Document PDF

© 2015 The Journal of Risk and Insurance. Vol. 83, No. 4, 1007–1043 (2016).
DOI: 10.1111/jori.12080
INTERACTIONS BETWEEN RISK TAKING, CAPITAL, AND

REINSURANCE FOR PROPERTY–LIABILITY INSURANCE FIRMS
Selim Mankaı̈
Aymen Belgacem
ABSTRACT
Financial theory has long recognized the structural relationship between
capital and risk. This article posits reinsurance usage as a new endogenous
decision variable and analyzes its effect on this decision mix from a sample
of U.S. property–liability insurance firms. Empirical results obtained from a
simultaneous equation model confirm the mutual interactions among
capital, reinsurance and risk taking. Risk taking is positively related to
capital, which highlights the effectiveness of regulatory mechanisms and the
relevance of the capital buffer hypothesis. Reinsurance is negatively
associated with capital, for which it displays a substitutive effect. These
results seem to vary with the insurers’ level of capitalization, affiliation with
a group, size, and organizational form. Unlike other decision variables, the
capital ratio is adjusted to its target level.
INTRODUCTION
To reduce the likelihood of their failure, insurance firms have always been subject to
various constraints related to risk taking and capital holding. In this respect, capital
adjustments are generally made through earnings retention or new shares issuance.
Hoerger, Sloan, and Hassan (1990) and Garven and Lamm-Tennant (2003) demons-
trate that reinsurance, which involves ceding part of the assumed underwriting risk,
may affect these decisions by reducing loss volatility and acting as contingent capital.
More recently, the role of reinsurance has become even more important in view of
regulatory developments that are more amenable to defining capital requirements in
terms of some of its qualitative aspects (Eling and Holzm€ uller, 2008; Scordis and
Steinorth, 2012). Understanding the relationship between capital, risk, and
reinsurance is of great significance for regulators, who must craft prudential rules
Selim Mankaı̈ is at the IPAG Business School. Aymen Belgacem is at the University of Orleans,
Technology Institute of Bourges, Laboratoire d’Economie d’Orleans. Selim Mankaı̈ can be
contacted via e-mail: selim.mankai@udamail.fr.
1007
1008 THE JOURNAL OF RISK AND INSURANCE
to regulate insurers’ solvency.1 Shareholders are also concerned with the possible
transmission of shocks to capital resulting from unanticipated losses. Negative shocks
often entail forced sales of assets, which adversely affect the firm value.
Among the relationships between the three decision variables considered in this
article, those between capital and risk are by far the most discussed in the literature.
Several hypotheses related to moral hazard, agency costs, and regulatory pressures
have been posited to explain their mutual interactions. The first hypothesis, based on
agency costs and buffer capital theory, predicts a positive relationship between
capital and risk. An alternative hypothesis, based on information asymmetry,
predicts a negative relationship. The conflicting predictions of these two hypotheses
paved the way for an active empirical research. Shrieves and Dahl (1992) were the first
to examine this relationship for U.S. banks. In a subsequent article, Cummins and
Sommer (1996) investigate the issue for nonlife insurance firms and provide empirical
support for a positive relationship between capital and risk. Baranoff and Sager (2002)
empirically explore these interactions in the case of life insurance, finding a positive
(negative) relationship between capital and asset (liability) risk. Shim (2010) also
confirms the positive relationship between these variables.
While the association between capital and risk has been extensively studied in the
literature, few articles to date have explored the relationship between risk and
reinsurance usage (e.g., Cole et al., 2011), and even fewer, the joint interactions
between the three decision variables. Reinsurance mitigates underwriting and
solvency risks and enables insurers to increase their capacity to underwrite new
business. There are several reasons to believe that reinsurance is endogenously
influenced by the choice of capital and risk, and vice versa. In this respect, MacMinn
(1987) and Plantin (2006) find that the reinsurance ratio is determined together with
the capital structure. Dionne and Triki (2004) document a strong positive relationship
between leverage as a subscription risk indicator and reinsurance demand. In the
same vein, Shiu (2011) focuses on the endogenous nature of reinsurance and finds that
it is positively related to leverage, and vice versa. All these results support the
hypothesis of interdependence between reinsurance, capital, and risk.
This study analyzes capitalization policy and its relationship with risk taking and
reinsurance usage. More specifically, we aim to determine the nature of adjustments
between these decisions and how they move to their target levels. The contribution of
this article to the existing literature is twofold. First, we attempt to fill the gap in the
literature by examining interactions among these three decision variables instead of
studying them in pairs. We believe that much is to be gained through a joint analysis
acknowledging the simultaneity associated with decisions made in this regard. The
second contribution is an extensive empirical analysis of the interactions following
1
The theoretical literature often develops conflicting arguments about the effects of regulatory
pressure (Scannella, 2012). Strict regulation may create distortions in the operations of solvent
firms. By contrast, permissive regulations may lead to high risk exposure, threatening the
creditworthiness of insurers. This situation frequently implies a slowdown in innovation,
inefficient investment strategies, or passive capital accumulation, and to encourage good risk
management practices.
INTERACTIONS BETWEEN RISK TAKING, CAPITAL, AND REINSURANCE 1009
several factors, such as the level of regulatory pressure, group affiliation, firm size,
and organizational form. We show that capital, risk, and reinsurance interact in both
directions and vary according to the transversal factors. We also provide empirical
evidence that the capital ratio moves slowly toward a target level.
This article is organized as follows. The second section sketches the theoretical
underpinnings of the interactions between risk taking, capitalization, and reinsurance
and develops a set of hypotheses. The third section identifies and discusses the main
determinants of each decision. The fourth section presents the econometric model and
estimation technique. The fifth section describes the data set and provides summary
statistics. The sixth section reports and discusses empirical results and provides some
robustness checks. The seventh section provides concluding remarks.
THEORETICAL BACKGROUND AND RESEARCH HYPOTHESIS

In this section, we briefly outline the theoretical literature that highlights the
interactions between capital, reinsurance usage, and risk taking and we posit our
hypotheses.
Interactions Between Capital and Risk

Interactions between capital and risk have been the focus of active research,
particularly in the banking sector. Since its introduction in 1994, the risk-based capital
(RBC) system appears to have steadily reinforced the interdependence between these
two decisions for insurance firms. Several theoretical arguments related to agency
costs, moral hazard, and regulatory pressures have been proposed to explain such
interactions.
The literature makes contradictory predictions regarding the nature of such
interactions. The first hypothesis refers to insurers’ incentives to exploit guaranty
funds, which provide last-resort protection when insurers are unable to fulfill their
contractual commitments. Failure costs are borne by all guaranty fund members.
When members’ contribution to the fund are not correlated with actual risk, that is,
when premiums are flat fee rather than risk based, adverse incentives can result for
insurers to increase risk and reduce capital (Cummins, 1988). However, the
importance of this hypothesis is tempered by the fact that the coverage of the
guaranty fund is less complete than in the banking sector and by the growing
effectiveness of supervisory mechanisms and market discipline (Cheng and Weiss,
2012). Therefore, the incentive to take excessive risk for nonlife insurers is restricted.
Another explanation for a negative relationship between capital and risk for U.S.
insurers may be due to flaws in the RBC formula. Cheng and Weiss (2012) argue that
some risks are overweighed, while others are underweighted. This discrepancy
encourages insurers to rearrange their portfolios by choosing assets or lines of
business with low weights. Thus, insurer aggregate risk may increase while capital
requirements decrease, resulting in a negative relationship.
The second hypothesis suggests a positive relationship between these two variables.
According to the capital buffer theory, insurers hold excess capital above the
regulatory minimum as a guaranty against unanticipated extreme losses and to avoid
regulatory costs (Shim, 2010). Most empirical evidence supports the positive
relationship between capital and risk for both nonlife and life insurance firms
(Cummins and Sommer, 1996; Baranoff and Sager, 2002; Shim, 2010). The relationship
between capital and risk is more important for insurers with relatively low capital
levels. These firms are expected to respond more extensively in terms of capital
adjustment to risk taking because of regulatory pressure and market discipline.
Fonseca and Gonzalez (2010) argue that self-regulation encouraged by market
discipline can further affect the recapitalization decision. Insurers exposed to high
market discipline are pushed into adjusting their capital more extensively. Based on
the recent empirical results, our first hypotheses are as follows:
H1A: Capital and risk adjustments are positively related.

H1B: Capital and risk adjustments are more highly positively related for low-capitalized
insurers than for high-capitalized insurers.
Interactions Between Capital and Reinsurance

The analysis of the relationship between capital and reinsurance is poorly
documented in the financial literature. Stulz (1996) and Adiel (1996) demonstrate
that reinsurance affects the solvency and can function as off-balance-sheet capital. In
addition, reinsurance increases the insurer’s surplus and strengthens its underwrit-
ing capacity (Graham and Rogers, 2002; Aunon-Nerin and Ehling, 2008; Zou and
Adams, 2008; Cole et al., 2011). Under the renting capital hypothesis, reinsurance is
typically viewed as a substitute for capital (Armstrong and Dror, 2007). It is also
considered a risk transfer tool that reduces the required capital on the insurer’s
balance sheet by enabling the use of capital ‘‘rented’’ from the reinsurer. Using
traditional sources of capital, such as corporate debt, contingent capital, or new
equity, in response to a shock is typically an expensive and difficult operation. Thus,
insurers will choose reinsurance rather than holding more capital as it allows them to
maintain an acceptable level of solvency. Cummins et al. (2008) show that the optimal
use of reinsurance can improve shareholder value by substituting for equity, which
thereby reduces the cost of capital and increases returns from underwriting activities.
When the benefit of the risk mitigation exceeds the expected return sacrifice,
reinsurance increases the value of the company, leading to the following hypothesis
H2: There is a negative relationship between capital and reinsurance in both directions.
Interactions Between Reinsurance and Risk

One way to reduce volatility and insolvency risk is to use reinsurance, particularly
when the capital level moves closer to solvency constraints (Adams, 1996).
Transferring risk relaxes capital restrictions and allows insurers to expand their
capacity to issue new policies. Highly leveraged insurers are more exposed than
others to the risk of insolvency and are expected to use more reinsurance, suggesting
positive interactions (Shiu, 2011). In this regard, Froot, Scharstein, and Stein (1993)
argue that risk management techniques such as reinsurance enhance the market value
of insurance firms by enabling managers to realize positive net present value (NPV)
projects when risk capacity is binding. However, the impact of reinsurance usage on
risk taking is not clear. De Haan and Kakes (2010) assume that this relationship
depends on the capital level. Based on this discussion, we propose the following
hypothesis:
H3: There is a positive relationship between risk and reinsurance in one direction.
Impact of Transversal Factors

The neoclassical theory of the firm identifies several factors that may affect the main
financial decisions of insurers.
Size. Firm size plays an important role in influencing the insurer’s risk appetite
through its effect on investment opportunities, reinsurance usage, and the firm’s
access to capital. Large firms are generally subject to lower information asymmetry
between managers and potential investors, which reduces the cost of capital (Smith,
1977). Moreover, large firms are likely to have a better qualitative and geographical
allocation of risks and hold proportionally less capital than small firms. Due to
economies of scale in risk management and their greater ability to raise capital in the
short run, large firms are expected to require less capital to operate and to undertake
greater risk (Titman and Wessels, 1988). Numerous studies document that firm size
negatively affects the demand for reinsurance (Hoerger, Sloan, and Hassan, 1990;
Powell and Sommer, 2007). Small insurers depend more on reinsurance usage
because they do not have economies of scale and scope, and they have higher
financing costs when raising external funds for at least two main reasons. First, the
direct costs of financial failure are not proportional to the size of a company (Warner,
1977). Second, raising capital in financial markets is an expensive undertaking. Cole
and McCullough (2006) view firm size as an inverse measure of bankruptcy costs and
find a negative relationship between this variable and the demand for reinsurance.
Thus, large insurers will rely less on reinsurance to expand their underwriting
capacity. Thus, we formulate the following hypothesis:
H4: Large insurers hold less capital, use less reinsurance, and take more risk than small
insurers.
Organizational Form. There are two main forms of organization in the insurance
industry: mutual firms and stock firms. In mutual organizations, customers provide
capital, bear risk, and own the residual value of the firm. In stock organizations, the
shareholders bring capital and receive the residual value, whereas the risk is shared
between shareholders and policyholders. The implications of the pecking order
theory vary, depending on organizational form. Stock firms have greater access to the
financial markets. Mutual firms have more difficulties in raising capital than stock
insurers. Agency costs also vary with organizational form. The shareholders–
policyholders and shareholders–managers conflicts may impact the choice of capital
level. Managers have an incentive to maximize their perquisite consumption and to
protect their own human capital (Mayers and Smith, 1990). However, shareholders
have more control over managers in stock firms than in mutual firms. Harrington and
Niehaus (2002) find that mutual insurers tend to hold more capital than stock insurers
and are more likely to manage risk through reinsurance. Indeed, Cole and
McCullough (2006) note that mutual firms, which have less access to capital
markets in cases of catastrophic loss, use more reinsurance. Thus, we formulate the
following hypothesis:
H5: Mutual insurers hold more capital, use more reinsurance, and take less risk than stock
insurers.
Group Affiliation. Group membership is a potential factor that may affect

interactions among the decisions of interest. Cheng and Weiss (2012) argue that
capital and risk for insurers within a group might be determined strategically at the
group parent level. Insurers belonging to a group generally hold less capital and have
incentives to take more risk, as they have access to internal resources provided by
other group members. Cummins and Sommer (1996) find that group-affiliated
insurers are associated with greater risk relative to single unaffiliated insurers. It is
also easier for affiliated insurers to obtain capital injections from parent firms when
their capital levels become insufficient. Powell, Sommer, and Eckles (2008) find
evidence that internal capital markets play a significant role in investment behavior.
Powell and Sommer (2007) note the importance of reinsurance transactions with
affiliated insurers, hypothesizing that reinsurance from internal sources should be
more cost effective than from external markets. Mayers and Smith (1990) and Cole
and McCullough (2006) find evidence supporting a positive and significant
correlation between the use of reinsurance and group membership. This result is
confirmed by Powell and Sommer (2007), who argue that transactions within a
group have lower asymmetric information costs than those with nongroup firms
and, thus, lower monitoring costs. Reinsurers have incentives to monitor the
underwriting risk of the insurers with whom they do business (Cole et al., 2011). The
relationship between risk and reinsurance is more significant for affiliated than
for nonaffiliated reinsurers since the monitoring costs will be higher for the
latter. Accordingly, decisions of interest are made differently following group
membership:
H6 Insurers belonging to a group hold less capital, take more risk, and use more reinsurance
than nongroup firms.
CONTROL VARIABLES
In addition to the transversal variables assumed to affect capital, risk, and reinsurance
simultaneously, the empirical literature identifies several specific factors that may
individually have an impact on these variables. Here, we present the most important
factors to be considered in the empirical model developed below.
Determinants of Capital
Performance. Firms with high profitability generally have sufficient internal funds
that may be transformed into capital. According to the pecking order theory, firms
prefer internal sources of funding (such as retained earnings) to external financing

(i.e., debt or new equity) because internal sources of financing are less expensive and
more efficient (Park and Pincus, 2001). Alternatively, in the presence of asymmetric
information, the use of external funding may convey negative information to the
market about the firm’s value. Thus, we expect performance to positively affect
capital levels. Following Titman and Wessels (1988) and Fama and French (2002), we
use return on assets as an indicator of this variable.
Cost of Capital. The cost of holding capital is an important determinant of capital

levels (Estrella, 2004). Various costs, such as agency and information asymmetry
costs, are related to the use of equity capital. The level of capital is expected to be
inversely related to such costs. The cost of holding capital is difficult to measure in
practice. Similar to Ayuso, Perez, and Saurina (2004) and Jokipii and Milne (2008), we
approximate it as the average of positive returns on equity (ROE) over the most recent
5 years.
Information Symmetry. Signal theory suggests that information asymmetry along

with the opacity of the insurance industry are important determinants of the
capitalization level (Pottier and Sommer, 2006; Morgan, 2002). Insurers with volatile
incomes are likely to use retained earnings rather than external capital to cover future
losses or cope with external shocks. Insurers can overcome information asymmetry
by building capital stock during periods of high profitability. We use the volatility of
ROE as a measure of this variable (Cummins and Nini, 2002; Grubisic and Leadbetter,
2007).
Growth Opportunities. Insurers with better growth opportunities are expected to

hold more capital. The literature uses the book-to-market ratio or changes in R&D
expenses to measure this variable. Hovakimian, Opler, and Titman (2001) and Fama
and French (2002) use the book-to-market ratio as a proxy for a firm’s growth
opportunities (GROWTH). This proxy is defined as the ratio of the book value of total
assets minus the book value of equity and debt plus the ratio of market value of equity
and debt to the book value of total assets (Barclay and Smith, 1995; Rajan and
Zingales, 1995). However, due to a lack of data, we use an alternative measure,
following Carayannopoulos and Kelly (2004), namely, the past growth (over 5 years)
in premiums.
Exposure to Extreme Risk. Exposure to extreme risks is likely to influence the level of
capital. Zanjani (2002) demonstrates that companies insuring heavily against natural
disasters have higher capital levels than those less exposed to such events. The level of
exposure to extreme risk is measured in our model as the proportion of direct
premiums written on property insurance in Eastern coastal states and on earthquake
insurance (Powell and Sommer, 2007).
Liquidity Risk. Insurers with a large share of liquid assets are more likely to be exempt
from regulatory constraints than others (Meyer et al., 2013). The low risk associated
with these assets allows for easy adjustments of capital levels. Therefore, insurers
with more liquid assets are expected to have less capital and take more risks. De
Ceuster and Masschelein (2003) find that the major source of illiquidity is the asset–
liability mismatch, which may encourage insurance firms to hold more capital. In this
study, liquidity risk is measured by the ratio of liabilities to liquid assets.
Deficit. This variable is used to test for the existence of a preference order among
funding sources for insurers. It may reflect the need for external financing when
internal cash flows are exhausted. According to the pecking order theory, firms with
high deficits have higher leverage and lower capital as they prefer to finance
investments internally (De Bie and De Haan, 2007). Thus, the financial deficit has a
negative impact on the capital ratio. In the same way, this variable may have a
mechanical negative effect on capital since it would deplete it. The deficit variable is
measured as the total amount of cash dividends, investments, and change in working
capital minus internal cash flow.
Determinants of Risk Taking

Previous theoretical and empirical work suggests that risk taking is affected by
various factors, including the following.
Business Mix. The business mix is the degree of centering on a firm’s core business. A
high concentration of premiums in certain lines of business exposes insurers to
significant risk. However, high concentration may also reflect specialization and
better risk pricing. Thus, we expect a positive relationship between the degree of
centering and risk taking. This variable is approximated by the Herfindahl index of
the four major branches of nonlife insurance business, namely, short- and long-term
personal insurance and short- and long-term commercial insurance.
Loss Volatility. When the loss ratio is highly volatile, the insurer has less certainty
about the future value of losses and may thus need to reduce risk taking to guard
future possible insurance payouts. However, Hoerger, Sloan, and Hassan (1990)
argue that the increased volatility of claims generates increased risk. This factor is
approximated by the standard deviation of the loss ratio gross of reinsurance over the
last 3 years.
Geographic Concentration. An insurer may reduce its overall risk by holding a

portfolio whose components are not perfectly correlated across regions and/or
activities. Empirical studies find that in the banking sector, diversification is
associated with moderate risk taking (Hughes, Lang, and Mester, 1996; Deng,
Elyasiani, and Mao, 2007), which is a hypothesis based on the benefits of cost
reduction and income synergy (Saunders and Cornett, 2007). Alternatively,
diversification may be associated with higher risk taking because of the agency
problem and competition. The degree of diversification is measured by the
Herfindahl index percentages of direct premiums written by geographical area.
Determinants of Reinsurance
Researchers have documented that reinsurance can be affected by several firm-
specific factors.
Business Mix. Mayers and Smith (1990) examine the effects of the composition of a
firm’s portfolio of activities on the demand for reinsurance. They observe that an
increased concentration of activities increases the volatility of cash flows and the
risk of bankruptcy. Reinsurance might be a solution to the risk of insolvency arising
from this source. Moreover, Shortridge and Avila (2004) and Cole and McCullough
(2006) demonstrate that this factor reflects the degree of centering on the core
business. By contrast, the economic benefits of specialization can reduce demand for
reinsurance.
Leverage. Highly leveraged insurers are exposed to a higher likelihood of

insolvency and thus higher expected bankruptcy costs (Shiu, 2011). Moreover, it is
more difficult for them to raise capital in financial markets at low cost. Thus,
reinsurance may act as a substitute for capital and reduce the probability of
insolvency (Garven and Lamm-Tennant, 2003). Higher leveraged insurers have more
incentives to increase their reinsurance demand. Thus, we expect a positive
relationship between leverage and reinsurance.
Geographic Concentration. Geographical concentration reflects the degree of

diversification of an insurer across states. Cole and McCullough (2006) find a
negative relationship between geographic concentration and reinsurance and
between lines-of-business concentration and reinsurance. Garven and Lamm-
Tenant (2003) and Mayers and Smith (1990) also find a negative relationship using
a different measure of diversification. This variable is measured by the Herfindahl
index of the percentage of direct premiums written by geographic area.
METHODOLOGY
Model
In this section, we investigate the relationships between capital, reinsurance, and risk
adjustments using a simultaneous equations model. Note that the observed variations
of these variables are both discretionary and caused by factors exogenous to the
insurer (Shrieves and Dahl, 1992; Jacques and Nigro 1997; Shim, 2010). To account for
this behavior, we introduce lagged variables as partial adjustment components
(cf. the Appendix). The model has the following specifications:
DCapi;t ¼ l0 fCap Capi;t 1 þ l 1 DReinsi;t þ l2 DRiski;t þ l3 Perfi;t þ l4 Cost capi;t

þ l5 Extremei;t þ l6 Assyi;t þ l7 Lqt riski;t þ l8 Growthi;t þ l9 Deficiti;t
þ l10 Stocki;t þ l11 Groupi;t þ l12 Sizei;t þ Yeart þ ei;t ð1Þ
DReinsi;t ¼ b0 fReins Reinsi;t 1 þ b1 DCapi;t þ b2 DRiski;t þ b3 Levi;t þ b4 Mixi;t

þ b5 Higeoi;t þ b6 Stocki;t þ b7 Groupi;t þ b8 Sizei;t þ Yeart þ ui;t ð2Þ
DRiski;t ¼ g 0 fRisk Riski;t 1 þ g 1 DReinsi;t þ g 2 DCapi;t þ g 3 Loss voli;t þ g 4 Mixi;t

þ g 5 Higeoi;t þ g 6 Stocki;t þ g 7 Groupi;t þ g 8 Sizei;t þ Yeart þ vi;t ; ð3Þ
TABLE 1
Variable Definitions and Endogeneity
Expected Signs
DWH Test
Variable Description DCap DReins DRisk ofEndogeneity
Dependent variables
Capital Cap Ratio of surplus to (þ/) (þ/) Endogenous
total admitted
assets
Reinsurance Reins Ratio of premiums () (þ) Endogenous
ceded to direct
premiums written
plus reinsurance
premiums assumed
Volatility of asset Risk1 Volatility of the asset (þ) (þ) Endogenous
and liability to liability ratio
Ratio of risky Risk2 Ratio of risky assets (þ) (þ) Endogenous
assets and and liabilities
liabilities
Control variables
Regulatory Reg 1 if firm’s net Exogenous
pressure premium to surplus
ratio 300 percent,
0 otherwise
Performance Perf Return on assets ratio (þ) Endogenous
(ROA)
Cost of capital Cost_cap Average of positive () Exogenous
ROE over the last
5 years
Exposure to Extreme Proportion of direct (þ) Endogenous
extreme risk premiums written
for property
insurance in
Eastern coastal
states and for
earthquake
insurance
Information Assy The standard (þ) Exogenous
asymmetry deviation of the
firm’s ROE over the
last 5 years
Business mix Mix Herfindahl index of (þ/) (þ) Exogenous
short and long tails
of personal and
commercial lines
(Continued)
TABLE 1
Continued
Expected Signs
DWH Test
Variable Description DCap DReins DRisk ofEndogeneity
Liquidity risk Lqt_risk Ratio of liabilities to (þ) (þ) Endogenous

liquid assets
Loss volatility Loss_vol Standard deviation of (þ) Exogenous
the loss ratio, gross
of reinsurance over
the last 3 years
Leverage Lev Ratio of direct (þ) Endogenous
business written to
surplus
Geographic Higeo Herfindahl index of (þ) (þ/) Exogenous
diversification direct premium
written across
geographic areas
Growth Growth Percentage growth of (þ/) Exogenous
opportunities written premiums
Deficit Deficit Financial deficit ratio (þ) Exogenous
calculated as cash
dividends plus
investments plus
change in working
capital minus
internal cash flow
Stock vs. mutual Stock 1 if the insurer is a () Exogenous
stock firm, 0 if it is a
mutual firm
Group affiliation Group 1 if the insurer is a () Exogenous
member of a group
in year t, 0
otherwise
Size Size The logarithm of total () Exogenous
assets
Time effects Year Annual temporal
effect
Note: This table describes the variables used in the empirical model. Endogeneity tests for dependent and
control variables not reported are available from the authors upon request.
where ei;t , vi;t , and ui;t are error terms. Cap, Reins, and Risk denote the capital ratio,
reinsurance ratio, and risk taking, respectively. The definitions of the control variables
are listed in Table 1. Time effects are included in the model to capture systematic
events and high losses associated with manmade and natural disasters. In capturing
interactions between capital, risk, and reinsurance, simultaneous equations models
have shown better performance than estimating the equations individually. The
system of equations considers two specifications, which alternatively exclude and
include the regulatory pressure mechanisms. Following Shim (2010), we introduce

further interaction between the degree of insurer capital and the key variable
adjustments.
Variable Construction
Capital (Cap). Because most of the insurers in our sample are not listed, it is not
possible to determine their market values. Similar to Cummins and Sommer (1996)
and Shim (2010), we proxy the capital ratio by the book value of the surplus divided
by the total value of assets.
Reinsurance (Reins). This variable is given, for each insurer, by the ratio of
reinsurance premiums ceded to total premiums written, which include direct
premiums written and reinsurance premiums assumed (Garven and Lamm-Tennant,
2003; Cole and McCullough, 2006; Powell and Sommer, 2007).
Risk Taking (Risk). Unlike capital and reinsurance measures, which have
straightforward definitions, risk assessment remains an open issue. Several
measures have been proposed in the literature, especially in the banking sector:
the standard deviation of the loss ratio (Meyers, 1989), insurer’s exposure to external
factors as captured by market betas (Eling and Schuhmacher, 2007), and equity
volatility, which reflects both external and firm-specific factors (systematic and
nonsystematic risk). Shim (2010) argues that those measures do not fully capture the
complex risk profile of the insurance business. An alternative measure proposed in
the literature (Cummins and Sommer, 1996; Rime, 2001; Shim, 2010) is the proportion
of risky assets and liability. Based on portfolio theory, we measure total risk by the
volatility of the asset-to-liability ratio. According to Cummins and Sommer (1996) and
Shim (2010), an insurer’s asset–liability volatility can be expressed as follows:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s ¼ s 2A þ s 2L 2s A:L ; ð4Þ
where s A and s L are the volatility measures of the insurer’s assets and liabilities,
respectively, and s A;L is the covariance of the logarithms of the assets and liability
values. Let us denote the proportion of assets of asset type i in the investment portfolio
by xi and the proportion of liabilities from business line j by yj. The respective
volatilities of the asset and liability portfolios and the covariance of the logarithms of
the liability and asset returns are given as follows:
X
N X
N
s 2A ¼ xi xj rAi Aj s Ai s Aj ð5Þ
i¼1 j¼1
X
M X
M
s 2L ¼ yi yj rLi Lj s Li s Lj ð6Þ
i¼1 j¼1
X
N X
M
s AL ¼ xi yj rAi Lj s Ai s Lj ; ð7Þ
i¼1 j¼1
where s Ai and s Lj denote the volatilities of the log of asset type i and the log of
liabilities in business line j, respectively. The parameter reflects the correlation
between the log of the ith asset and the log of the liabilities in the jth business
line, whereas N is the number of asset categories and M is the number of lines of
business.
To measure asset–liability volatility, one must first define different lines of activities
and asset categories. Following Shim (2010), we aggregate each insurer’s lines of
business into 12 categories: homeowners/farmowners, auto physical damage, auto
liability, commercial multiple peril, special property, fidelity/surety, accident,
health, financial guaranty, medical malpractice, workers’ compensation, other
liability, special liability, and miscellaneous liability. The types of assets are classified
into 7 categories: stocks, government bonds, corporate bonds, real estate, mortgages,
cash and other invested assets, and noninvested assets. A second risk measure is
used to assess the robustness of our results and it is defined as the mean of two
conventional risk measures. The first is asset risk approximated by the value of
investments in equities and real estate divided by total invested assets. The second
component is underwriting risk, which is defined as the proportion of premiums
written in risky lines (commercial auto liability, allied lines, earthquake, surety, theft,
inland marine, fire, international, boiler and machinery, reinsurance, and medical
malpractice occurrence).
Estimation Methodology
Estimating the equations individually ignores the problem of the potential
endogeneity that violates the condition of no correlation between exogenous
variables and the error terms (Baltagi, 2005). To address this problem and to avoid
biased ordinary least squares (OLS) estimates, we use the three-stage least squares
(3SLS) technique (Shim, 2010; Rime, 2001; Jacques and Nigro, 1997). This
methodology also treats ‘‘suspicious’’ endogeneity of some control variables using
instrumental variables. First, we estimate the model using the OLS and 3SLS methods
and compare the results using the Hausman test. As we find significant differences in
the estimates, the endogeneity problem is confirmed.
To examine the interactions between capital, risk, and reinsurance, it is crucial
to determine whether such variables are endogenous or exogenous. It is also
important to check the exogeneity of the control variables. For that purpose, the
Durbin–Wu–Hausman (DWH) test is performed for all variables that are
individually regressed against all the exogenous variables and instrumental
variables. The residuals obtained from the first stage are then added as an additional
independent variable in their respective equation. The results of the DWH test are
summarized in Table 1. According to the test, all left-hand-side variables in the
model are endogenously determined. Some of the control variables (exposure to
extreme risks, performance, liquidity risk, and leverage) are endogenous. Therefore,
we take this result into account by treating them as endogenous when we estimate
the model.
Regarding the choice of instruments, we identify them according to the extant
literature and the Sargan–Hansen test. In the model, we use the largest number
possible for which the Sargan statistic for overidentification restrictions is still
satisfied. Specifically, we choose a large list of variables used in the literature (Shim,
2011; Jokipii and Milne, 2011; Shiu, 2012). We first examine the OLS results to identify
variables that affect one of the three variables of interest but not the remaining two.
We find that the two-period-lagged risk and reinsurance variables, the economic
growth rate variable, and the variable for the ratio of unrealized gains (with no lag
and one period lagged) are valid instruments. Second, in order to improve our
estimations, we add additional exogenous variables used in the model as
supplementary valid instruments. Prior to estimation, we verify whether the time
series are stationary based on the Levin–Lin–Chu test.
DATA
Sample
The data used in this study are collected from the National Association of Insurance
Commissioners (NAIC) annual statement database for U.S. property–liability
insurers from 1999 to 2008. The sample is limited to solvent insurers reporting
positive values of admitted assets, gross and net premiums written, equity capital,
and ceded reinsurance premiums. We retain only active insurers with no regulatory
actions in process. After applying these sample screens, our final sample consists of
12,511 year-firm observations. Our sample thus accounts for 82 percent (85 percent) of
the entire U.S. property–liability market in terms of total assets in the year 1999 (2008).
We use an unbalanced data panel to allow for a comprehensive evaluation of the
property–liability market. The sample includes firms that entered or left the market
during the study period. In addition to the NAIC database, we also collected data
from DataStream to estimate asset returns.2
Descriptive Statistics and Correlation Analysis

Table 2 presents descriptive statistics of all variables used in this study. The mean and
median of the capital ratio are approximately 42 and 38 percent, respectively. These
statistics are (39 and 34 percent) and (2.7 and 0.4 percent) for the reinsurance ratio and
risk level, respectively. The distributions of the two risk measures are positively
skewed. With regard to the diversification profiles measured by the two Herfindahl
indices (Higeo and Mix), we note small variations across insurers. Approximately
74 percent of insurers are stock firms and 73 percent are group affiliated.
Before conducting the regression analysis, we first considered the possibility of multi-
collinearity among independent variables, which may lead to biased estimates.
Table 3 presents the Pearson correlation between the variables included in the
regression model, which indicates that the correlation between capital and risk is
positive and confirms the hypothesis that insurers adjust their capital levels upward
following increases in risk. Note that correlations between all exogenous variables
classified in Table 1 are modest and do not exceed 0.3. Nonetheless, we report some
2
We use the S&P 500 Global Index, Barclays Capital, U.S. 20-year Treasury bond, Dow Jones
Corporate Bond Index, Merrill Lynch Mortgage, U.S. Real Estate & Rental & Leasing, and the
3-month U.S. Treasury bill as proxies for asset returns.
TABLE 2
Descriptive Statistics
Standard 5 Percent 95 Percent

Mean Median Deviation Quantile Quantile Skew Kurt
Cap 0.420 0.381 0.169 0.211 0.764 0.977 0.667
Reins 0.394 0.345 0.283 0.024 0.899 0.617 0.778
Risk1 0.027 0.004 0.192 0.0009 0.071 12.608 17.747
Risk2 0.017 0.001 0.174 0.0004 0.014 17.163 31.493
Perf 0.024 0.026 0.049 0.054 0.091 0.057 14.694
Cost_cap 0.056 0.062 0.110 0.001 0.203 0.950 13.740
Extreme 0.084 0.047 0.196 0 0.549 2.958 8.520
Assy 0.079 0.049 0.104 0.009 0.241 5.512 52.209
Lqt_risk 0.746 0.733 0.514 0.255 1.190 18.164 55.039
Mix 0.331 0.278 0.321 0.000 0.931 0.590 0.931
Loss_vol 1.461 0.754 5.076 0.241 3.606 21.185 50.094
Up 1.742 1.285 1.639 0.109 5.207 1.960 4.554
Higeo 0.388 0.377 0.095 0.264 0.500 0.018 1.634
Growth 0.122 0.049 0.646 0.395 0.763 1.571 61.047
Deficit 0.057 0.056 0.157 0.159 0.295 1.064 10.801
Stock 0.740 1 0.439 0 1 1.096 0.799
Group 0.731 1 0.443 0 1 1.042 0.914
Size 18.510 18.418 1.815 15.721 21.668 0.298 0.069
Reg 0.018 0.000 0.133 0 1 7.238 50.397
Note: This table reports the mean, median, standard deviation, 5th percentile, 95th percentile,
skewness, and kurtosis of all variables. The data are collected from the NAIC and Datastream
databases covering U.S. property–liability insurance firms from 1999 to 2008. Cap, capital;
Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky assets and liabilities;
Reg, regulatory pressure; Perf, performance; Cost_Cap, cost of capital; Extreme, exposure to
extreme risk; Assy, information asymmetry; Mix, business mix; Lqt_risk, liquidity risk;
Loss_vol, loss volatility; Lev, leverage; Higeo, geographic diversification; Growth, growth
opportunities; Deficit, deficit; Stock, stock versus mutual; Group, group affiliation; Size, size of
the insurer.
significant correlation between endogenous variables, which confirms our choice to

treat them as endogenous in the estimation procedure. Another notable result
outlined in Table 3 is the high correlation between the two risk measures. Consistent
with standard econometric practice, we assess the degree of multicollinearity among
the independent variables using the variance inflation factor (VIF). All VIF values
reported in Table 4 range between 1.01 and 2.2. Thus, no multicollinearity is detected
among the explanatory variables.
RESULTS
This section reports and discusses empirical results. We estimate the model first for
the full sample and then for various subsamples by following the level of capital,
affiliation with a group, firm size, and organizational form. In the second phase, we
perform some robustness tests.
1022
TABLE 3
Correlation Coefficient Matrix
THE JOURNAL
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
OF
1. Cap 1
RISK
2. Reins 0.012 1
3. Risk1 0.046* 0.029* 1
4. Risk2 0.033* 0.013 0.243* 1
5. Perf 0.206* 0.059* 0.0023 0.032* 1
6. Cost_cap 0.090* 0.084* 0.0113 0.0141 0.6368* 1
7. Extreme 0.127* 0.072* 0.034* 0.0162 0.0008 0.052* 1
AND INSURANCE
8. Assy 0.238* 0.017 0.022 0.0048 0.1704* 0.400* 0.0736* 1

9. Lqt_risk 0.504* 0.149* 0.0261 0.0109 0.1363* 0.079* 0.0492* 0.1339* 1
10. Mix 0.072* 0.050* 0.068* 0.0232 0.0336* 0.053* 0.4085* 0.013 0.0955* 1
*
11. Loss_vol 0.084* 0.084* 0.021 0.007 0.0785* 0.103* 0.015 0.0906 0.022 0.0066 1
12. Lev 0.384* 0.460* 0.0503* 0.0212 0.1121* 0.111* 0.0288 0.1759* 0.3954* 0.095* 0.028 1
13. Higeo 0.155* 0.183* 0.0541* 0.0286 0.0015 0.054* 0.180* 0.055* -0.0728* 0.0499* 0.0121 0.0017 1
14. Growth 0.031* 0.047* 0.0154 0.0009 0.0819* 0.025 0.003 0.0249 0.0531* 0.0028 0.0950* 0.067* 0.0107 1
* * *
15. Deficit 0.145* 0.044* 0.0058 0.009 0.0255 0.058* 0.0212 0.0621 0.1306 0.001 0.0157 0.1260 0.0022 0.392* 1
16. Stock 0.106* 0.215* 0.0414* 0.0012 0.0434* 0.048* 0.1614* 0.0395* 0.0973* 0.2312* 0.0316* 0.0965* 0.237* 0.0389* 0.0363* 1
17. Group 0.117* 0.230* 0.0067 0.0436* 0.0081 0.020 0.0477* 0.0284 0.0798* 0.1193* 0.0201 0.0261 0.317* 0.0001 0.023 0.302*
18. Size 0.375* 0.042* 0.0031 0.0823* 0.0602* 0.157* 0.1195* 0.0418* 0.1586* 0.067* 0.0457* 0.1094* 0.479* 0.024 0.027 0.0833* 1 1
19. Reg 0.157* 0.056* 0.0129 0.0094 0.0878* 0.101* 0.0225 0.1558* 0.1021* 0.0833* 0.0151 0.2115* 0.0503* 0.0474* 0.048* 0.022 0.038* 0.033* 1
Note: Cap, capital; Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky assets and liabilities; Reg, regulatory pressure; Perf, performance; Cost_Cap, cost of capital;
Extreme, exposure to extreme risk; Assy, information asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol, loss volatility; Lev, leverage; Higeo, geographic diversification;
Growth, growth opportunities; Deficit, deficit; Stock, stock versus mutual; Group, group affiliation; Size, size of the insurer.
* represents statistical significance at the 1 percent level.
TABLE 4
Variance Inflation Factors
Eq DCap Eq DReins Eq DRisk1

DReins 1.1 DCap 1.02 DReins 1.01
DRisk1 1 DRisk1 1 DCap 1.02
Capt1 1.65 Reinst1 1.47 Risk1t1 1.01
Perf 1.86 Mix 1.1 Loss_vol 1.01
Cost_cap 2.06 Lev 1.34 Mix 1.07
Extreme 1.06 Higeo 1.43 Higeo 1.39
Assy 1.27 Stock 1.22 Stock 1.2
Lql_risk 1.38 Group 1.38 Group 1.32
Growth 1.34 Size 1.51 Size 1.46
Deficit 1.24
Stock 1.16
Group 1.32
Size 1.44
Mean VIF 1.38 Mean VIF 1.28 Mean VIF 1.16
Note: Cap, capital; Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky
assets and liabilities; Reg, regulatory pressure; Perf, performance; Cost_Cap, cost of capital;
Extreme, exposure to extreme risk; Assy, information asymmetry; Mix, business mix; Lqt_risk,
liquidity risk; Loss_vol, loss volatility; Lev, leverage; Higeo, geographic diversification;
Growth, growth opportunities; Deficit, deficit; Stock, stock versus mutual; Group, group
affiliation; Size, size of the insurer.
Full Sample
The main hypotheses tested in this paper are related to the existence of mutual
interactions between risk taking, capital, and reinsurance. Table 5 displays the full
sample estimation of equations system ((1)–(3)) using asset–liability volatility
(Risk1) as a risk measure. A closer look at the individual equations’ coefficients
indicates that most of the results are consistent with expectations. First, let us
consider the baseline results without regulatory pressure mechanisms in the
regressions. Overall, the coefficients related to the capital equation are significant,
except those related to exposure to extreme risks and the transversal factors. The risk
variable is positive and statistically significant, which is consistent with the capital
buffer hypothesis and suggests that an increase in risk taking leads to positive
adjustments of capital as a guaranty against unanticipated extreme losses (Jokipii
and Milne, 2011). Second, the negative sign of the reinsurance variable supports the
original hypothesis regarding the substitutability effect between these two variables.
This result is consistent with the findings of Stulz (1996) and Armstrong and Dror
(2006), who put forward the fact that reinsurance can serve as off-balance-sheet
capital that reduces the capital requirement and allows new policies to be issued. The
speed of adjustment toward the desired capital ratio, captured by the lagged
depended variable, is low (0.0389) and can be explained by the presence of various
adjustment costs.
TABLE 5 1024
Estimation Results of the Simultaneous Eequations Using Asset–Liability Volatility (Risk1) as a Risk Measure
Model I Model II
DCapt DReinst DRisk1t 4Capt 4Reinst 4Risk1t

THE JOURNAL
OF
Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat Coef t-stat
RISK
DCapt – – 0.556*** 5.91 0.276*** 3.7 – – 0.553*** 4.03 0.279*** 3.74

DReinst 0.180*** 3.72 – – 0.505*** 26.4 0.181*** 3.82 – – 0.516*** 24.24
DRisk1t 0.354*** 2.94 1.979*** 19.96 – – 0.319** 2.68 1.934*** 18.12 – –
Capt1 0.0389* 1.9 – – – – 0.0242 1.14 – – – –
Reinst1 – – 0.0044 0.49 – – – – 0.0062 0.68 – –
AND INSURANCE
Risk1t1 – – – – 0.0029 0.44 – – – – 0.0039 0.53

Regt1 Capt1 – – – – – – – – 0.347** 2.6 0.197*** 2.78
Regt1 Reinst – – – – – – 0.0567*** 3.33 – – 0.007 0.52
Regt1 Risk1 – – – – – – 0.174 0.69 0.017 0.06 – –
Perf 0.553*** 22.34 – – – – 0.540*** 21.97 – – – –
Cost_cap 0.0568*** 4.98 – – – – 0.0511*** 4.45 – – – –
Extreme 0.0159 0.53 – – – – 0.0063 0.22 – – – –
Assy 0.0576*** 4.37 – – – – 0.0584*** 4.46 – – – –
Lqt_risk 0.013 *** 4.08 – – – – 0.012*** 3.52 – – – –
Mix – – 0.354*** 4.8 0.187*** 5.27 – – 0.369*** 4.96 0.204*** 5.48
Loss_vol – – – – 0.00008 0.36 – – – – 0.0001 0.43
Lev – – 00006 0.44 – – – – 0.0009 – – –
Higeo – – 0.0681 1.33 0.0353 1.34 – – 0.0701 1.41 0.0375 1.41
Growth 0.029*** 9.16 – – – – 0.029*** 9.65 – – – –
Deficit 0.107*** 12.62 – – – – 0.107*** 12.87 – – – –
Stock 0.0021 0.69 0.0665*** 4.18 0.0349*** 4.42 0.0017 0.58 0.0676*** 4.33 0.037*** 4.61
Group 0.0025 1.05 0.0017 0.15 0.0013 0.23 0.0025 1.1 0.0031 0.28 0.0024 0.4
Size 0.0001 0.12 0.005* 1.71 0.0028* 1.72 0.0001 0.18 0.00577* 1.89 0.003* 1.92
(Continued)
TABLE 5
Continued
Model I Model II
DCapt DReinst DRisk1t 4Capt 4Reinst 4Risk1t
2001.year 0.018*** 2.77 0.106*** 5.89 0.0540*** 6.02 0.016* 2.51 0.104*** 5.94 0.0545*** 6.08
2002.year 0.0222*** 3.35 0.0868*** 4.96 0.043*** 5.03 0.0208*** 3.22 0.0849*** 4.99 0.043*** 5.03
2003.year 0.0272*** 7.85 0.023 1.38 0.0116 1.33 0.0280*** 8.31 0.0223 1.39 0.0114 1.32
2004.year 0.0168*** 4.98 0.015 0.93 0.00764 0.9 0.0178*** 5.4 0.0163 1.03 0.0083 0.97
2005.year 0.0185*** 5.09 0.011 0.64 0.00583 0.61 0.0195*** 5.53 0.0119 0.67 0.0060 0.63
2006.year 0.0234*** 6.44 0.0166 0.9 0.00823 0.85 0.0240*** 6.84 0.0165 0.93 0.0083 0.87
2007.year 0.0170*** 4.84 0.0074 0.41 0.00374 0.39 0.0175*** 5.15 0.0081 0.47 0.0042 0.44
Intercept 0.0034 0.13 0.310*** 3.52 0.163*** 3.63 0.0084 0.3 0.327*** 3.78 0.179*** 3.91
x2 1730.25 453.4 963.62 887.21 399.26 901.86
p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Sargan–Hansen 42.740 52.666
statistic
p-value 0.2748 0.1047
Note: This table reports the results of the 3SLS estimation for the full sample, without (Model I) and with (Model II) interaction terms. The Sargan–
Hansen test evaluates the validity of the instruments. Cap, capital; Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky assets
and liabilities; Reg, regulatory pressure; Perf, performance; Cost_Cap, cost of capital; Extreme, exposure to extreme risk; Assy, information
INTERACTIONS BETWEEN RISK TAKING, CAPITAL,
asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol, loss volatility; Lev, leverage; Higeo, geographic diversification; Growth,
AND
growth opportunities; Deficit, deficit; Stock, stock versus mutual; Group, group affiliation; Size, size of the insurer; Year, time effects. *, **, and ***
represent statistical significance at the 10, 5, and 1 percent levels, respectively.
REINSURANCE
1025
Regarding control variables, performance appears to have a significant and positive

effect on capital, confirming pecking order theory. Because of its lower cost, profitable
firms seem to rely more on retained earnings to raise capital. The cost of capital
variable is negative and significant, which highlights the importance of this variable
on the use of capital. An increase in information asymmetry, as measured by the
volatility of ROE, drives insurers to hold more capital. To some extent, this variable
may reflect an alternative measure of risk and thereby confirms the positive
relationship between this variable and capital. The coefficients of the liquidity risk
and the deficit variables are negative and significant. We also note the presence of
significant temporal effects, which may be explained by the impact of various
macroeconomic shocks or extreme loss periods. The introduction of regulatory
pressure mechanisms in the second specification slightly alters the estimation results,
and only the coefficient on Regt1 Reinst1 is positive and statistically significant.
These results indicate that highly leveraged companies in the previous period tend to
increase their levels of reinsurance to adjust their capital levels to avoid regulatory
costs.
Table 5 also presents the estimation results for the equation of the ratio of reinsurance.
The x2 statistic is significant at the 1 percent level, which illustrates the validity of the
estimation. Positive adjustments of risk taking or capital generate an increase in the
ratio of reinsurance. Unlike the capital equation, we find that capital is not a substitute
for reinsurance. This result can be explained by the imperfect substitutability between
the two variables in the sense that reinsurance can play the same role as capital, whereas
the reverse is not the case. For example, capital does not reduce the volatility of losses.
Estimation results show also that the partial adjustment factor is not significant and
illustrate that the reinsurance policy is not adjusted to its target level. Consistent with
the argument that the economic benefits of specialization can reduce the demand for
reinsurance, the coefficient of the business mix variable is negative. Stock firms appear
to use less reinsurance than mutual firms. This result supports the idea that stock firms
have easier access to external funding and are less dependent on reinsurance than
mutual firms. Regulatory pressure does not change the pattern of results.
The last part of Table 5 presents the estimation results for the risk equation. The x2
statistic is significant at the 0.001 level. An asymmetric relationship between capital
and risk adjustments is notable. A positive capital adjustment leads to lower risk.
Although surprising, this result can be explained by the fact that capital adjustment
as a result of a stricter capital-based regulation may reduce risk taking (Mailath
and Mester, 1994). Note that a positive adjustment of reinsurance increases this risk.
This finding is consistent with the view of Shiu (2011) and Aunon-Nerin and Eling
(2008), who document that the heavy use of reinsurance leads to high risk taking. The
partial adjustment factor is not significant, which indicates that risk taking is not
adjusted to its target level. The Mix variable, which reflects the degree of
concentration of the main underwriting branches and lines of business, is positive
and significant. As the activities of the company become more concentrated, the
company’s risk taking increases in the short term. As expected, stock firms are
positively related to risk taking. The Sargan–Hansen test shows the validity of the
selected instrumental variables used to address the endogeneity issue for models I
and II.
Subsample Analysis
To reduce the effect of sample heterogeneity and the aggregation bias that it may
imply, we divide the overall sample into subsamples following the transversal factors
discussed previously. The first distinction is based on the RBC ratio, that is, the ratio of
observed capital to regulatory capital. We estimate model I by selecting two
subsamples: the bottom third of firms (low-capitalized firms) and the top third (high-
capitalized firms). The estimation results (displayed in Table 6) demonstrate that low-
capitalized insurers adjust capital, risk, and reinsurance more extensively than high-
capitalized insurers. Most coefficients of the key variables are significant and higher
for low-capitalized than for high-capitalized firms. In the former, the relationship
between capital adjustment and risk remains positive. The coefficient is significant
with a relatively high value, indicating a high sensitivity of the capital to risk.
Adjustments of reinsurance appear to have a greater impact on capital. Reinsurance
seems to be a more outstanding substitute for capital for these companies. All the
dependent variables are adjusted to their target levels, with speeds greater than those
observed in the overall sample, which is also the case for high-capitalized firms.
The second distinction made in our entire sample is based on group affiliation. Table 7
provides the results of the two estimations and shows that the methodology used in
this article and the hypotheses tested are more consistent with affiliated insurers than
with nonaffiliated insurers. First, the estimation results of the former are similar to
those from the entire sample. Second, most of the variables used in the model—and
some of those of interest—are nonsignificant for nonaffiliated insurers. This result can
be explained by the fact that internal market plays a significant role. As affiliated
insurers can better control decisions regarding the capital level, the risk level, and the
reinsurance level (Cheng and Weiss, 2012; Powell, Sommer, and Eckles, 2008), the
interactions among the three decision variables are more pronounced for affiliated
than for nonaffiliated insurers. Furthermore, the results show that reinsurance usage
for affiliated insurers seems to serve as a substitute for capital, which is not the case for
nonaffiliated insurers that have more difficulties in using reinsurance since the
monitoring costs are more important for them (Cole et al., 2011).
Table 8 provides estimation results following the size of the firm. We consider two
subsamples related to the lower-third and upper-third groups in term of size. The
empirical results from our first set of equations generally provide evidence that large
insurers tend to purchase less reinsurance due to their stronger financial ability than
small firms. This result can be explained by the fact that size may influence the
insurer’s risk appetite through its effect on investment opportunities and firm’s access
to capital. Moreover, large firms are likely to have a better allocation of risk and hold
less capital than small firms.
Table 9 distinguishes between stock and mutual insurers and sheds light on some
more interesting results. First, we confirm the links between the three variables of
interest (i.e., capital, risk, and reinsurance) in at least one way for both stock and
mutual insurers. The signs of interactions remain the same as in the first estimation
(Table 5). Second, both the capital adjustment and the reinsurance relationship with
risk are more pronounced for mutual firms than for stock insurers. This result is not in
line with the pecking order theory (c.f. the second section), but confirms the findings
TABLE 6 1028
Estimation Results Following the Capital Level
Low Cap High Cap
DCapt DReinst DRisk1t DCapt DReinst DRisk1t

THE JOURNAL
OF
RISK
4Capt – – 0.821* 2.28 0.073 0.77 – – 0.119 0.93 0.041 0.33

4Reinst 0.385*** 3.16 – – 0.150*** 3.8 0.161** 2.03 – – 0.701* 6.42
4Risk1t 1.334** 1.98 3.25*** 4.54 – – 0.543*** 2.6 0.888*** 5.43 – –
Capt1 0.260** 2.2 – – – – 0.172*** 5.51 – – – –
Reinst1 – – 0.238** 3.4 – – – – 0.0719*** 3.73 – –
AND INSURANCE
Risk1t1 – – – – 0.46* 1.65 – – – – 0.514*** 3.02

Perf 0.549 3.86 – – – – 0.503*** 7.92 – – – –
Cost_cap 0.099*** 2.25 – – – – 0.081** 2.16 – – – –
Extreme 0.082 0.61 – – – – 0.0247 0.48 – – – –
Assy 0.006 0.16 – – – – 0.178*** 3.85 – – – –
Lqt_risk 0.128*** 3 – – – – 0.04*** 3.17 – – – –
Mix – – 0.353 1.48 0.0608 1.47 – – 0.169** 1.98 0.197*** 2.84
Loss_vol – – – – 0.0017** 2.23 – – – – 0.0018 1.42
Lev – – 0.026*** 2.61 – – – – 0.005* 1.71 – –
Higeo 0.1 0.67 0.0304 0.73 0.012 0.2 0.127* 1.93
Growth 0.008 0.47 – – – – 0.0162*** 2.81 – – – –
Deficit 0.083 1.63 – – – – -0.125*** 4.85 – – – –
Stock 0.029** 2.38 0.072* 1.72 0.0038 0.37 0.003 0.4 0.0273 1.32 0.046*** 2.58
Group 0.005 0.45 0.02 0.71 0.001 0.16 0.0014 0.24 0.0117 0.9 0.015 1.34
Size 0.0035 1.35 0.007 0.93 0.0013 0.62 0.0049*** 3.08 0.008** 2.35 0.008** 2.35
2001.year 0.056 1.56 0.126** 2.08 0.057*** 4.01 0.0327*** 2.98 0.054*** 2.72 0.063*** 3.47
2002.year 0.071* 1.65 0.186*** 2.81 0.0213 0.96 0.017 1.26 0.031 1.63 0.011 0.58
2003.year 0.0023 0.13 0.045 0.93 0.0142 1.07 0.016* 1.71 0.01 1 0.63 0.0127 0.68
(Continued)
TABLE 6
Continued
Low Cap High Cap
DCapt DReinst DRisk1t DCapt DReinst DRisk1t
2004.year 0.006 0.36 0.015 0.31 0.005 0.45 0.009 1.04 0.004 0.22 0.0087 0.47
2005.year 0.001 0.05 0.025 0.5 0.008 0.59 0.014 1.42 0.013 0.67 0.0044 0.22
2006.year 0.0017 0.09 0.0103 0.21 0.0122 0.86 0.0189 2.17 0.009 0.56 0.018 1.12
2007.year 0.0172 0.76 0.005 0.1 0.008 0.5 0.0136 1.52 0.0012 0.07 0.009 0.55
Intercept 0.250** 2.94 0.294 1.12 0.038 0.67 0.196*** 4.75 0.266*** 2.86 0.210* 2.33
x2 108.67 43.63 122.95 364.45 77.13 81.34
p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
statistic
p-value 0.1392 0.2530
Hausman test x2 29.21
p-value 0.014
Note: This table reports the results of the 3SLS estimation following the buffer level. Following our sample split, an insurer is considered to be
high capitalized if RBC ratio 6.5 and low capitalized if RBC ratio 2.5 The Sargan–Hansen test evaluates the validity of the instruments. The
null hypothesis of this test is that overidentifying restrictions are valid. The null hypothesis of the Hausman test is that the difference in
coefficients between the two subsamples is not systematic. Cap, capital; Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky
assets and liabilities; Reg, regulatory pressure; Perf, performance; Cost_Cap, cost of capital; Extreme, exposure to extreme risk; Assy, information
AND
asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol, loss volatility; Lev, leverage; Higeo, geographic diversification; Growth,
growth opportunities; Deficit, deficit; Stock, stock versus mutual; Group, group affiliation; Size, size of the insurer; Year, time effects.*, **, and
*** represent statistical significance at the 10, 5, and 1 percent levels, respectively.
REINSURANCE
1029
TABLE 7 1030
Estimation Results Following Group Affiliation
Affiliated Insurers Nonaffiliated Insurers
4Capt 4Reinst 4Risk1 4Capt 4Reins 4Risk1t

THE JOURNAL
OF
RISK
4Capt – – 0.406*** 2.67 0.204*** 2.56 – – 0.435* 1.9 0.250** 2.04

4Reinst 0.195*** 2.88 – – 0.507*** 22.27 0.312 0.68 – – 0.503*** 8.25
4Risk1t 0.913*** 5.81 1.952*** 12.9 – – 0.417*** 3.15 1.702*** 5.68 – –
Capt1 0.189*** 11.71 – – – – 0.111** 2.22 – – – –
Reinst1 – – 0.0035 0.35 – – – – 0.02 0.38 – –
AND INSURANCE
Risk1t1 – – – – 0.0048 0.05 – – – – 0.158 1.08

Perf 0.697*** 18.5 – – – – 0.382*** 8.06 – – – –
Cost_cap 0.0696*** 4.82 – – – – 0.0393 1.28 – – – –
Extreme 0.0954* 1.68 – – – – 0.0256 0.23 – – – –
Assy 0.03 1.61 – – – – 0.0024 0.08 – – – –
Lqt_risk 0.03*** 8.88 – – – – 0.0273** 2.06 – – – –
Mix – – 0.379*** 3.26 0.199*** 3.09 – – 0.174 1.35 0.0558 1.14
Loss_vol – – – – 0.00007 0.21 – – – – 0.0012 0.39
Lev – – 0.00045 0.16 – – – – 0.00854 1 – –
Higeo – – 0.0296 0.47 0.0145 0.4 – – 0.0303 0.31 0.00053 0.01
Growth 0.013*** 2.88 – – – – 0.101 1.37 – – – –
Deficit 0.122*** 10.6 – – – – 0.0562 1.81 – – – –
Stock 0.0015 0.27 0.0401** 2.19 0.0210** 2.14 0.00752 1.42 0.0556 1.58 0.0259* 1.85
Group – – – – – – – – – – – –
Size 0.003*** 2.85 0.00074 0.21 0.00034 0.18 0.0033 1.44 0.00918 1.29 0.00262 0.73
2001.year 0.065*** 5.07 0.159*** 6.51 0.0818*** 7.29 0.0023 0.15 0.0374 1.28 0.0204 1.35
2002.year 0.0580*** 5.29 0.0899‘‘* 4.03 0.0463*** 3.84 0.0042 0.23 0.0636** 2.13 0.0335** 2.22
2003.year 0.0171*** 2.8 0.0455** 2.23 0.0232** 2.18 0.0181 1.26 0.0423 1.42 0.0195 1.33
(Continued)
TABLE 7
Continued
Affiliated Insurers Nonaffiliated Insurers
4Capt 4Reinst 4Risk1 4Capt 4Reins 4Risk1t
2004.year 0.0099* 1.66 0.0299 1.5 0.0152 1.46 0.00705 1 0.022 0.76 0.0101 0.7
2005.year 0.0127* 1.93 0.019 0.85 0.00961 0.83 0.00674 0.76 0.0142 0.46 0.00853 0.54
2006.year 0.0091 1.37 0.0364 1.61 0.0184 1.56 0.00195 0.23 0.0426 1.36 0.0246 1.57
2007.year 0.0106 1.63 0.0243 1.1 0.0124 1.07 0.00594 0.51 0.0412 1.35 0.0245 1.58
Intercept 0.137*** 5.75 0.208* 1.9 0.107* 1.8 0.125*** 3.96 0.247 1.38 0.0586 0.65
x2 183.60 732.88 672.46 449.84 45.14 120.50
p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
statistic
p-value 0.1005 0.1419
p-value 0.0011
Note: This table reports the results of the 3SLS estimation according to group affiliation. The Sargan–Hansen test evaluates the validity of the
instruments. The null hypothesis of the Hausman test states that the difference in coefficients between the subsamples is not systematic. Cap,
capital; Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky assets and liabilities; Reg, regulatory pressure; Perf, performance;
Cost_Cap, cost of capital; Extreme, exposure to extreme risk; Assy, information asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol,
loss volatility; Lev, leverage; Higeo, geographic diversification; Growth, growth opportunities; Deficit, deficit; Stock, stock versus mutual;
AND
Group, group affiliation; Size, size of the insurer; Year, time effects. *, **, and *** represent statistical significance at the 10, 5, and 1 percent levels,
respectively.
REINSURANCE
1031
TABLE 8 1032
Estimation Results for Small and Large Sized Firms
Small Large
4Capt 4Reinst 4Risk1t 4Capt 4Reinst 4Risk1t

THE JOURNAL
OF
RISK
4Capt – – 0.509*** 4.87 0.175* 1.72 – – 0.668* 1.7 0.174 1.4

4Reinst 0.408*** 5.26 – – 0.462*** 6.39 0.495*** 5.45 – – 0.289*** 13
4Riskt 1.568*** 6.65 1.496*** 11.38 – – 1.303** 2.85 3.529*** 4.73 – –
Capt1 0.0463 1.08 – – – – 0.0136 0.35 – – – –
Reinst1 – – 0.0791*** 5.69 – – – – 0.0095 0.68 – –
AND INSURANCE
Risk1t1 – – – – 0.0109 0.72 – – – – 0.0628 0.43

Perf 0.532*** 10.73 – – – – 0.506*** 8.98 – – – –
Cost_cap 0.0597* 3.03 – – – – 0.0578*** 3.67 – – – –
Extreme 0.00146 0.15 – – – – 0.00812 0.61 – – – –
Assy 0.0557** 2.14 – – – – 0.0679*** 4.2 – – – –
Lqt_risk 0.0239*** 2.78 – – – – 0.00508 1.56 – – – –
Mix – – 0.0254 0.84 0.0131 0.54 – – 0.0296 0.41 0.00852 0.37
Loss_vol – – – – 0.00160** 2.51 – – – – 0.000034 0.1
Lev – – 0.0112*** 4.94 – – – – 0.000763 0.48 – –
Higeo – – 0.0842* 1.74 0.0121 0.45 – – 0.0468 0.73 0.0109 0.56
Growth 0.00948 1.2 – – – – 0.0457*** 7.1 – – – –
Deficit -0.230*** 8.71 – – – – -0.0300** 2.91 – – – –
Stock 0.00162 0.22 0.0062 0.44 0.0112 0.95 0.0000195 0 0.0083 0.43 0.00262 0.44
Group 0.0103 1.38 0.00926 0.92 0.0044 0.52 0.0103 1.54 0.0266 0.95 0.00743 0.86
Size 0.0057 1.11 0.00509 0.8 0.00068 0.13 0.000663 0.39 0.00479 0.65 0.00088 0.35
2001.year 0.000726 0.06 0.01 18 0.7 0.00602 0.41 0.146** 2.98 0.365*** 4.22 0.104*** 10.73
2002.year 0.00285 0.24 0.00721 0.43 0.00307 0.21 0.152‘‘ 2.88 0.405*** 4.5 0.108*** 5.87
2003.year 0.0517*** 3.78 0.0047 0.28 0.0060 0.4 0.0332‘‘* 4.94 0.00556 0.19 0.00191 0.21
(Continued)
TABLE 8
Continued
Small Large
4Capt 4Reinst 4Risk1t 4Capt 4Reinst 4Risk1t
2004.year 0.0304** 2.32 0.0141 0.82 0.0017 0.12 0.0144* 2.27 0.00117 0.04 0.000142 0.02
2005.year 0.0391*** 2.75 0.00974 0.51 0.0106 0.63 0.0181* 2.42 0.00006 0 0.000071 0.01
2006.year 0.0311** 2.18 0.0152 0.79 0.00239 0.14 0.0341*** 4.59 0.00957 0.28 0.00295 0.28
2007.year 0.0287* 2.04 0.00574 0.3 0.0015 0.09 0.0226*** 3.68 0.00814 0.29 0.00232 0.27
Intercept 0.0818 0.79 0.125 1.15 0.0073 0.08 0.048 1.25 0.0979 0.61 0.0201 0.39
x2 301.81 180.84 58.65 500.27 156.26 178.34
p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
statistic
p-value 0.1063 0.3995
p-value 0.0000
Note: This table reports the results of the 3SLS estimation according to firm size. The Sargan-Hansen test evaluates the validity of the instruments.
The null hypothesis of the Hausman test states that the difference in coefficients between the subsamples is not systematic. Cap, capital; Reins,
reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky assets and liabilities; Reg, regulatory pressure; Perf, performance; Cost_Cap,
cost of capital; Extreme, exposure to extreme risk; Assy, information asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol, loss
volatility; Lev, leverage; Higeo, geographic diversification; Growth, growth opportunities; Deficit, deficit; Stock, stock versus mutual; Group,
AND
group affiliation; Size, size of the insurer; Year, time effects. *, **, and *** represent statistical significance at the 10, 5, and 1 percent levels,
respectively.
REINSURANCE
1033
TABLE 9 1034
Estimation Results for Stock and Mutual Firms
Stock Mutual
4Capt 4Reinst 4Riskt 4Capt 4Reinst 4Riskt

THE JOURNAL
OF
RISK
4Capt – – 0.595*** 4.59 0.334*** 3.67 – – 0.194 0.7 0.0746 0.70

4Reinst 0.246*** 5.52 – – 0.604*** 15.07 0.371*** -6.88 – – 0.377*** 21.5
4Reinst 0.468*** 3.91 1.625*** 13.29 – – 0.334** 2.21 2.669*** 6.81 – –
Capt1 0.0562** 2.42 – – – – 0.164*** 13.94 – – – –
Reinst1 – – 0.0213* 1.69 – – – – 0.0061 0.58 – –
AND INSURANCE
Risk1t1 – – – – 0.00992 0.89 – – – – 0.002 0.03

Perf 0.534*** 16.22 – – – – 0.569*** 19.24 – – – –
Cost_cap 0.0594*** 4.01 – – – – 0.002 0.15 – – – –
Extreme 0.00626 0.17 – – – – 0.012 1.24 – – – –
Assy 0.0555*** 3.47 – – – – 0.0542*** 3.76 – – – –
Lqt_risk 0.0146*** 4.47 – – – – 0.089*** 13.99 – – – –
Mix – – 0.270*** 3.91 0.204*** 4.95 – – 0.0302 1.07 0.011 0.98
Loss_vol – – – – 0.000262 0.63 – – – – 0.0001 0.18
Lev – – 0.00296 1.56 – – – – 0.00035 0.2 – –
Higeo – – 0.0104 0.22 0.000342 0.01 – – 0.00392 0.04 0.0007 0.02
Growth 0.0277*** 8.41 – – – – 0.082*** 8.38 – – – –
Deficit 0.115*** 12.06 – – – – 0.101*** 10.48 – – – –
Stock – – – – – – – – – – – ––
Group 0.00384 1.2 0.0147 1.33 0.00743 0.98 0.0074** 3.21 0.001 0.05 0.0009 0.01
Size 0.000859 0.79 0.0026 0.9 0.00208 1.06 0.0020** 3.13 0.0009 0.16 0.0004 0.17
2001.year 0.0236*** 3.3 0.0738*** 4.3 0.0471*** 4.22 0.0086 0.68 0.156*** 3.76 0.0582*** 4.55
2002.year 0.0257*** 3.85 0.0561*** 3.38 0.0348‘‘* 3.23 0.0363** 2.66 0.192*** 4.64 0.0719*** 5.12
2003.year 0.0254*** 5.4 0.0259* 1.67 0.0152 1.41 0.0338*** 7.7 0.00479 0.14 0.0017 0.13
(Continued)
TABLE 9
Continued
Stock Mutual
4Capt 4Reinst 4Riskt 4Capt 4Reinst 4Riskt
2004.year 0.0142*** 3.1 0.0115 0.75 0.00633 0.6 0.0217*** 5.54 0.013 0.39 0.00501 0.38
2005.year 0.0167*** 3.46 0.00749 0.45 0.0037 0.32 0.0192*** 4.29 0.00224 0.06 0.00103 0.07
2006.year 0.0212*** 4.41 0.0119 0.7 0.00588 0.5 0.0221*** 5.1 0.0143 0.36 0.00559 0.36
2007.year 0.0160*** 3.42 0.00728 0.44 0.0038 0.33 0.0185*** 4.25 0.00734 0.18 0.00261 0.17
Intercept 0.0292 0.96 0.037 0.59 0.0221 0.52 0.143*** 8.88 0.0295 0.21 0.0122 0.22
x2 968.80 236.53 368.14 1604.26 59.41 612.88
p-value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
statistic
p-value 0.1544 0.7121
p-value 0.0000
Note: This table reports the results of the 3SLS estimation according to the organizational form. The Sargan–Hansen test evaluates the validity of
the instruments. The null hypothesis of the Hausman test states that the difference in coefficients between the subsamples is not systematic. Cap,
capital; Reins, reinsurance; Risk1, volatility of assets and liabilities; Risk2, risky assets and liabilities; Reg, regulatory pressure; Perf, performance;
Cost_Cap, cost of capital; Extreme, exposure to extreme risk; Assy, information asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol,
loss volatility; Lev, leverage; Higeo, geographic diversification; Growth, growth opportunities; Deficit, deficit; Stock, stock versus mutual;
AND
Group, group affiliation; Size, size of the insurer; Year, time effects. *, **, and *** represent statistical significance at the 10, 5, and 1 percent levels,
respectively.
REINSURANCE
1035
of Cummins, Phillips, and Smith (2001) that mutual firms are more risk averse and
respond more rapidly to capital and regulatory requirements and tend to manage
risk through reinsurance. We also find empirical support for the organizational
form theory; in other words, mutual companies have more limited access to
external capital, so they have a greater demand for reinsurance than their stock
counterparts. As shown by a Hausman test, all differences between the two
subsamples in Tables 6–9 are systematic.
Robustness Checks
In this section, we conduct additional robustness tests covering other aspects of the
empirical analysis, such as the risk measurement, the sample structure, and the choice
of instrumental variables. It is possible that the previous results outlined regarding
the interactions between the key variables are sensitive to the construction of the risk
measure. We provide new estimates of the model using the alternative risk measure
(Risk2) defined above and reported in Table 10. We confirm the interactions initially
obtained, which remain significant in all cases and have the same sign.
The second robustness test is related to the sample structure. Indeed, we use an
unbalanced panel such that the final sample contains missing years that may affect the
quality of the results. This consideration is particularly important when there are
several consecutive years missing in the sample. To resolve this issue, we introduce a
data screen requiring that an insurer appears in the sample for at least k consecutive
periods. This condition was not applied in the original sample. We find that when k is
up to 5, all interactions between capital, risk, and reinsurance remain significant.
Similarly, when k is greater than 5 (i.e., 5 consecutive years), some of the relationships
become insignificant because of information loss. A third robustness test concerns the
choice of instrumental variables, a central step in the estimation process using the
3SLS method. Therefore, we reestimate the model with other instruments, including
those proposed by Shim (2010), such as the deviation of earnings over the medium
term, the size over the last 5 years, and the average loss. Overall, we find that the
results remain stable and significant.
CONCLUSION
This article analyzes interactions between capital, reinsurance, and risk taking for
property–liability insurance firms. Our main purpose is to determine the nature of
adjustments between these decision variables. Such issues are of particular
importance to regulators interested in how insurers have responded to prudential
rules. To this end, we estimate a simultaneous equations model to identify the
potential links between these variables while controlling for endogeneity.
Empirical results based on a sample of U.S. property–liability insurance firms are
consistent with our theoretical hypotheses that predict the existence of significant
relationships between the variables of interest, supporting the view that they are
jointly determined. In their efforts to maximize firm value, insurers simultaneously
adjust various decision variables. The adjustments between risk and capital are
positive. This result, found only in one direction, provides interesting insights into the
effects of regulation on insurer behavior. By contrast, reinsurance is negatively
TABLE 10
Estimation Results of the Simultaneous Equations Using the Alternative Risk Measure
4Capt 4Reinst 4Risk2t
Coef t-stat Coef t-stat Coef t-stat

4Capt – – 0.745** 3.8 0.277** 3.26
4Reinst 0.868*** 5.25 – – 0.362*** 8.74
4Risk2t 4.052** 2.88 1.713** 2.52 – –
Capt1 0.176* 1.92 – – – –
Reinst1 – – 0.0377 2.05 – –
4Risk2t1 – – – – 0.354 1.44
Perf 1.185*** 4.95 – – – –
Cost_cap 0.190*** 3.5 – – – –
Catastrophe 0.0444 0.57 – – – –
Assy 0.128** 2.27 – – – –
Lqt_risk 0.0434 0.79 – – – –
Mix – – 0.0033 0.17 0.0009 0.1 1
Loss_vol – – – – 0.00018 0.26
Lev – – 0.0007 0.17 – –
Higeo – – 0.0476 1.5 0.0208 1.13
Growth 0.017 1.18 – – – –
Deficit 0.244*** 3.9 – – – –
Stock 0.0093 0.58 0.0043 0.28 0.0018 0.28
Group 0.0114 0.71 0.0105 0.66 0.0011 0.17
Size 0.0027 0.64 0.0021 0.54 0.0054 1.37
2001.year 0.646*** 2.85 0.261** 2.35 0.159*** 15.2
2002.year 0.676*** 2.85 0.287** 2.48 0.107** 2.46
2003.year 0.0146 0.63 0.0225 0.97 0.0096 0.96
2004.year 0.0015 0.07 0.0236 1.03 0.0088 0.9
2005.year 0.0101 0.39 0.0151 0.59 0.0065 0.6
2006.year 0.0067 0.26 0.0285 1.1 0.0113 1.01
2007.year 0.0082 0.33 0.0161 0.64 0.0066 0.61
Intercept 0.141 1.12 0.0833 1.02 0.115 1.5
x2 297.21 57.92 1147.37
p-value 0.000 0.003 0.000
Sargan–Hansen statistic 27.36
p-value 0.125
Note: This table reports the results of the 3SLS estimation using the alternative risk measure. The
Sargan-Hansen test evaluates the validity of the instruments. Cap, capital; Reins, reinsurance;
Risk1, volatility of assets and liabilities; Risk2, risky assets and liabilities; Reg, regulatory
pressure; Perf, performance; Cost_Cap, cost of capital; Extreme, exposure to extreme risk;
Assy, information asymmetry; Mix, business mix; Lqt_risk, liquidity risk; Loss_vol, loss
volatility; Lev, leverage; Higeo, geographic diversification; Growth, growth opportunities;
Deficit, deficit; Stock, stock versus mutual; Group, group affiliation; Size, size of the insurer;
Year, time effects. *, **, and *** represent statistical significance at the 10, 5, and 1 percent levels,
respectively.
associated with capital, for which it acts as a substitute. The capital ratio is slowly
adjusted to its target level. Furthermore, the results show that for low-capitalized
insurers, capital, risk, and reinsurance adjustments are more extensive than for their
high-capitalized counterparts. Interactions between the variables of interest are also
sensitive to group affiliation, size, and organizational form.
Regulatory pressures play an active role in moderating risk taking and capital to
acceptable levels. Adjustments depend on the level of capital held in excess of the
minimum requirement. Low-capitalized insurers reconstruct a buffer capital by
raising equity or by reducing risk. In contrast, adequately capitalized insurers
maintain their capital reserves and increase both their risk and reinsurance levels. In
both cases, higher capital ratios may prevent moral hazard and mitigate
informational frictions between policyholders and shareholders. However, these
results cannot tell us whether insurers operate efficiently.
APPENDIX
The observed changes in the insurer’s capital, reinsurance, and risk-taking levels are
the sum of two components, a discretionary adjustment and a change caused by
factors exogenous to the insurer. Formally, the three equations are as follows:
DCapi;t ¼ wCapDCapdi;t1 þ vi;t ðA1Þ
DReinsi;t ¼ fReins DReinsdi;t1 þ ui;t ðA2Þ
DRiski;t ¼ fRisk DRiskdi;t1 þ ei;t ; ðA3Þ
where DCapi;t , DReinsi;t , and DRiski;t are the observed changes in capital, reinsurance,
and risk levels, respectively, for insurer i at time t. The variables DCapdi;t1 , DReinsdi;t1 ,
and DRiskdi;t1 represent discretionary adjustments, whereas vi;t , ui;t , and ei;t are
exogenous adjustments of capital, reinsurance, and risk, respectively. The partial
adjustment model supposes that insurers may not be able to adjust their desired
capital, risk, and reinsurance levels instantaneously. Thus, discretionary changes are
proportional to the differences between target levels and the levels observed in the
last period:

DCapdi;t ¼ fCap Capi;t Capi;t1

DReinsdi;t ¼ fRein Reinsi;t Reinsi;t ðA5Þ

DRiskdi;t ¼ fRisk Riski;t Riski;t1 ðA6Þ
The coefficients fCap , fReins , and fRisk measure the speed of adjustments. Capi;t ,
Reinsi;t , and Riski;t are target capital, risk, and reinsurance levels, respectively.
Substituting Equations (A4), (A5), and (A6) into (A1), (A2), and (A3) yields the
following:

DCapdi;t ¼ fCap Capi;t Capi;t1 ðA7Þ

DReinsi;t ¼ fReins Reinsi;t Reinsi;t1 þ ui;t ðA8Þ

DRiski;t ¼ fRisk Riski;t Riski;t1 þ ei;t : ðA9Þ
The target levels of capital, risk, and reinsurance are not observable. As documented
in Flannery and Rangan (2006), Rime (2001), and Shim (2010), these target levels may
differ across insurers or over time and depend on the insurer’s specific characteristics.
Thus, the target levels of capital, risk, and reinsurance can be written as follows:
Capi;t ¼ r:Xi;t ðA10Þ
Reinsi;t ¼ v:Yi;t ðA11Þ
Riski;t ¼ u:Zi;t : ðA12Þ
Substituting Equations (A13), (A14), and (A15) into (A10), (A11), and (A12), we obtain
this final form of the structural model:
DCapi;t ¼ fCap Capi;t1 þ fCap r:Xi;t þ vi;t ðA13Þ
DReinsi;t ¼ fReins Reinsi;t1 þ fReins v:Yi;t þ ui;t ðA14Þ
DRiski;t ¼ fRisk Riski;t1 þ fRisk u:Zi;t þ ei;t : ðA15Þ
REFERENCES
Adams, M., 1996, The Reinsurance Decision in Life Insurance Firms: An Empirical
Test of the Risk-Bearing Hypothesis, Accounting and Finance, 36: 15-30.
Adiel, R., 1996, Reinsurance and the Management of Regulatory Ratios and Taxes in
the Property-Casualty Insurance Industry, Journal of Accounting and Economics, 22(1-
3): 207-240.
Armstrong, J., and D. M. Dror, 2006, Do Micro Health Insurance Units Need Capital or
Reinsurance? A Simulated Exercise to Examine Different Alternatives, The Geneva
Papers, 31: 739-761.
Aunon-Nerin, D., and P. Ehling, 2008, Why Firms Purchase Property Insurance,
Journal of Financial Economics, 90: 298-312.
Ayuso, J., D. Perez, and J. Saurina, 2004, Are Capital Buffers Pro-cyclical? Evidence
From Spanish Panel Data, Journal of Financial Intermediation, 13: 249-264.
Baltagi, B. H., 2005, Econometric Analysis of Panel Data, 3rd edition (New York: John
Wiley. & Sons).
Baranoff, E. G., and T. W. Sager, 2002, The Relations Among Asset Risk, Product Risk,
and Capital in the Life Insurance Industry, Journal of Banking and Finance, 26(6): 1181-
1197.
Barclay, M., and C. Smith, 1995, The Maturity Structure of Corporate Debt, Journal of
Finance, 50(2): 609-631.
Carayannopoulos, P., and M. Kelly, 2004, Determinants of Capital Holdings:
Evidence From the Canadian Property/Casualty Insurance Industry, Journal of
Insurance Regulation, 23(2): 45-65.
Cheng, J., and M. A. Weiss, 2012, The Impact of RBC Requirements in Property-
Liability Insurance, Working Paper, Shanghai Jiao Tong University.
Cole, C. R., E. He, K. A. McCullough, A. Semykina, and D. W. Sommer, 2011, An
Empirical Examination of Stakeholder Groups as Monitoring Sources in Corporate
Governance, Journal of Risk and Insurance, 78: 703-730.
Cole, C. R., and K. A. McCullough, 2006, A Reexamination of the Corporate Demand
for Reinsurance, Journal of Risk and Insurance, 73(1): 169-192.
Cummins, J. D., 1988, Risk-Based Premiums for Insurance Guaranty Funds, Journal of
Finance, 43: 823-839.
Cummins, J. D., G. Dionne, R. Gagne, and A. Nouira, 2008, The Costs and Benefits of
Reinsurance, Working Paper, Temple University.
Cummins, J. D., Z. Feng, and M. A. Weiss, 2012, Reinsurance Counterparty
Relationships and Firm Performance In the U.S. Property-Liability Insurance
Industry, Working Paper, Temple University.
Cummins, J. D., and G. P. Nini, 2002, Optimal Capital Utilization by Financial Firms:
Evidence From the Property-Liability Insurance Industry, Journal of Financial
Services Research, 21: 15-53.
Cummins, J. D., R. D. Phillips, and S. D. Smith, 2001, Derivatives and Corporate Risk
Management: Participation and Volume Decisions in the Insurance Industry,
Journal of Risk and Insurance, 68(1): 51-91.
Cummins, J. D., and D. W. Sommer, 1996, Capital and Risk in Property-Liability
Insurance Markets, Journal of Banking and Finance, 20: 1069-1092.
De Bie, T., and L. De Haan, 2007, Market Timing and Capital Structure: Evidence for
Dutch Firms, De Economist, 155(2): 183-206.
De Ceuster, M., and N. Masschelein, 2003, Regulating Banks Through Market
Discipline: A Survey of the Issues, Journal of Economic Surveys, 17: 749-766.
De Haan, L., and J. Kakes, 2010, Are Non-Risk Based Capital Requirements for
Insurance Companies Binding? Journal of Banking and Finance, 34(7): 1618-1627.
Deng, S., E. Elyasiani, and C. X. Mao, 2007, Diversification and Cost of Debt of Bank
Holding Companies, Journal of Banking and Finance, 31: 2453-2473.
Dionne, G., and T. Triki, 2004, On Risk Management Determinants: What Really
Matters? Working Paper, HEC Montreal.
Eling, M., and I. Holzm€ uller, 2008, An Overview and Comparison of Risk-Based
Capital Standards, Journal of Insurance Regulation, 26(4): 31-60.
Eling, M., and F. Schuhmacher, 2007, Does the Choice of Performance Measure
Influence the Evaluation of Hedge Funds, Journal of Banking and Finance, 31: 2632-
2647.
Estrella, A., 2004, The Cyclical Behaviour of Optimal Bank Capital, Journal of Banking
and Finance, 28(6): 1469-1498.
Fama, E. F., and K. R. French, 2002, Testing Trade-Off and Pecking Order Predictions
About Dividends and Debt, Review of Financial Studies, 15: 1-33.
Fonseca, A. R., and F. Gonzalez, 2010, How Bank Capital Vary Across Countries: The
Influence of Cost of Deposits, Market Power and Bank Regulation, Journal of Banking
and Finance, 34: 892-902.
Froot, K., D. Scharstein, and J. Stein, 1993, Risk Management: Coordinating Corporate
Investment and Financing Policies, Journal of Finance, 48(5): 1629-1658.
Garven, J. R., and J. Lamm-Tennant, 2003, The Demand for Reinsurance: Theory and
Empirical Tests, Insurance and Risk Management, 71: 217-238.
Graham, J. R., and D. A. Rogers, 2002, Do Firms Hedge in Response to Tax Incentives,
Journal of Finance, 57: 815-839.
Grubisic, E., and D. Leadbetter, 2007, The Determinants of Capital in the P&C
Insurance Industry, Working Paper, ARIA Annual Meeting, Quebec City.
Harrington, S. E., and G. Niehaus, 2002, Capital Structure Decisions in the Insurance
Industry: Stocks Versus Mutuals, Journal of Financial Services Research, 21: 145-163.
Hoerger, T. J., F. A. Sloan, and M. Hassan, 1990, Loss Volatility, Bankruptcy, and the
Demand for Reinsurance, Journal of Risk and Uncertainty, 3: 221-245.
Hovakimian, A., T. Opler, and S. Titman, 2001, The Debt-Equity Choice, Journal of
Financial and Quantitative Analysis, 36: 1-24.
Hughes, J. P., W. W. Lang, and L. J. Mester, 1996, Safety in Number? Geographic
Diversification and Bank Insolvency Risk, Working Paper, Federal Reserve Bank of
Philadelphia.
Jacques, K., and P. Nigro, 1997, Risk-Based Capital, Portfolio Risk, and Bank Capital:
A Simultaneous Equation Approach, Journal of Economics and Business, 49: 533-
547.
Jokipii, T., and A. Milne, 2008, The Cyclical Behaviour of European Bank Capital
Buffers, Journal of Banking and Finance, 32(8): 1440-1451.
Jokipii, T., and A. Milne, 2011, Bank Capital Buffer and Risk Adjustment Decisions,
Journal of Financial Stability, 7(3): 165-178.
MacMinn, R. D., 1987, Insurance and Corporate Risk Management, Journal of Risk and
Insurance, 54: 658-677.
Mailath, G. J., and L. J. Mester, 1994, A Positive Analysis of Bank Closure, Journal of
Financial Intermediation, 3: 272-299.
Mayers, D., and C. W. Smith, 1990, On the Corporate Demand for Insurance: Evidence
From the Reinsurance Market, Journal of Business, 63: 19-40.
Meyer, T., P. Cornelius, C. Diller, and D Guennoc, 2013, Mastering Illiquidity: Risk
Management for Portfolios of Limited Partnership Funds (New York: John Wiley &
Sons).
Meyers, G. G., 1989, Analysis of the Capital Structure of an Insurance Company,
Proceedings of the Causality Actuarial Society, 76: 147-171.
Morgan, D. P., 2002, Rating Banks: Risk and Uncertainty in an Opaque Industry,
American Economic Review, 92: 874-888.
Park, C. W., and M. Pincus, 2001, Internal Versus External Equity Funding Sources
and Earnings Response Coefficients, Review of Quantitative Finance and Accounting,
16(1): 33-52.
Plantin, G., 2006, Does Reinsurance Need Reinsurers, Journal of Risk and Insurance,
73: 153-168.
Pottier, S. W., and D. W. Sommer, 2006, Opaqueness in the Insurance Industry: Why
Are Some Insurers Harder to Evaluate Than Others? Risk Management and Insurance
Review, 9: 149-163.
Powell, L. S., and D. W. Sommer, 2007, Internal Versus External Capital Markets in the
Insurance Industry: The Role of Reinsurance, Journal of Financial Services Research,
31(2): 173-189.
Powell, L. S., D. W. Sommer, and D. L. Eckles, 2008, The Role of Internal Capital
Markets in Financial Intermediaries: Evidence From Insurer Groups, Journal of Risk
and Insurance, 75: 439-461.
Rajan, R. G., and L. Zingales, 1995, What Do We Know About Capital Structure Some
Evidence From International Data, Journal of Finance, 50(5): 1421-1460.
Rime, B., 2001, Capital Requirements and Bank Behaviour: Empirical Evidence for
Switzerland, Journal of Banking and Finance, 25: 1137-1178.
Saunders, A., and M. M. Cornett, 2007, Financial Institutions Management: A Risk
Management Approach, 6th edition (New York: McGraw-Hill).
Scannella, E., 2012, Capital Regulation and Italian Banking System: Theory and
Empirical Evidence, International Journal of Economics and Finance, 4(2): 31-43.
Scordis, N., and P. Steinorth, 2012, Value From Hedging Risk with Reinsurance,
Journal of Insurance Issues, 35(2): 210-231.
Shim, J., 2010, Capital-Based Regulation, Portfolio Risk and Capital Determination:
Empirical Evidence From the US Property-liability Insurers, Journal of Banking and
Finance, 34: 2450-2461.
Shiu, Y. M., 2011, Reinsurance and Capital Structure: Evidence From the United
Kingdom Non-Life Insurance Industry, Journal of Risk and Insurance, 78(2): 475-494.
Shortridge, R., and S. Avila, 2004, The Impact of Institutional Ownership on the
Reinsurance Decision, Risk Management and Insurance Review, 7(2): 93-106.
Shrieves, R., and D. Dahl, 1992, The Relationship Between Risk and Capital in
Commercial Banks, Journal of Banking and Finance, 16: 439-457.
Smith, C. W., 1977, Alternative Methods for Raising Capital: Rights Versus
Underwritten Offerings, Journal of Financial Economics, 5: 273-307.
Stulz, R. M., 1996, Rethinking Risk Management, Journal of Applied Coporate Finance,
9(3): 8-24.
Titman, S., and R. Wessels, 1988, The Determinants of Capital Structure Choice,
Journal of Finance, 43: 1-19.
Warner, J. B., 1977, Bankruptcy Costs: Some Evidence, Journal of Finance, 32: 337-347.
Zanjani, G., 2002, Pricing and Capital Allocation in Catastrophe Insurance, Journal of
Financial Economics, 65(2): 283-305.
Zou, H., and M. B. Adams, 2008, Debt Capacity, Cost of Debt, and Corporate
Insurance, Journal of Financial and Quantitative Analysis, 43: 433-466.
Copyright of Journal of Risk & Insurance is the property of Wiley-Blackwell and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for
individual use.

Document PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Document PDF

Uploaded by

Copyright:

Available Formats

© 2015 The Journal of Risk and Insurance. Vol. 83, No. 4, 1007–1043 (2016).

INTERACTIONS BETWEEN RISK TAKING, CAPITAL, AND

THEORETICAL BACKGROUND AND RESEARCH HYPOTHESIS

Interactions Between Capital and Risk

H1A: Capital and risk adjustments are positively related.

Interactions Between Capital and Reinsurance

Interactions Between Reinsurance and Risk

Impact of Transversal Factors

Group Afﬁliation. Group membership is a potential factor that may affect

prefer internal sources of funding (such as retained earnings) to external financing

Cost of Capital. The cost of holding capital is an important determinant of capital

Information Symmetry. Signal theory suggests that information asymmetry along

Growth Opportunities. Insurers with better growth opportunities are expected to

Determinants of Risk Taking

Geographic Concentration. An insurer may reduce its overall risk by holding a

Leverage. Highly leveraged insurers are exposed to a higher likelihood of

Geographic Concentration. Geographical concentration reflects the degree of

DCapi;t ¼ l0 fCap Capi;t 1 þ l 1 DReinsi;t þ l2 DRiski;t þ l3 Perfi;t þ l4 Cost capi;t

DReinsi;t ¼ b0 fReins Reinsi;t 1 þ b1 DCapi;t þ b2 DRiski;t þ b3 Levi;t þ b4 Mixi;t

DRiski;t ¼ g 0 fRisk Riski;t 1 þ g 1 DReinsi;t þ g 2 DCapi;t þ g 3 Loss voli;t þ g 4 Mixi;t

Liquidity risk Lqt_risk Ratio of liabilities to (þ) (þ) Endogenous

include the regulatory pressure mechanisms. Following Shim (2010), we introduce

Descriptive Statistics and Correlation Analysis

Standard 5 Percent 95 Percent

significant correlation between endogenous variables, which confirms our choice to

8. Assy 0.238* 0.017 0.022 0.0048 0.1704* 0.400* 0.0736* 1

Eq DCap Eq DReins Eq DRisk1

DCapt DReinst DRisk1t 4Capt 4Reinst 4Risk1t

DCapt – – 0.556*** 5.91 0.276*** 3.7 – – 0.553*** 4.03 0.279*** 3.74

Risk1t1 – – – – 0.0029 0.44 – – – – 0.0039 0.53

DCapt DReinst DRisk1t 4Capt 4Reinst 4Risk1t

Regarding control variables, performance appears to have a significant and positive

Low Cap High Cap

DCapt DReinst DRisk1t DCapt DReinst DRisk1t

4Capt – – 0.821* 2.28 0.073 0.77 – – 0.119 0.93 0.041 0.33

Risk1t1 – – – – 0.46* 1.65 – – – – 0.514*** 3.02

Low Cap High Cap

DCapt DReinst DRisk1t DCapt DReinst DRisk1t

Affiliated Insurers Nonaffiliated Insurers

4Capt 4Reinst 4Risk1 4Capt 4Reins 4Risk1t

4Capt – – 0.406*** 2.67 0.204*** 2.56 – – 0.435* 1.9 0.250** 2.04

Risk1t1 – – – – 0.0048 0.05 – – – – 0.158 1.08

Affiliated Insurers Nonaffiliated Insurers

4Capt 4Reinst 4Risk1 4Capt 4Reins 4Risk1t

4Capt 4Reinst 4Risk1t 4Capt 4Reinst 4Risk1t

4Capt – – 0.509*** 4.87 0.175* 1.72 – – 0.668* 1.7 0.174 1.4

Risk1t1 – – – – 0.0109 0.72 – – – – 0.0628 0.43

4Capt 4Reinst 4Risk1t 4Capt 4Reinst 4Risk1t

4Capt 4Reinst 4Riskt 4Capt 4Reinst 4Riskt

4Capt – – 0.595*** 4.59 0.334*** 3.67 – – 0.194 0.7 0.0746 0.70

Risk1t1 – – – – 0.00992 0.89 – – – – 0.002 0.03

4Capt 4Reinst 4Riskt 4Capt 4Reinst 4Riskt

4Capt 4Reinst 4Risk2t

Coef t-stat Coef t-stat Coef t-stat

DCapi;t ¼ wCapDCapdi;t1 þ vi;t ðA1Þ

DReinsi;t ¼ fReins DReinsdi;t1 þ ui;t ðA2Þ

DRiski;t ¼ fRisk DRiskdi;t1 þ ei;t ; ðA3Þ

Capi;t ¼ r:Xi;t ðA10Þ

Reinsi;t ¼ v:Yi;t ðA11Þ

Riski;t ¼ u:Zi;t : ðA12Þ

DCapi;t ¼ fCap Capi;t1 þ fCap r:Xi;t þ vi;t ðA13Þ

DReinsi;t ¼ fReins Reinsi;t1 þ fReins v:Yi;t þ ui;t ðA14Þ

DRiski;t ¼ fRisk Riski;t1 þ fRisk u:Zi;t þ ei;t : ðA15Þ

DCapt – – 0.556* 5.91 0.276* 3.7 – – 0.553* 4.03 0.279* 3.74

4Capt – – 0.406* 2.67 0.204* 2.56 – – 0.435* 1.9 0.250** 2.04

4Capt – – 0.595* 4.59 0.334* 3.67 – – 0.194 0.7 0.0746 0.70

Capi;t ¼ r:Xi;t ðA10Þ

Reinsi;t ¼ v:Yi;t ðA11Þ

Riski;t ¼ u:Zi;t : ðA12Þ