1

Institutions, Behaviour, and the Propensity to Hold Income Protection Insurance by Country of Residence *+Gordon L Clark, *Sarah McGill, and *#Juncal Cuñado. *Smith School of Enterprise and the Environment, Oxford University, South Parks Rd., Oxford OX1 3QY, UK; +Department of Banking and Finance, Monash University, Caulfield VIC 3145, Australia; #Department of Economics, University of Navarra, 31080 Pamplona, Spain. Abstract. In this paper, we report the results of a survey on the individual propensity to hold income protection insurance across 11 countries and focus on whether it is meaningful to assume that observed preferences for such products are driven by factors such as respondents’ socio-demographic and economic circumstances or country of residence. We also ask whether it is meaningful to associate observed preferences for income protection insurance to blocs of countries such as the Americas and continental Europe. While country-specific effects dominate the empirical analysis, there is evidence of common preferences across higher-paid individuals within and between blocs of countries. Implications are drawn from our results for understanding the relationship between national institutions, labour market practices and individual behaviour. Keywords. Behaviour, income protection, national institutions. JEL Codes. D12, G22, I31 Introduction Low- and middle-income households are increasingly vulnerable to shortfalls in expected earnings. A recent UK report suggested that lower-income households typically have less than a month’s financial ‘reserves’ available to meet existing financial commitments should earned income suddenly cease (Legal & General 2014). It has also been shown that younger households are particularly vulnerable – those aged 45 years or under had, on average, little more than a week of reserves before they would have to claim state benefits, or rely on others to meet their normal commitments. These households also use short-term loans or credit cards to cover ongoing commitments, thereby exacerbating high levels of indebtedness and low savings. Case studies suggest that what counts as a ‘commitment’ includes housing rent or mortgage payments, the purchase of food and clothing, transportation costs, and the like – that is, the costs of living as well as the costs of remaining in the workforce. In Lusardi et al.’s (2011, 2) study of ‘household financial fragility’ – their capacity to access emergency funds from any source – the authors sought to determine the extent to which Americans could obtain US$2000 within 30 days. It was found that about 25 per cent of Americans surveyed reported that they could not do so while another 20 per cent would have done so by “selling or pawning possessions or taking payday loans.” Their research came in the aftermath of the global financial crisis and was concerned with how households deal with a financial shock rather than with the immediate consequences of an unexpected loss of earned income. Nonetheless, their results, like those from the UK, suggest that the long-term well-being of a significant segment of the working population is directly related to how they cope with or insure against events that would undercut their capacity to earn a living. Those most at risk in the UK and the USA are people on lower incomes, younger working families with children, those with limited financial skills and an inability to save even a modest portion of their current income.1

1. In a related study, Lusardi et al (2009) utilized survey data from the USA, the UK, Canada, France, and Germany

with comparable results, but some significant differences by country of residence.

2

At one level, this paper is an analysis of an important economic and social issue – the vulnerability of individuals and households to interruptions in earned income. At another level, it is an argument in favour of the persistence of nation-state institutions, which elaborates on Christopherson’s (2002) argument to the effect that there are persistent differences between countries in terms of their labour market practices (comparing the Anglo-Saxon world with continental Europe). We provide an empirical perspective on these issues without ending up with an argument that everything and everywhere is different. Instead, we find that residents of countries that have historically provided comprehensive welfare are less likely to report holding income protection insurance because of the various public short- and long-term benefits available to those who are vulnerable to unexpected interruptions to earned income. By contrast, residents of countries that historically provided little in the way of collective insurance for losses of earned income are more likely to report that they self-insure or purchase related insurance products to maintain their standard of living. Two related hypotheses inform our analysis. First, we assume that country-specific institutional factors determine the overall propensity to hold income protection insurance. Second, we test for commonalities between individuals across countries in terms of their propensity to hold income protection products hypothesizing that, whatever their country of residence, the demand for income protection is also determined by shared socio-demographic characteristics. Utilizing variables such as a respondent’s age, gender, education, household size, whether the primary or secondary income earner in the household, employed full-time, part-time or self-employed, we also test whether financial literacy is more important than having prior experience (or knowing someone who has had experience) of income loss. Lusardi and Mitchell (2014) show that age, gender and income are also related to financial literacy and hence the capacity of individuals to respond to changes in their financial circumstances. Recent research into the circumstances of the bottom half of the workforce in OECD countries has taken rather different routes. There are studies of local communities and their attachment to financial institutions (Leyshon et al. 2004, 2006); there are studies of financial literacy within and between countries (Lusardi and Mitchell 2014); and there are a small number of large-scale surveys of selected countries designed to elicit national differences in household financial behaviour (Lusardi et al. 2015). Local studies have many virtues, including recognition of the rich texture of everyday life (Morduch and Schneider 2017). But the findings from these studies are difficult to generalize. Equally, cross-national studies run the risk of superficiality if one model or set of expectations dominates the interpretation. Our analysis relies on a unique database created through a survey of consumer finance across 11 countries, utilizing established panels of survey respondents managed by a commercial provider. See the Appendix for a brief assessment of the issues involved in the design and implementation of our survey. The survey focused on respondents from the UK, Germany, Switzerland, the USA, Mexico, Brazil, Hong Kong,2 Malaysia and Australia (administered in 2016). As such, our survey is unusual for its geographical scope and because the questions posed link the propensity to hold income protection products with individual and household attributes, including financial literacy and health status. We begin with the argument that institutional differences between countries and groups of countries are entrenched in their respective histories and geography, then move on to recent debates about the homogenizing force of globalization, the hegemony of neoliberalism, and the status of the individual. This is followed by a brief discussion of the survey instrument and the econometric methods used to establish patterns in survey responses and test for the significance of country effects on the demand for income protection products. Our results are reported in two sections, one dealing with country

2. Throughout, we refer to Hong Kong as a ‘country’. We do so for convenience. While we recognize that it is

formally part of the People’s Republic of China, we are also mindful of its differences from the mainland, especially concerning its legacy of UK benefit systems.

3

effects and the other with commonalities and differences between individuals by country. In the final section of the paper, we summarize the results and draw implications for future research. Welfare-State Regimes For many historians and political scientists, the defining moment of the post-Second World War era was the establishment of the welfare state. Esping-Andersen (1989: 1) noted that the welfare state is (or was) “a novel phenomenon in the history of capitalist societies,” coming as it did after the Great Depression of the 1930s, the continental war in Europe, and the Pacific war with Japan. His study sought to address the causes and consequences of the growth of the welfare state, while recognizing that its shape and functions varied between countries and groups of countries. He acknowledged the significance of inherited institutions, political coalitions, and country-specific forms of economic development, especially the role of corporations in economic development. He took as axiomatic the role of government in maintaining full employment and in underwriting income. Given the limits of insurance markets, he suggested that collective insurance was an important function of the welfare state. Esping-Andersen emphasized countries’ common social factors despite the variations in their political traditions and economic institutions. Specifically, he suggested that underwriting the welfare of groups previously at the margins of markets and states had taken on a newfound importance: the political mobilization of the hitherto marginalized working classes was an important factor in all countries even if framed through the lens, for example, of liberal democracy or social democracy. In many countries, he suggested, garnering the political support of these new and emerging middle classes was vital in realizing the ambitions of those who sought to underwrite the welfare of those at the margins of markets. Based on this analysis, he argued that Western countries could be grouped into three clusters or types of welfare states – the ‘liberal’ welfare state, the ‘corporatist’ welfare state, and the ‘social democratic’ welfare state. Archetypical examples of each included the USA, Canada, and Australia in the first cluster; Austria, France, Germany, and Italy in the second; and the Scandinavian countries in the third. Esping-Andersen’s model was greeted with some scepticism, but it came to dominate the debate. He was able to strike a balance between features of market capitalism and the history and geography of states. I n this respect, his approach influenced the ‘varieties of capitalism’ (VoC) research programme (Hall and Soskice 2001) in which three related threads of argument can be discerned. First, as Esping-Andersen found a way to reconcile democracy with capitalism, the VoC programme was able to give life to democracy without retreating to scepticism. Second, as Esping-Andersen recognized the significance of corporations and industry groups, the VoC programme fashioned a robust defence of Rhenish capitalism as opposed to liberal capitalism. Third, as Esping-Andersen provided a way of understanding the distribution of income in clusters of countries, the VoC programme offered an explanation of why certain groups are more privileged than others in various countries. Less recognized was Esping-Andersen’s (1999) contribution to understanding the changing status and significance of the welfare state. Whereas his first book sought to explain the logic and emergence of the post-war welfare state and the coexistence of different forms of the welfare state, his second book sought to explain the emerging crisis in “contemporary welfare regimes” (Esping-Andersen 1999: 5). Instead of arguing that the welfare state in its various country-specific forms was riven by internal contradictions, an argument found most recently in Streeck (2014), he suggested that a set of global forces had emerged to challenge the viability of the welfare state – specifically, slow economic growth, global integration, and the claims of resource-rich economies on energy prices that had made it difficult to sustain the ‘egalitarian ideals’ of welfare regimes. Also referenced was ‘Euro-sclerosis’ – the combination of deindustrialization and the discounting of the wage premium claimed by manufacturing workers in continental Europe.

4

At issue was the capacity of nation-states to respond to these pressures and, in particular, whether one type more than others can accommodate ageing populations along with lower than expected rates of economic growth, and political claims in favour of greater individual and household responsibility for future welfare. Whereas the significance attributed to welfare states, the varieties of capitalism, and related forms of regime theory have somewhat lost favour in the face of a focus on the costs and consequences of twenty-first-century financialization (Boyer 2000; Dixon 2014), there remains a strong commitment to understanding the particularities of institutional change and transformation privileging historiography and geography over regime theory (Carnes and Mares 2007). Of particular interest to us in this paper is the continuity of institutions and traditions, and their resilience in the face of economic and political crises (Ebbinghaus 2011; Engelen 2003). Behaviour and Institutions As noted above, macroscopic analyses of national regimes of accumulation and social welfare tend to privilege inherited institutions and traditions over individual preferences. To the extent that individual preferences are discussed, it is often assumed that preferences are framed by institutions both in terms of what is expected and what is possible. Inevitably, the institutional fabric of a society varies by the level of economic development, and the means by which societies regulate behaviour through state-sponsored organizations or norms and conventions (Brennan et al. 2014). Therefore, individual behaviour is always ‘embedded’ in that there are limits to the capacity of individuals to float free of societal commitments and obligations (Grabher 1993; Granovetter 1985). Three objections can be made to this approach. One is that it over determines behaviour, leaving little possibility for individual aspirations. While ‘over determination’ has lost favour across the social sciences in the past couple of decades, the status of individual volition remains highly contested, especially in circumstances where the ‘person’ is objectified in culture, markets and the law (Phillips 2013). In this respect, it is arguable that being embedded can be thought to coexist with individual aspirations where national institutions allow for individual discretion over a range of issues and options. A less demanding argument is that, given the existence of formal and informal inherited institutions, individuals can and do defect from social norms and conventions when seeking to realize their own interests (Clark and Wójcik 2007). This can take various forms, including the exploitation of ambiguities in societal expectations, or circumventing expectations reaching beyond inherited institutions to options hitherto not considered legitimate. By contrast, liberal theory presumes that societies are formed out of individual preferences and in cooperation around common goals (Rawls 1971). This argument has many versions, most of which are founded on biological, economic, or moral principles. It is less important to distinguish between these strands of thought than to note, analytically, that when theorists go from the individual to society they do so by suggesting that individuals make a calculation about the costs and benefits of joining others to achieve certain goals and objectives (Olson 1965). Many economists have focused on the design and implementation of incentives and sanctions to realize desirable outcomes that are of mutual benefit. This particular argument is entirely commonplace. However, few economic geographers are inclined to discount the significance of the institutional context in framing behaviour whatever the incentives and sanctions used to encourage certain kinds of behaviour (Bathelt and Glückler 2111; Gertler 2017). All three arguments provide an action space for individual behaviour. To illustrate, national welfare systems may be designed to leave certain issues and options unresolved and open to individual discretion; individuals can find ways of defecting from norms and conventions to realize their own aspirations; and individuals’ aspirations can be the product of their institutional setting (the action space provided by incentives and sanctions) and their sense of self-realization (related to their desired

5

place in society). If we accept that most, if not all, social systems contain ambiguities about the applicability of norms and conventions and are incomplete in the sense that inherited institutions do not exhaust the available set of individual options, we should expect to see behaviour that is country specific along with behaviour that is not entirely determined by country of residence. To the extent that social position is shared across countries we might expect to see similar behaviour across countries, even if they differ historically and geographically (Rodrik 2013). Yet another argument reinforces the significance of individual behaviour relative to institutional tradition. It comes in various guises, but is most often conceptualized as neoliberalism (see Harvey 2007). There are three parts to this argument. First, it is suggested that, some years ago, national institutions, including regulatory regimes, were very important in setting boundaries on individual discretion, the costs of defection, and the benefits of common solutions to individual aspirations. Second, it is suggested that differences between countries in terms of the opportunities available to individuals in realizing their aspirations and preferences have been eroded in the face of globalization, the impoverishment of nation-states, and the hegemony of the market over the state. Third, and most important, it is suggested that many people hitherto sheltered from the market by public institutions have been forced into the market to meet their basic needs. In this sense, individual behaviour has been conceptualized as ‘coerced’ rather than just the expression of self-interest (see Preda 2009). There is considerable debate about the meaning and significance attributed to the concept of neoliberalism, including its translation into observed reality (see Storper 2016).3 Moreover, critics suggest that the priority attributed to neoliberalism discounts the continuing significance of political movements and the persistence of the nation-state (see, for example, Rodrik 2013). We are not about to adjudicate this debate through the empirical research reported here. But, it is apparent that, at the limit, there are two competing claims about the relationship between individual behaviour and institutions. Those who favour the strong version of embeddedness would hypothesize that country-specific differences in institutions dominate behaviour, whereas advocates of the neoliberal thesis would hypothesize that country-specific differences in welfare institutions have been so eroded that there are marked similarities between individuals across countries in terms of their expectations and behaviour. Patterns of Income Protection Insurance In sum, we expect to see similarities between countries in relation to the propensity of their residents to hold income protection insurance if only because some countries such as those found in continental Europe have shared certain types of institutions, democratic impulses, and the like. Even so, it would be remiss to concentrate on nation-state variations in the propensity to hold income protection insurance, thus ignoring the existence of commonalities between individuals – for example, the possibility that neoliberalism has driven lower-paid people whatever their country of residence to take up this type of insurance because of shared labour market uncertainties (see McDowell et al. 2008, 2009; Strauss 2017; Weil 2014). Survey instrument Underpinning our project is a survey of working individuals (currently employed or out of work for fewer than three months) aged 25–60 in 11 countries. Relying on a commercial market research provider that maintains panels of respondents across most of the countries in this study, we sought

3. For our purposes, it is sufficient to suggest that neoliberalism refers to the systematic discounting of nation-

state institutions and traditions, replacing social solidarity with individual interest while monetizing the moral commitments made between national institutions and citizens, employers and employees, and communities and individuals. At the limit, neoliberalism is a dystopian world of fractured alliances driven by self-interest (Davies 2014).

6

men and women by age, participation in the workforce across a full range of occupations, and across their country-specific distribution of earned income. Countries included in this survey were Germany, Italy, Spain, Switzerland, and the UK (Europe); the United States, Mexico and Brazil (the Americas); Hong Kong and Malaysia (East Asia); and Australia. These countries were chosen for three reasons. First, following Christopherson (2002) among others, it was useful to compare Anglo-Saxon countries with continental European ones while adding others that could disrupt the implications drawn from such ‘standard’ comparisons. Second, to ensure comparability, it was appropriate to seek out countries in which the provider had established panels. Third, these countries were also of interest to the sponsor of the research (the sponsor underwrote the costs of the survey). The survey contained 57 questions, some of which were routed according to respondents’ answers. Professionals translated it into the local language(s) of each of the 11 countries and then volunteer native speakers (all of whom were offered a modest incentive) translated it back to ensure consistency and accuracy across countries. The survey instrument produced 11,584 complete responses. Its first section posed a series of questions to determine the respondents’ eligibility to participate in the project. The second section asked questions about the respondents’ awareness, knowledge, and experience of income protection insurance. Since one aim of the survey was to test awareness of such insurance, a key question (#6) posed to eligible respondents was as follows: “Do you personally have insurance (beyond obligatory government benefits), which would protect your income against any of the following types of risk? Please select one answer per types of risk.” Answer options were “Yes”, “No”, and “Don’t know”.4 The next sections of the survey posed questions on the respondents’ perceived personal financial and health risk situation; their beliefs on the role of governments and employers in providing income protection; and their attitudes towards, and trust in, various institutions. The subsequent sections tested respondents’ attitudes to financial risk, their willingness to pay for income protection insurance, and their financial literacy. The last two sections of the questionnaire asked about further socio-demographic characteristics and their employment situation.5 Data collection To ensure global reach and robust, timely data collection, we engaged a private survey research firm to administer our survey. The method used for the survey research is called CAWI (Computer Assisted Web Interviewing). CAWI engages respondents through a network of panel providers, thus allowing both global reach and access to local knowledge and expertise in panel management. All the panels with which the research firm engages are actively managed, proprietary to the providers, and built over periods of usually at least a decade. The providers’ panel management practices are compliant with market research industry standards (ESOMAR), and data protection and privacy laws. Panel members are individuals who participate freely in online surveys after enrolling voluntarily as a standing panel member. Recruitment takes place either through open enrolment, where individuals sign up to participate as a standing panel member, or through invitations targeted at people who share

4. Of the 11,584 survey respondents, 8.66 per cent (1003) indicated that they did not know whether they held

income protection insurance. Countries with lower than the medium proportion of “don’t knows” included Germany (5.41 per cent), Hong Kong (7.80 per cent), Italy (6.58 per cent), Malaysia (5.02 per cent), Mexico (4.96 per cent), and the UK (5.60 per cent). Those above the medium were Australia (11.54 per cent), Brazil (8.91 per cent), Spain (11.98 per cent), Switzerland (17.48 per cent) and the USA (10.64 per cent). In the next section, it is shown that predicting those who “don’t know” can be allied with predicting those who carry income protection insurance.

5. The full text of the survey is available from the authors.

7

demographic characteristics of interest to the panel providers. Recruitment is conducted via online marketing channels and direct email. Enrolment in the panel then takes place via a double opt-in registration process: prospective panellists first complete a registration form, including a consent form, and then confirm their registration by clicking on a customised link sent to them via an automatically generated email. To take part in a specific survey, panel members receive invitations via email, which include basic participation instructions and information about the approximate length of the survey. These invitations are designed to be ‘non-leading’: panellists receive no advance information about the topic or contents of the survey, only an estimated time of completion. Members of a panel receive rewards for participating in surveys according to a structured incentive scheme. The size of the reward offered as an incentive is determined by the length and complexity of the survey; in many cases it is also determined by the socio-demographic profile of panellists needed for a specific survey. For instance, highly qualified professionals would generally receive greater incentives for completing a survey related to their profession than for completing one about general consumer purchasing decisions. However, the incentives provided for a given study are identical for all respondents. Only panellists who complete a survey successfully receive incentives. These are tailored to local laws and preferences but generally include a ‘points’ programme, gift cards, vouchers, charitable contributions, and prize draws. Point’s programmes, in which panellists can redeem points for other types of incentives, are standard in most countries. To ensure that the target audience for a given study is selected from available panel members, a screener section is normally included at the beginning of a survey (as was the case here; see discussion below). To ensure an otherwise representative sample within this target audience, a quota system based on nationally representative demographic variables such as age, gender, region, and income level is generally used in sample selection. It is worth noting that whereas panels in developed countries are not significantly skewed on these variables, in developing countries they are skewed towards urban areas where internet access is higher. [Insert Table 1 About Here]

Summary statistics Table 1 above presents the main variables used in the empirical analysis, together with the summary statistics.6 The initial step in the analysis was to undertake an ANOVA based on the means of respondents holding protection insurance by country (see Table 2). Here, the null hypothesis was that the means were equal across countries. This was augmented, as indicated below, by a test of difference on the predicted margins of means made available through the logit model of all countries and variables with respect to the dependent variable. [Insert Table 2 About Here]

Table 2 also provides a count of the number of survey respondents by country. It has been observed elsewhere that a representative sample of the UK population would be in the order of 1200 respondents (Clark et al. 2012). It is arguable that the number of survey respondents by country is less sensitive to the total population of each country than might have been expected. In this respect, the number of respondents from countries with large populations such as the USA and Brazil are such that these countries may be underrepresented in the analysis, whereas countries with relatively smaller populations such as Australia and Switzerland may be overrepresented in it. Nonetheless,

6. We acknowledge that there are bound to be differences in the interpretation and salience of these questions

and the rationale implicit therein. However, inspection of the raw data did not reveal systematic differences in the median and distribution of responses by question and/or by countries.

8

sensitivity analysis of our results by country and specific sets of questions indicated that no country or group of countries was so unusual as to dominate the statistical results. Analytical steps The second step in the analysis was to estimate a ‘reference’ model where the dependent variable was whether or not a survey respondent held such a product. Explanatory variables included (a) socio-demographic factors such as gender, age, income, whether the respondent was the household’s primary income earner, household size, and years of education; (b) employment factors, notably work status (full-time, part-time, or self-employed); and (c) other factors, including financial literacy using a set of questions based on Lusardi and Mitchell (2014).7 Other variables included respondents’ self-assessed health status, whether they had previously experienced loss of income for a range of physical or mental health-related reasons and/or the premature death of a household breadwinner, and whether they knew someone who had experienced loss of income for similar reasons, and; (d) country of residence, by estimating a logit model.8 The third step in the analysis was to consider specific countries. The reference logit model was estimated for each country by comparing these results against the estimates for its notional group. The results of this analysis were used to distinguish, on a variable-by-variable basis, between the country case and the universal case and its group members. Since the estimated coefficients cannot be directly compared across countries, we focused on the signs on the coefficients and the relative significance of variables within each country model. This three-step analysis allowed us to test two related hypotheses. The first was whether there is a universal model that explains the propensity to hold income protection – in other words, one model that is sufficiently robust to discount the value of a country-by-country or group of countries mode of explanation. The second and related hypothesis was that the propensity to hold income protection insurance is always specific to a country – that is, specific to the institutional context in which survey respondents make decisions about whether to purchase income protection insurance. These two hypotheses are two sides of the same coin. As we show in the following sections, distinguishing one from the other provides insight into the commonalities and differences between respondents by country. Country Effects and Model Estimates Through analysing the variance of the mean percentage of respondents holding income protection insurance in each country, it was determined that there were significant differences across countries and evidence for the existence of three or four groups of countries. Based on an inspection of the data provided in Table 2, Brazil, Germany, Italy, Spain, and the UK could be one group of countries (0.18–0.29); Australia, Switzerland, and Mexico (0.30–0.37) could be a second group; the USA (0.45) appears as an outlier relative to its near neighbour and European countries; and Hong Kong and Malaysia (each 0.67) could be a fourth group of countries. Reference model

7. To measure financial literacy, respondents were asked to answer the three questions underpinning Lusardi

and Mitchell’s (2014) well-known program: one was related to interest rates, another to inflation rates, and the last one was on risk diversification. The variable used in the empirical analysis was the number of correct answers provided by each of the respondents.

8. It is possible that financial literacy and the propensity to hold insurance are co-determined, being mirror images of one another. However, financial literacy as measured by Lusardi and Mitchell is not domain specific and is framed by reference to three concepts of financial theory. Morduch and Schneider (2017: 90–1) noted that when applied to individual circumstances, it is difficult to show the relevance of financial literacy as formally conceived (see also Clark 2014).

9

A logit model using the entire database was estimated with the dependent variable being the propensity to hold income protection insurance (whether the respondent held such a policy or not). Following Lusardi et al. (2009, 2011) among others, four sets of explanatory variables were utilized – individual socio-demographic characteristics; employment status; financial literacy and experience; and country of residence. Different specifications of the model produced some differences in the significance of country effects. The first model was estimated with only socio-demographic and country of residence variables, whereas the second one included the socio-demographic, status of employment, and country of residence variables.9 These results demonstrate that an individual’s characteristics and particular circumstances mediate the country of residence effect. The results of estimating our reference model utilizing all factors and variables across all respondents and countries are reported in Table 3.10 [Insert Table 3 About Here]

Groups of countries In Table 3, it is apparent that variables such as respondents’ age, income, and work status were found to be statistically significant, as were experience variables – notably, those other than financial literacy. Significant country effects were also found. Following the preliminary analysis of variance of the proportion of respondents holding income protection insurance by country, the hypothesis of equal means across countries was tested on the predicted margins derived from the logit model. The analysis of margins measures the probability that an ‘average’ individual in each country will have income protection insurance. Using 95 per cent confidence intervals, we tested whether the estimated margins were different for each pair of countries. The logit results largely confirmed the preliminary analysis: Hong Kong and Malaysia formed one group of countries; Australia, Mexico, Spain, Switzerland and the USA formed another; Germany, Italy, and the UK formed a third group; and Brazil remained a statistical outlier. These results suggest that whatever the significance of respondents’ characteristics, the propensity to hold income protection insurance is significantly associated with the group to which their country belongs. Within group country-specific results To understand better the commonalities and differences in the results within a group of countries, the reference model was re-estimated for each country in that group. These results are displayed in Tables 4–6. We should recall that the reference model including all countries and variables, including country of residence, produced results showing that the estimated parameters on almost all explanatory variables were statistically significant. Not found to be significant were the parameters on respondents’ gender and years of education. Considering the country specific results, fewer statistically significant parameters and sets of variables were dropped out of the analysis. For example, in the Australian case, the parameters on respondents’ gender, household status (primary or secondary wage earner), household size, years of education, gender of the primary wage earner, financial literacy, and self-reported health status were found not to be significant. Statistically significant parameters were found on respondents’ age, income, work status (of limited significance) and whether the respondent had experienced or knew someone who had experienced a loss of income.

9. Full details are available from the authors. 10. To explain the significance of country effects, the Social Safety Protection index was obtained from the World

Economic Forum and included as an explanatory variable. The results suggest that national social safety systems are a factor behind these cross-country differences. However, we are not confident that this variable is sufficiently robust in terms of its definition and the data supporting the index to warrant inclusion in the reported results.

10

[Insert Table 4 About Here]

Between country groups results In the previous discussion, we focused on commonalities and differences within groups of countries found to be more alike, statistically speaking, than other countries included in the survey. It is nonetheless arguable that these five countries are quite unlike one another in terms of their institutional heritage, cultures, and systems of government. It is plausible, moreover, to suggest that Australia and the USA share a British heritage and Mexico and Spain share a language and certain cultural traditions. If so, Switzerland is different. It is arguable that this group is a statistical artefact. Tables 5 and 6 compare the results of the country-specific estimates for the UK and two other countries in the small European group with those of the East Asian group, most notably Hong Kong and Malaysia. At issue, in the first instance, was whether Australia and the USA share statistically significant explanatory variables with the UK, notwithstanding marked differences in terms of the proportion of survey respondents holding income protection insurance. As in most cases, it was found that the most important explanatory variables for the UK were income and having experienced a loss of income due to some physical impairment. In addition, age (like Australia), financial literacy (like the USA), and the experience of others (like Australia) were also statistically significant for the UK. But unlike Australia and the USA, in the UK it was found that household size was statistically significant. [Insert Table 5 About Here]

We also compared the results for the European group of countries with the group containing Australia and correlated countries. Once again, income was most important for Germany and Italy (see also Table 7). On experience, however, there were mixed results. Rather than one’s own experience being significant, for Germany it was the experience of others whom respondents knew personally that was significant. In the Italian case, experience of any kind was significant. Also unusual was the finding that gender was significant in both countries (that is males were more likely to hold insurance) while household size was significant in Italy, matching the result for Switzerland and the USA. These differences could be thought marginal, given the importance of income across all countries (see Table 5). But differences between countries by age, gender, household size, and whether the respondent was the primary wage earner separately and in combination reflect significant institutional differences between countries, notwithstanding similar levels of development (Christopherson 2002). [Insert Table 6 About Here]

Likewise, we compared the results for the Anglo-Saxon countries with Hong Kong and Malaysia. In Table 6, it is shown that neither age nor gender was significant for either of the latter two countries. In Hong Kong, being the primary and/or sole wage earner was important, but not significant for Malaysia. Unlike most other countries, in Hong Kong and Malaysia being healthier than average had a positive effect on holding income protection insurance, as was having experienced a loss of income due to some impairment. In both cases, however, at issue was the experience of some loss of income due to mental ill health rather than a physical cause. Also different from most other country results was the finding that financial literacy was a significant and positive predictor of holding income protection insurance in both countries. Finally, seemingly unrelated to any other country or group of countries was Brazil. Also included in Table 6 are the results for estimating the reference model for this country. Once again, it was found that income and experience were important along with being in full-time employment and being the primary household wage earner. As for experience, it was found that mental rather than physical health was significant.

11

Synthesis of Results In Table 7, we provide a summary of the statistically significant findings by explanatory variable across countries and groups of countries. Here, a distinction is made between levels of significance for the estimated parameters on each explanatory variable, namely 99 per cent for entries in bold, 95 per cent for entries in italics, and 90 per cent for entries in neither bold nor italics. At the highest level of significance, income was the single most important predictor of whether a respondent, regardless of country or group of countries, held income protection insurance. The lower their income against the base case of a high level of income, the less likely respondents were to hold income protection insurance. Note that in Germany, the USA, and East Asia the parameter on this variable scored a lower level of statistical significance, and was not significant for Switzerland. [Insert Table 7 About Here]

Experience of loss of income was shown to be an important predictor of holding income protection insurance. However, as Table 5 shows, personal experience of a loss of income due to physical impairment was only weakly statistically significant in Italy, although more significant in the UK and Germany. By contrast, in the middle group of countries it was uniformly significant, albeit at the 95 per cent level of confidence. As noted above, the European and mixed groups of countries clearly differed from the East Asian ones in terms of the importance attributed to the experience of mental illness in driving the demand for income protection insurance. Here though, for Malaysia the parameter on this variable was not significant. Others’ experience of loss of income due to impairment was significant in all countries. Other explanatory variables were also significant, but country-specific in terms of importance. The only exception was East Asia, where being healthier than average was weakly significant for both Hong Kong and Malaysia in terms of predicting the proportion of those holding income protection insurance. At the same time, there were some surprising results. For example, in the UK (strong), the USA (medium), and Germany (weak), it was found that financial literacy was negatively associated with the propensity to hold income protection insurance, but in Hong Kong and Malaysia, by contrast, financial literacy was positively associated with it. Being younger was either weakly or moderately significant as a predictor of holding income protection insurance in three countries, thereby joining Germany and the UK to Australia. Surprisingly, perhaps, gender was significant in its own right for only Germany and Italy. Furthermore, it was found that when respondents’ ‘primary’ or ‘secondary’ wage earner status was excluded from the estimated model, gender remained insignificant for the USA and Hong Kong. By contrast, it was significant in Italy and more significant when we exclude the ‘primary/secondary wage earner’ variable. To understand the significance of the results, consider the commonalities and differences between the group of European countries and the mixed group of countries. Here, as noted above, there are statistically significant differences between the two groups in terms of the average propensity of their respondents to hold income protection insurance. Where they differ is in the significance attributed to having experienced loss of income due to physical impairment. By implication, what joins the respondents of the mixed group of countries is this kind of shared experience. More broadly, this type of experience could be significant for all respondents because their government and/or employer-provided benefits deliver limited coverage in this area or leave it to the discretion of the individual. With respect to Hong Kong and Malaysia, it is possible that each, in its own way, offers adequate government and employer benefits in the event of physical impairment. Respondents had the option to identify instances when their income was interrupted due to any of a range of physical or mental health problems. Respondents identifying a past experience of mental impairment as a key driver in holding income protection insurance could reflect fundamental shortfalls in the benefit systems of their respective countries. Alternatively, it is possible that the significance attributed to mental health

12

reflects the hurly-burly of everyday life, especially for younger respondents seeking success in highly competitive environments.11 Implications and Conclusions In this paper, we have used a cross-national survey of employed people to assess the determinants of the propensity to hold income protection insurance. The paper was framed with respect to the argument that nation-states remain important as the geographical lens through which to understand individuals’ welfare and prospects. We also sought to determine the extent to which there are commonalities amongst respondents notwithstanding their different countries of residence. In doing so, our paper is relevant to the on-going debate about the significance or otherwise of neoliberalism and the extent to which working people in developed economies have been exposed to the vicissitudes of global financial markets and local labour markets (see Harvey 2007; Langley 2008). Our paper is not a direct test of the statistical significance of neoliberalism – given the challenges inherent in the implementation of the survey across countries it is appropriate to be somewhat circumspect about the implications to be drawn from our study. Nonetheless, the results offer insights into the terms of the debate about neoliberalism (see Weller and O’Neill 2014), along with findings that deserve greater scrutiny. Take, for example, our search for commonalities between countries in terms of their residents’ propensity to hold income protection insurance. We should recall that an important thread in social science research over the past 25 years has been around the theme of nation-state regimes of accumulation, governance, and welfare. Esping-Andersen (1990, 1999) argued that welfare states should be understood in terms of their shared traditions as well as their differences that give rise to very different institutions and policy practices. Hall and Soskice (2001) argued that nation-states could be grouped according to ‘varieties of capitalism’, an argument that economic geographers including Clark and Wójcik (2007), Peck and Theodore (2011), and more recently by Dixon (2014) have analysed and developed. The results showed that it is meaningful to distinguish between two groups of countries – European and East Asian – along with a third group, which we concluded was a statistical artefact rather than the expression of shared institutional and/or national policy principles and practices. We found stark differences between groups of countries in terms of the propensity of individuals to hold income protection insurance represented at the low end of the distribution by Germany (0.183) and at the high end by Hong Kong (0.675). In between, the USA’s propensity was (0.448). Residents of countries at the low end of the distribution might possibly either believe in, or have experienced, the integrity of government programmes and policies that underwrite their incomes, or feel that they cannot afford to pay the premiums for income protection insurance. At the high end of the distribution, it seems that our respondents make their own provision for income protection for specific reasons, including the expected value of government benefits relative to their income and the fact that they are able to pay the requisite premiums. In framing the debate on institutions and behaviour vis-à-vis income protection, we referred to recent research in the UK and USA on the vulnerability of many employed lower paid workers and their families to interruptions in employment. We found that in terms of the propensity to hold income protection insurance, the UK (0.214) was aligned with Germany and Italy rather than the mixed group, which included the United States (0.448). One explanation of the finding might be that many of our UK respondents, referencing the heritage of the UK welfare state rather than current circumstances, systematically underestimate their vulnerability. Another explanation may be that the rapid growth in UK employment since 1995, notwithstanding the costs and consequences of the global financial

11. Inspecting the data for outliers that might have biased the results revealed that the proportions of younger

and full-time employed respondents in Hong Kong and Malaysia were higher than in the other countries. Further details are available from the authors.

13

crisis, has encouraged the average private sector employee to overestimate the probability of remaining employed. Our UK respondents also tend to believe that the NHS will carry the burden associated with any health-related interruption to earned income. By contrast, it is arguable that many of our USA respondents understand only too well the consequences of significant interruptions in earned income. While country of residence and the association of one country with a group of others were meaningful in terms of understanding variation in the propensity to hold income protection insurance, we also found that individual attributes were important in predicting who will and who will not hold income protection insurance. For example, being the primary wage earner of a multi-income household, having average or higher earned incomes, being younger, and having experienced a significant health-related event were statistically important across the countries represented in the survey. There were, as indicated above, some differences between countries on these issues. Nonetheless, experience is important, as are shared socio-demographic attributes, in driving the propensity to hold income protection insurance. We draw three implications from these results. First, any country- or regime-specific explanation of the propensity to hold income protection insurance would be incomplete without reference to the shared attributes of individuals across countries. Second, those who hold income protection insurance are in the middle to upper echelons of wage earners (by country). Our respondents are obviously dissimilar in terms of their earned incomes if we were to benchmark those incomes against the US dollar. However, they share similar social and labour market positions within the countries represented in our sample. It might be argued that this is evidence of a universal model of human behaviour that discounts geography in favour of the shared attributes of individuals. Even so, taking into account shared attributes leaves a significant country effect. We should also consider the finding that the lower a respondent’s income, the lower the propensity to hold income protection insurance. The most obvious interpretation is that holding insurance is an issue of affordability: given a constrained budget and limited capacity to reallocate a portion of current income to insure against a risk that might seem remote, one response might be to hope for the best. By this logic, risk aversion is mediated by individuals’ presumed capacity to ‘manage’ such an issue (cf. Kahneman and Tversky 1979). In the cases of Germany and Italy, however, it is arguable that the lack of significance of this type of experience suggests that government and/or employer-provided benefits provide sufficient insurance against such an event that it is not salient to many of those countries’ respondents (Bordalo et al. 2012). In this regard, it is notable that even in the case of the UK the weak significance of experience could well represent reliance, especially among older respondents, on the NHS and remnants of the welfare state. Third, we were surprised by the significance of experience over financial literacy. Most important was the experience of the respondent, less important was the experience of someone they knew. Having suffered the consequences of such an event, or having seen others affected, is consistent with behavioural notions of regret (looking backwards) and loss aversion (looking forward) (Kahneman and Tversky 1979). It is also consistent with theories of learning and behaviour in which experience is assumed to trump formal knowledge (Arrow 1962; Gertler 2003). This is one explanation for the lack of significance, in most cases, of financial literacy. Instead of assuming that it ipso facto establishes the relevance of an issue or product like income protection insurance, financial literacy is, at best, a tool or instrument through which to make a decision after the salience of that decision has been established by experience. Appendix Any survey of household behaviour and expectations is a challenging project. There are trade-offs to be made between the desired scope of a survey (the number of questions asked) and maintaining the attention of the

14

respondent (answering each question); between the desired sophistication of the questions posed and the comprehension of the average respondent; and often also between the specificity of the questions posed and the particularities of each respondent – the more specific a question and the more focused it is on a specific aspect of household behaviour, the more likely some respondents will find it irrelevant. More often than not, researchers have an overriding interest in respondents completing their surveys: incomplete responses can bring into doubt the plausibility of statistical analysis and the implications drawn thereof. Many surveys of household behaviour and expectations tackle readily recognized issues, such as how respondents earn and spend their incomes. These types of studies have an advantage in that respondents typically have a direct and material interest in their current incomes and commitments. These issues are largely routine and are rarely disrupted by unanticipated events. The evidence suggests that, in these circumstances, average respondents have a good grasp of their spending patterns. Even so, given the upfront costs involved in launching a survey and analysing its data, these types of surveys are normally one-offs. Studies that follow respondents week-in and week-out focusing on flows of income and changing commitments to consumption and savings (Samphantharak and Townsend 2010) are much rarer. It is not surprising that the preferred mode of data collection in these situations is to focus on small groups of respondents using diaries and repeated interviews to collect (unrepresentative) data (Morduch and Schneider 2017). Our study is more ambitious than most. Whereas many surveys of household behaviour and expectations focus on a specific population in a particular place, ours holds constant the desired population (employed individuals aged between 25 and 64) and seeks evidence of variations in responses within and between country-specific groups of respondents. The survey instrument was designed to be sufficiently simple to apply to different age cohorts and country circumstances. At the same time, recognizing that there are differences between countries’ employment benefits and welfare institutions, the survey was designed to focus on an aspect of long-term welfare that all individuals and households would recognise. Just as the World Bank’s Living Standards and Measurement Study is sensitive to cross-country variations in people’s circumstances, we were sensitive to the implementation and translation of the survey across countries and languages.12 Simplicity and transparency were important in ensuring that translation was, as much as possible, effective and meaningful. There is, however, a further complication: whereas many surveys of household behaviour and expectations focus on routine matters related to monthly income and expenditure, ours focused on the responses our respondents might deploy, or have deployed, in relation to a rare event – an unexpected and significant interruption to their earned income (due to a health-related situation) that may or may not have been previously experienced. We found that the incidence of respondents having experienced such an event is relatively low within each country and across the countries included in our sample. Nonetheless, this type of event can have significant long-term consequences for individual and household welfare. Two implications follow. First, following a cohort using a high-frequency survey instrument and/or diaries was inappropriate to this type of issue. Second, acknowledging and bringing to the forefront of empirical analysis the existence of variations between respondents by socio-demographic characteristics and by country was the over-riding priority of the research programme (something rarely found in other studies).

References Arrow, K. J. (1962) ‘The economic implications of learning by doing’, Review of Economic Studies, 29

(3): 155–73. Bathelt, H. and Glückler, J. (2011) The Relational Economy: Geographies of Knowing and Learning,

Oxford: Oxford University Press. Bordalo, P., Gennaioli, N. and Shleifer, A. (2012) ‘Salience theory of choice under risk’, Quarterly

Journal of Economics, 127 (3): 1243–85. Boyer, R. (2000) ‘Is a finance-led growth regime a viable alternative to Fordism? A preliminary

analysis’, Economy and Society, 29(1): 111–45. Brennan, G., Eriksson, L., Goodin, R. E. and Southwood, N. (2014) Explaining Norms, Cambridge:

Cambridge University Press.

12. http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/EXTLSMS/0,,contentMDK:23636522~pagePK:64168445

~piPK:64168309~theSitePK:3358997,00.html

15

Carnes, M. E. and Mares, I. (2007) ‘The welfare state in global perspective’, in Boix, C. and Stokes, S. C. (eds) The Oxford Handbook of Comparative Politics, Oxford: Oxford University Press, 868–85.

Christopherson, S. (2002) ‘Why do national labor market practices continue to diverge in the global economy? The “missing link” of investment rules’, Economic Geography, 78(1): 1–20.

Clark, G. L (2014) ‘Roepke Lecture in Economic Geography: Financial literacy in context’, Economic Geography, 90(1): 1–23.

Clark, G. L. (2017) Behaviour in context, in Clark, G. L., Feldman, M., Gertler, M. S. and Wójcik, D. (eds) The New Oxford Handbook of Economic Geography, Oxford: Oxford University Press (forthcoming).

Clark, G. L. and Wójcik, D. (2007) The Geography of Finance: Corporate Governance in the Global Marketplace, Oxford: Oxford University Press.

Clark, G. L. Strauss, K. and Knox-Hayes, J. (2012) Saving for Retirement: Intention, Context, and Behavior, Oxford: Oxford University Press.

Davies, W (2014) The Limits of Neoliberalism, London: Sage Publications. Dixon, A (2014) The New Geography of Capitalism: Firms, Finance, and Society, Oxford: Oxford

University Press. Ebbinghaus, B. (ed.) (2011) The Varieties of Pension Governance: Privatization of Pensions in Europe,

Oxford: Oxford University Press. Engelen, E. (2003) ‘The logic of funding European pension restructuring and the dangers of

financialization’, Environment and Planning A, 35(8): 1357–72. Esping-Andersen, G. (1990) The Three Worlds of Welfare Capitalism, Cambridge: Polity Press. Esping-Andersen, G. (1999) Social Foundations of Postindustrial Economies, Oxford: Oxford University

Press. Gertler, M. S. (2003) ‘Tacit knowledge and the economic geography of context, or the undefinable

tacitness of being (there)’, Journal of Economic Geography, 3 (1): 75–99. Gertler, M. S. (2017) ‘Institutions, geography and economic life’, in Clark, G. L., Feldman, M., Gertler,

M. S. and Wójcik, D. (eds) The New Oxford Handbook of Economic Geography, Oxford: Oxford University Press (forthcoming).

Grabher, G. (ed.) (1993) The Embedded Firm: On the Socioeconomics of Industrial Networks, London: Routledge.

Granovetter, M. (1985) ‘Economic action and social structure: The problem of embeddedness’, American Journal of Sociology, 91(3): 481–510.

Hall, P. A. and Soskice, D. W. (2001) Varieties of Capitalism: The institutional Foundations of Comparative Advantage, Oxford: Oxford University Press.

Harvey, D. (2007) A Brief History of Neoliberalism, Oxford: Oxford University Press. Kahneman, D. and Tversky, A. (1979) ‘Prospect theory: an analysis of decision under risk’,

Econometrica, 47: 263–91. Langley, P. (2008) The Everyday Life of Global Finance: Saving and Borrowing in Anglo-America,

Oxford: Oxford University Press. Legal & General (2014) On the Brink: Deadline to the Breadline Report, London: Legal & General. Leyshon, A., Burton, D., Knights, D., Alferoff, C. and Signoretta, P. (2004) ‘Towards an ecology of retail

financial services: understanding the persistence of door-to-door credit and insurance providers’, Environment and Planning A, 36(4): 625–45.

Leyshon, A., Burton, D., Knights, D., Alferoff, C. and Signoretta, P. (2006) ‘Walking with moneylenders: the ecology of the UK home-collected credit industry’, Urban Studies, 431: 161–86.

Lusardi, A. and Mitchell, O. S (2014) ‘The economic importance of financial literacy: theory and evidence’, Journal of Economic Literature’, 52(1): 5–44.

Lusardi, A., Schneider, D. J., and Tufano, P. (2009) ‘Households of risk: a cross-country study of household financial risk’, mimeo.

Lusardi, A., Schneider, D. J. and Tufano, P. (2011) ‘Financially fragile households: evidence and implications’, Working paper 17072, Cambridge MA: National Bureau of Economic Research.

16

Lusardi, A., Schneider, D. J. and Tufano, P. (2015) ‘The economic crisis and medical care use: comparative evidence from five high income countries’, Social Science Quarterly, 96(1): 202–13.

McDowell, L., Batnitzky, A. and Dyer, S. (2008) ‘Internationalization and the spaces of temporary labour: the global assembly of a local workforce’, British Journal of Industrial Relations, 46(4): 750–70.

McDowell, L., Batnitzky, A. and Dyer, S. (2009) ‘Precarious work and economic migration: emerging immigrant divisions of labour in Greater London’s service sector’, International Journal of Urban and Regional Research, 33(1): 3–25.

Morduch, J. and Schneider, R. (2017) The Financial Diaries: How American Families Cope in a World of Uncertainty, Princeton: Princeton University Press.

Olson, M. (1965) The Logic of Collective Action: Public Goods and the Theory of Groups, Cambridge MA: Harvard University Press.

Peck, J. and Theodore, N. (2012) Fast Policy: Experimental Statecraft at the Thresholds of Neoliberalism, Minneapolis: University of Minnesota Press.

Phillips, A. (2013) Our Bodies, Whose Property? Princeton: Princeton University Press. Preda, A. (2009) Framing Finance: The Boundaries of Markets and Modern Capitalism, Chicago:

University of Chicago Press. Rawls, J. (1971) A Theory of Justice, Cambridge MA: Harvard University Press. Rodrik, D. (2013) ‘Roepke lecture in economic geography: who needs the nation‐state?’ Economic

Geography, 89(1): 1–19. Samphantharak, K. and Townsend, R. M. (2010) Households as Corporate Firms: An Analysis of

Household Finance using Integrated Household Surveys and Corporate Financial Accounting, Cambridge: Cambridge University Press.

Storper, M. J. (2016) ‘The neo-liberal city as idea and reality’, Territory, Politics, Governance, 4 (2): 241–63.

Strauss, K. (2017) ‘Precarious work and winner take-all economies’, in Clark, G. L., Feldman, M., Gertler, M. S. and Wójcik, D. (eds) The New Oxford Handbook of Economic Geography, Oxford: Oxford University Press (forthcoming).

Streeck, W. (2014) Buying Time: The Delayed Crisis of Democratic Capitalism, London: Verso Press. Weil, D. (2014) The Fissured Workplace: Why Work Became So Bad for So Many and What Can Be Done

to Improve It, Cambridge MA: Harvard University Press. Weller, S. and O’Neill, P. (2014) ‘An argument with neoliberalism: Australia’s place in a global

imaginary’, Dialogues in Human Geography, 4(2): 105–30.

17

Table 1. Variable description and basic statistics Variable Description Mean Std. dev. Min. Max.

Demand for insurance

Do you have insurance Binary variable (it takes the value 1 if the respondent does have insurance; 0, otherwise)

0.36 0.48 0 1

Socio-economic variables

Gender Binary variable( it takes the value 1 for male respondents, 0 for females)

0.48 0.50 0 1

Age (groups) 25–35 years 36–45 years 46–55 years 56–60 years

0.35 0.29 0.26 0.10

0.48 0.45 0.44 0.30

0 0 0 0

1 1 1 1

Income* Lower than average Equal than average Higher than average

0.36 0.39 0.25

0.47 0.48 0.46

0 0 0

1 1 1

Household size # people household 2.99 1.41 1 7

Education # years 14.79 4.86 0 25

Work status

Work status Full-time Part-time Self-employed Not employed

0.65 0.18 0.13 0.04

0.48 0.39 0.34 0.19

0 0 0 0

1 1 1 1

Health

Self-reported health status

Healthier than average Same as healthy Less healthy

0.26 0.62 0.12

0.44 0.49 0.32

0 0 0

1 1 1

Financial literacy

Financial literacy Correctly responded Q1 (interest rates) Correctly responded Q2 (inflation rates) Correctly responded Q3 (risk diversification) All three correct At least 1 do not know

0.66 0.55 0.74 0.38 0.20

0.47 0.50 0.44 0.49 0.40

0 0 0 0 0

1 1 1 1 1

Experience of loss of income

Experience (Have you ever experienced a loss of income due to any of these causes?)

Physical cause Emotional cause

0.10 0.25

0.30 0.43

0 0

1 1

Do you know someone who has experienced a loss of income due to illness or disability

Binary variable (it takes the value 1 if the respondent knows someone; 0, otherwise)

0.54 0.50 0 1

Country of residence

Country of residence Australia Brazil Germany Hong Kong Italy Malaysia Mexico Spain Switzerland UK USA

0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.11 0.09

0.28 0.28 0.29 0.29 0.29 0.29 0.28 0.29 0.28 0.31 0.29

0 0 0 0 0 0 0 0 0 0 0

1 1 1 1 1 1 1 1 1 1 1

* Respondents were divided into three different groups depending on their income relative to the mean income in their country of residence. Source: authors

18

Table 2. Propensity to hold insurance by country of residence.

Mean Std. dev. n

Australia 0.308 0.462 912

Brazil 0.247 0.431 940

Germany 0.183 0.387 980

Hong Kong 0.675 0.468 958

Italy 0.235 0.424 965

Malaysia 0.667 0.471 983

Mexico 0.371 0.483 978

Spain 0.286 0.452 911

Switzerland 0.334 0.472 850

UK 0.214 0.411 1180

USA 0.448 0.498 924

F=140.85***

Source: Authors. F is the statistic used to test the null hypothesis that all the means are equal; *** means significant at a 1 per cent level.

Table 3. Demand for Insurance (estimated coefficients and marginal effects)

Explanatory

variables

Indicator Model 1 Model 2 Model 3

Coeff. Marginal effects Coeff. Marginal effects Coeff. Marginal effects

Gender Male 0.070 (1.26) 0.014 (1.27) 0.058 (1.14) 0.011 (1.15) 0.068 (1.10) 0.013 (1.10)

Age 25–35 years

36–45 years

46–55 years

0.577*** (4.75)

0.343*** (4.42)

0.145* (1.71)

0.112*** (4.97)

0.067*** (4.52)

0.028 (1.71)

0.533*** (4.55)

0.312*** (4.34)

0.131 (1.62)

0.103***(4.76)

0.060***(4.46)

0.025 (1.62)

0.412*** (4.51)

0.285*** (4.29)

0.144* (1.75)

0.077***(4.66)

0.053***(4.31)

0.027* (1.75)

Income Lower than av.

Equal

–1.045*** (–6.47)

–0.499*** (–4.44)

–0.203*** (–6.88)

–0.097***(–4.61)

–0.978*** (–6.13)

–0.477*** (–4.21)

–0.190***(–6.47)

–0.092***(–4.35)

–0.997*** (–6.61)

–0.485*** (–4.74)

–0.186***(–6.98)

–0.091***(–4.89)

Primary/

secondary

wage

earner

Only wage

Primary wage

Equal wages

Second wage

0.748**(2.31)

0.789**(2.48)

0.525 (1.59)

0.188 (0.60)

0.145**(2.30)

0.153**(2.47)

0.102 (1.58)

0.036 (0.59)

0.588*(1.92)

0.633*(2.14)

0.389 (1.24)

0.101 (0.35)

0.114*(1.91)

0.123*(2.13)

0.075 (1.23)

0.020 (0.35)

0.527*(1.90)

0.565**(2.05)

0.371 (1.26)

0.182 (0.66)

0.098*(1.88)

0.105**(2.04)

0.069 (1.25)

0.032 (0.65)

Household

size

0.099** (2.34) 0.019**(2.37) 0.099** (2.29) 0.019**(2.32)

0.061* (1.90) 0.011* (1.93)

Education # years 0.001 (0.15) 0.0002 (0.15) 0.002 (0.20) 0.0003 (0.20) 0.009 (1.39) 0.002 (1.36)

Work status Full-time

Part-time

Self-employed

0.319* (1.72)

0.121 (0.49)

0.075 (0.42)

0.062* (1.73)

0.023 (0.49)

0.015 (0.42)

0.347* (1.85)

0.124 (0.52)

0.065 (0.35)

0.065* (1.85)

0.023 (0.53)

0.012 (0.35)

Financial

literacy

–0.076 (–1.22) –0.014 (–1.24)

Health Healthier

Same as

healthy

0.402***(5.23)

0.220***(3.12)

0.075***(5.11)

0.041***(3.06)

Experienced

loss of

income

Physical

Mental

0.636***(6.45)

0.352***(5.64)

0.119***(6.69)

0.066***(4.63)

Know

someone

0.465***(10.00) 0.087***(10.57)

Country

effects

Yes Yes Yes Yes Yes Yes

Number

obs

10060 10060 10060 10060 10060 10060

Pseudo R2 0.1285 0.1285 0.1298 0.1298 0.1554 0.1554

Source: Authors. Standard errors are clustered by country. *, ** and *** mean significant at the 10 per cent, 5 per cent and 1 per cent significant levels.

19

Table 4. Demand for Insurance by Country (estimated coefficients and marginal affects)

Explanatory variables Indicator Country

Australia Mexico Spain Switzerland US(A)

Coeff. Marginal effects Coeff. Marginal effects Coeff. Marginal effects Coeff. Marginal effects Coeff. Marginal effects

Gender Male 0.202 (1.19) 0.039 (1.19) –0.110 (–0.71) –0.023 (–0.71) 0.137 (0.83) 0.026 (0.83) 0.243 (1.29) 0.051 (1.30) 0.046 (0.29) 0.009 (0.29)

Age 25–35 years

36–45 years

46–55 years

0.563** (2.03)

0.335 (1.22)

0.367 (1.39)

0.108** (2.05)

0.064 (1.22)

0.070 (1.39)

0.429 (1.41)

0.315 (1.01)

0.289 (0.93)

0.088 (1.42)

0.065 (1.01)

0.059 (0.93)

0.260 (0.92)

0.310 (1.14)

–0.163 (–0.58)

0.049 (0.92)

0.058 (1.14)

–0.031 (–0.58)

0.461 (1.64)

0.046 (0.16)

0.200 (0.71)

0.097 (1.65)

0.010 (0.16)

0.042 (0.71)

0.149 (0.55)

–0.036 (–0.13)

–0.300 (–1.20)

0.027 (0.55)

–0.006 (–0.13)

–0.057 (–1.20)

Income Lower than av.

Equal

–1.370*** (–4.40)

–0.695** (–3.69)

–0.262***

(–4.57)

–0.133***

(–3.80)

–1.397*** (–7.08)

–0.903*** (–5.10)

–0.286*** (–2.45)

–0.185* (–1.76)

–0.94*** (–2.93)

–0.569* (–1.92)

–0.176***

(–2.93)

–0.107* (–1.92)

–0.051 (–0.20)

0.0581 (0.30)

–0.001 (–0.20)

0.012 (0.30)

–0.570** (–2.45)

–0.453** (–2.51)

–0.109** (–2.45)

–0.086** (–2.51)

Primary/

secondary wage earner

Only wage

Primary wage

Equal wages

Second wage

–0.148 (–0.34)

–0.247 (–0.55)

–0.538 (–1.15)

–0.637 (–1.39)

–0.028 (–0.34)

–0.047 (–0.55)

–0.103 (–1.15)

–0.122 (–1.39)

0.306 (0.46)

0.216 (0.33)

0.198 (0.30)

–0.141 (–0.21)

0.063 (0.46)

0.044 (0.33)

0.041 (0.30)

–0.029 (–0.21)

1.319 (1.60)

0.967 (1.17)

0.865 (1.05)

0.857 (1.05)

0.247 (1.60)

0.181 (1.17)

0.162 (1.05)

0.161 (1.05)

1.180 (1.04)

1.124 (0.99)

1.415 (1.24)

1.471 (1.29)

0.250 (1.04)

0.238 (0.99)

0.300 (1.24)

0.311 (1.29)

2.065* (1.91)

2.300** (2.13)

2.084* (1.91)

1.632 (1.50)

0.392* (1.91)

0.436** (2.13)

0.395* (1.91)

0.307 (1.50)

Household size 0.02 (0.39) 0.004 (0.39) –0.012 (–0.22) –0.003 (–0.22) 0.163** (2.31) 0.031** (2.31) 0.180** (2.38) 0.038** (2.38) 0.200*** (2.97) 0.038*** (2.97)

Education # years 0.013 (0.70) 0.002 (0.71) 0.030 (1.63) 0.006 (1.63) 0.020 (1.20) 0.004 (1.20) 0.028 (1.51) 0.006 (1.51) –0.010 (–0.59) –0.002 (–0.59)

Work status Full-time

Part-time

Self-employed

–0.454 (–1.17)

–0.873** (–2.26)

–0.791* (–1.92)

–0.087 (–1.17)

–0.167** (–2.28)

–0.151* (–1.93)

0.809 (1.26)

1.057 (1.60)

0.385 (0.60)

0.166 (1.26)

0.217 (1.60)

0.079 (0.60)

0.003 (0.01)

–0.322 (–0.70)

–0.109 (–0.23)

0.000 (0.01)

–0.060 (–0.70)

–0.20 (–0.23)

–0.075 (–0.13)

–0.091 (–0.15)

0.049 (0.08)

–0.016 (–0.13)

–0.019 (–0.15)

0.010 (0.08)

0.587 (1.23)

0.048 (0.10)

0.044 (0.08)

0.110 (1.23)

0.007 (0.10)

0.008 (0.08)

Financial literacy –0.046 (–0.53) –0.009 (–0.53) –0.128 (–1.58) –0.026 (–1.58) –0.09 (–1.07) –0.017 (–1.07) –0.052 (–0.57) –0.011 (–0.57) –0.280*** (–

3.21)

–0.053*** (–

3.21)

Health Healthier

Same as healthy

0.345 (1.32)

0.106 (0.45)

0.066 (1.33)

0.020 (0.45)

0.819** (2.05)

0.513 (1.29)

0.168** (2.05)

0.105 (1.29)

–0.038 (–0.14)

0.008 (0.03)

–0.007 (–0.14)

0.002 (0.03)

0.324 (1.13)

0.353 (1.35)

0.069 (1.13)

0.075 (1.35)

0.248 (0.86)

–0.195 (–0.73)

0.047 (0.86)

–0.038 (–0.73)

Experienced loss of income Physical

Mental

0.633** (2.26)

0.439** (2.39)

0.121** (2.29)

0.084** (2.41)

0.772*** (2.67)

–0.092 (–0.53)

0.158*** (2.67)

–0.019 (–0.53)

0.951*** (3.26)

0.435** (2.31)

0.178*** (3.26)

0.081** (2.31)

0.912** (2.49)

0.130 (0.65)

0.193** (2.53)

0.028 (0.65)

0.673** (2.07)

0.398** (2.07)

0.128** (2.07)

0.074** (2.07)

Know someone 0.617*** (3.69) 0.118*** (3.80) 0.302* (1.95) 0.062* (1.95) 0.505*** (2.93) 0.095*** (2.93) 0.459*** (2.71) 0.097*** (2.71) 0.419** (2.59) 0.080** (2.59)

Number obs 844 844 928 928 894 894 773 773 918 918

Pseudo R2 0.0992 0.0992 0.0989 0.0989 0.0735 0.0735 0.0424 0.0424 0.1850 0.185

*, ** and *** mean significant at the 10 per cent, 5 per cent and 1 per cent significant levels.

Source: Authors

20

Table 5. Demand for Insurance by Country: Germany, Italy and the UK (estimated coefficients and marginal effects)

Explanatory

Variables

Indicator Germany Italy UK

Model 3 Model 3 Model 3

Coeff. Marginal Effects Coeff. Marginal Effects Coeff. Marginal Effects

Gender Male 0.547*** (2.65) 0.070*** (2.65) 0.408** (2.26) 0.065** (2.26) –0.223 (–1.28) –0.031 (–1.28)

Age 25–35 years

36–45 years

46–55 years

0.895*** (2.64)

0.495 (1.41)

0.140 (0.41)

0.114*** (2.64)

0.063 (1.41)

0.018 (0.41)

0.029 (0.10)

0.058 (0.21)

0.046 (0.16)

0.005 (0.10)

0.009 (0.21)

0.007 (0.16)

0.822** (2.40)

0.685** (1.99)

0.603* (1.77)

0.116** (2.40)

0.097** (1.99)

0.085* (1.77)

Income Lower than av.

Equal

–0.834** (–2.31)

–0.423** (–1.97)

–0.107** (–2.31)

–0.054** (–1.97)

–1.87*** (–6.59)

–0.845*** (–4.01)

–0.300*** (–

6.59)

–0.135*** (–

4.01)

–1.50*** (–4.77)

–0.511** (–2.43)

–0.211*** (–4.77)

–0.072** (–2.43)

Primary/

secondary

wage earner

Only wage

Primary wage

Equal wages

Second wage

–1.121 (–0.91)

–0.919 (–0.75)

–0.915 (–0.74)

–1.573 (–1.25)

–0.143 (–0.91)

–0.117 (–0.75)

–0.117 (–0.74)

–0.201 (–1.25)

–0.136 (–0.28)

–0.008 (–0.02)

–0.403 (–0.80)

–0.244 (–0.49)

–0.022 (–0.28)

–0.001 (–0.02)

–0.064 (–0.80)

–0.039 (–0.49)

–0.186 (–0.23)

–0.182 (–0.23)

–0.879 (–1.05)

–1.082 (–1.34)

–0.026 (–0.23)

–0.026 (–0.23)

–0.124 (–1.05)

–0.153 (–1.34)

Household

size

0.073 (0.83) 0.009 (0.83) –0.128 (–1.64) –0.020 (–1.64) 0.208*** (3.10) 0.029*** (3.10)

Education # years –0.008 (–0.38) –0.001 (–0.38) –0.012 (–0.67) –0.002 (–0.67) –0.028* (–1.67) –0.004* (–1.67)

Work status Full-time

Part-time

Self-employed

1.137 (1.42)

0.817 (0.98)

1.206 (1.42)

1.145 (1.42)

0.104 (0.98)

0.154(1.42)

0.851 (1.26)

0.875 (1.24)

0.930 (1.36)

0.136 (1.26)

0.140 (1.24)

0.149 (1.36)

0.303 (0.34)

–0.011 (–0.01)

0.020 (0.02)

0.043 (0.34)

–0.001 (–0.01)

0.003 (0.02)

Financial

literacy

–0.296*** (–

2.88)

–0.038*** (–

2.88)

–0.138 (–1.52) –0.022 (–1.52) –0.348*** (–4.11) –0.049*** (–4.11)

Health Healthier

Same as healthy

0.216 (0.60)

0.509 (1.62)

0.028 (0.60)

0.065 (1.62)

0.331 (1.00)

–0.037 (–0.12)

0.053 (1.00)

–0.006 (–0.12)

0.691** (2.42)

0.425 (1.62)

0.097** (2.42)

0.060 (1.62)

Experienced

loss of income

Physical

Mental

0.420 (1.30)

0.441** (2.00)

0.054 (1.30)

0.056** (2.00)

0.704* (1.88)

0.641*** (2.63)

0.113* (1.88)

0.102*** (2.63)

0.976*** (3.85)

0.006 (0.03)

0.138*** (3.85)

0.001 (0.03)

Know

someone

0.856*** (4.37) 0.109*** (4.37) 0.424** (2.45) 0.068** (2.45) 0.523*** (3.17) 0.074*** (3.17)

Number obs 953 953 929 929 1130 1130

Pseudo R2 0.1361 0.1361 0.1003 0.1003 0.1627 0.1627

*, ** and *** mean significant at the 10 per cent, 5 per cent and 1 per cent significant levels.

Source: Authors

21

Table 6. Demand for Insurance by Country: Hong Kong, Malaysia, Brazil (estimated coefficients and marginal effects)

Explanatory

Variables

Indicator Hong Kong Malaysia Brazil

Model 3 Model 3 Model 3

Coeff. Marginal Effects Coeff. Marginal Effects Coeff. Marginal Effects

Gender Male –0.239 (–1.54) –0.046 (–1.54) –0.009 (–0.06) –0.002 (–0.06) 0.117 (0.64) 0.020 (0.64)

Age 25–35 years

36–45 years

46–55 years

–0.105 (–0.24)

0.261 (0.60)

–0.146 (–0.33)

–0.020 (–0.24)

0.051 (0.60)

–0.028 (–0.33)

0.051 (0.10)

0.045 (0.09)

0.047 (0.09)

0.011 (0.10)

0.009 (0.09)

0.010 (0.09)

0.606 (1.43)

0.318 (0.73)

0.504 (1.14)

0.102 (1.43)

0.053 (0.73)

0.084 (1.14)

Income Lower than av.

Equal

–0.436** (–2.31)

–0.132 (–0.63)

–0.085** (–2.31)

–0.026 (–0.63)

–0.630** (–2.21)

–0.220 (–0.72)

–0.131** (–2.21)

–0.046 (–0.72)

–1.216*** (–4.14)

–0.766** (–2.52)

–0.204*** (–4.14)

–0.128** (–2.52)

Primary/

secondary wage

earner

Only wage

Primary wage

Equal wages

Second wage

2.472*** (3.08)

2.413*** (3.05)

2.295*** (2.87)

1.974** (2.45)

0.482*** (3.08)

0.470*** (3.05)

0.447*** (2.87)

0.385** (2.45)

0.472 (0.91)

0.396 (0.78)

0.501 (0.97)

0.356 (0.70)

0.098 (0.91)

0.083 (0.78)

0.104 (0.97)

0.074 (0.70)

0.327 (0.79)

0.829** (2.06)

0.277 (0.65)

–0.196 (–0.45)

0.055 (0.79)

0.139** (2.06)

0.046 (0.65)

–0.033 (–0.45)

Household size 0.003 (0.04) 0.001 (0.04) –0.060 (–1.28) –0.013 (–1.28) 0.086 (1.31) 0.014 (1.31)

Education # years 0.019 (1.32) 0.004 (1.32) 0.029* (1.89) 0.006* (1.89) –0.004 (–0.22) –0.001 (–0.22)

Work status Full-time

Part-time

Self-employed

0.993 (1.04)

0.668 (0.67)

0.894 (0.87)

0.193 (1.04)

0.130 (0.67)

0.174 (0.87)

0.710 (1.36)

0.641 (1.06)

0.289 (0.53)

0.148 (1.36)

0.134 (1.06)

0.060 (0.53)

0.786** (2.05)

0.544 (1.25)

0.132 (0.34)

0.132** (2.05)

0.091 (1.25)

0.022 (0.34)

Financial literacy 0.220*** (2.75) 0.043*** (2.75) 0.221*** (2.76) 0.046*** (2.76) 0.047 (0.49) 0.008 (0.49)

Health Healthier

Same as healthy

0.572** (2.22)

0.259 (1.15)

0.111** (2.22)

0.051 (1.15)

0.447 (1.63)

0.468** (2.22)

0.093 (1.63)

0.098** (2.22)

0.063 (0.22)

0.021 (0.08)

0.011 (0.22)

0.003 (0.08)

Experienced loss of

income

Physical

Mental

0.221 (1.09)

0.548*** (2.98)

0.043 (1.09)

0.107*** (2.98)

0.208 (0.73)

0.295 (1.64)

0.043 (0.73)

0.061 (1.64)

0.446* (1.70)

0.613*** (3.11)

0.075* (1.70)

0.103*** (3.11)

Know someone 0.618*** (4.00) 0.121*** (4.00) 0.256* (1.74) 0.053* (1.74) 0.476** (2.28) 0.080** (3.17)

Number obs 942 942 958 958 791 791

Pseudo R2 0.0867 0.0867 0.0435 0.0435 0.1242 0.1242

*, ** and *** mean significant at the 10 per cent, 5 per cent and 1 per cent significant levels. Source: Authors

Table 7. Summary of Significant Findings

Explanatory

Variables

Countries

Germany Italy UK Australia Mexico Spain Switzerland USA Hong

Kong

Malaysia Brazil

Age +ve

Younger

+ve

Younger

+ve

Younger

Income +ve +-ve +-ve +-ve +-ve +-ve +-ve +-ve +-ve +-ve

Primary etc. +ve +ve +ve

Gender

Wage

Earner

+ve

Male

+ve

Male

Household

Size

+ve +ve +ve +ve

Work Status -ve PT -ve PT +ve FT

Financial

Literacy

-ve -ve -ve +ve +ve

Health +ve +ve

Healthier

+ve

Healthier

Experience +ve

Mental

+ve

Physical

Mental

+ve

Physical

+ve

Physical

Mental

+ve

Physical

Mental

+ve

Physical

Mental

+ve

Physical

+ve

Physical

Mental

+ve

Mental

+ve

Mental

Physical

Others’

Experience

+ve +ve +ve +ve +ve +ve +ve +ve +ve +ve +ve

Source: Authors