Does Insurance Promote Economic Growth?

Evidence from OECD Countries

Author(s): Damian Ward and Ralf Zurbruegg
Source: The Journal of Risk and Insurance, Vol. 67, No. 4 (Dec., 2000), pp. 489-506
Published by: American Risk and Insurance Association
Stable URL: http://www.jstor.org/stable/253847 .
Accessed: 04/06/2014 19:24

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

American Risk and Insurance Association is collaborating with JSTOR to digitize, preserve and extend access
to The Journal of Risk and Insurance.

http://www.jstor.org

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
?DTheJournalof Risk and Insurance, 2000, Vol. 67, No. 4, 489-506

PROMOTEECONOMIC
DOES INSURANCE GROWTH?
EVIDENCEFROMOECD COUNTRIES
Damian Ward
Ralf Zurbruegg

ABSTRACT
This articleexamines the short- and long-run dynamic relationshipsexhib-
ited between economic growth and growth in the insurance industry for
nine OECDcountries.This is achieved by conductinga cointegrationanaly-
sis on a unique set of annual data for real GDP and total real premiums
issued in each countryfrom 1961to 1996.Causalitytests arealso conducted,
which account for long-run trends within the data. The results from the
tests suggest that in some countries,the insuranceindustry Grangercauses
economic growth, and in other countries,the reverseis true. Moreover,the
results indicate that these relationshipsarecountryspecific and any discus-
sion of whether the insuranceindustry does promoteeconomic growth will
be dependent on a number of national circumstances.

INTRODUCTION
The objective of this article is to examine the causal relationship that potentially ex-
ists between growth in the insurance industry and economic growth. Moreover, by
recognizing that the wider economic benefits of insurance are conditioned by na-
tional regulations, economic systems and culture, it is argued that an examination of
the inter-relationship between insurance and economic growth needs to be conducted
on a per country basis. Such a study is important because in contrast to the available
evidence on the economic importance of banks, typified by the work of Levine and
Zervos (1998), little is known about insurance. Given that insurance not only facili-
tates a myriad of economic transactions through risk transfer and indemnification
but is also seen to promote financial intermediation, it is surprising that rigorous, in-
depth research of this kind is not more prominent.
Despite the apparent lack of literature on the role of insurance, the work by Outreville
(1990, 1996) is notable for identifying links between an economy's financial develop-
ment and insurance market development. Unfortunately, much as Outreville's work
is a substantial empirical contribution to academics' understanding of the link be-
tween the economy and the insurance industry, a number of issues still remain unre-
DamianWardis a lecturerin economics at the University of BradfordManagementCentre,
Bradford,West Yorkshire,U.K. Ralf Zurbruegg is a senior lecturer in finance, School of
Bankingand Finance,University of New South Wales,Sydney,Australia.The authorswish
to thank two anonymous refereesfor helpful comments.
489

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
490 THEJOURNALOF RISKAND INSURANCE

solved. First, as Patrick (1966) discusses, economic growth can be either supply-led
through growth in financial development, or alternatively, financial development can
be demand-led through growth in the economy. This aspect of understanding the
causal relationships between insurance market growth and economic development
is still lacking.' Second, while Outreville was rightly concerned with understanding
the importance of insurance within the development process, his findings are limited
to the developing economies of the world and may not be applicable to the devel-
oped world. Finally, in recent advances in the macroeconomic growth literature, such
authors as Arestis and Demetriades (1997), Demetriades and Hussein (1996), and
Pesaran, Haque, and Sharma (2000) have pointed out that it is important to accom-
modate the potential for causal relationships to differ in size and direction across
countries. For if this heterogeneity is not acknowledged, then any results will only be
applicable to the average country and in consequence one would be left to question
the relevance or existence of a representative country within the analysis.
Importantly, with respect to the issue of heterogeneity, empirical and theoretical evi-
dence exists to suggest that the role of insurance in the economy may be varied across
countries. From a demand perspective, Beenstock, Dickinson, and Khajuria (1986)
and Browne and Kim (1993) found that the role of the state in providing insurance
services was a determinant of the demand for life insurance, as was the level of edu-
cation and the age dependency ratio, all of which are likely to differ across countries.
Furthermore, according to Hofstede (1995), the level of insurance within an economy
will depend on the national culture and the willingness of individuals to use insur-
ance contracts as a means of dealing with risk.
Through the recognition of these points, it is to be expected that the relationship be-
tween insurance market development and economic growth may be varied across
countries. Therefore, it is the aim of this article to advance Outreville's (1990, 1996)
work by analyzing the long- and short-run causal relationships between economic
growth and insurance market development. This is carried out on a country-by-coun-
try basis in order to allow for the scope for the causal relationships to differ in size
and direction across countries. Nine leading OECD countries are examined with real
GDP used as the measure of economic activity and real total written premiums as a
measure of insurance activity. Although it is acknowledged that total written premi-
ums aggregates life and non-life business into the same analysis, the data series for
total premiums was substantially longer than that available for life and non-life busi-
ness and therefore more suitable for an examination of causality. Nevertheless, within
the context of causality testing, aggregation of life and non-life insurance may not be
too unreasonable. If one views the key economic benefits of insurance as risk transfer,
indemnification, and financial intermediation, then the benefits of risk transfer and
indemnification are likely to be major characteristics of non-life and health insur-
ance, while financial intermediation is a primary aspect of life insurance. Thus, an
aggregated approach will embrace all of these ideas within the same analysis.
While being mindful of the data used, the results from this study do support the view
that the relationship between insurance market development and economic devel-
opment do differ across countries. There is a clear need for policy makers to be made

1 See Outreville (1990), p. 491.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE
ECONOMIC FROMOECDCOUNTRIES491
GROWTH?EVIDENCE

aware of this and for future research to verify the reasons for such diversity.
The next section of this article details the possible links between insurance and eco-
nomic development and highlights the potential for these relationships to differ across
economies. The third section then discusses the econometric methodology to be em-
ployed for the empirical results that are discussed in the fourth section. The final
section provides a summary and concluding remarks.

THE WIDER ECONOMICIMPORTANCE

OF INSURANCE

BenefitsFromRiskTransferand Indemnification
Insurance services are capable of generating significant productive impact within an
economy. For example, the offering of risk transfer and indemnification services aids
risk-averse individuals in purchasing large-expense items, such as automobiles and
real estate. This leads to insurance coverage having positive externalities in terms of
increased purchases, profits, and employment both within and alongside the insurance
sector. In addition, insurance facilitates innovation within an economy by offering to
underwrite new risks. For instance, without access to product liability insurance, firms,
particularly pharmaceuticals, would be unwilling to develop and market highly ben-
eficial products.
Further positive externalities from insurance are in its potential to reduce risk in the
economy. For example, risk-taking individuals such as smokers and dangerous auto-
mobile drivers place physical and human productive capital at risk. However, the
more risk an individual poses to the stock of accumulated capital, the higher the cost
of insurance. As the price of insurance rises, individuals face increased incentives to
modify their behavior, which hopefully provides beneficial effects for the accumu-
lated productive capital.
Unfortunately, the benefits to be derived from the purchasing and pricing of insur-
ance may be constrained by national circumstances. Following Fukuyama (1995), the
economic benefits derived from insurance are likely to be conditional on the cultural
context of a given economy. For example, insurance will offer important economic
benefits when activities are generally seen as risky and risks are optimally managed
through insurance contracts rather than by other risk control or transfer mechanisms.
In contrast, where societies are indifferent to risk, or alternative risk transfer mecha-
nisms such as the family are seen as important, then the characteristics of the social
fabric of a given society will reduce the potential benefits from insurance.
Fukuyama (1995) puts these differences down to the culturally determined levels of
trust within an economy. The U.S., U.K., Japan, and Germany are seen as high-trust
economies, which enable economic transactions to be undertaken beyond the bound-
aries of the extended family. In contrast, France and Italy are characterized as low-trust
societies, in which individuals find it difficult to transact with unknown individuals.
With particular regard to insurance, Hofstede (1995) develops a similar taxonomy by
characterizing countries as low- or high-group societies. Low-group societies with an
emphasis on the individual will tend to support market-based means of dealing with
risk such as insurance. High-group societies, with an emphasis on collective mecha-
nisms, will tend to further the role of families and government within the risk man-
agement process. The cultural context of insurance can be extended further to in-

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
492 THEJOURNALOF RISKAND INSURANCE

clude consumer reactions to taxation. In particular, where attitudes to authority, gov-

ernment, and the rule of law are weak, the willingness to indulge in tax evasion will
be high. In consequence, any incentives to purchase tax-efficient insurance products
will be much reduced. In summary, therefore, if the different socio-cultural determi-
nants of insurance purchasing decisions are taken into account, then the ability of
insurance to promote economic growth will likely vary across economies.

BenefitsFromInsuranceas a FinancialIntermediary
An additional benefit from the insurance industry stems from its activities as a finan-
cial intermediary and as such the development of the insurance market has signifi-
cant implications for the accumulation of productive capital within an economy. Sig-
nificantly, the process of capital accumulation has been seen by Pagano (1993) to
become more efficient as the development of the financial intermediary function
improves. This view can be illustrated in the first instance by the recognition that
increased competition between intermediaries will improve productive efficiency.
Yet evidence provided by Cummins, Tennyson, and Weiss (1999) and Hardwick (1997)
for life insurers both in the U.S. and U.K. suggests that productive inefficiency is both
high and persistent. Second, improved financial intermediation services enable in-
vestors to access diversified investment portfolios, which facilitates a willingness to
invest in high-risk, high-productivity projects. Furthermore, liquidity in the financial
markets is enhanced by financial intermediation that will limit the waste of economic
resources because of stakeholders' early monetary realization of asset holdings. As
investigated by Levine and Zervos (1996), Arestis and Demetriades (1997), and Levine
and Zervos (1998), liquidity aids economic growth by facilitating a smooth flow of
funds to capital-accumulating projects. Finally, insurance companies as institutional
investors in corporations not only aid capital allocation but also further enhance their
investments through increased levels of monitoring. Therefore, as examined by
Conyon and Leech (1994), institutional investors may improve the productive poten-
tial of the projects that they choose to fund.
However, it is important to temper these arguments by reference to the existence and
role of regulation in protecting policyholder interests. Unfortunately, the position and
purpose of regulation in the insurance-economic growth link is difficult to deter-
mine. Angerer (1993) discusses how regulation facilitates the transacting of insur-
ance services while at the same time constraining insurance companies' activities.
Moreover, the form of regulation differs across countries. In the U.K., for example,
prudential regulation is given the greater emphasis with little, if any, monitoring of
premium rates or coverage. Instead, the market is allowed to determine the price of
insurance. However, reliance on the market can be seen to introduce volatility into
the system. Until recently, this position was taken by Japan, which operated a "con-
voy system," in which the regulator set rates at a level that enabled the weakest com-
pany to remain solvent. However, according to Hayakawa et al. (2000), such a regu-
latory structure produced a fail-proof safety net for the industry and provided few
incentives to innovate or be cost effective. Indeed, according to Hayakawa et al. (2000),
the loss ratios for the Japanese automobile insurance sector have been persistently
below 60 percent for the last 20 years, whereas in 1994 they were 71.9 percent in the
U.K., 86.6 percent in France, and 66.9 percent in the U.S. These figures clearly high-
light the extent to which Japanese insurance companies have been protected from

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE FROMOECDCOUNTRIES493
GROWTH?EVIDENCE
ECONOMIC

competition by regulatory structure. The regulation of business activities is also likely

to influence the functioning of the insurance industry. With respect to financial inter-
mediation, LaPorta, Lopez-de-Silanes, Scheifer, and Vishney (1997,1998) have shown
that the character and origin of the legal system within an economy substantially
affect laws relating to bank credit, the enforcement of credit contracts, and the stan-
dards for corporate information disclosure. Not surprisingly, therefore, Levine, Loayza,
and Beck (2000) found that such factors influence the rate of the banking sector's
development. Although the literature on the legal system and insurance market de-
velopment is less developed, Browne, Chung, and Frees (2000) have recently shown
that a country's legal system is a significant determinant of the demand for automo-
bile and general liability insurance.
Further ambiguity can also be found in the effect that insurance market development
will have on the saving rate and the availability of investment funds. This is because
financial intermediation, through portfolio diversification, reduces risk and thereby
diminishes the need to save. This problem is further exacerbated by the improved
availability of credit, leading to a reduction in the precautionary saving motive by
improving the accessibility to third parties' liquid funds. Similar arguments can be
raised from the mere existence of insurance, as it deals with the treatment of non-
diversifiable risks. Hence, insurance itself acts as a disincentive to save because the
major determinants of the precautionary motive to save, such as death, retirement,
unemployment, loss of home, car,and even mechanical failure of domestic consumables,
can all be transferred as risks to insurance companies.
As a consequence, it is difficult to assert the form of the relationship between insur-
ance companies, financial intermediation, and economic growth. First, this is because
although developments in financial intermediation may bring about improvements
in the capital accumulation process, such gains may be outweighed by the potential
reduction in the domestic savings rate. Second, the importance of financial interme-
diation via insurance companies and in particular through stock markets is likely to
vary according to the way in which capital is traditionally raised in a particular
economy. As examined by Arestis and Demetriades (1997), where banks are the tra-
ditional source of funding, the role of stock markets in the capital accumulation
process is likely to be reduced. This point is reflected by the Comite European des
Assurance (1999) measurement of the level of insurance firms' investments as a per-
centage of GDP. For instance, in 1996 the level of investment by the French insurance
industry was 40.4 percent of GDP, Italy 12.4 percent, Switzerland 75.5 percent, and
the U.K. 91.4 percent. Third, while the role of the stock market may well be important
in a country, the position of insurance companies in the market may not be influen-
tial. Indeed, in markets with a global position, overseas financiers could play a much
more important role in the process of capital accumulation.
Finally, it is important to fully recognize the negative externalities of the insurance
industry and in particular the problem of moral hazard. In an obvious sense, the
provision of insurance alters the risk-taking behavior of the insured, which ultimately
leads to an increase in the loss rate on accumulated productive capital. In a less obvi-
ous sense, recent research by Ruser (1998) has suggested that the moral hazard effect
can be quite pronounced when employers are required to take out workers compen-
sation insurance. Such insurance is associated with longer periods of illness, the re-

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
494 THEJOURNALOF RISKAND INSURANCE

porting of more severe injuries and the apportionment of non-work-related injuries

to the workplace. While all these limit economic output, the more important aspect
of moral hazard in the context of this study is that workers compensation insurance
facilitates worker absence, which unfortunately has the potential to reduce the acqui-
sition rate of firm-specific capital. As a consequence, firms offering benefits to work-
ers under workers compensation insurance have been found by Butler, Gardner, and
Gardner (1998) to suffer lower levels of productivity. Therefore, the impact of insur-
ance on economic growth may well differ across economies depending on the design
of workers compensation schemes and the incentives to display moral hazard type
behavior, as well as the importance of firm-specific human capital in the production
process.
In consideration of the preceding discussion, when one begins to investigate the link
between insurance market development and economic development, it is evidently
important to recognize the cultural context of economic exchange, the role of regula-
tion, the relative importance of other financial intermediaries, and the effects of moral
hazard. Undoubtedly, all of these factors are likely to differ across economies and
thereby affect the form and degree of causality between insurance and economic
growth.
From a more practical standpoint, attempts to seek patterns of causality in the mac-
roeconomic growth literature using cross-sectional data were, until recently, typified
by the work of Barro (1991) and King and Levine (1993), whose indicators of eco-
nomic development were averaged over a number of decades for each cross-sec-
tional unit. These values were then regressed on initial starting values for measures
of financial development. However, problems of inference and model mis-specifica-
tion have also been associated with this approach. Pesaran et al. (2000), Demetriades
and Hussein (1996), and Arestis and Demetriades (1997) have all discussed a number
of these issues. Of most consequence, by analyzing cross-sectional units, the esti-
mated parameters are usually restricted to be constant across all countries, thereby
failing to allow the causal relationships to vary across economies.
Therefore, it is evident from the recent macroeconomic growth literature that when
assessing the importance of insurance in the process of economic growth, it is neces-
sary to recognize the heterogeneous nature of the countries under investigation. Un-
fortunately, under the cross-sectional approaches adopted by Outreville (1990,1996),
it is more difficult to identify the country-specific patterns of causality. In recognition
of this, this study employs country-specific cointegration and causality tests, as in
Demetriades and Hussein (1996) and Arestis and Demetriades (1997), in order to
understand the dynamic and causal relationships that insurance sector growth has
on overall economic growth. This is not without its own faults but will undoubtedly
allow for country-specific effects to be more readily uncovered.

METHODOLOGYAND DATA
The tests presented in the empirical section are initially based on constructing a re-
duced-form bivariate vector autoregression (VAR) to test for Granger causality. Fol-
lowing Granger (1969), one can consider two stationary series {x,}and {yj of length T
and subsequently estimate the following system of vector autoregressive (VAR) equa-
tions to test for causality:

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE
ECONOMIC FROMOECDCOUNTRIES495
GROWTH?EVIDENCE

xt = A(L)xt + B(L)yt + Glzt + EX't (1)

Yt= C(L)xt + D(L)yt + G2zA+ ?y (2)

where A(L), B(L), C(L), and D(L) are all lag polynomials with roots outside the unit
circle and z being a set of exogenous variables to the system. The regression error
terms, ?x,t, and 8y,t, are also assumed to be mutually independent and individual
processes. The test for whether y strictly Granger causes x involves a standard joint
test (F-test) on whether lagged coefficients of y have significant linear predictive power
on x. The null hypothesis is that of no linear causality, implying in Equation (1) that
the coefficients of B(L) are not jointly significantly different from zero. Similarly, in
the case of testing whether x causes y, the test will be conducted on the coefficients
contained in the lag polynomial C(L) to see whether they are jointly significantly
different from zero. If both B(L) and C(L) joint tests for significance show they are
different from zero, the series are bi-directionally related.
However, as is common with other macroeconomic time series, where the variables
are not stationary and contain cointegrative trends, a re-parameterization of the model
becomes necessary. Specifically, as Engle and Granger (1987) point out, if a linear
combination of two or more nonstationary series exists, the nonstationary series are
cointegrated, indicating a long-run equilibrium relationship between the variables.
In such cases, an error-correction term should be imposed upon a restricted VAR so
that the cointegrating behavior of the endogenous variables ensures they converge to
a long-run equilibrium. However, in the short run, a number of dynamic effects may
still take place. To allow for this, the restricted VAR can be expressed as:

p
Au/U= flu1+X>F1Aut+F+s,(3
lt1 l -r/fi + Fv, + ?ti (3)
i=l

where

p p
rL= Ei -I, r= -E Fj (4)
i=1 j=i+l

and ut = are now nonstationary I(1) variables, v, are any number of deter-
(Xt, Yt)'
ministic variables, and p is the number of lags in the polynomials. From this an error-
correction model can be derived, given that both xt and y, are integrated processes of
order 1 and a linear combination exists such that Hl = a/3' and f3'u, is stationary
where /B contains the cointegrating vector and a the adjustment parameter within
the restricted VAR. Also, with the error-correction model any tests on causality are
now performed on the first differences of the series.
The preceding discussion on the importance of accounting for stationarity in the se-
ries and any long-run relationships between them cannot be ignored if statistically
correct estimates are to be obtained. As Granger (1988) argues, a failure to include

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
496 THEJOURNALOF RISKAND INSURANCE

necessary cointegrating equations into a model can lead to erroneous results. There-
fore, in this article, when proceeding to test for causality between insurance premi-
ums and GDP where they are cointegrated, an error-correction framework will be
used. Although Sims, Stock, and Watson (1990) argue that it may not be necessary to
do so for series that contain a unit root in order to conduct causality tests, a more
common and accepted technique is to follow a restricted VAR approach with an er-
ror-correction mechanism.2 Furthermore, an error-correction model will also enable
a test for the long- and short-run dynamics between the insurance market and eco-
nomic growth, as there are now two possible avenues of causation, specifically, through
either the first differences of the lagged dynamic terms of either series, or alterna-
tively, the lagged cointegrating vector fl'u,-l for where a ? 0.

To proceed with the above tests, the following empirical section first conducts Phillips-
Perron (1988) unit root tests that can account for higher-order serial correlation in a
series.3 Once the number of unit roots has been determined, a Johansen (1991)4
cointegration trace test is conducted. From equations (3) and (4), and if each series
consists of nonstationary I(1) variables, then a test for cointegration is a test to iden-
tify the cointegrating rank, r, of the coefficient matrix II . The method to identify the
rank is to estimate an unrestricted VAR and then test to see whether the necessary
restrictions to be imposed on the reduced rank of fl are rejected. For the case of two
time series, there is the possibility of one cointegrating vector. This will lead to either
H(r) = 0 or H (r) < 1 . If a cointegrating relationship is noticed between any series, the
causality tests are based on the error-correctionmodel with the estimated cointegrating
equation imposed upon the VAR; otherwise, the VAR approach detailed in equations
(1) and (2) are applied on stationary series. The actual statistical tests then conducted
on the estimated models are (i) causality tests using the usual joint test for signifi-
cance of the lagged dynamic terms in each equation, (ii) a simple t-test, where appro-
priate, on the significance of the error-correction terms in the restricted VAR from the
cointegrating relationship (i.e., H: a = 0), and (iii) a joint test of significance for both
the error-correction component and the lagged dynamic terms.5 These tests will al-
low for a detailed examination of the relationship the insurance market has with
economic growth, both in the short term and long run.
The data used in this study come from various sources. For the insurance industry,
the value of total written premiums from the United States, United Kingdom, Aus-
tria, Australia, Canada, Switzerland, France, Italy, and Japan were graciously pro-
vided by Swiss Re. However, it is recognized that total insurance premiums may not
be an accurate measure of total insurance market output. In particular, as suggested
by Browne and Kim (1993), total insurance premiums fail to account for different

2 Also, a number of criticisms have been raised using the former approach of an unrestricted
VAR in levels where the series contain unit roots, such as by Toda and Phillips (1993), who
point out problems with the asymptotic theory underlying the process.
3Unlike other tests, such as the Augmented Dickey-Fuller procedure, this method has shown
to be robust to a wide variety of heteroskedasticity and serial correlation of unknown form.
4See also Johansen and Juselius (1990) and Johansen (1995) for further references.
IThis test also is a test for strong exogeneity as noted by Charemza and Deadman (1992).

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE
ECONOMIC FROMOECDCOUNTRIES497
GROWTH?EVIDENCE

market forces in various countries and make comparisons difficult. Furthermore, they
fail to account for regulatory effects on pricing, nor do they encapsulate total risk
transfer as would a measure of insurance coverage. However, as Beenstock et al.
(1986) argue, in order to attain some level of understanding, one has to work with
available data, while, as we suggest, recognizing its weaknesses.
In terms of economic development, real GDP was taken as an acceptable proxy. The
data for this variable were extracted from annual OECD National Accounts. In total,
35 years ranging from 1961 through 1996 were used for this study.
Because there are only 35 observations per country, the question of whether the re-
sults suffer from small sample bias may arise because the time series is not exces-
sively long. However, it is the length of the estimation period rather than the fre-
quency of observations that is paramount for examining cointegrative relationships
(see Hakkio and Rush, 1990). As such, the results should provide reliable evidence of
any long-run trends between the series under review.
Furthermore, a number of explanatory variables were also collected for possible in-
clusion into any of the testing procedures that are conducted: namely, changes in
private saving rates, the general government budget surplus, population size, the
general government level of current expenditure, and youth plus old age dependency
ratios, measured as the proportion of the total population under 16 and over 65 years
of age, respectively. These variables were chosen for their potential to influence and
increase the explanatory power of the regressions. Empirical support for the use of
these variables can be found in Beenstock et al. (1986), Browne and Kim (1993), Pesaran
et al. (2000), and Outreville (1990, 1996). The data for the various measures were
extracted from both OECD and World Development Bank data sets. Finally, when
considering whether to include any of these exogenous variables within the VAR
system, Hausman6 (1978) specification tests for exogeneity are conducted to deter-
mine whether they are endogenous to the system.

EMPIRICALRESULTS
Table 1 presents Phillips-Perron unit root test results on the levels and first differ-
ences of the logarithm of real GDP and total real premiums written for each country.
The tabulated results show that all series are nonstationary in levels but stationary in
their first differences, to at least the 5 percent significance level.
The preceding results imply that the causality tests should be conducted using the
first differences of the series to ensure stationarity. The same must apply for the addi-
tional exogenous variables to be incorporated into the estimation system. Unfortu-
nately, this precludes a number of the variables mentioned earlier. Furthermore, most
of the variables examined also showed signs that they were either endogenously re-
lated to either GDP growth or changes in real premiums or, alternatively, did not

6
The Hausman test involves obtaining fitted values from reduced form equations of the
variables that are to be tested for exogeneity, and then these are regressed within the full
estimation system specified in the model. If the coefficients associated with the fitted values
are significant, then these variables cannot be treated as exogenous.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
498 THEJOURNALOF RISKAND INSURANCE

significantly enhance the explanatory power of the regressions. The selection criteria
for choosing which variables to include were based on ensuring that the maximum
amount of explanatory power was given to the causality models without risking the
possibility of adding redundant parameters. For this reason each variable was tested
for exogeneity and then subsequently tested for whether its incorporation into the
system of equations was justified. Given the above factors, the choice of including
additional explanatory variables became very limiting. Of all the possible variables,
changes in the savings rate for each country, along with population size changes,
seemed to be most beneficial. Although for each country different variables were
sometimes found to be endogenous, and in other countries not, it was decided to
standardize the potential list of exogenous variables to those that seemed to perform
best for all countries. Table 2 only provides test results for exogeneity and signifi-
cance in explanatory power to an unrestricted VAR for population growth and changes
in savings rates. The test for significance of the exogenous variables was based on an
F-test for joint significance if both variables were not endogenous; otherwise, a simple
t-test was performed. The number of lags within the VAR was set to four, as this was
found to be the most appropriate lag length determined by log-likelihood tests.7

TABLE1
Phillip-Perron
UnitRootTests

Country Real Premiums Real GDP A Real Premiums A Real GDP

Australia -0.1684 -3.1923 -5.1421a -50145a

Austria -1.2795 -0.9475 -4.0559b _5.3531a
Canada -0.1660 -1.1504 -4.5058a -4.7508a
France -1.6372 -2.0420 -6.0841a -4.4617a
Italy -1.0742 -1.0182 -4.1492b -5.3083a
Japan -3.0558 -1.7379 -4.9856a -3.7109b
Switzerland -3.3634 -2.1168 -4.7017a -3.5762b
United Kingdom -1.9967 -2.3636 _4.7777a -3.7863b
United States 0.0251 -3.0419 -3.9846b -4.2755a
The tests above are conducted on logarithmicvalues for both series where the t-statisticis
correctedfor heteroskedasticityand autocorrelationof unknown form using the Newey-
West consistent estimate with a time trend. The number of truncationlags was three,based
on a floor function from the Newey-West estimatorand the number of observations.
aIndicatessignificance at the 99 percent confidence level.
bIndicates significance at the 95 percent confidence level.

' Not all countries had the same optimal lag length, as some showed a smaller number would
suffice. However, given previous research indicating that the Johansen cointegration tests
and subsequent error-correction models conducted in this study are sensitive to the lag
length used (see Banerjee, Dolado, Galbraith, and Hendry, 1993) but are less sensitive to an
over-parameterization than an under-parameterization (see Cheung and Lai, 1993), it was
decided to choose the most common long lag length observed for the nine countries.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE FROMOECDCOUNTRIES499
GROWTH?EVIDENCE
ECONOMIC

From the results of Table 2, it is apparent that even for these two variables, either one
or both of the series are not exogenous in all cases. Also, even when they are, there is
no general pattern of consensus across countries for when they significantly enhance
the estimation procedure of the unrestricted VAR. Of all the countries, only for Aus-
tria and France were both variables considered viable to add to the system, based on
the F-statistic being significant at the 10 percent significance level. For Italy and Ja-
pan, changes in population size were incorporated, as the series were exogenous and
did lead to a significant improvement in the model. For all other countries, none of
the variables were either exogenous or statistically important at the 10 percent sig-
nificance level to the VAR.8

TABLE2
Exogeneityand SignificanceTestsfor the ExplanatoryVariables

Dependent Real Real Real Real Real Real

Variables Premiums GDP Premiums GDP Premiums GDP
Test for Exogeneity of Test for Exogeneity of Test for Significance of
the Variable Population the Variable Savings the Exogenous Variable(s)
Country
Australia -0.0707 -1.2099 -0.4381 -1.3620 0.2625F 1.131 6F
Austria -0.1643 0.4185 -0.4066 0.7209 2.2494Fc 1.7974F
Canada -0.9940 -1.6015 -0.4918 -1.8278c -0.1134T -0.4385T
France 1.5367 -0.8715 0.5883 -1.4799 3.3915Fb 2.9433Fc
Italy 0.8946 -1.2988 -0.7592 1.7682c 1.6878Tc 1.2803T
Japan 0.7638 1.6258 -0.6731 -2.1210b 0.5043T -2.4116Tb
Switzerland 0.0983 -1 .6594c -1.1131 -2.0264b _
United Kingdom -0.0802 -2.0825b 0.1143 2.309b - -

United States -0.4133 -0.7924 -0.7253 -1.2867 0.4924F 0.2774F

The above figures for the Hausman test of exogeneity are t-statistics. Significant t-statistics
indicate endogeneity. The tests for the significance of the explanatory variables on the
system are either single t-statistics or Wald F-tests on the joint restriction from including
the additional variables into an unrestricted VAR when they are not determined to be
endogenous, at the 10 percent critical level. First differences of the logarithmic values for
each series are used in the VAR to ensure stationarity in testing.
F
Indicates an F-statistic
T
Indicates a t-statistic
bIndicates significance at the 95 percent confidence level
cIndicates confidence at the 90 percent level

8
Results from using any of the other alternate exogenous variables discussed in the
methodology section did not significantly alter the results of the causality tests. In fact,
even for those included in this study, the results are not generally different than if no
consideration was made for exogenous variables. The inclusion in this study for any
exogenous series is merely to show and provide a more complete estimation framework to
examine the insurance-economic growth relationship.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
500 THEJOURNALOF RISKAND INSURANCE

The next step of the analysis, having identified those exogenous variables to be in-
cluded in subsequent testing, is to conduct the Johansen cointegration tests. The trace
statistics for the tests are presented in Table 3 and are based on there being a deter-
ministic trend within the series as the unit root tests noted trends with drift in each
series. The results indicate that some, but not all, countries reveal a long-run relation-
ship between economic growth and growth in the insurance industry. Those with no
cointegrative relationship include Austria, Switzerland, the U.K., and the U.S. For
Australia, Canada, France, Italy, and Japan, at the 1 percent critical value, the hy-
pothesis of no cointegrating relationship (a rank of zero) is rejected in favor of accept-
ing no more than one cointegrating relationship. However, a precautionary note must
be made that at the 5 percent significance level, the possibility of a mis-specification
due to more than one cointegrating relationship existing cannot be rejected for France,
Italy, and Japan. Although rejected at the 1 percent critical value, this can still indi-
cate that the modeling of the cointegrating relationship has been mis-specified, to a
small degree, for these countries. A possible explanation lies in the fact that a time
trend may exist within the cointegrating relationship. A test for a single cointegrating
relationship with a time trend is not rejected at the 5 percent level for these countries.
However, for the purposes of this article, this matter is not considered further as a
necessary reason for why these three countries should also incorporate a time trend
as it is neither easy to theoretically justify9 nor, at the 1 percent critical value, is this
concern statistically important.

TABLE3
JohansenCointegrationTraceTests

Rank Ho: r =O Ho :r<1

Country
Australia 23.2237a 3.2961
Austria 13.0459 4.8974b
Canada 20.0453a 3.7159
France 23.9238a 5.7323b
Italy 25.0149a 6.3062b
Japan 20.0519a 5.6746b
Switzerland 12.4128 0.4665
United Kingdom 6.3642 2.1082
United States 11.7149 0.0532
The above figures are trace statistics for the hypothesis of either a rank of zero or a rank of
no more than one, incorporatinga four period lag structure.Criticalvalues are obtained
from Osterwald-Lenum(1992)where and b indicate rejectionof the null hypothesis at the
a

99 percent and 95 percent confidence levels, respectively.

Now that long-run relationships have been identified for each country, the actual
causality test results are presented. However, the first set of statistics tabulated in
9 It is worth mentioning that even with a time trend in the cointegratingrelationship,the
results from the causality tests are not dramaticallydifferent.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE FROMOECDCOUNTRIES501
GROWTH?EVIDENCE
ECONOMIC

Table 4 are from rudimentary causality tests in levels. Although the results will be
statistically spurious, as the asymptotic properties of the results are uncertain from
not accounting for stationarity and cointegration between the series, they should,
nevertheless, show the most basic causal trends within the data series. Furthermore,
as discussed previously, the use of VAR in levels is not completely unjustified10 and
involves no manipulation of the data whatsoever.

TABLE4
CausalityTestsfromVectorAutoregressionsin Levels
Ho:Real
Premiums H: Real GDP
Do Not Cause Does Not Cause
Country Real GDP Real Premiums

Australia 0.7392 1.2542

Austria 1.0315 1.2123
Canada 3.5205b 0.6233
France 0.8106 1.5191
Italy 2.5753c 2.3740c
Japan 4.2986a 0.2890
Switzerland 1.9274 0.4345
United Kingdom 0.8784 0.8166
United States 0.3451 1.3402

All figures are F-statisticsand the lag length for the VARsis set to four where logarithmic
values are taken for each series.
a Indicates significanceat the 99 percent
confidence level
bIndicates significance at the 95 percent confidence level
c Indicates significanceat the 90 percent confidence level

Examining Table 4, only in three countries are notable causal relationships evident.
In Canada and Japan, at the 5 percent and 1 percent significance levels respectively,
there would seem to be evidence of growth within the insurance market leading to
growth in the economy at large. Apart from these two countries, the only other case is
Italy, where a bidirectional relationship exists between the economy and the insur-
ance sector. This bicausal relationship is, however, weaker than for the other two
countries as it is significant only at the 10 percent level. For all other countries, there
is no trace of any feedback from one market to another.
The final set of results presented are in Table 5, detailing the causality test results for
the first differences of the series when the VAR is restricted to incorporate a long-run
error-correction term and any further exogenous variables determined in Table 2.

10 See Sims et al. (1990) for a further discussion.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
502 THEJOURNALOF RISKAND INSURANCE

Broadly speaking, the results are not that different from the test results reported in
Table 4, although now the avenues open to explaining causation between insurance
and GDP growth have increased. Specifically, there are now potentially both short-
and long-run dynamics that can be driving the interaction between the insurance
market and GDP growth. For example, growth in the insurance sector can potentially
have an effect on economic growth via the short-run dynamics of the lagged pre-
mium terms in the restricted VAR, through the long-run equilibrium relationship
between the markets, or both. In the case of Australia, both short- and long-run dy-
namics play a part in determining a significant causal relationship from insurance
growth to GDP growth to at least the 10 percent significance level. Furthermore, there
is also evidence, at the 5 percent level, that GDP growth affects insurance premium
growth from the joint test for significance of both the short- and long-run factors.
However, there is no evidence of short-run causation from GDP growth to the insur-
ance market. Nevertheless, particularly in the case of Australia, it would seem that
without accounting for the cointegrative relationship between the series, no causa-
tion would be noted, as was the case in Table 4.

TABLE5
CausalityTestsFromthe Error-Correction
Models

Real Premiums Real GDP

Hypothesis Do Not Cause Does Not Cause
Real GDP Real Premiums
Country
Short-Run Long-Run Both Short-Run Long-Run Both
Australia 2.1940c 1.9954b 1.8880 1.8343 3.4139a 2.6896b
Austria 1.1731 1.5286
Canada 4.6038a _3.3332a 6.0997a 1.6024 -1.6855c 1.4690
France 0.1954 1.2728 1.3453 1.2229 3.2695a 3.8500b
Italy 1.6178 1.6552c 2.3736c 1.3699 2.6155a 2.4425c
Japan 3.0068b 1.8656c 4.0378a 0.4558 3.2909a 2.6772c
Switzerland 0.9556 1.7262
United Kingdom 0.1797 1.3848
United States 0.6362 0.7426
The figures for noting significant short-runand both short- plus long-run dynamics are
F-statisticsfrom a joint test for significanceof the relevant coefficients.The test for any
long-run component is a t-test for the significanceof the speed of adjustmentcoefficient,
a . A lag structureof four periods is used for each case.
a
Indicatessignificanceat the 99 percent confidence level
bIndicates significanceat the 95 percent confidence level
c Indicatessignificanceat the 90 percent confidence level

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 PROMOTE
DOESINSURANCE FROMOECDCOUNTRIES503
GROWTH?EVIDENCE
ECONOMIC

For Canada the results are generally the same as the results provided in Table 4.
There is again evidence of growth in premiums to cause growth in GDP, but this time
at the 1 percent significance level. Both short- and long-run dynamics play a role and
are jointly significant. However, feedback from economic growth to the insurance
market is weaker as only the error-correction term is significant at the 10 percent
level. For France an interesting result emerges showing a causal effect from the
economy to the insurance market. However, although the error-correction term is
statistically significant at the 1 percent level for the real premium equation, the same
is not true on the GDP growth side. This result indicates a possible mis-specification
within the modeling of the relationship between the series as one would expect the
error-correction terms to be significant in both equations of the VAR, given that the
cointegration tests indicate a strong linear relationship between the series, as illus-
trated in Table 3. One explanation for this is that a structural change within the series
has led to a change in the relationship between the variables. Although to statistically
test for structural breaks is not really feasible given the small sample size that this
study investigates, cointegration and subsequent causality tests were conducted tak-
ing a subsample of the years between 1961 and 1990. The reason for omitting the
1990s data is that France, as most of the other countries under review, went through
a period of financial deregulation and a partial move away from state provision of
health and life insurance. It may well be that the change in the financial sector during
the 1990s has changed the previous relationship within the country. In fact this seems
to be the case, as the cointegrative relationship is stronger for this subsample period
in France and shows a stronger relationship, at the 1 percent significance level, from
GDP growth to the insurance market. Furthermore, significant error-correction terms,
at the 10 percent level, in both the GDP and premium estimation procedure were
reported. There is, however, no significant difference in the results for any short-run
effects, so the results tabulated in Table 5 do not really reveal anything less than the
subsample results that were subsequently estimated.
For Italy, the bicausal relationship observed in Table 4 translates into significant joint
short- and long-run dynamics from the insurance market to GDP and back again.
However, no purely short-run causation is noted. For Japan, the results are the same
as in the case of Italy but with the addition of a significant short-run causal relation-
ship from the insurance market to GDP growth. Hence, a stronger unidirectional re-
lationship emerges from real premiums to GDP than vice versa.
Finally, for Austria, Switzerland, the U.K., and U.S., no causality is noted in either
direction, which is basically what was observed in Table 4, and possibly expected,
given no cointegrative relationship seemed to exist between the series.
The authors' analysis does not permit testing the specific reasons why such differ-
ences in the causal relationship exist. However, some interesting comparisons can be
made. For example, it is worth noting that Switzerland, the U.K., and the U.S., all
countries with no causal relationship, possess high measures of insurance density
and/or penetration. In contrast, a causal relationship does exist for Japan, a country
that has similar high measures of density and penetration but that is also highly regu-
lated. Furthermore, a causal relationship appears to exist for Italy, a country with low
measures of insurance density and penetration and a national culture that, according
to Fukuyama (1995) and Hofstede (1995), is unlikely to support the insurance sector.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
504 THEJOURNALOF RISKAND INSURANCE

These results may seem anomalous except when one looks at the rate of change in
insurance penetration. According to Swiss Re (1997) Japan, in 1995, was ranked third
in the world on this measure and Italy, thirtieth. But since 1980 Japan's measure of
insurance penetration has increased by 150 percent, with Italy's increasing by 92 per-
cent. Over the same period, insurance penetration in the U.S. increased by only 19
percent. Therefore, in accord with the empirical analysis that has been conducted, the
importance of insurance to an economy is not necessarily in its level of activity. Rather,
the significance may in part derive from the rate of structural change in an economy
that results from an increase in the level of insurance provision.
To summarize, it would seem that if no consideration is made for cointegration be-
tween the insurance market and GDP, evidence of causation is sparse for the nine
countries. However, once countries that exhibit a cointegrative relationship are al-
lowed to display both short- and long-run dynamics, the inter-relationships between
growth in the economy and the insurance sector become more prominent. However,
it is important to recognize that the results are conditional on the data set used. Spe-
cifically, while it is possible to assert that differences in the insurance-economic link
do exist across countries, such divisions may only reflect variation in the proportion-
ate level of life and non-life insurance in each country. Importantly, such differences
are lost in the aggregate total insurance measure employed in this study.

CONCLUSION
This study has sought to advance the work on the importance of insurance within the
domestic economy. An important thrust of this article is that the causal relationships
between economic growth and insurance market development may well vary across
countries. This is because the influence of insurance market development, while chan-
neled through indemnification and financial intermediation, is tempered by country-
specific factors. Attitudes to risk and the appropriateness of insurance as a risk man-
agement technique are likely to be culturally determined and therefore different across
economies. Also, the stringency of the regulatory framework differs between coun-
tries. Furthermore, insurance may be dominated by other modes of financial inter-
mediation in the capital allocation process and thereby constrained in its ability to
promote economic growth. From this, there is a pressing need for a fuller empirical
understanding of why the relationship varies across countries. This is certainly a com-
plex question. However, suspicion surrounds the nature of the cultural, regulatory,
and legal environment; the improvement in financial intermediation; and the moral
hazard effect of insurance.

REFERENCES
Angerer, A., 1993, Insurance Supervision in OECD Member Countries, Policy Issues
in Insurance (Paris: OECD Publications).
Arestis, P., and P. Demetriades, 1997, Financial Development and Economic Growth:
Assessing the Evidence, EconomicJournal, 107(442): 783-799.
Banerjee, A., J.Dolado, J. W. Galbraith, and D. F. Hendry, 1993, Co-integration,Error-
Correction,and the EconometricAnalysis of Non-Stationary Data (Oxford: Oxford
University Press).

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
 DOESINSURANCE
PROMOTE FROMOECDCOUNTRIES505
GROWTH?EVIDENCE
ECONOMIC

Barro, R. J., 1991, Economic Growth in a Cross Section of Countries, QuarterlyJournal

of Economics, 106(2): 407-443.
Beenstock, M., G. Dickinson, and S. Khajuria, 1986, The Determination of Life Premi-
ums: An International Cross-Section Analysis 1970-1981, Insurance, Mathematics
and Economics,5: 261-270.
Browne, M. J., and K. Kim, 1993, An International Analysis of Life Insurance De-
mand, Journalof Risk and Insurance, 60(4): 616-634.
Browne, M. J., J. Chung, and E. W. Frees, 2000, International Property-Liability Insur-
ance Consumption, The Journalof Risk and Insurance, 67(1): 73-90.
Butler, R. J., B. D. Gardner, and H. H. Gardner, 1998, More Than Cost Shifting: Moral
Hazard Lowers Productivity, Journalof Risk and Insurance, 65(4): 671-688.
Charemza, W. W., and D. F. Deadman, 1992, New Directions in EconometricPractice
(Cambridge, U.K.: Edward Elgar Press).
Cheung, Y. W., and K. S. Lai, 1993, Finite Sample Sizes of Johansen's Likelihood Ratio
Tests for Cointegration, Oxford Bulletin of Economicsand Statistics, 55(3): 313-328.
Comite European des Assurance, 1999, EuropeanInsurance in Figures-Complete 1997
Data (CEA).
Conyon, M. J., and D. Leech, 1994, Top Pay, Company Performance and Corporate
Governance, Oxford Bulletin of Economicsand Statistics, 56(3): 229-247.
Cummins, J. D., S. Tennyson, and A. Weiss, 1999, Consolidation and Efficiency in the
U.S. Life Insurance Industry, Journalof Bankingand Finance, 23(2-4): 325-357.
Demetriades, P. O., and K. A. Hussein, 1996, Does Financial Development Cause Eco-
nomic Growth? Time-Series Evidence From 16 Countries, Journalof Development
Economics,51(2): 387-411.
Engle, R. F., and C. W. J. Granger, 1987, Co-integration and Error Correction: Repre-
sentation, Estimation, and Testing, Econometrica,55 (2): 251-276.
Fukuyama, F., 1995, Trust: The Social Virtues and the Creation of Prosperity (London:
Hamish Hamilton).
Granger, C. W. J.,1969, Investigating Causal Relations by Econometric Models and
Cross-Spectral Methods, Econometrica,37(3): 424-438.
Granger, C. W. J., 1988, Some Recent Developments in a Concept of Causality, Journal
of Econometrics,39(1, 2): 199-211.
Hakkio, C., and M. Rush, 1990, Cointegration: How Short is the Long Run? (research
working paper: 90:08, Federal Reserve of Kansas City).
Hardwick, P., 1997, Measuring Cost Inefficiency in the U.K. Life Insurance Industry,
Applied Financial Economics, 7(1): 37-44.
Hausman, J. A., 1978, Specification Tests in Econometrics, Econometrica,46(6): 1251-
1272.
Hayakawa, H., P. S. Fischbeck, and B. Fischhoff, 2000, The Japanese Automobile In-
surance Industry: Regulation and Market Structure, Journal of Insurance Regula-
tion, 18(3): 385-403.
Hofstede, G., 1995, Insurance as a Product of National Values, Geneva Papers on Risk
and Insurance, 20(4): 423-429.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions
506 THEJOURNALOF RISKAND INSURANCE

Johansen, S., 1991, Estimation and Hypothesis Testing of Cointegration Vectors in

Gaussian Vector Autoregressive Models, Econometrica,59(6): 1551-1580.
Johansen, S., 1995, Likelihood-based Inference in Cointegrated Vector Autoregressive
Models (Oxford: Oxford University Press).
Johansen, S., and K. Juselius, 1990, Maximum Likelihood Estimation and Inference
on Cointegration-With Applications to the Demand for Money, Oxford Bulletin
of Economicsand Statistics, 52(2): 169-210.
King, R. G., and R. Levine, 1993, Finance and Growth: Schumpeter Might be Right,
QuarterlyJournalof Economics, 108(3): 717-737.
Laporta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny, 1997, Legal Determi-
nants of External Finance, Journalof Finance, 54(6): 1131-1150.
Laporta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny, 1998, Law and Fi-
nance, Journalof Political Economy, 106(6): 1113-1155.
Levine, R., N. Loayza, and T. Beck, 2000, Financial Intermediation and Growth: Cau-
sality and Causes, Journalof Monetary Economics,46(1): 31-77.
Levine, R., and S. Zervos, 1996, Stock Market Development and Long-Run Growth,
WorldBank EconomicReview, 10(2): 323-339.
Levine, R., and S. Zervos, 1998, Stock Markets, Banks, and Economic Growth, Ameri-
can EconomicReview, 88(3): 537-558.
Osterwald-Lenum, M., 1992, A Note with Quantiles of the Asymptotic Distribution
of the Maximum Likelihood Cointegration Rank Test Statistics, Oxford Bulletin of
Economicsand Statistics, 54(3): 461-472.
Outreville, J. F., 1990, The Economic Significance of Insurance Markets in Developing
Countries, Journalof Risk and Insurance, 57(3): 487-498.
Outreville, J. F., 1996, Life Insurance Markets in Developing Countries, Journalof Risk
and Insurance, 63(2): 263-278.
Pagano, M., 1993, Financial Markets and Growth: An Overview, EuropeanEconomic
Review, 37(1): 613-622.
Patrick, H., 1966, Financial Development and Economic Growth in Underdeveloped
Countries, EconomicDevelopmentand Cultural Change, 14(1): 174-189.
Pesaran, M. H., N. U. Haque, and S. Sharma, 2000, Neglected Heterogeneity and
Dynamics in Cross-Country Savings Regressions, in: J. Krishnakumar and E.
Ronchetti, eds., Panel Data Econometrics-Future Direction:Papersin Honour of Pro-
fessor Pietro Balestra (Amsterdam: Elsevier Science), 53-82.
Phillips, P. C. B., and P. Perron, 1988, Testing for a Unit Root in a Time Series Regres-
sion, Biometrika,75(2): 335-346.
Ruser, J.W., 1998, Does Workers' Compensation Encourage Hard To Diagnose Inju-
ries?, Journalof Risk and Insurance, 65(1): 101-124.
Sims, C. A., J. H. Stock, and M. W. Watson, 1990, Inference in Linear Time Series
Models With Same Unit Roots, Econometrica,58(1): 113-144.
Swiss Reinsurance Company, 1997, World Insurance in 1995, in: Sigma (Zurich: Swiss
Re).
Toda, H., and P.C. B. Phillips, 1993, VectorAutoregressions and Causality, Econometrica,
61(6): 1367-1393.

This content downloaded from 128.230.234.162 on Wed, 4 Jun 2014 19:25:00 PM

All use subject to JSTOR Terms and Conditions