Journal of Economic Surveys

THE RELATIONSHIP BETWEEN RELATIVE RISK AVERSION AND THE LEVEL OF EDUCATION: A SURVEY AND IMPLICATIONS FOR THE DEMAND FOR LIFE INSURANCE

Authors


  • *[Correction added on 10 November 2014, after first online publication: affiliation of the author changed].

Abstract

The expected utility hypothesis has been widely used in the construction of economic models and numerous difficulties are encountered in attempting to take into account preferences toward risk in a real-world setting. More recently, attention has focused on comparative studies in economics and business in an international framework and problems related to the hypothesis of relative risk aversion (RRA). One ambiguous hypothesis is the relationship between the level of RRA and the level of education, which has been found either positive or negative. From a causality point of view, it may be argued that investors with a high level of education are less risk averse, but it may also be argued that less risk-averse individuals choose to pursue a higher level of education. The purpose of this paper is to survey the empirical literature on this subject. It provides evidence that risk aversion is negatively correlated with higher education and human development. The results have important implications for macroeconomic empirical studies and the demand for financial assets and more specifically on the demand for life insurance. Assuming the same degree of RRA for utility-maximizing consumers should be limited to homogeneous samples.

1. Introduction

Nearly all theoretical and empirical work on the demand for life insurance takes Yaari (1964, 1965) and Hakansson (1969) as a starting point. The demand for insurance is properly considered within the context of the consumer's lifetime allocation process (Fisher, 1973; Campbell, 1980; Lewis, 1989; Bernheim, 1991). Within this framework, the consumer maximizes lifetime utility and a variety of variables are used to represent the possible outcome of the decision being represented. Demand is a function of wealth (or total assets), expected income, expected rate of returns on alternative choices, and subjective discounting functions to evaluate these choices. It is also assumed here that each utility-maximizing household has the same degree of relative risk aversion (RRA).

Since Pratt (1964) and Arrow (1965) independently derived the measure of absolute and relative risk aversion, the concept of RRA has been used as a postulate in many theoretical or empirical models (Eeckhoudt, 2012). If, as wealth is increased across households, a greater (smaller) proportion of wealth is held in the form of risky assets, households are said to exhibit decreasing (increasing) RRA, that is, they are relatively less (more) risk averse.

However, as shown by Meyer and Meyer (2005), variations in the way the outcome variable of a risky choice is defined or measured, significantly alter the RRA measure determined by the decision maker. In addition, various studies also frequently define or measure wealth or income in different ways. Measures of wealth, for instance, often exclude the value of human capital. Although the measure of RRA is invariant to the unit in which the outcome variable is measured, this elasticity measure is sensitive to what is included or excluded when defining or measuring a variable (Meyer and Meyer, 2005). For example, Karni and Zilcha (1985, 1986) addressed the problem of measurement of risk aversion and studied the implications of differences in risk aversion for the optimal choice of life insurance coverage. Cleeton and Zellner (1993) detailed the relationship between insurance demand and income when there is a change in the degree of risk aversion.

Although the issue of measurement is not considered in this survey, the empirical literature on risk aversion contains a vast array of estimates undertaken in various contexts, with different types of data, different populations, different models and assumptions regarding utility, and different estimation techniques. While macroeconomic and asset-pricing approaches to estimation of risk aversion are based on rational expectations and life cycle theories of consumption, surveys or measures using questionnaires require no assumption on the form of the individual utility function (Kagel and Roth, 1995). As a consequence, the results have been quite diverse. A number of observers have been disconcerted by this lack of consistency across studies but Meyer and Meyer (2006) in analyzing the measurement of risk aversion conclude that when existing estimates of risk aversion are converted to a common basis, much of the reported discrepancy disappears.

The earliest studies have used questionnaires to recover individual investor's risk preferences (Lease et al., 1974; Lewellen et al., 1977). Attention has also focused on the conditions under which it is possible in principle to recover these individual preferences by observing their behavior (Cohn et al., 1975). One of the earliest and most quoted studies of risk aversion and wealth is by Friend and Blume (1975). Their measure of risk aversion depends on the individual investor's portfolio allocation between risky and risk-free assets. They find evidence of decreasing relative risk aversion (DRRA), that is, individuals invest a larger proportion of their wealth in risky assets as wealth increases. When wealth is defined to include the value of houses, cars, and human capital, they find evidence of constant relative risk aversion (CRRA). Following these approaches, several studies have presented evidence concerning RRA (for example, Dybvig and Polemarchakis, 1981; Siegel and Hoban, 1982; Ferson, 1983; Morin and Suarez, 1983; Bellante and Saba, 1986; Mankiw and Zeldes, 1991; Riley and Chow, 1992; Barsky et al., 1997; Halek and Eisenhauer, 2001; Brav et al., 2002; Chetty, 2006).1

The methodologies and contexts have been varied and the results also. Numerous difficulties are encountered in attempting to measure preferences toward risk in a real-world setting. Experiments conducted by Binswanger (1981), Bar-Shira et al. (1997), Yesuf and Bluffstone (2009), and Tanaka et al. (2010) and using farmers as subjects, are a few attempts to measure directly risk aversion in a real-world environment. Estimates of the value of RRA range from less than 1.0 (Hansen and Singleton, 1982, 1983; Keane and Wolpin, 2001; Belzil and Hansen, 2004; Brodaty et al., 2007), to well over 30 (Hansen and Singleton, 1983; Blake, 1996; Palacios-Huerta, 2006).2 At a macroeconomic level, it has been shown by Szpiro (1986), that it is possible to obtain an aggregate measure of risk aversion by observing the behavior of individuals toward their demand for insurance.3

Following Cohn et al. (1975), a number of studies have analyzed RRA when controlling for the effects of other individual characteristics, such as family characteristics (Siegel and Hoban, 1991; Schooley and Worden, 1996; Dohmen et al., 2011), gender (Levin et al., 1988; Bajtelsmit and Bernasek, 1996, 2001; Palsson, 1996; Schooley and Worden, 1996; Powel and Ansic, 1997; Scheiber and Shoven, 1997; Hersch, 1998; Jianakoplos and Bernasek, 1998; Halek and Eisenhauer, 2001; Croson and Gneezy, 2009; Lin, 2009; Dohmen et al., 2011; Charness and Gneezy, 2012), age (Morin and Suarez, 1983; Bellante and Saba, 1986; Riley and Chow, 1992; Palsson, 1996; Bajtelsmit and Bernasek, 2001; Halek and Eisenhauer, 2001; Bellante and Green, 2004; Lin, 2009, Dohmen et al., 2011), religion (Barsky et al., 1997, Halek and Eisenhauer, 2001, Dohmen et al., 2011), and education. However, the effects of education on risk-taking are mixed. For example, Riley and Chow (1992) find that risk aversion decreases with education and Jianakoplos and Bernasek (1998) find that risk aversion increases with education, without any significant difference between women and men.

While it can be argued that these traits may affect one's risk aversion, it may also be that one's risk aversion affects these lifestyle choices. The causality links have not been explored in the literature. It may for example be argued that wealthy people with a high level of education are less risk averse, but it may also be argued that less risk averse individuals choose to pursue a higher level of education and consequently are more wealthy. Dohmen et al. (2011) also find that higher parental education has a significant positive impact on the willingness to take risks.

From the point of view of life insurance demand, the link between RRA and the level of education or human capital accumulation in empirical studies remains controversial. In many papers, the level of education proxies the level of risk aversion and most of the empirical papers have verified a positive and significant relationship between insurance demand and the level of education (Outreville, 2013).

The motivation of this paper is not to reconcile the conflicting evidence but to survey in the context of the empirical literature, this relationship between the level of risk aversion and the level of education. The paper claims to illustrate the negative relationship between the level of RRA and the level of education. It distinguishes between papers that consider the impact of risk aversion on education and papers that consider the impact of education on risk aversion without solving the causality issue (which is not the objective of the paper). Because of the paucity of data related to this issue, this survey reports on an empirical analysis based on a limited sample of countries that provides support to this claim and an extension is proposed to analyze the broader concept of human capital development. The results have important implications for macroeconomic empirical studies on the demand for life insurance or other financial assets available to the consumers.4 Most of the empirical studies of the demand for insurance using education as a proxy for risk aversion find a positive relationship that is inconsistent with the evidence presented in this survey. The paper claims that this is an important issue for future lines of research.

The paper is structured as follows. Section 'The Relationship between Risk Aversion and the Level of Education' provides an assessment of the relationship between risk aversion and the level of education. Section 'Risk Aversion in Macroeconomic Empirical Studies on Life Insurance Demand' illustrates this relationship in macroeconomic empirical studies on the demand for life insurance. Section 'RRA and the Human Development Index (HDI): Further Tests' presents further empirical tests on the relationship between RRA, education, and human development indices. The paper concludes with remarks and research implications.

2. The Relationship between Risk Aversion and the Level of Education

2.1 Risk Aversion as a Determinant of Education Demand

Most recent rational choice theories of educational decision making can be traced back to Boudon (1974) and Gambetta (1987) and the rational choice perspective offers an alternative to the theories on educational inequality, such as the cultural reproduction theory of Bourdieu (1984). One of the most influential is the theory of RRA (Goldthorpe, 1996; Breen and Goldthorpe, 1997; Morgan, 1998; Breen, 1999). According to the RRA theory individuals are utility-maximizing agents and it is a rational choice theory because agents are assumed to make educational decisions in light of the expected benefits and costs of these decisions.

With respect to the education process, acquiring schooling should be unambiguously viewed as a risky investment. Shaw (1996) in an empirical analysis based on the Survey of Consumer Finances demonstrates that risk aversion lowers the investment in human capital. Higher education attendance requires sacrificing present consumption and absorbing substantial psychic costs in return for future rewards, but successful grade achievement is rarely a certain outcome. For this reason, the probability of losing the investment paid up front cannot be ignored and may act as a strong disincentive.

Individual differences in educational attainment can be explained by differences in marginal costs and returns and by differences in the degree of risk aversion. Brunello (2002) hypothesizes that the causal relation runs from absolute risk aversion to educational attainment, not vice versa. Using a survey containing detailed information on individual attitudes toward risk collected by the Bank of Italy – Italian Survey on Household Income and Wealth (SHIW) – he demonstrates the negative impact of risk aversion on the level of education. The less risk-averse individuals are more likely to obtain higher education. Similar result using the same survey for a different period of time is presented by Guiso and Paiella (2006).

However, existing empirical studies provide mixed support for the RRA theory (Jaeger and Holm, 2012). Belzil and Leonardi (2007) using the same Italian survey show that parents background variables dominate in schooling decisions (as demonstrated in Cameron and Heckman, 1998; Belzil and Hansen, 2002; Davies et al., 2002) as opposed to differences in attitudes toward risk. They report that risk aversion varies at different grade levels: the relationship is negative at lower levels of schooling and reverts to positive for higher education. Van de Werfhorst and Hofstede (2007) using a Dutch survey on households show that cultural capital (parents’ background) has no effect on long-term schooling ambitions and that RRA strongly affects these ambitions at higher levels of education. There is no statistically significant impact at the primary or secondary school level.5

The review of the recent literature on the relationship between RRA and the level of education tends to support the view that more risk-averse individuals have a lower tendency to pursue a university education. The work of Brown et al. (2006) based on a US panel study of income supports this hypothesis. Guiso and Paiella (2008) looking at the Bank of Italy SHIW data for 1995 show that risk aversion is higher among the least educated people and that parents’ background does not affect significantly the degree of risk aversion. Holm and Jaeger (2008) analyze data from the Danish Youth Longitudinal Study and find strong evidence that RRA affects the decision to pursue higher levels of education. That is, if the costs associated with university education and the risk of failure, outweigh the benefits of moving into a higher social class, there is little incentive to pursue higher education. Brodaty et al. (2007) use a survey on the ‘Generation 92’ developed by the CEREQ in France to find that the enrollment in higher education is very sensitive to risk aversion. Hartlaub and Schneider (2012) looking at data from the German Socio-Economic Panel Study (SOEP) report that risk-averse students have a lower tendency to pursue a university education and, instead prefer vocational training.

2.2 Risk Aversion Explained by the Level of Education

The results on the effects of education on risk-taking are mixed. In the earliest study, Riley and Chow (1992) find that financial risk aversion appears to decrease with education. However, Jianakoplos and Bernasek (1998) find the opposite – that risk aversion increases with education, without any significant difference between women and men. A similar result is reported by Hersch (1996) in the analysis of the National Medical Expenditure Survey in 1987. A common concern in interpreting the results of these studies is that education, income, and wealth tend to be highly correlated.

A survey of the literature shows that most recent papers support the hypothesis of a negative relationship between RRA and the level of education (Table 1). Bellante and Green (2004), using the nationwide AHEAD (Asset and Health Dynamics Among the Oldest Old) study conducted by the University of Michigan Institute for Social Research, find that high school graduates exhibit a greater tendency to accept risk than those without a high school diploma, but a lesser tendency than those possessing a college degree. Harrison et al. (2007) using experiments in the field across Denmark find that individuals with postsecondary or higher education are associated with risk-taking attitudes (lower risk aversion) than those with less education. Lin (2009) with data from the 2003 Survey of Family Income and Expenditure (SFIE) in Taiwan finds that household heads with more years of education have a lower risk aversion. Kapteyn and Teppa (2011) find a similar result with households’ data from the CentERpanel representative of the Dutch population.

Table 1. Testing the Level of Education as a Determinant of Risk Aversion
  Nonsignificantly #      
NegativePositive      from 0      
Riley and Chow (1992)Hersch (1996)Barsky et al. (1997)
Bellante and Green (2004)Jianakoplos and Bernasek (1998)Halek and Eisenhauer (2001)
Harrison et al. (2007)  
Lin (2009)  
Kapteyn and Teppa (2011)  

To temper these results, Barsky et al. (1997) suggest that the relationship is probably nonlinear and that depending of the number of years of schooling the sign and significance of the relationship may change. Halek and Eisenhauer (2001) also find a nonsignificant relationship. In the context of financial risk-taking, Bayer et al. (2009) argue that a more relevant effect to measure would be access to financial knowledge rather than education in general. They find that measures of savings activity are significantly higher when employers offer retirement seminars and the effects are greater for lower paid employees than for higher paid employees. Similar results are proposed by Van Rooij et al. (2011) and Clark et al. (2012).

3. Risk Aversion in Macroeconomic Empirical Studies on Life Insurance Demand

3.1 Review of the Literature

In the insurance literature, the level of RRA is hypothesized to be positively correlated with insurance consumption in a nation.6 However, it has been shown how it is difficult to recover RRA at a country level and most of the empirical studies are based on data available in some OECD countries. There are only a few studies available concerning emerging countries (Lin, 2009 in Taiwan). Comparing risk aversion among countries in cross-section data studies is also an impossible task and the only exception of a measure of RRA across countries is a paper by Szpiro (1986) retrieving data from the aggregate demand for insurance. Using the same set of data, Szpiro and Outreville (1988) provide some explanation of the factors that may explain the variation of RRA across countries.

From the point of view of insurance demand, all empirical papers have used proxies for risk aversion following Lewis (1989) who suggested that a number of variables may explain international differences in risk aversion and insurance demand. In many papers, the level of education proxies the level of risk aversion (Browne and Kim, 1993; Ward and Zurbruegg, 2002; Beck and Webb, 2003) and most of the empirical papers with only a few exceptions have verified a positive and significant relationship between insurance demand and the level of education (Outreville, 2013) (Table 2).

Table 2. Education as a Proxy for Risk Aversion in the Demand for Life Insurance
PositiveNonsignificant (or even negative)
Truett and Truett (1990); Browne and Kim (1993); Ward and Zurbruegg (2002); Hwang and Gao (2003); Hwang and Greenford (2005); Li et al. (2007); Arena (2008); Curak et al. (2009); Han et al. (2010); Lee and Chiu (2012).Outreville (1996); Beck and Webb (2003); Millo and Carmeci (2012); Feyen et al. (2013).

However, the link between risk aversion, the level of education, or human capital accumulation remains largely hypothetical. It has been argued that a higher level of education may lead to a greater degree of risk aversion and greater awareness of the necessity of insurance (Browne and Kim, 1993; Browne et al., 2000; Hwang and Gao, 2003). Truett and Truett (1990) also recognized that more educated people may have also a stronger desire to protect dependants, that is, a higher intensity of the bequest motive. A different view is proposed by Millo and Carmeci (2012). They find from survey data a negative relationship and relate education to the capacity to manage risks implying that better educated people are better able to diversify their portfolio.

An alternative risk aversion proxy is the uncertainty avoidance index proposed by Hofstede (1995) as a determinant of the demand for insurance. Based on survey data, this index is constructed using employee attitudes toward the extent to which company rules are strictly followed, the expected duration of employment with current employers, and the level of workplace stress. This is formally tested by Park et al. (2002) and Esho et al. (2004), who found no significant statistical relationship. However, intercountry differences are likely to reflect differences in the degree of RRA and therefore affect the demand for insurance as well as a country's health status may alter the structure of utility functions (Viscusi and Evans, 1990).

3.2 Direct Empirical Tests

The demand for insurance may differ according to country-specific variables including human capital endowment. Human capital endowment indices have been developed by international organizations such as UNESCO, United Nations Development Programme (UNDP), and UNCTAD but have not been used in empirical papers examining the demand for insurance. Haiss and Sumegi (2008) tested a human capital index which takes into account education and other variables such as the UNCTAD's innovation capability index proposed in UNCTAD (2005) but found it negative and nonsignificant.

Outreville and Szpiro (1998)7 looking at a cross-section of developing and developed countries, provide empirical evidence that aversion toward risk is negatively correlated with higher education. Their analysis relies on estimated figures for the level of RRA based on the work of Szpiro (1986). Data are available for 22 countries (Appendix A) to examine the relationship between risk aversion (RRA) and socioeconomic factors which could explain differences in the degree of risk aversion. They argue that higher education leads to lower risk aversion that in turn leads to more risk-taking by skilled and well-educated people. To our knowledge, the hypothesis concerning the relationship between risk aversion and the level of education has never been verified in any cross-section study. The level of education can be proxied by the percentage of the labor force with higher education (usually tertiary education) relative to the population as proposed originally by Baldwin (1971).

Figure 1 is a replicated figure using the data set presented in Outreville and Szpiro (1998). It shows is a negative slope in the relationship between RRA and the level of higher education (HighEDUC).

Figure 1.

Relationship between RRA and the Level of Education (22 countries).

Using the same set of data, the estimated relationship with ordinary least squares (OLS) gives the following results:

display math(1)
display math(2)

Most empirical studies on the demand for insurance have used the percentage of population enrolled in third-level education (HighEDUC) as a proxy for the level of education (Outreville, 1990, 1996; Truett and Truett, 1990; Browne et al., 1993, 2000; Ward and Zurbruegg, 2002; Hwang and Greenford, 2005; Li et al., 2007; Han et al., 2010; Park and Lemaire, 2011). Following Mankiw et al. (1992), some studies have suggested the proportion of the population enrolled in secondary education (Esho et al., 2004; Arena, 2008; Curak et al., 2009). Testing directly the level of secondary education, we find a similar negative coefficient but less significant.

Considering the debate mentioned in the previous section on the nonlinearity of the role of education, and using the same set of data, we have reconstructed a proxy variable (AvgEDUC) taking into account both the level of secondary education (weighted 1/3) and the level of tertiary education (weighted 2/3) (Appendix A). It was not possible because of high multicollinearity among the variables to test directly in a single equation, both levels of education to verify the hypothesis of nonlinearity.

The estimated relationship with OLS gives the following results which are very similar to the previous ones:

display math(3)
display math(4)

4. RRA and the Human Development Index (HDI): Further Tests

The UNDP publishes since 1990 indicators on human capital and development indices. The most critical ones are to lead to a long and healthy life, to be educated and to enjoy a decent standard of living. Human development is measured by UNDP as a comprehensive index – called the HDI – reflecting life expectancy, knowledge, and command over the resources to enjoy a decent standard of living. See Folloni and Vittadini (2010) for a survey on Human Capital measurement.

Life expectancy (LIFEXP) is one of the principal indicators of human resources development. The point made by Ram and Schultz (1979) is that longer life expectancy leads to greater incentives for human capital accumulation. For the second key component – knowledge – education figures are only a crude reflection of access to education. Rapid improvements in basic education have sharply increased the knowledge of people in developing countries. Several developing countries have adult literacy rates above 90%, comparable to the rates in industrialized countries. Skilled and well-educated people have generally better access to information and are more likely to behave as less risk-averse people. Higher education (AvgEDUC) leads to lower risk aversion and higher savings as claimed in this paper.

Equally important is the effect of income on the health status of a country. Viscusi and Evans (1990) claim a higher marginal utility of income when people are in good health relative to bad health. The health status (HEALTH) is defined and calculated by the UNDP as the population with access to safe water or by the population with access to health services. The urbanization rate of the population (URBAN) is also considered as a control variable in the analysis.

Appendix B reports the coefficients of the bivariate regressions. If we regress each of the explanatory variables on all the remaining variables, the correlation coefficients of the auxiliary regressions give a measure of the degree of multicollinearity.

The estimated relationship between RRA and HDI with OLS gives the following result:

display math(5)

Replacing the variable HDI by its components, the estimation by generalized least squares (GML) leads to the following equation:

display math(6)

With z-statistic in [] and convergence achieved after one iteration

With R2 equivalent = 0.39

Although there is high multicollinearity among some of the components of the HDI index (Appendix B), the coefficient associated with higher education remains statistically significant along with the urbanization rate. These results provide more support to the negative relationship between RRA and the level of education and lead to a strong reject of empirical results showing that the demand for insurance is positively related to the level of education based on a hypothetical positive relationship between RRA and the level of education.

5. Concluding Remarks and Research Implications

Although there is considerable information available on the determinants of the demand for insurance there are several issues that still require further attention. The purpose of this paper is to compare the degree of RRA with the level of education. The review of the recent literature on the relationship between RRA and the level of education tends to support the view that higher education leads to lower risk aversion, which in turn leads to more risk-taking by skilled and well-educated people.

In the insurance literature, the level of RRA is hypothesized to be positively correlated with insurance consumption in a nation. However, most of the empirical papers have verified a positive and significant relationship between insurance demand and the level of education which would imply a positive correlation between RRA and education, that is, a higher level of education may lead to a greater degree of risk aversion and greater awareness of the necessity of insurance.

The evidence presented in this paper is based on a survey of studies that have examined empirically the relationship between the level of risk aversion and the level of education. Replication of a paper providing data on a cross-section of developed and developing countries, suggests that there is a negative and significant relationship between RRA and the level of education in a country. This result, of course, is not definitive partly because of possible measurement errors, small sample size, and more fundamentally, because the evidence points only to association between variables, and not to the nature of the causal links among these variables. Typically, the present value of family members’ future lifetime allocation process is positively related to income and education.

However, assuming the same degree of RRA for utility-maximizing consumers should be limited to homogeneous samples. In macroeconomic and cross-section studies, this hypothesis does not hold and it cannot be postulated that there is a positive correlation between risk aversion and the level of education. Obviously, the positive and significant relationship between insurance demand and the level of education is reflecting another issue, not risk aversion. The limited empirical results presented in this paper point to multicollinearity among variables related to human and economic development and the level of education is clearly highly correlated with these variables.

Considering the growing interest for the insurance sector in emerging and developing countries, there exists much room for further theoretical and empirical research in this area of insurance economics.

Notes

  1. 1

    Following the methodology designed by Holt and Laury (2002) a number of papers have also investigated individual's risk attitudes in laboratory experiments. These studies mainly report on factors related to risk attitudes and do not focus necessarily on education. For this reason, they are not surveyed in this paper.

  2. 2

    A number of observers have been disconcerted by this lack of consistency across studies. Gollier (2001, pp. 424–425) has remarked, ‘It is quite surprising and disappointing for me that almost 40 years after the establishment of the concept of risk aversion by Pratt and Arrow, our profession has not been able to attain a consensus about the measurement of risk aversion’.

  3. 3

    In contrast to the abundant literature on relative risk aversion, there has been relatively little work done in estimating prudence (Dynan, 1993; Merrigan and Normandin, 1996, Eisenhauer and Halek, 1999; Eisenhauer, 2000; Eisenhauer and Ventura, 2003; Ebert and Wiesen, 2011). There are even fewer studies on higher order risk attitudes like temperance (Deck and Schlesinger, 2010).

  4. 4

    This focus is clearly on life insurance but it could be generalized to the demand for all financial assets as part of a basket of securities available to the consumer. By considering this approach, the analysis ignores the corporate demand for insurance. The insurance literature has paid insufficient attention to the fundamental differences between individual and corporate purchasers. Although risk aversion is at the heart of the demand for insurance by individuals, it provides an unsatisfactory framework from the corporate finance point of view.

  5. 5

    As mentioned by a referee, it is important to point out that measurement of education varies considerably across studies.

  6. 6

    Shortly after the seminal papers of Pratt and Arrow on the measures of risk aversion, Smith (1968) and Mossin (1968) applied their analysis to insurance purchases. Ehrlich and Becker (1972) combined expected utility theory and indifference curve analysis within the context of state preference, to derive a theory of demand for insurance. Further papers by Schlesinger (1981) and Doherty and Schlesinger (1983) investigated the effect of risk aversion on optimal insurance coverage.

  7. 7

    Unpublished working paper quoted in Browne et al. (2000) and Esho et al. (2004).

Appendix A

Data

CountryPeriodRRAHDIHighEDUCAvgEDUC
Australia1950–802.630.97828.876.8
Austria1966–842.060.96131.370.3
Brazil1962–823.320.78411.927.9
Denmark1950–801.470.97129.179.1
Finland1950–802.360.96736.785.7
France1950–841.900.97429.774.2
Germany1950–792.230.96734.282.2
Guatemala1975–842.640.5928.633.6
Ireland1970–800.960.96127.072.5
Israel1950–802.930.95735.775.2
Japan1952–802.930.99635.082.5
Korea, Rep.1969–842.060.90316.562.0
Malaysia1972–833.270.8007.637.1
Mexico1966–833.780.87619.346.3
Norway1950–791.850.98334.180.1
Portugal1960–832.180.89912.135.6
Singapore1965–833.920.89913.348.3
Spain1955–801.800.96529.678.1
Sweden1950–792.800.98728.873.8
Switzerland1950–802.160.98631.779.7
USA1950–801.860.96155.5104.5
Venezuela1967–841.340.86127.551.5
Mean 2.380.91726.566.2
Standard Deviation 0.760.09511.420.3

Data source: Szpiro and Outreville (1988), UNDP, Human Development Indicators (1990).

Appendix B

Measures of Human Development

Dependent variable: RRA
IndependentBivariate Regression Auxiliary Regression
VariableCoefficientt-valueCoefficient
HDI−2.581.520.91
AvgEDUC−0.022.300.79
LIFEXP−0.061.530.88
HEALTH−0.011.270.90
URBAN0.031.320.31

Ancillary