The relationship between insurance and economic growth in Europe: a theoretical and empirical analysis

Abstract

The role of insurance companies, although growing in importance in financial intermediation, has received less attention than bank and stock markets and if so, mainly as a provider of risk transfer in single country or very heterogeneous samples. We investigate both the impact of insurance investment and premiums on GDP growth in Europe. We conduct a cross-country panel data analysis from 1992 to 2005 for 29 European countries. We find a positive impact of life insurance on GDP growth in the EU-15 countries, Switzerland, Norway and Iceland. For the New EU Member States from Central and Eastern Europe, we find a larger impact for liability insurance. Furthermore our findings emphasise the impact of the real interest rate and the level of economic development on the insurance-growth nexus. We argue that the insurance sector needs to be paid more attention in financial sector analysis and macroeconomic policy.

Appendix 1: Review of empirical studies on the insurance-growth nexus

Beenstock et al. (1988) apply cross-section analysis to 1970–1981 data covering 12 countries. They regress premiums for property liability insurance (PLI) onto gross national product (GNP), income and interest rate development. They find that premiums are correlated to interest rate and GNP; the marginal propensity to insure (short- & long-run) rises with income per capita and is always higher in the long run. Beenstock et al. (1988) argue that insurance consumption is not affected by economic cycles or cyclical income variations.

Outreville (1990) conducts a cross-section analysis of PLI premiums for the years 1983 and 1984 for 55 developing countries as related to GDP, insurance price and other macroeconomic figures. The results are similar to Beenstock et al.’s (1988) and support the significance of income and financial development (M2/GDP). Other explanatory variables do not seem to be important. Problems in the investigated countries are insufficient demand for insurance services and hence the resulting unbalanced portfolio of the insurer.

Browne and Kim (1993) conduct a cross-sectional analysis of life insurance consumption per capita for 45 countries for the years 1980 and 1987. Income, dependency and social security expenses are positively correlated while inflation is negatively correlated and significant in both years. Religious origins––i.e. a country being Muslim––is always negatively connected to insurance consumption and so the findings support the works of Hofstede (1995, 2004) and Fukuyama (1995) in their reasoning that social back-up influences insurance demand.

Outreville (1996) investigates the correlation of life insurance premiums to GDP and other factors for the year 1986 for 48 developing countries. The results contradict his former work (Outreville 1990) by showing no significance for real interest rate or financial development (M2/GDP). Only the income elasticity is similar to those found in earlier works (Beenstock et al. 1988, Outreville 1990 and Browne and Kim 1993).

Zhi Zhuo (1998) focuses on China, conducting a cross-regional study for 1995 and a time-series analysis for the period 1986 to 1995. In accordance with other findings both the cross-regional and the time-series analysis show that GDP per capita and the consumer price index are significantly correlated with insurance consumption. Furthermore the children-dependency ratio is important, whereas the level of education is not causally related.

Browne et al. (2000) apply a panel model to motor vehicle (MV) and general liability (GL) insurance in the OECD over the 1986–1993 period. Income and the legal system are positively correlated to insurance consumption, while loss probability and wealth are negatively correlated. The number of foreign firms in the market and higher risk aversion increase MV consumption. Browne et al. (2000) argue that income affects insurance consumption. The correlation with risk aversion is statistically insignificant for MV consumption and negatively connected in the cross-sectional model for GL consumption.

Catalan et al. (2000) analyse the Granger causality of insurance assets for 14 OECD and five developing countries over the 1975–1997 period vis-à-vis GDP growth (among others). According to their analysis, contractual savings seem to have some connection to market capitalisation (MC) and value traded (VT) in the majority of countries. The correlation between MC and pension funds shows similar links to its connection to contractual savings, but the pension funds-VT nexus is mixed. In Catalan et al.’s (2000) analysis, nine OECD countries support the life insurance-MC link while the results for the developing countries are mixed. Evidence for the life insurance-VT connection is not so strong in OECD countries, whereas the majority of non-OECD countries show this link. The impact of the non-life business is almost equal to the impact of the life business for MC and less so for VT. The authors favour contractual savings institutions over other institutional investors (e.g. non-life insurance) and so they recommend an appropriate sequencing of the financial institutions’ development.

Ward and Zurbruegg (2000) analyse the Granger causality between total real insurance premiums and real GDP for nine OECD countries over the period 1961 to 1996. For two countries (Canada, Japan) the authors found that the insurance market affected GDP and for Italy they found a bidirectional relationship. The results for the other countries showed no connection. The results from the Error-Correction model depicted similar results and added Australia and France to the group of countries providing evidence for some kind of connection

Beck and Webb (2002) apply a cross-country and a time-series analysis for the relation between life insurance penetration, density and percentage of private savings and GDP as the dependent variables and GDP, real interest rate, inflation volatility and others as the explanatory variables. Strong evidence was found for influences of GDP, old dependency ratio, inflation and banking sector development. From the group of additional explanatory variables, anticipated inflation, the real interest rate, secondary enrolment and the private savings rate were found to be significant. When analysing the share of life insurance in private savings, the results suggest that the ratio decreases with an increasing saving rate although the saving rate has a positive coefficient. This could be due to the preference of households to limit life insurance expenses and transfer additional income to other forms of saving.

Park et al. (2002) concentrate on the link between insurance penetration and GNP and some socio-economic factors adopted from Hofstede (1983). The results of the analysis of the cross-sectional data from 38 countries in 1997 show significance for GNP, masculinity, socio-political instability and economic freedom. Deregulation was found to be a process capable of facilitating growth in the insurance industry, supporting the expectations of Kong and Singh (2005).

Webb et al. (2002) use a Solow-Swan model and incorporate both the insurance and the banking sector, with the insurances divided into property/liability and life products. Their findings indicate that financial intermediation is significant. When split into the three categories, the banking and life sector remain significant for GDP growth, while property/liability insurances lose their importance. Furthermore, results show that a combination of one insurance type and banking has the strongest impact on growth.

Lim and Haberman (2003) concentrate on the Malaysian life insurance market. While the interest rate for savings deposits and price enter significantly in the equation, the positive sign for the interest rate puzzles the authors. This could be in line with findings by Webb et al. (2002) that the best results are obtained when insurance and the banking sector are combined in the estimates. Price elasticity is found to be more than even.

The work of Davis and Hu (2004) is special in terms of the direction of the regression and the variable setup. The authors test for causality between output per worker (OW) as the dependent variable and pension fund assets (PFA) and capital stock per worker (CS) on the explanatory side with data spanning 43 years from 1960 to 2003 for 18 OECD countries and 20 East & Middle East (EME) countries. The results give evidence for PFA and CS having a positive and significant effect on OW. The dynamic heterogeneity findings of the models support the OLS results in the long run. The co-integration test suggests that PFA and CS are co-integrated with OW. The findings also show that PFA development has a stronger impact on OW in EME countries than in OECD countries.

Zou and Adams (2004) provide insights into the Chinese property insurance market for the years 1997 to 1999. Due to market regulation and the specialities of the Chinese market, this work is more suitable to provide evidence for the law-and-finance view of La Porta et al. (1998) or the socio-political decision model of Hofstede (1995). The results show a tendency for companies that are highly leveraged or have physical-assets intensive production to consume property insurance, while partial state-ownership or a possible tax-loss carry-forward decreases demand. Increased managerial or foreign ownership and better growth options facilitate demand, while the size of the company enters inversely.

Esho et al. (2004) focus on the legal framework alongside the GDP-Property-Causality Insurance Consumption (PCI) link. The causality analysis is based on data from 44 countries from 1984 to 1998 and includes OLS and fixed-effects estimations and a GMM estimation on panel data. No matter which methodology is used, real GDP and the strength of the property rights in a country are positively correlated to insurance consumption. The insurance demand is significantly connected to loss probability, but the link with risk aversion is rather weak. The price only shows a slight, negative impact if investigated with the GMM estimator. Although the data set showed major differences between the developments of countries of different legal origins (PCI per capita, GDP, PCI price, etc.), no evidence was found for the legal origin being a significant indicator for PCI consumption. In contrast to other sectors, the importance of property rights suggests that the legal environment facilitates demand for insurance.

Boon (2005) investigates the growth supportive role of commercial banks, stock markets and the insurance sector for Singapore. The author’s findings show short- and long-run causality running from bank loans to GDP, and a bi-directional relationship between capital formation and loans. GDP growth seems to enhance stock market capitalisation in the short run and market capitalisation enters significantly when determining capital formation in the long run. Total insurance funds affect GDP growth in the long run and capital formation in the short and the long run.

Kugler and Ofoghi (2005) focus on the UK and analyse the long-run relationship and Granger causality between insurance premiums and economic growth for the period 1966–2003. Johansen’s cointegration test shows a long-run relationship between insurance and growth; the causality test indicates insurance-inducing growth for the majority of the analysed insurance products. Life, liability and pecuniary loss insurance do not cause growth in the short run.

Adams et al. (2005) conduct an analysis similar to Kugler and Ofoghi (2005) but focus on Sweden for the period of 1830–1998 and include additional variables like bank lending. Bank lending seems superior to insurance services and causes growth in the nineteenth century. In the twentieth century, the causality is reversed. Insurance seem to be more driven by economic growth.

Arena (2006) uses a GMM estimator to analyse panel data of 56 countries in the period 1976–2004. Total premiums, life and non-life premiums are regressed separately, together with stock market turnover and private credit. All three types of premium seem to be causal for GDP growth. Life insurance has more impact in high-income countries, while non-life is significant for growth in both country groups.