non-life insurance industry.
Based upon Table 4, the average y is .53 and average (B is .268. These
imply that the partial adjustment coefficient is .53 and the average
target payout ratio is 26.8 percent. The estimated y of .53 also im-
plies that it takes non-life insurance firms an average of about 2 years
to adjust their dividend payments to desirable levels.

-16-

Fama and Babiak [6] have shown that Lintner's [12] model without the

constant term has the greatest forecasting power. Therefore, equation (5)

without a and a. is also calculated for the 61 firms. It is found that
o 4

—2
the R associated with Lintner's model without the constant term are gen-
erally lower than with the constant term. This is due to the fact that the
capacity ratio is important for 17 firms.

III. SUR AND AGGREGATE BEHAVIOR FOR SHAREHOLDER DIVIDENDS
In the final section of the paper, a pooled time series and cross
sectional simultaneous equation model will be constructed using the
following extensions of equation (3), above:

Dti " *i + W + cl(D(t-l)l>

Dt2 = A2 + VEt2> + C2(D(t-l)2) (6)

D_ = A + B (E„ ) + C (D. .. )
tn n n tn n (t-l)n

where the equations are generated for each year and the subscripts l...n
refer to each of the insurers studied.

Zellner's [15] seemingly unrelated regression method is now used
to simultaneously estimate these equation systems. The strength of this
method rests in its ability to consider the effects of both time and firm
behavior upon dividend policy.

We would anticipate that dividend behavior would vary by the
ownership arrangement of the insurer. Based upon the ownership arrange-
ment the OLS residuals associated with the shareholder dividend behavior

Total Insurers

.73

Group i

.92

Group ii

.69

Group iii

.62

-17-

equation are used to estimate three variance-covariance matrices and
three correlation coefficient matrices. It is found that the relation-
ships among OLS residuals within each group are relatively strong. This
implies that Zellner's SUR method can be used to improve the efficiency

of the estimated shareholder dividend behavior relationship.

-2
The R values under the regressions without policyholder dividends

-2
are generally lower than the R values for the total dividends presented

in the third section of the paper. This is demonstrated by the following:

-2 -2

R for Total Dividends R for Shareholder Dividends

.64
.89
.50
.60

These results imply that the earnings payout decision is made on
the basis of total policyholder and shareholder dividends rather than
considering these dividends separately. It is also found that SUR pro-
vides a more accurate estimation of the regression coefficients than
the OLS method.

Overall, we found that the prior dividend was the most important
variable explaining the level of current shareholder dividends under the
SUR technique. This variable was significant at the 5 percent level for
100, 68 and 63 percent of the group i, ii, and iii insurers respectively.
Next, the capacity ratio was found to be important in explaining share-
holder dividends for 53, 44, and 44 percent of the group i, ii, and iii
insurers, respectively. Similar percentages for current earnings were
59, 32, and 50 percent respectively.

-18-

The aggregated results for the equation (5) regression without the
dummy variable presented differ significantly from the disaggregated
results. This may be due to aggregation bias and/or the equal weighting
procedure in the individual company case. Also the correlation coeffi-
cient between the lagged dividend and the capacity ratio approximated a
negative .8.

IV. SUMMARY AND CONCLUDING REMARKS

We have examined the shareholder and policyholder dividend policies
of a large sample of non-life insurers over 1955-75 in terms of some
widely accepted financial models. These models were applied using four
definitions of income and three insurer ownership groups. Adjustments
were also made for capacity ratios and the partial and complete dividend
freezes in the early 1970 's.

We found based upon a count of the significant t-values for the a„
coefficient in Table 4a for each earnings measure, that unrealized cap-
ital gains and losses were viewed as a transitory non-life insurance in-
come component in the earnings payout decision. This probably follows
from the widely accepted accounting practice of treating unrealized re-
sults as a surplus adjustment rather than an income component. In ad-
dition, we found that the total average earnings payout ratio was higher
than 50 percent for statutory income definitions. The average share-
holder dividend yield for all of insurers for 1955-75 was 3.38 percent.
However, this yield fluctuated widely over time. This was the result
of non-life insurance common stock prices tending to move in concert
with the overall market during this period.

-19-

Lintner's dividend forecasting model was modified and applied to
the non-life insurance company sample. Regressions were run for both
total policyholder and shareholder dividends and the shareholder divi-
dends alone. On an overall basis, the ranking of the explanatory power
of the regressors was (1) the prior year's dividend, (2) the current

earnings, and (3) the capacity ratio. We also found a significant dif-

-2
ference in the R among the three ownership groups. It was also

determined that it takes approximately two years for non-life insurers

to adjust their total earnings payouts to desired target levels using

Lintner's formulation. The dummy variable for the partial and total

dividend freeze was also tested and found not to be important for the

great majority of insurers.

The cross-sectional and temporal average capacity ratios and
their coefficients of variation were calculated for the sample non-life
insurers. Over 50 percent of the insurers had average 1955-75 capacity
ratios ranging from .3 to .5. Wide variations in the temporal coeffi-
cients of variation in the capacity ratios of the individual insurers
were found. This resulted from different compositions of underwriting
and investment portfolios and varying premiums written/ capital and
surplus ratios among the insurers.

The SUR is superior to the OLS regression method if the OLS resi-
duals among the firms within the group are correlated. We found that
the SUR technique was superior to the OLS in estimating the dividend
determination behavior relationships for each of the three ownership
groups. This implies that there was some behavioral similarity in each
of these groups. The SUR findings were otherwise consistent with the
other findings in the paper.

-20-

The capacity ratio dominated the aggregate dividend determination
model. This result was at variance with the other results reported in
the paper. This may be due to aggregation bias.

The purpose of this research is to show how the finance theory and
technique used in the industrial firms can be used to do dividend deci-
sion of non-life insurance industry. The complication of dividend de-
cision for the non-life insurance industry relative to that of indus-
trial firms is the interaction relationships among different income
measure, ownership and capacity ratio and alternative definition of
dividend decision as is dictated in Figure 1. From the theoretical
analysis and empirical investigation of this paper it is found that
dividend decision rules used in the industrial firms can generally be
used to help the dividend decision for non-life financial managers.
However, the unique nature of the definitions and interaction relation-
ship of non-life insurance industries (see Figure 1). Some modifica-
tion of the dividend behavior decision model for industrial firms may
well be beneficial from the viewpoints of corporate finance theory
and practices.

-21-

Footnotes

Other possible mechanisms for controlling the level of retained
earnings are: (1) decrease premiums written, and (2) reinsurance for
all or portion of the existing portfolio.

2
Foster [6] has investigated the impacts of these four alterna-
tive earnings on the market value of property-liability companies. One
of the referees has argued that the use of statutory earnings as a
dividend decision factor is unrealistic since management would recog-
nize the need to adjust earnings before such a decision was made.
However, alternative earning measures have different implications on
the earnings power of a non-life insurance firm. Therefore, different
income measures should lead to a difference in valuation approach for
non-life insurance companies as discussed by Foster [6].

3
It would be interesting to see the effect, if any, upon dividend

policy of a change in the common stock ownership of the firm. This will

be a subject for future research.

4
The "follow" or "fit" implies that the power of forecasting the div-
idends payment over time. The specific formulation and test of these
arguments are the adjusted coefficient of determination as defined in
the following section.

The measurement of the capacity ratio used in this paper does not
take into account the portfolio characteristics of the sample firm's
assets and liabilities. See Stone [15] for detail. The authors are
grateful to one of the referees for supplying these helpful comments.

M/E/66

-22-

REFERENCES

1 . AICPA Insurance Auditing Tasks Force Discussion Memorandum, Accounting
for Property and Liability Insurance Companies (New York: American
Institute for Certified Public Accountants, November 26, 1975).

2. Brittain, J. A. "The Tax Structure and Corporate Dividend Policy,"
American Economic Review, (May, 1964), pp. 272-87.

3. Brittain, J. A. Corporate Dividend Policy, (Washington: The Brookings
Institution, 1966).

4. Dhrymes, P. and M. Kurz. "Investment, Dividends, and External Finance
Behavior of Firms," in R. Ferber, ed. , Determinants of Investment Be-
havior, New York, 1967.

5. Fama, E. F. and H. Babiak. "Dividend Policy: An Empirical Analysis,"
J. American Statist. Assn., Dec. 1968, pp. 63, 1132-61.

6. Foster, F. "Valuation Parameters of Properties-Liability Companies,"
The Journal of Finance (June 1977), pp. 823-35.

7. Forbes, S. W. "Growth Performances of Non-Life Insurance Companies,"
The Journal of Risk and Insurance (September, 1970), pp. 341-60.

8. , "Profitability in the Non-Life Insurance Industry: 1955-

74" CPCU Annals (June, 1977), pp. 126-34.

9. Gordon, M. J. "Dividends, Earnings, and Stock Prices," Review of
Economics and Statistics, (1959), pp. 99-105.

10. Haugen, R. A. and Kroncke, C. 0. "Rate Regulation and the Cost of
Capital in the Insurance Industry" Journal of Financial and Quanti-
tative Analysis (December, 1971), pp. 1283-1305.

11. Lee, Cheng F. "Functional Form and the Dividend Effect in the Electric
Utility Industry," Journal of Finance, (December, 1976), pp. 1481-86.

12. Lintner, J. "Distribution of Incomes Among Dividends, Retained Earn-
ings and Taxes," American Economic Review, (May, 1956), pp. 97-113.

13. Nye, D. J. A Simulation Analysis of Capital Structure in a Property
Insurance Firm (Homewood, 111.: Richard D. Irwin, Inc., 1975).

14. Stone, J. M. "A Theory of Capacity," (part I, part II), The Journal of
Risk and Insurance (June, September, 1973), pp. 231-243, pp. 339-855.

15. Zellner, A. "An Efficient Method of Estimating Seemingly Unrelated
Regressions and Tests for Aggregation Bias," J. of American Statist.
Assn., 1962, pp. 57, 348-68.

-23-

TABLE 1

AGGREGATE EARNINGS PAYOUT RATIOS

Insurer Group (i)

Insurer Group (ii)

Insurer Group (iii)

CO.

Elb

E2C

E3d

E4e

CO.

El

E2

E3

E4

CO.

El

E2

E3

E4

11

.83

.96

.72

.81

01

.89

.97

.85

.92

02

.55

.65

.51

.60

14

.67

.63

.64

.60

03

.53

.56

.47

.49

05

.14

.08

.11

.07

17

.82

.91

.79

.88

04

.74

.76

.68

.70

06

.83

.72

.78

.68

24

.75

1.61

.63

1.13

07

.62

.63

.57

.58

08

.83

1.14

.80

1.10

30

.21

.28

.16

.19

09

.58

.40

.51

.37

13

.67

.74

.64

.69

33

.71

.84

.58

.66

10

.53

.75

.53

.74

15

.66

.55

.49

.43

36

.99

1.22

.83

.99

12

.93

1.17

.78

.95

16

.36

.33

.35

.31

38

.73

.79

.57

.61

18

.96

1.05

.83

.89

19

.23

.35

.23

.33

40

-.00

-.00

-.00

-.00

23

1.13

1.15

1.12

1.14

20

.64

.55

.60

.51

44

.59

.85

.53

.72

25

.86

.80

.69

.65

21

.67

.88

.70

.93

45

.15

.15

.13

.13

31

.33

.35

.31

.33

22

.06

.06

.04

.04

46

.45

.28

.47

.29

34

.61

.66

.60

.65

26

.58

.62

.55

.58

48

.52

.42

.49

.40

37

.86

.82

.82

.79

27

.84

.66

.72

.59

56

.73

.65

.50

.46

39

.21

.20

.20

.19

28

-1.21

-.92

-2.14

-1.37

58

.37

.37

.31

.30

41

.32

.34

.27

.28

29

.55

.38

.47

.34

60

.77

.67

.68

.60

42

.97

1.26

.96

1.24

32

.11

.12

.10

.10

61

.49

.29

.44

.28

43

.37

.76

.80

.71

35

.77

.71

.71

.66

50

.50

.42

.45

.38

47

.33

.32

.27

.26

52

.80

.86

.80

.86

49

.35

.32

.30

.28

53

.92

.80

.85

.75

51

.33

.30

.30

.28

54

.65

.76

.62

.71

55

.55

.50

.45

.42

57

.30

.30

.27

.27

59

.24

.30

.22

.26

Average .61

.68

.53

.57

,66 .69 .61

.64

.41 .43 .33

,37

a = ownership groups (i), (ii), and (iii) are defined in the text.
b = total 1955-75 dividends •=■ total 1955-75 earnings as measured by definition
c = total 1955-75 dividends * total 1955-75 earnings as measured by definition
d = total 1955-75 dividends $ total 1955-75 earnings as measured by definition
e = total 1955-75 dividends t total 1955-75 earnings as measured by definition

(A)

in

text.

(B)

in

text.

(C)

in

text.

(D)

in

text.

-24-

TABLE 2
NON-LIFE INSURER DIVIDEND YIELDS AND CAPACITY RATIOS

Annual Average

Capacity

Sample

Year

Dividend Yield (%)

Mean

Ratio

C.V.

Size

1955

2.90%

.4249

.4552

22

1956

3.37

.4040

.3988

23

1957

3.63

.3760

.4282

25

1958

3.38

.3960

.3874

26

1959

3.04

.4192

.6171

28

1960

3.58

.3903

.3941

28

1961

2.79

.4202

.3796

29

1962

2.70

.4010

.3850

28

1963

2.36

.4064

.3889

29

1964

2.48

.4093

.3997

30

1965

3.14

.4007

.3894

30

1966

3.14

.3742

.3979

28

1967

3.43

.3844

.3858

28

1968

3.30

.3872

.3597

26

1969

4.40

.3595

.4412

19

19 70

4.27

.3571

.4439

18

1971

2.90

.3766

.4033

19

19 72

2.72

.3960

.4006

21

1973

3.68

.3472

.4502

20

1974

5.57

.2794

.5845

21

1975

6.03

.2799

.5858

20

Overall

Average

3.38%

F-test results for average dividend yields among years 1955-75,
F = 5.3006 (significant at 1 percent level).

-25-

Table 3a

Frequency Distribution of 1955 - 1975 Coefficient Variation
for Average Capacity Ratio

AT

AND LESS

LEAST

THAN

0

- 0.1

0.1

- 0.2

0.2

- 0.3

0.3

- 0.4

0.4

- 0.5

Number of Firms

Percentage

8

10.25

25

32.05

32

41.03

9

11.54

3

3.85

1.4 - 1.5 1 1.28

78 100%

-26-

Table 3b
Frequency Distribution of Average 1955 - 1975 Capacity Ratio

AT

AND LESS

LEAST

THAN

0.1

-

0.2

0.2

-

0.3

0.3

-

0.4

0.4

-

0.5

0.5

-

0.6

0.6

-

0.7

0.7

-

0.8

0.8

-

0.9

Number of Firms

Percentage

2

2.56%

15

19.23%

36

46.15%

14

17.95%

4

5.13%

4

5.13%

2

2.57%

1

1.28%

78

100%

-27-

TABLE 4

EMPIRICAL RESULTS FOR EQUATION (7)-
INCCME MEASUREMENT (A)

Co

. No.a

a

a,

a„

ao

a/

R2

S.E.b

0

1

2

3

4

(01)

28856

48347

1.0732

-.1909

-1477.9

.8705

22487

(1.03)c

(2.73)

(7.99)

(-1.64)

(-2.32)

(02)

-2 088.7

5311.6

.0615

1.0043

-88.1867

.9572

2179.9

(-.41)

(1.48)

(.84)

(9.13)

(-.63)

(03)

-33600

-22349

.444

.4477

1615.8

.4771

18474

(-.87)

(-1.79)

(2.02)

(1.55)

(1.41)

(04)

41880

10161

.377

-.3301

-913.30

.8093

4100.4

(3.51)

(1.99)

(3.68)

(-1.12)

(-3.18)

(05)

111.92

21.052

-.0146

.5626

1.9058

.6159

103.67

(.36)

(.26)

(-.97)

(3.37)

(.19)

(06)

2224.4

89.114

-.012

.6742

-42.8317

.7374

499.14

(2.07)

(.17)

(-.11)

(2.83)

(-1.28)

(07)

4606.9

-2050.4

.1777

-.1215

-26.82

.9107

474.73

(4.69)

(-5.05)

(4.04)

(-.83)

(-1.51)

(08)

-7201.8

957.27

-1.2438

.2049

224.87

.4194

2474.7

(-2.23)

(.57)

(-2.26)

(.97)

(2.56)

(09)

3209.8

1464.7

.1395

.9153

-144.21

.7183

1406.4

(.48)

(.66)

(1.32)

(2.29)

(-.79)

(10)

-271.55

116.32

.0601

.2545

14.329

.5125

135.36

(-.79)

(1.58)

(1.57)

(1.48)

(1.33)

(11)

3651.3

-149.76

.1194

.7052

-103.72

.9865

632.16

(1.31)

(-.22)

(13.95)

(5.71)

(-1.39)

(12)

483.36

-342.8

-.0194

.4561

6.182

.5171

232.98

(.73)

(-2.53)

(-.47)

(1.61)

(.31)

(13)

371.99

-360.08

.1098

.7322

2.5613

.8936

294.38

(.48)

(-1.23)

(1.97)

(3.98)

(.13)

(14)

-1146

-464.32

-.0279

.706

93.125

.9771

190.12

(-1.28)

(-2.1)

(-.4)

(8.87)

(2.66)

(15)

2595.05

-1705.75

-.1866

-.2681

-1.2436

.3308

1137.2

(1.29)

(-2.61)

(-1.09)

(-.62)

(-.02)

(16)

266.56

-11.34

.0031

.3117

.1542

.5612

29.8

(3.76)

(-.56)

(.42)

(4.10)

(.10)

(17)

-6628

3664.6

.0089

1.1095

54.7184

.9825

1572.4

(-3.22)

(2.28)

(.35)

(16.55)

(1.89)

(18)

25089

7552.2

.4611

.1089

-633.14

.4185

9842.5

(2.22)

(.92)

(2.27)

(.5)

(-1.88)

(19)

-234.61

511.6

.0602

.3044

-.0705

.1690

737.46

(-.09)

(.97)

(.27)

(1.25)

(-.00)

(20)

632.62

278.37

.0731

.5974

-21.454

.8193

204.76

(2)

(2.31)

(2.1)

(4.42)

(-2.75)

(21)

-474.03

-83.623

-.0611

-.2239

21.8

.3824

336.77

(-1.28)

(-.44)

(-.45)

(-.85)

(2.64)

(22)

-158.59

820.79

.0563

-.1037

-13.14

.75

294.31

(-.18)

(1.68)

(3.29)

(-.52)

(-.59)

-28-
TABLE 4 (con't.)

Co. No.

a

0

al

a2

a3

a4

R2

S.E.

(23)

236.36

-132.37

.2276

.6218

-.5987

.7821

174.42

(.94)

(-1.04)

(3.19)

(4.45)

(-.04)

(24)

76.332

-30.198

.0146

.7806

14.1488

.7048

386.83

(.12)

(-.12)

(.64)

(4.84)

(.75)

(25)

133.92

12.326

.0223

.3131

-1.3485

.75

12.874

(5.27)

(1.09)

(2.98)

(3.00)

(-2.73)

(26)

1011.7

12.994

.0561

.1079

-21.728

.3014

101.96

(2.98)

(.21)

(.76)

(.57)

(-2.05)

(27)

5649.4

946.63

.4176

.0973

-148.02

.7476

1004.5

(2.9)

(.90)

(2.68)

(.33)

(-2.7)

(28)

8211.5

-3162.3

.2192

1.1228

-119.34

.8404

2004.8

(4.58)

(-2.72)

(3.29)

(5.77)

(-3.03)

(29)

-1563.5

-290.22

.2386

-.2357

54.328

.3096

219.41

(-2.36)

(-1.41)

(.78)

(-1.11)

(2.41)

(30)

22.086

-5.1426

.0515

.8264

-.0806

.9866

42.299

(.37)

(-.10)

(1.54)

(5.42)

(-.03)

(31)

-12.87

-32.086

.13279

.3581

3.5981

.6896

141.45

(-.04)

(-.19)

(1.86)

(1.83)

(.42)

(32)

178.68

-113.31

.0212

.7986

-1.1135

.9422

92.144

(1.01)

(-.89)

(.93)

(5.38)

(-.73)

(33)

129.26

-16.705

.0661

.5398

-1.6399

.9359

18.196

(1.66)

(-.98)

(3.73)

(2.66)

(-1.11)

(34)

135.03

-193.11

.1684

.5969

6.5116

.9463

189.98

(.45)

(-.92)

(1.16)

(1.78)

(.56)

(35)

2240.8

3069.2

.8689

.2253

-161.74

.3504

6055.8

(.15)

(.48)

(1.30)

(.80)

(-.6)

(36)

59.841

5.7237

.0182

.7153

-.3979

.8378

14.71

(1.26)

(.48)

(1.44)

(3.44)

(-1.04)

(37)

-20111

8985.1

1.1034

.4233

182.74

.6978

7045.5

(-1.3)

(1.59)

(4.08)

(2.32)

(.57)

(38)

-7459.8

1379.9

-.0318

1.6049

195.75

.881

715.82

(-1.93)

(1.69)

(-.67)

(4.49)

(1.75)

(39)

799.85

-739.81

.0031

.0157

19.998

.9635

194.45

(6.78)

(-6.59)

(.81)

(.32)

(19.84)

(40)

7.3184

-42.759

0

.8356

1.7461

.7457

44.817

(.12)

(-1.72)

(-.61)

(5.36)

(.92)

(41)

1016.4

-908.3

.3012

-.5897

-2.7769

.8896

268.97

(1.48)

(-1.62)

(3.08)

(-1.18)

(-.43)

(42)

317045

122943

.3593

-.4354

-5716

.4084

109098

(2.16)

(1.10)

(.68)

(-1.61)

(-2.00)

(43)

-989.77

351.41

2.1274

.174

-91.979

.8914

2095.4

(-.18)

(.21)

(9.03)

(1.35)

(-1.1)

(44)

4.7857

2.1733

.0333

.8963

-.0282

.9353

4.3958

(.48)

(.33)

(1.69)

(10.1)

(-.15)

(45)

823.23

-125.00

.0017

.7098

-8.739

.9362

76.1949

(2.02)

(-1.57)

(.54)

(4.98)

(-1.25)

(46)

27.2701

6.0099

.07781

.8345

-.3367

.9463

11.9968

(.32)

(.65)

(2.33)

(8.11)

(-.3)

-29-

TABLE 4 (con't.)

Co. No.

a

0

al

a2

a4

a4

I2

S.E.

(47)

637.75

-397.18

.0201

-.4872

-5.936

.3842

205.65

(2.4)

(-2.09)

(.21)

(-1.20)

(-1.00)

(48)

-171.78

198.29

.0038

1.1509

-1.8817

.984

126.47

(-.54)

(1.37)

(.22)

(15.86)

(-.27)

(49)

4.7876

3.1862

.0242

.9569

.0307

.9821

17.961

(.12)

(.16)

(3.99)

(12.35)

(.04)

(50)

-53.344

-3 7.445

.0229

.6896

4.4049

.542

150.27

(-.22)

(-.26)

(.66)

(2.21)

(.68)

(51)

-32.542

5.29

.0198

1.0743

.6195

.9766

25.401

(-.57)

(.15)

(2.02)

(10.62)

(.72)

(52)

3676

-1260.8

.5737

.089

-71.872

.82 66

1026.1

(2.67)

(-1.47)

(3.41)

(.48)

(-1.34)

(53)

50783

-14548

-.4155

-.3623

-543.13

.5344

9389.4

(4.34)

(-1.88)

(-1.51)

(-1.77)

(-2.66)

(54)

-744.65

-757.48

-.0171

-.0399

56.1391

.0209

2652.4

(-.2)

(-.31)

(-.07)

(-.16)

(.53)

(55)

855.25

-146.2

.1232

.6583

-13.2631

.9582

265.47

(1.3)

(-.45)

(5.00)

(4.74)

(-1.24)

(56)

300.33

-40.63

.0843

1.0597

-7.668

.9382

67.948

(1.72)

(-.47)

(1.62)

(5.97)

(-2.09)

(57)

-126.11

-153.88

-.0592

.7387

17.3201

.5249

192.4

(-.26)

(-.92)

(-1.83)

(3.34)

(1.44)

(58)

6.1672

-37.944

.0402

.8567

1.3717

.8506

40.326

(.07)

(-.8)

(.81)

(4.55)

(.67)

(59)

93.2801

8.5132

.024

.8533

-1.0067

.912

61.563

(.73)

(.15)

(.66)

(4.69)

(-.54)

(60)

-9381.5

9615.2

.093

1.1848

-42.876

.988

2029.8

(-2.74)

(3.77)

(1.87)

(17.65)

(-.45)

(61)

-356.07

-202.36

-.0227

1.1427

16.443

.9845

119.6

(-.75)

(-1.37)

(-1.13)

(17.95)

(1.37)

See Appendix A for company names.

Standard error of the estimate.

"All values in parentheses are t-values.

-30-

TABLE 5

-2

F-TEST FOR SIGNIFICANT DIFFERENCE IN R
AMONG OWNERSHIP GROUPS

Ownership Group
(i)a
(ii)
(iii)

Overall
Average

—2

Average R

.917705
.690487
.623534

.731858

F Value

9.0361

See text for the definition.

Significant at .038% level.

-31-

TABLE 6
AVERAGE 1955-75 RATIOS OF POLICYOWNER DIVIDENDS TO TOTAL DIVIDENDS

INA

HARTFORD ACCIDENT AND INDEMNITY

AETNA CASUALTY AND SURETY

FEDERAL

AMERICAN STATES

ROYAL INDEMNITY

WESTCHESTER FIRE

PBOVIDENCE WASHINGTON

GOVERNMENT EMPLOYEES

PEERLESS

EMPLOYERS FIRE

EMPLOYERS CASUALTY

UNITED PACIFIC

AMERICAN GENERAL

RELIANCE

AFFILIATED FM

CONNECTICUT INDEMNITY

STATE FARM FIRE AND CASUALTY

CALIFORNIA COMPENSATION AND FIRE

HANOVER

AMERICAN POLICYHOLDERS

GLOBE INDEMNITY

PACIFIC

TRI STATE

CIVIL SERVICE EMPLOYEES

WEST AMERICAN

AMERICAN AUTOMOBILE

AMERICAN DRUGGISTS

THE AMERICAN INSURANCE COMPANY

BITUMINOUS CASUALTY CORPORATION

THE CINCINNATI INSURANCE COMPANY

THE CONTINENTAL INSURANCE COMPANY

HARBOR INSURANCE COMPANY

PACIFIC EMPLOYERS

PHOENIX INSURANCE COMPANY

REPUBLIC INSURANCE COMPANY

SOUTH CAROLINA INSURANCE COMPANY

TRINITY UNIVERSAL

UNITED FIRE AND CASUALTY

UNITED STATES FIDELITY & GUARANTY

WESTERN CASUALTY AND SURETY

Mean

.0170
.0979
.3133
.0409
.0033
.1010
.0824
.1106
.2366
.1461
.1544
,7680
.1237
.0379
.0259

1.000
.1981

1.000
.9440
.0418
.7301
.0912
.0879
.2190
.1139

1.000
.7339
.0694
.3692
.2413
.0096
.0183
.6176
.6764
.1001
.0041
.0025
.0301
.1718
.0356
.0712

.0237
.0986
.4016
.0288
.0102
.1070
.0991
.2899
.0808
.2194
,2227
.0341
.2962
.0255
.0445

0
.3853

0
.0360
.0627
.0799
.0956
.2163
.3574
.1858

0
.4040
.0813
.4306
.1379
.0207
.0334
.4633
.1208
.1313
.0107
.0086
.4426
.2490
.0428
.0878

-32-

FIGURE 1

OWNERSHIP ARRANGEMENTS, EARNINGS MEASURES, AND
DIVIDEND DEFINITIONS AVAILABLE FOR ANALYSIS

Ownership
Group

Earnings
Definition"

Dividend Definition

ii

iii

(1) = with policyowner

dividends

(2) = without policyowner
dividends

bee text for definitions.

Figure 1 summarizes the combinations of potential ownership arrange-
ments, earnings definitions, and dividend definitions that could be
tested in this study.

-33-

APPENDIX A

Co. No.

Company List

Co. No.

Company List

01 INA (ii) 49

02 Hartford Accident & Indemnity (iii) 50

03 AETNA Casualty & Surety (ii) 51

04 Federal (ii) 52

05 American States (iii) 53

06 Royal Indemnity (iii) 54

07 Westchester Fire (ii) 55

08 Calvert Fire (ii) 56

09 Ohio Casualty (ii) 57

10 Providence Washington (ii) 58

11 Government Employees (i) 59

12 Peerless (ii) 60
•13 Employers Fire (iii) 61

14 Employers Casualty (i)

15 United Pacific (iii)

16 National Casualty (iii)

17 American General (i)

18 Reliance (ii)

19 American Credit Indemnity (iii)

20 Affiliated FM (iii)

21 Connecticut Indemnity (iii)

22 State Farm Fire & Casualty (iii)

23 California Compensation & Fire (ii)

24 Hanover (i)

25 Utah Home Fire (ii)

26 American Policyholders (iii)

27 Globe Indemnity Co. (iii)

28 Pacific (Earlier Guarantee) (iii)

29 Tri State (iii)

30 American Bankers (i)

31 Civil Service Employees (ii)

32 West American (iii)

33 Excelsior Insurance Company of New York (i)

34 Republic Indemnity (ii)

35 American Automobile (iii)

36 American Druggists Insurance Co. (i)

37 The American Insurance Co. (ii)

38 American Reinsurance (i)

39 Bituminous Casualty Corp. (ii)

40 Carolina Casualty (i)

41 The Cincinnati Insurance Co. (ii)

42 The Continental Insurance Co. (ii)

43 Fidelity and Deposit Co. of Maryland (ii)

44 Firemens Insurance Company of Washington D.C. (i)

45 General Reinsurance Corp. (i)

46 Germantown Insurance Co. (i)

47 Harbor Insurance Co. (iii)

48 The Hartford Steam Boiler Inspection and Insurance Co

Hawkeye-Security Insurance Co. (iii)

Interstate Fire & Casualty Co. (ii)

Northeastern Insurance Co. of Hartford (iii)

Pacific Employers (ii)

Phoenix Insurance Co. (ii)

Reinsurance Corporation of New York (ii)

Republic Insurance Co. (ii)

South Carolina Insurance Co. (i)

Trinity Universal Insurance Co. (ii)

United Fire & Casualty Co. (i)

United Fire Insurance Co. (ii)

United States Fidelity and Guaranty Co. (i)

Western Casualty and Surety Co. (i)

(i)

Ownership arrangements are in the parentheses following the company names,
(i), (ii), and (iii) ownership arrangements are defined in the text.

These

.t\v