Private spanner in public works? The corrosive effects of private insurance on public life

Contemporary societies are not only “risk societies”, but also insurance societies. While the shift of systemic risks from the community to the individual is a distinctive trait of modernity, research on the consequences of this process has focused almost exclusively on welfare state responses aimed at re-collectivizing societal risks. Individual-level reactions associated with the need for a private safety net against the uncertainty brought by risk societies have been largely overlooked. What happens to a society and its individuals when private insurance becomes commonplace? Focusing on Germany, we use the data of the German Socio-Economic Panel (1984–2018) to investigate the attitudinal antecedents and consequences of contracting private insurance. As one of the most important sources of private welfare, life insurance attracts risk-averse individuals who are highly concerned with public economic affairs and see the market-based solutions of conservative parties as the best way to safeguard their economic security. While short-term attitudinal effects are absent, a longitudinal approach reveals that becoming insured gradually increases economic security but also entails withdrawal from public life and aversion to parties that support social redistribution. The loss of dynamism of a society may thus be related not only to public welfare but also to a private institution at the heart of the financial markets, which moreover has


| INTRODUCTION: THE UNKNOWN WORLD OF PRIVATE ECONOMIC SECURITY
The rapid social, economic, and technological change characterizing modern societies has led sociologists to refer to them as "risk societies" (Beck, 1992).Not only are they faced with a growing variety of uninsurable risks, but they also increasingly shift the risk burden to individuals, often exacerbating risk inequality in light of welfare state retrenchment (Hacker, 2006).The financial crisis has laid bare the cracks in the risky financial system with the failure of banks and re-insurance companies such as AIG.In addition to the health and social implications of COVID-19, its ongoing negative economic effects are skewed toward the most risk-exposed social strata (Witteveen, 2020).A growing sociology of risk has therefore pointed to rising numbers of risks and their unequal distribution (e.g., Beck, 1992;Busemeyer & Iversen, 2020;Giddens, 1991).
Yet modern societies can equally be described as "insurance societies".With the number of risks, the number and size of insurance products has grown.There is almost no societal or personal risk that insurance companies do not provide coverage for.In addition, the higher the risk awareness in the population, the higher the demand for insurance.Contemporary societies spend more on private insurance than ever before (SwissRe, 2020), and ever larger parts of the world population are now covered by private insurance contracts (OECD, 2020).To date, however, insurance has largely been studied with the state as the "ultimate risk manager" (Moss, 2004), and particularly with the welfare state as an insurance buffer against capitalist risk, not just a redistributive tool (Hacker & Rehm, 2022).
Hence, public insurance and its impact on the individual, society, and politics have been studied in extenso by political science to explain welfare state path dependencies through feedback mechanisms (Montanari, 2001;Svallfors, 2020) and by economists to explain the crowding out of private savings (Alessie et al., 2013).The private financial security people have increasingly been acquiring, with its political and societal effects, has largely gone below the radar, however, despite the fact that private insurance has penetrated modern societies and witnessed a similarly spectacular growth as banks, mortgage debt, stock markets, or other elements of financialized capitalism.Even in the study of financialization (Van der Zwan, 2014), the sociology of insurance is very much unknown and certainly dwarfed by the sociology of risk and welfare.While there have been studies on the attitudinal and political consequences of other forms of private wealth-most notably housing assets (André et al., 2018;Ansell, 2019;Hadziabdic & Kohl, 2021) and total private wealth (Nadeau et al., 2011)-the effects of life insurance assets are widely overlooked.
This paper delves into the attitudinal antecedents and consequences of a "private insurance society".Insurance is not only among the largest assets held by private households but also one of the key outcomes of the creative future orientation which actors need to be equipped with in modern societies (Bourdieu, 1977).Conceiving and personally acting against the risks provoked by the freedoms and burdens associated with rising individualism is intricately enmeshed with an individual's future outlook (Beckert and Bronk, 2018), with imaginaries of good life and death, and family values (Zelizer, 1983).This becomes particularly visible in insurance, with its long-lasting contracts that accompany policyholders across decades of their life course.Insurance is praised as a provider of a sense of security and an enabler of risk-taking and simultaneously decried as a big disincentivizer that kills innovative trends through moral hazard.What happens, then, to a society and its individuals when they are increasingly lulled into the cocoon-like embrace of private insurance?
We answer this question by focusing on Germany, which has high-quality longitudinal data since 1984 covering private insurance status, assets, and a diverse set of attitudes on a yearly basis.As an insurance country, Germany privatizing, welfare-eroding effects.The paper argues for a more general sociology of insurance.

K E Y W O R D S
attitudes, Germany, panel data, private insurance, risk, welfare is both similar and dissimilar to other countries: it is similar because it has a mature insurance industry which has followed similar trends over time across recent decades, but it is clearly different in that its life insurance sector lags behind G7 levels-by about 36% in 2018 (GdV, 2020); and unlike Anglophone countries of "Maritime" insurance tradition (Albert, 1993), in particular, its tax-subsidized products are a long-term and low-risk investment.Life and other types of insurance have a long history in Germany and have seen unprecedented growth over recent decades, despite current and historical levels being lower than those of the United States, which we use here for comparison for it being one of the most privately insured countries belonging to the Anglo-Maritime as opposed to Germany's "Alpine" insurance culture (cf. Figure 1). 1 By about 1980, when the micro-data of the German Socio-Economic Panel (SOEP) we use set in, both life and non-life insurance had recovered from the two war and hyperinflation shocks and almost doubled in size by the time of the Global Financial Crisis (Borscheid, 2012).In peak times, Germans spent more than 7% of annual GDP on insurance, total insurance assets reached 50% of GDP and life insurance savings up to 30% (SwissRe, 2020).It has only declined in recent years due to aging societies paying in less but taking out more and, more recently, due to the low interest rate environment (GdV, 2020).
This intensive growth was preceded by an extensive growth of life insurance policies among the population, reaching on average one policy per inhabitant by the 1980s (cf. Figure 1).Most policies involve both premature death insurance and old-age pension insurance.In 2020, insurers or pension providers held around 40% of private households' financial wealth in these relatively illiquid contracts, a larger item than bank deposits and much more than stocks (Bundesbank, 2021).Insurance has become a de facto third pillar of private welfare.What are the micro implications of this increasing insurance penetration in German society?
Relying on SOEP panel data between 1984 and 2018, we examine the way general and economic risk perceptions, political attitudes, and opinions on the role of the state as a risk redistributor act as precursors of the decision to become insured and whether the experience of becoming insured does itself modify these attitudes.A descriptive analysis confirms existing priors: insurance demand is highest among risk-averse, politically active conservatives leaning toward private welfare.Individuals self-selecting into insurance worry about general economic and financial risks, but not their personal risks, as they are already sufficiently well-off, a finding which may reflect the particular nature of German life insurance products across the last decades.Fixed effects models show that almost none of the self-selection effects is accompanied by a short-term causal one.Yet a dynamic analysis reveals that the causal structure of the impact of insurance is more intricate and long-term.Individuals purchase insurance in their F I G U R E 1 Growth of life and non-life insurance in Germany in the long run.The left-hand panel refers to West Germany between 1949 and 1990 and the whole of Germany otherwise; the right-hand panel refers to the territory of the historical German Democratic Republic.Sources: Borscheid and Drees (1988), SwissRe (2020), Destatis (1994), and Kohl ( 2022) younger years and go through decades of contractual savings in the long-term expectation of eventually relying on this private safety net.This gradually makes them less politically interested, more conservative, and strongly in favor of private welfare arrangements and against public ones.Rather than acting as a safety net from which to take on more risks, insurance slowly draws individuals into the risk-averse private spheres of life.Insurance, we argue, is one micro-mechanism behind the depoliticization and privatization trends that have appeared over recent decades.
The paper thus makes three overall contributions: First, it puts a new twist on most economists' studies on a public-private crowding out by pointing to a reverse private-public crowding out.Second, it extends political science's purely cross-sectional work on political effects of private assets by confirming many findings in a longitudinal design, by zooming in on insurance assets, and by extending the outcomes to more than just political attitudes.This larger set of societal implications of insurance is, third, meant to center the finding more broadly in an understudied sociology of insurance which so far has tended to look at the insurance sector and its organizations (Vargha, 2015), but not so much at the persons insured.
The paper is structured as follows: In the next section, we situate the sociology of insurance at the crossroads of risk sociology and the political sociology of welfare attitudes.After presenting the data and our methodology, we deliver the empirical findings in three moments by describing the evolution of life insurance in Germany, by focusing on the average attitudinal differences between insured and non-insured, and by describing the way the beliefs of the insured change dynamically.The discussion elaborates theoretically on the "privatizing" effect we attribute to insurance.We conclude the paper by sketching out future lines of insurance sociological research.

| THE ORIGINS AND CONSEQUENCES OF PRIVATE INSURANCE: THEORETICAL FRAMEWORK
Insurance, though rarely itself an object of study in major sociological theories, intersects very much with a number of broader concepts in general sociology.First, traditional communities, or Tönnies' (2005Tönnies' ( [1887]]) Gemeinschaft, still relied very much on collective institutions to redistribute risks, while modern societies (Gesellschaft) saw the rise of the more abstract communities of risk created by modern insurance companies.Not coincidentally, Tönnies was the first classical sociologist to situate a "sociology of insurance" at the heart of analyzing modern societies (Tönnies, 1917).Second, this shift evidently goes along with changes at the individual level, where self-reflexive actors only start demanding insurance once they rationally anticipate and imagine future events in their life course (Beckert, 2016;Giddens, 1991).It was, third, the individualization of risk in combination with the rise of complex, uninsurable, or systemic risks which led Ulrich Beck to speak of "risk societies", in which man-made risks prevail and the technology of risk-managing institutions dialectically produces additional risks (Beck, 1992).
Finally, historically, insurance can be understood as one instance in the rationalization of modern life, which Max Weber describes in his Protestant Ethic (Weber, 1984(Weber, [1920]]).Not coincidentally, insurance emerged in the Protestant territories within the German lands (Müller-Armack, 1981), and life insurance was used from early on by Protestant clergymen's families, before it spread among the educated urban middle-class professionals (Borscheid, 2012).
Where traditional classes such as farmers or craftsmen could always fall back on their real estate or tangible assets, life insurance became the arrangement for modern employees to insure premature death to the benefit of widows and orphans and, given rising life expectancy, old age to their own benefit (Borscheid, 1987).
Life insurance then trickled down into broad strata of the population through aggressive marketing (Mcfall & Dosworth, 2009) and through the sale of so-called small or industrial life insurances which initially only covered burial costs (Eriksson, 2015).In the industry, the classic saying is that insurance is a product that needs to be sold, that is, customers do not automatically develop demand for insurance, might not even know about it, and have the ability and willingness to pay for more insurance than they actually buy.Already in the nineteenth century, therefore, insurance companies set up widespread agency systems to sell insurance door to door, in public lectures, and by advertisement (Mcfall & Dosworth, 2009).
While there is hardly any sociological literature on who buys insurance in the present day, this is clearly a long-standing question of economic and industry research in Germany and elsewhere.This literature reveals that despite the trickle down, life insurance still is a product for middle-and upper-income groups (Enz, 2000), also thanks to advantageous tax treatment in the income tax code, making it the most important private pension product for the self-employed and the most important private pension supplement for civil servants in the German context (Sauter et al., 2015).It has also been a gendered product: traditionally, life insurance has been used by male breadwinners to make provision for their family, while women started to take out policies only later (Noelle-Neumann, 1985), but still to a lesser degree (Bulmahn, 2003).The age structure of life insurance customers follows an inverted U, peaking in the 30s and 40s; among different professional groups, the self-employed have the broadest coverage, followed by civil servants and highly qualified employees, while unqualified workers have the lowest coverage rates (Bulmahn, 2003).Some economic studies find that households with more dependents tend to have more life insurance (Browne & Kim, 1993), also driven by bequest motives (Sauter, 2014).
Life insurance companies have sold different kinds of products, where the classic policy insuring premature death or burial costs has gradually been overtaken and complemented by those functioning as old-age savings vehicles (investment, group insurance), either leading to a lump sum on retirement (whole life) or a regular private pension (pension insurance).At the start of our sample period in Germany, the classic whole life insurance made up more than 80% of all premium payments (cf. Figure 2).These policies, often including insurance for premature death, are tax-subsidized savings vehicles to build private wealth for old age through regular contributions, with a government-guaranteed basic interest rate (which has fallen from 3% in the 1980s to 0.25% today), and a mandatory annual share of additional profits (GdV, 2020).As German insurers are regulated to invest conservatively in mortgages and fixed-income securities, these additional profits are lower and less volatile than those of their stock-investing Anglophone counterparts (Van der Heide, 2020).A similar product has also been offered by employers as group insurance, with premiums directly deducted from pay cheques and eventually resulting in occupational pensions-a tax-privileged second pillar of the pension system-much smaller than in the Anglophone context (Klein, 2010).
Overall, this low-risk, government-subsidized product was partially crowded out by private pension insurance products, which started to flourish after Germany's belated turn toward more private pensions in the 1998 reform and currently make up 31% of all premiums (cf. Figure 2).Private insurers successfully lobbied to make their more conservative products the key target of government pension savings subsidies rather than the less conservative investment fund industry (Röper, 2020).This so-called voluntary "Riester-Rente" held by about 25% of potential retirees also requires tax-subsidized monthly premiums and has a guaranteed interest rate, while promising a life-long supplementary pension from retirement age (GdV, 2020).
Since the 1960s, the sector has additionally set up specific investment schemes with a stronger focus on capital market investments, but much less so than in the UK, where life insurance has developed into a more speculative investment without guaranteed returns and direct profit-sharing such that the risk is carried entirely by individuals (Van der Heide, 2020).Overall, the function of insurance has moved away from insuring premature death to insuring the risk of outliving one's assets in old age and supplementing declining public pensions.In comparison to insurance investment in Anglophone contexts and to investment funds or direct stock investments in Germany, insurers have positioned themselves in the niche of a state-subsidized secure long-term lower-return investment product, originally geared toward the self-employed and gradually extended to civil servants and employees.In comparison to Germans' favorite form of financial wealth-deposits in commercial and savings banks accounts-insurers could market their long-term product as additional protection for the family and old age in case of adverse events.When asked about their motives to buy life insurance, the majority of Germans mainly pointed to security in old age and for descendants, and not to speculation and return on investment as in British surveys (Rüttler, 2006).
Private insurance is rarely the subject of social science research.Political science and economics have traditionally focused on feedback effects of the welfare state on recipients (Busemeyer et al., 2021), mainly in the domain of public insurance and its potential crowding-out effects (Andersson & Eriksson, 2015;Lehmann-Hasemeyer & Streb, 2017).The puzzle since the 1990s has been to explain why welfare state retrenchment was so resilient against the neoliberal onslaught starting in the 1980s (Naumann, 2018;Svallfors, 1997).
One key answer was to point to different path-dependent or feedback mechanisms that kept social expenditure levels high despite the general market-liberal trend (Pierson, 1996).Welfare states managed to create their own constituencies in the many beneficiaries of public programs who, even if initially against the welfare state, switched to welfare-supporting and pro-redistribution attitudes, voting for parties in favor of maintaining or promoting more welfare.Therefore, individual agency and attitudes entered the equation of this literature only as a way to identify the main factors that explain support for social redistribution (Jaeger, 2013).Economic studies accumulated evidence of a strong public pension provision crowding out or displacing private savings, particularly for higher-income and more educated households (Alessie et al., 2013).One serious shortcoming of this literature is that it mainly relies on cross-sectional household or macro time-series data but does not exploit the individual longitudinal dimension.
While the attitudinal effects of public welfare creating its own future constituencies and a public-private crowding out are very well studied, the actual effects of private wealth, as accumulated for instance through the privatization of countries' pension systems in the second and third pillar, are less well understood.Overall, wealth is found to have positive effects on well-being (Hochman & Skopek, 2013) and on conservative vote choice (Le Hay & Sineau, 2010), where riskier assets act as an amplifier when compared with non-riskier ones.While insurance products, particularly in Germany, are usually on the non-risky end, this is different for more speculative, investment-type products, which have historically been much more prevalent in the Anglophone context (Clark, 1999).Housing assets are generally associated with conservative and anti-welfare state attitudes (André & Dewilde, 2016).Even the anticipation of having wealth in the future, through one's own labor or inheritance, may increase risk affinity (Greenberg, 2013).
Whereas insurance studies are a broadly developed field of research, the sociology of insurance has rather been forgotten between a sociology of risk, on the one hand, and the sociology of public insurance (or welfare states), on the other.While insurance studies have some clear expectations as to which individuals choose insurance, sociologists of risk have rather focused on uninsurable risks or on insurers, but not on the insured.Some recent works in political science have mostly used cross-sectional data to associate the availability of private insurance options with attitudes toward public welfare (Busemeyer & Iversen, 2020).One finding suggests that while the middle classes might have supported public welfare states historically, when private insurance was still not a viable option, they have more recently turned against public welfare in contexts where private arrangements offer them a viable alternative (Busemeyer & Iversen, 2020).Other findings on the policy feedback effects of privatization policies show mixed results (Zhu & Lipsmeyer, 2015) and are based on cross-sectional analyses that cannot rule out selection bias (Bendz, 2017).
As authors in this field openly admit (Busemeyer & Iversen, 2020), cross-sectional studies are particularly inadequate for studying the effects of taking out private insurance on individual attitudes and behavior.Both life and property insurance contracts are among the longest-lasting (financial) contracts individuals sign in their life.Particularly life insurance is generally bought to cover the 2 decades or so of an individual's active working life during which premature death would leave their family without wage income.These whole life policies are then usually combined with retirement insurance, which provides either a lump sum payment at retirement age or an annual private pension throughout old age.In either case, this private pension aspect of life insurance means a contractual savings period of again several decades, during which the privately insured future pensioner receives no direct material return but invests in a distant future.If at all, we would not necessarily expect a one-time short-term effect of insurance, but rather a long-term socialization into life as a privately insured person.
Most insurance-related effect studies have produced two potentially contradictory expectations of how individuals behave in reaction to taking out insurance, also known as the "moral hazard" problem.It refers to the idea that taking out insurance makes individuals change their behavior-in its original meaning with the goal of tricking the insurance through insurance fraud (Rowell & Connelly, 2012).Since Kenneth Arrow (1968), the term has referred more neutrally to the incentives insured individuals would have to expose themselves to higher levels of risk.For insurers, this always meant an information asymmetry problem because without taking into account "moral hazard", the premium charged would be too low to cover all claims, leading to potential failures of insurance markets (Finkelstein, 2014).In more sociological terms, by contrast, this strand of moral hazard research would suggest that taking out private insurance actually increases the willingness of individuals to take risks.In other words, insurance paradoxically enables further risk-taking by providing individuals with a background security.
Liberal economics, in turn, has developed quite a contradictory expectation of how public insurance in particular disincentivizes potentially rent-seeking beneficiaries to take any further risk and become passive pensioners of the state (Sinn, 1996).While also being a moral hazard effect, the expectations of the private and public insurance effect are asymmetric: where the former is usually said to increase levels of risk-taking, the latter is rather seen as destroying any individual endeavor to move beyond receiving one's public rent.
In summary of these different strands of literature, we would expect that taking out insurance rather increases one's support for private welfare arrangements to the detriment of public welfare arrangements, but this effect is not necessarily short-term; if at all, it is long-term.It might also be stratified strongly by the usually unequal provision of private welfare and might also lead the insured to vote for political parties favoring private welfare arrangements.
Expectations of whether this makes insured individuals more risk-taking or less are not necessarily clear-cut.Sociological research on private insurance is generally under-developed and it is hence a good moment to provide a fresh investigation of what it does to an individual to take out insurance in early life.

| The German Socio-Economic Panel: Data
We examine the link between life insurance and individual attitudes using the data of the German Socio-Economic Panel (SOEP, 2021), a representative household panel survey that covers the period between 1984 and 2018.Our key independent variable-having a life insurance policy-is available in all survey years.While the question is asked at the household level, we focus only on the individuals most likely to have experienced attitudinal effects as a consequence of the personal decision to become insured.In most cases these are individuals working full-time, as insurance requires regular premium payments.We therefore restrict our analyses to this segment of the German population.Spill-over effects to other household members are possible, but they are not the main focus of this paper.
Regarding the dependent variables, we first consider a self-evaluated measure of risk avoidance.We then take into account two dimensions measuring the level of concern respondents exhibit in terms of general economic issues (economic development; stability of financial markets) and one variable for worries regarding their personal financial situation.A third group of dependent variables focuses on political attitudes: a dimension of political involvement (interest in politics); the preference for the two main social democratic (SPD) and conservative (CDU) political parties and for a market-liberal party (FDP), all of which are characterized by an important heterogeneity in their stance on social redistribution; and a scale composed of various dimensions operationalizing the respondent's opinion on the need for state intervention, as opposed to letting private forces run their course, in various policy domains linked to family, work, and care. 2 In the models presented below, we also consider a set of standard control variables: gender, age class, education, foreign national status, region, couple status, number of children in the household, and time dummies.The latter are meant to capture the main societal trends experienced by both insured and non-insured respondents.Consistent with our theoretical expectations about the presence of potential heterogeneous attitudinal effects between individuals with different levels of insurance assets relative to income, we stratify holders of insurance assets in three tertiles based on the ratio between insurance assets and household income.The wording and exact coding of all variables is available in Table 1.Descriptive statistics on all variables are available in Supporting Information S1: Table A1, Appendix A.

| Average and dynamic effects: Empirical strategy
Examining the link between life insurance and individual attitudes, we first set out to characterize the attitudes of the individuals who become insured and understand to what extent differences between them are the consequence of becoming insured; we do this by deriving estimates that solve some key endogeneity issues of cross-sectional models used so far in existing research.We consider the following functional form: with A representing an attitudinal dimension; α an intercept term assumed to be constant across individuals and time; β the coefficient associated with the binary life insurance status I (non-insured versus insured); δ the vector of estimates associated with the set of control variables C; ν a time-invariant error term varying only between individuals; µ an error term varying both across individuals and time; i and t two indexes representing individuals and time-periods, respectively.
Linked to the first goal, we begin by estimating an Ordinary Least Squares (OLS) model without controls (setting C as the null vector) that gives the average attitudinal difference between non-insured and insured individuals.In an effort to understand whether these attitudinal differences can be linked to a (causal) effect of life insurance or whether individuals with specific attitudes are more likely to become insured to begin with, the dominant cross-sectional approach attempts to control for potential self-selection issues by adding a set of observable control variables.We do this in a second model by including the control variables we described in the previous section, again using OLS as the estimator.Working with panel data, we improve this strategy by considering a third model in which we estimate specification (1) through fixed effects.In addition to a set of observed controls, this model also allows us to partial out the impact of all unobservable omitted variables that do not vary over time, such as personality traits, represented by the time-invariant error term ν.
All analyses are performed only on the first observed transition to life insurance during respondents' participation in the survey.This leads us to focus only on switches from non-insured to insured status and avoiding considering the opposite switch (from insured to non-insured) as having the opposite effect.In order to account for potential serial correlation and heteroscedasticity issues, we use robust clustered standard errors with the individual as cluster unit.
While the fixed effects model is a strong improvement on cross-sectional models, we are still exposed to a potential time-varying endogeneity issue.Although our models include controls for changing individual characteristics and time dummies that capture general societal changes and shocks such as financial crizes, we cannot exclude that additional changing idiosyncratic circumstances may be associated with time-varying omitted variables or reversed causality we are not controlling for.Instead of trying to solve the issue through approaches such as instrumental variables that rely on untestable assumptions, we consider a second set of models by exploiting panel data to also explore how becoming insured influences attitudes dynamically.In order to do that, we split the binary life insurance status into a series of dummies, each one representing a specific moment in the trajectory of an individual becoming insured: or more years after having become insured.We intentionally focus in more detail on the post-insurance phase and operationalize the pre-insurance phase through only two dummies (3 years or more before becoming insured; 1 or 2 years before).On the one hand, this is justified for theoretical reasons, as insurance usually requires aggressive sales marketing and is not a product customers anticipate buying years earlier.We therefore exclude long-run anticipation effects.On the other hand, our empirical explorations also confirm that pre-insurance years are not associated with any relevant attitudinal trends.
We estimate this model again through fixed effects, which implies that the dynamic evolution we observe is only linked to time-varying variables, the hypothesis being that the attitudinal effects of becoming insured vary with the duration of insurance.Being the point furthest away from the switch to life insurance, "3 years or more before becoming insured" makes sense as the benchmark against which the estimates associated with the rest of the trajectory are measured.All individuals included in this analysis are observed at least in the year immediately before and immediately after taking out insurance, while the number of observations decreases moving away from the transition (cf.Supporting Information S1: Table A2, Appendix A).In order to decrease the multicollinearity between consecutive dummies and to increase statistical power, we grouped together pairs of consecutive trajectory years.

| Life insurance in Germany: Descriptive analysis
We begin by examining the proportion of individuals who have taken out a life insurance policy in Germany since 1984.As we described above, since SOEP data give information only about the presence of at least one insured person within every household, we provide figures on both the general population and on full-time wage-earners, that is, the individuals who are most likely to be the actual insurance holders.Relying on cross-sectional weights for all the descriptive analyses of this sub-section, 3 over the whole period under examination , slightly more than half of Germans (50.4%) belong to a household of which at least one member is insured, while the proportion reaches almost two-third (64.4%) for individuals working full-time.This is because retirees have already cashed in their insurance, while paying regular insurance premiums requires regular income flows.Looking at how the proportion of insured households and full-time working individuals has evolved over time in Germany in Figure 3, the proportion of insured individuals in 1984 (all: 56.6%; full-time: 68%) is higher than the average of the whole period and exhibits a slow decrease 4 until the beginning of the twenty-first century.In 2018, slightly more than one-third of households have at least one insured member (37%), while the proportion is around 50% of individuals working full-time (49.4%).The downward trend reflects demographic aging and early-retirement waves, as more older people drop out of their life contracts than young people contract into them.Only in recent years, the growing unattractiveness of long-term insurance contracts in a low-interest environment might have played out in the aggregate numbers.
Studying how insurance is associated with age in Figure 4 shows that the proportion of individuals belonging to an insured household is relatively high for individuals aged 18 or under (all: 63.9%; full-time: 63.9%), as many are still part of their parents' household.With most respondents gradually leaving the (insured) family nest, the figures decrease over the following years, with a bottom at the age of 26 (51%) for the general population and at the age of 28 (57%) for full-time workers.With the stabilization of the professional and family trajectory, the numbers increase again, peaking at the age of 42 (66.2%)among all individuals and 1 year later (69.6%) among those working full-time.
By this age, the majority of individuals have finalized plans about insuring their loved ones in causa mortis and about insuring their own old age.These figures then gradually decrease, as no new contracts set in and first policies reach their term through drop-outs or early retirement, with a clear fall directly at average retirement age.The numbers decrease even further afterward, with a floor of around 10% insured reached among people aged 80 or over: by this time, all insurance plans insuring life have come to term, while some plans insuring death (or burial) persist until the bitter end.
Information also available on the size of assets related to life insurance between 2002 and 2017 at the individual level for certain benchmark years reveals that among the 38.2% of respondents who have some insurance assets, the median/average value of these assets is 9410/19,324 euros, reaching a peak of 17,000/32,928 euros at the age of 64, immediately before policies are cashed in after retirement.Adopting as a simple measure of the level of inequal distribution of these assets the ratio between assets owned by the third and the first quartile, we find that those in the richest quartile have 5.7 times (20,000/3500) more assets than those in the poorest quartile.Computed than financial (held by 47.1% with a ratio of 35,000/5000 = 7) and tangible assets (held by 7.4% with a ratio of 13,000/2000 = 6.5), much less skewed toward the wealthy when considering the value of owned businesses (held by 4.2% with a ratio of 150,000/10,000 = 15), but more inegalitarian when taking into account the market value of primary and secondary homes (held by 41.5% with a ratio of 305,000/130,000 = 2.35).

| Selection or causal effects? Average treatment effects
We now focus on the average treatment effects of life insurance on individual attitudes using the three regression models introduced above in Equation (1).Since our interest lies primarily in the estimates associated with the changing effect of life insurance on individual attitudes as additional sources of endogeneity are controlled for, we plotted these estimates in Figure 5 and provide full regression models with all control variables in Supporting Information S1: Tables B1-B9, Appendix B. While the filling of the symbols representing the estimates indicates their statistical significance (white symbol: p-value ≥ 5%; black-filled symbol: p-value < 5%), in order to also visually compare the magnitude of the estimates, we re-scaled all dependent variables to a 0-10 scale and standardized the limits of the horizontal axis to [−0.7, 0.7].A few outlying estimates are not represented, but we provide their values while commenting on them.For every dependent variable, the first row of estimates refers to the full sample (indicated as "main" on the right-hand side of the two graphs in Figure 5), while the last three are linked to re-estimations within the three stratifications of the ratio between assets and income ("T1", "T2", "T3" on the right-hand side of the two graphs in Figure 5).
Starting with risk avoidance as the dependent variable, the descriptive model "OLS without controls" reveals that insured individuals are less likely to take risks (0.13, p < .001)than non-insured individuals.The differential is entirely related to pre-existing traits that make risk-averse individuals more likely to become insured, as the differential between the two groups decreases and becomes insignificant when controlling for observed (OLS with controls) and unobservable (FE with controls) confounding factors.In other words, life insurance, on average, does not appear to play any causal role.Looking at how estimates vary by assets-to-income ratio, we note that the self-selection of risk-averse individuals is clearly socially stratified, with individuals having the least assets being those who are most risk-averse (0.28, p < 0.001), followed by those with mid (0.22, p < .001)and large assets (0.18, p < .01).The estimates decrease when including observed controls, but remain still significant for the first two tertiles (low: 0.23, p < .001;mid: 0.13, p < .05).They become irrelevant and insignificant in all three groups only after controlling for unobservable heterogeneity with non-insured respondents.More generally, with the exception of two outlying estimates in the last dependent variable for two assets-to-income ratio categories we comment on below, Figure 5 shows that no other significant estimate in the "FE with controls" models is visible.In other words, when looking at average treatment effects, we find that life insurance plays almost no causal role on the attitudes we consider as dependent variables.
Therefore, for the remaining variables (aside from the last one), we focus our comments only on the first two models.
In the estimates related to worries about economic development, a relatively strong differential between insured and non-insured is visible (0.53, p < .001),decreasing but still remaining significant after introducing some observed controls (0.12, p < .001).Comparing the three assets-to-income ratio categories, we again see a stratified differential, with only wealthier individuals clearly worrying less than non-insured individuals about general economic development (−0.24, p < .001).While a small, barely significant differential is visible in the mid category (−0.14, p < .05), the lowest stratum is not significantly differentiated from non-insured workers.There is an important decrease in all three estimates in terms of magnitude and significance after observable controls are included.
An even more important differential compared to non-insured respondents is visible regarding the concerns insured individuals have about the stability of financial markets (0.67, p < .001), a self-selection effect primarily driven by individuals with a large assets-to-income ratio (0.87, p < .001)but also visible in a less pronounced way among those with a mid assets-to-income ratio (0.49, p < .001).Including observed controls leads to a large decrease in the estimates across all segments of the population, which remain significant only in the main sample (0.28, p < .001)and in the wealthiest tertile (0.37, p < .05).
If we move from general economic concerns to worries related to the personal financial situation, the picture changes.In this case, it is insured individuals who feel less at risk when compared to non-insured individuals (−0.34, p < .001), a descriptive differential that is clearly socially stratified, with individuals who have a large amount of assets having the largest advantage compared to non-insured individuals (−0.97, p < .001),followed by mid (−0.53, p < .001) and low (−0.16,p < .05)asset holders.All these descriptive estimates remain relatively stable after observed controls are included, turning to insignificance only when we control for individual fixed effects.
If we focus on a first type of political attitude, insured individuals are clearly more interested in politics to begin with when compared to non-insured individuals (0.62, p < 0.001), a differential almost entirely linked to observed controls such as the level of education and hence becoming insignificant in the second model.The differential is clearly stratified in the assets-to-income ratio categories, being most pronounced in the third most wealthy category (0.90, p < .001),less visible in the middle category (0.57, p < .001),and even smaller for individuals with a low amount of assets (0.17, p < .01).As in the main sample, the higher level of interest in politics of insured individuals is to a large extent linked to observable traits that set them apart from non-insured individuals.
In terms of partisan preferences, only a small (less than 1%) and barely significant higher propensity to support the SPD (0.0096, p < .05) is visible among insured individuals, a differential that is slightly more pronounced among mid-level assets holders (0.016, p < .05).Both estimates become clearly smaller and insignificant when observed controls are included.

F I G U R E 5 Average treatment effects of life insurance on attitudes toward risk and on political attitudes-Full sample and by insurance assets-to-income ratio tertiles
By contrast, a clear tendency to be more likely to feel close to the CDU appears among insured respondents (0.067, p < .001),which is halved but remains significant after inclusion of observable controls (0.036, p < .001).The conservative differential follows an expected pattern, with assets-rich individuals driving the selection effect (0.089, p < .001),followed by the middle category of individuals (0.060, p < .001),and insured individuals with low assets (0.037, p < .001).The differences decrease under observed controls but still remain relatively important, in particular for individuals with the highest assets-to-income ratio (0.077, p < .001).
Despite the small proportion of voters concerned, we also note a strongly significant higher propensity to support the FDP among insured when compared to non-insured individuals (0.010, p < .001),which is cut in half but still highly significant after we control for observable confounding factors (0.0050, p < .001).Similarly to CDU preferences, the differential is socially stratified, with assets-rich individuals leading the selection effect (0.017, p < .001),followed by the other two categories (0.0086, p < .01).Once again, the estimates decrease when observed controls are included, but they still remain substantively important.
Finally, the conservative stance of insured individuals is also associated with policy views linked to a preference for market-based solutions as opposed to state intervention (0.27, p < .001), a differential that remains still important after adding observable controls (0.16, p < .001)and that turns to insignificance only when controlling for unobservable time-invariant confounding factors.This policy-related conservative pattern is mainly visible among assets-rich individuals (0.23, p < .001), a differential that slightly increases when controlling for observed factors (0.28, p < .001).
A similar pattern with smaller estimates is visible in the other two categories.Interestingly, this is the only dependent variable for which we detect significant effects of insurance among mid-(1.24,p < .01)and high-asset holders (0.72, p < .05)even after partialling out the impact of unobservable time-invariant confounders in the fixed effects model.

| The lifetime influence of life insurance: Dynamic effects
In the previous sub-section, we have shown that taking out life insurance, on average, has virtually no causal effect on individual attitudes.Since the attitudes we consider may be insensitive to change, several years may be required to observe an impact.This is what we investigate by looking at how life insurance and attitudes are linked dynamically based on the specification described above in Equation ( 2).Since we are mainly interested in a specific set of estimates, that is, those associated with the dummy variables that represent the way the outcome variable varies during the life insurance trajectory, we have plotted these estimates in Figure 6.Full regression models with all control variables are available in Supporting Information S1: Table B10, Appendix B.
Since we are estimating the dynamic link between life insurance (operationalized through a series of dummy variables identifying specific moments of the pre-and post-insurance phase) and attitudinal outcome variables through a fixed effects model, the dynamic effects depicted in Figure 6 represent any time-varying changes observed in the dependent variables throughout the life insurance path of an individual that cannot be attributed to the control variables we included.This means that we are looking at how individual attitudes change immediately before and in the years after taking out a life insurance policy.Our theoretical framework makes us suppose that life insurance is associated with a process of multidimensional attitudinal change that concerns both attitudes toward risk and political views.The attitudinal trajectories allow us to track this bundle of changing attitudes.
Focusing on the dynamic effects on risk avoidance, we find that taking out life insurance does not present any visually relevant nor statistically significant dynamic effects.Concerns about economic development decrease continuously after individuals become insured (−2/−1 → 21/22: −0.69, p < .05).Probably because of statistical power issues (the dependent variable is available only in a few waves), the same consistent, significant pattern is absent for worries associated with the financial system, although a fluctuating decreasing trend is visible.The continuous calming effect of life insurance is also apparent when looking at worries regarding the personal financial situation (−3 → 21/22: −0.99, p < .05).
When it comes to political attitudes, taking out life insurance is associated with a strongly decreasing level of interest in politics (−3 → 23/24: −1.37, p < .001),matched by a less pronounced, albeit very consistent, decrease in the propensity to support the SPD (−3 → 21/22: −0.10, p < .05).The likelihood of feeling close to the CDU or the FDP does not exhibit any meaningful dynamic relationship to life insurance.

| DISCUSSION: THE PUBLIC COSTS OF PRIVATE VIRTUE
Zooming out again from these detailed findings, this discussion highlights three contributions: to the sociology of insurance broadly speaking (Vargha, 2015;Zelizer, 1978), to the pension insurance and "crowding out" literature (Alessie et al., 2013), and to political science literature on the effects of private wealth (Busemeyer & Iversen, 2020;Nadeau et al., 2011).

F I G U R E 6 Dynamic effects of life insurance
First and foremost, the findings seek to establish a more systematic sociology of insurance that goes beyond the existing science and technology studies of insurers by focusing on the long-term biographies of the insured individuals.In a social order in which the burden and consequences of societal risks are shifted toward the individual, the individualized and privatized safety net provided by life insurance represents both the indicator of and the reaction to this increasingly unstable social equilibrium.Focusing on individual-level data for Germany, we showed that life insurance is an asset that the majority of households and full-time workers acquired in the last 3 decades.As it is the outcome of a choice that most individuals make after having stabilized their professional and family trajectories, the assets associated with life insurance continuously increase until retirement and, along with private savings, represent one of the key resources individuals rely on to protect themselves against the financial risks modern societies increasingly expose them to.Life insurance is moreover just one proxy for the many other types of property insurance individuals contract into when purchasing a home and other tangible assets.In this discussion, we want to dive deeper into the question of who the individuals seeking insurance are and what consequences taking out insurance has on them.
Unlike the social insurance system, life insurance is the result of a fully privatized choice that is made by individuals who hold very specific attitudes toward the societal risks in which they are immerged.With a slightly higher tendency to avoid risk than the average citizen, insured individuals show a very strong aversion toward the hazards coming from the economic system as a whole, and financial markets in particular.This fear of general societal dangers they have no influence on is at odds with a clearly stronger feeling of security regarding their own economic situation.
The dichotomy between worried interest in the surrounding world and a positive view of circumstances determined by personal effort is also visible in the political attitudes of these individuals.Much more interested in politics than the rest of the population, insured individuals have a higher propensity to support conservative and market-radical parties that see in the private initiative related to market-clearing mechanisms a far preferable option to state-driven interventions.To fend off the dangers of a potentially uncontrollable external world, risk-averse individuals thus support political parties that promise to maintain the status quo, and use private insurance for all other unavoidable risks.Interestingly, with the exception of a relatively homogeneous tendency to be risk-averse and to be worried about the economy, all other attitudinal antecedents that lead individuals to self-select into life insurance are clearly socially stratified, being much more pronounced among wealthy individuals who can and do decide to invest a larger portion of their income in insurance assets.It is this worry directed toward risks external to them that leads insured individuals to look for a safety net that allows them to protect their privileged condition, which they see as the result of private meritocratic initiative and better protected by parties that act primarily to defend the status quo than entities that aim to redistribute risk across society.With the exception of the very illiquid owned house, life insurance assets appear, despite the particular German insurance market, to be the least inegalitarian form of assets, suggesting that they represent a response to trends that concern all social strata.These attitudinal findings hold under the standard control variables used to explain insurance demand in the literature.
While the attitudes of insured individuals clearly differentiate them from the rest of society, it appears that these differences have little to do with the impact of becoming insured, but almost exclusively with the antecedent reasons that lead individuals to become insured in the first place.Working with panel data, we were able to ascertain that self-selection into life insurance is related to both observable and unobservable factors.This is a distinction the existing purely cross-sectional studies on the topic have not been able to make regarding insurance (Busemeyer & Iversen, 2020), but also regarding the study of pensions and policy feedback effects (Larsen, 2019).While the tendency to support the SPD is the only outcome variable for which selection effects appear to play a marginal role, insured individuals not being significantly different from non-insured individuals, the cautious attitudes of insured individuals in terms of risk avoidance, general economic worries (economic development and financial markets), and interest in politics are mainly explained by observable controls.The higher likelihood of feeling close to parties such as the CDU or the FDP is explained in equal measure by observable and unobservable factors.Finally, the worries about personal financial situation and the preference for market-over state-based policies that distinguish insured individuals from the rest of the population can be traced back to unobservable personality traits one can only control for when working with panel data.This empirical pattern implies that the importance of classical social stratification variables (in particular education, age, and gender) as triggers of personal beliefs that motivate the decision to become insured varies heavily depending on the type of attitude.These variables have the strongest relevance for general risk-aversion attitudes, probably because stratified knowledge about general societal risks and the instability of the economic system plays a key role.Unobservable personality traits become more important when considering conservative partisan preferences, while they become the main attitudinal trigger when it comes to free-market policy preferences and the feeling of security about personal finances.
Our second contribution is to look at the actual consequences of private insurance, rather than seeing it as passively affected by state intervention, as is the case in much of the pension literature focusing on the "crowding out" of private alternatives (Alessie et al., 2013).The first interesting attitudinal finding is that dynamic changes are absent in the propensity of individuals to take risks.On the one hand, this could be seen as evidence that the moral hazard literature might be correct for certain risks (Finkelstein, 2014), for example, health risks when purchasing health insurance, but not necessarily when it comes to an individual's general risk perception.Individuals do not become risk-takers to exploit their insurance status.Much to the contrary, the risk profile of individuals seems to be relatively set from the start, after socialization experiences related to classical stratification variables.On the other hand, it is also possible that the positive effect of becoming objectively insured against risks is counterbalanced by an increased risk awareness provoked by the very fact of having a policy and receiving information about additional insurance products.We leave it to future research to test these alternative explanations.
While risk perception is not affected, the higher-than-average worries about the general economic situation that lead individuals to become insured gradually decrease after they have taken out a policy.The same applies to the already lower-than-average concerns about their personal economic situation.In other words, while the perception of being exposed to various kinds of hazards remains at similar levels, insurance works as a cognitive device to at least decrease the fears related to the objective dimension they are meant to insure, that is, financial security.In this sense, then, insurance does allow individuals to purchase a sense of security.
Existing economics research has focused very much on how public insurance would crowd out private insurance (Alessie et al., 2013), while our interest has been in the reverse direction of influence and in whether private provision weakens not only political support for public solutions but also any interest in the public sphere altogether.
In this regard, it is interesting to see that the higher propensity to be interested in politics that draws individuals into becoming insured gradually fades away after they become insured.While the already strong conservative and market-radical preferences remain unaffected, we also observe a relatively small, but highly consistent, negative effect on the likelihood to support the main social democratic party in Germany.Taken together, we can see these dynamic changes in political attitudes as signs of a cocoon effect, whereby insured individuals withdraw from involvement in public matters and become even more skeptical of political forces with inclusive societal views while still keeping their higher-than-average conservative convictions.Rather than becoming engaged in the public sphere and raising their voice to eliminate causes of potential risks, insured individuals tend to be locked by their insurance into a privatized solution: at similar levels of risk perception, insurance makes them feel safe enough to withdraw from public life and favor even more conservative market-driven policy solutions.Insurance, by abstracting from demanding solidary relationships, is thus one realization of individualism in market societies.
These findings also highlight our third contribution, which is to enrich existing political science research on political effects of private wealth in three ways: switching from cross-sectional to longitudinal study design, switching from total or housing assets to the understudied insurance assets, and considering insurance as independent variable.Our in-depth longitudinal results confirm Busemeyer and Iversen's (2020) cross-sectional findings for a number of OECD countries and are the mirror image of welfare state research: public welfare states entrench pro-welfare attitudes, and private welfare institutions entrench anti-public, pro-private welfare attitudes.The cross-sectional design is also predominantly employed in the public pension and policy feedback literature (Larsen, 2019), the literature on housing (Ansell, 2019) and on total wealth effects (Nadeau et al., 2011).Insurance wealth shares a number of privatizing and conservatizing effects with these studies on the effects of other components of private wealth.
In policy terms, private insurance thus contributes to two macro-trends characterizing Germany as much as other mature democracies: depoliticization and privatization.On the one hand, private insurance contributes to individuals retreating into private life (against public welfare, politically disaffected), which acts as a micro-mechanism behind the declining voter turnout, party membership, and political activism described in Germany (Schäfer, 2015).When first introduced in the nineteenth century, private insurance, as much as public insurance later on, was criticized for lacking social community ties, as it is based on contracts and abstract relationships and is missing organic solidarity when compared to direct charity or kinship-based payments in dense communities.Looking forward to relying on one's own private insurance, we find, does indeed give some weight to this old criticism.
On the other hand, private insurance acts as a micro-mechanism behind the general trends toward more privatized welfare solutions and public welfare retrenchment.Given private alternatives such as insurance or homeownership (Heide, 2021;Lux & Mau, 2018), individuals seem less and less motivated to support public welfare programs, regardless of in which domain.Private insurance can thus be part of a policy feedback loop: with the introduction of the third pillar of the pension system-in Germany as late as 1998 (Leifeld, 2016)-the increasing availability and necessity to supplement state with private pension and other welfare programs could thus increasingly undermine support for public alternatives and set welfare states on a long-term trend toward more privatization.

| TOWARD A SOCIOLOGY OF INSURANCE: CONCLUSION
Insurance is not very salient in social life, but it is permanently in the background and only becomes apparent at specific moments.Insurance contracts-we have focused here on life insurance, but this holds similarly for most property insurance-run over decades and have become part of the invisible armor modern individuals have equipped themselves with to fend off the risks of second modernity as traditional communities ceased to exist.Life insurance has emerged as a particular type of provision for human capital when income flows cease, either in case of death or old age.Yet households pay almost as much of their annual income, more than 3% of GDP, to insure their tangible capital, ranging from personal belongings to property or vehicles (SwissRe, 2020).Only data availability prevented us from investigating these more extended insurance domains in the German context.
Despite it being a background infrastructure, we show that taking out insurance, though not of immediate shock-like consequence, has a long-run biographical impact: it buys individuals economic security against a risky world that is beyond their control, with the benefit of allowing them to withdraw from political life and retreat into private welfare.Individuals do not become more entrepreneurial or lazy, as we find no moral hazard inciting the already risk-averse insured to take on more or less risk.Overall, the understudied phenomenon of private insurance penetration in modern societies can be seen as one micro-mechanism behind ongoing trends in declining political participation and the privatization of public welfare (Sennet, 1999).Part of the lack of dynamism in aging societies is not necessarily due to the soporific effects of public insurance, but probably to insurance arrangements more generally.
While our study focused on Germany for its rare quality of panel data covering insurance and many individual attitudes simultaneously over a long period, our results might travel to other countries of the Alpine insurance tradition.In these continental European countries around the Alps, insurance companies have long been more strongly regulated, collectivized risks to a larger extent, and invested in less-risky products (Albert, 1993), making insurance more of an old-age long-term savings vehicle than a short-term investment product.The introduction of investment products, where the individual bears the financial market risks, has also been belated and less developed there (cf. Figure 2) in comparison to the English-speaking Maritime countries.Germany has also been on the conservative side in the privatization of welfare services, its third-pillar pension reform being among the latest in the OECD world and still relying heavily on the first public pension pillar (Ebbinghaus, 2011).This could suggest that our findings might not be generalizable across the Maritime insurance countries, where insurance is much more in the form of investment products tied to riskier financial markets in competition with large pension and investment fund industries.
Our findings are not only restricted spatially but also with regard to the particular period of the last 3 decades during which the wealthy baby boomers started to retire and draw on their insurance assets.Since then, the premiums have declined for simple demographic reasons, but also due to the crisis of low-risk investment products-the core of the German insurance industry and of government-subsidized private pensions-as the guaranteed interest rate set for insurers by the government fell to 0.25% in 2022.The industry itself has reacted by selling more products without guarantees and more stock investment, but also alternative pension products such as ETF plans, investment funds, stocks, or housing.The literature on the wealth effect for political voting finds systematically different results depending on whether voters hold risky or non-risky assets (Nadeau et al., 2011).This could imply that the ongoing shift of insurance clients toward riskier assets could make them more active and supportive of market liberalization projects.
In the shadow of the sociology of risk and the political sociology of public welfare, a real sociology of insurancedespite some well-known classics calling for it (Tönnies, 1917;Zelizer, 1978)-has not gone beyond the few science and technology studies (Van der Heide, 2020;Vargha, 2015).Chan's comparative study of Chinese and Taiwanese life insurance cultures is a clear exception (Chan, 2012).But even the many recent studies on financialization, including the financialization of everyday life (Montgomerie, 2020), have focused on many financial phenomena of less macro-economic and political importance rather than on this sleeping financial insurance giant.Our modest ambition with this paper is to contribute to such an emerging sociology of insurance which could be modeled along the lines sketched here: it could borrow from welfare research in comparatively studying the emergence and consequences of different insurance regimes (following Albert, 1993), and it could borrow from financialization research in studying wealth stratification and its social and attitudinal consequences (Le Hay & Sineau, 2010).The study of potential trade-offs between different insurance institutions in a broad sense-the welfare state, private insurance companies, cooperatives, families, churches, networks, etc.-in providing functionally equivalent insurance functions in the different domains of life is a similarly fruitful object of study for such a sociology of insurance, as are the attitudinal consequences of these macro insurance institutions.Our paper has been just one example of this kind.
4 The spike observable in 1990 is related to the introduction of a non-fully representative sample for East Germany, in which insured individuals are overrepresented.In addition, since precise socio-demographic information on GDR households was not available, the weighting scheme was only approximative.The sample representativity issue was gradually solved over the following years with more reliable data on households in East Germany and with refreshment samples aimed at correcting the non-completely random one created in 1990.This implies that SOEP data are not a source of reliable descriptive statistics on East Germany in that period, while afterward the insurance patterns in both parts converge.That is why we preferred not to include a re-estimation for East Germany.
Private life insurance contributions by product line.Source: GdV (1980-2019) [Colour figure can be viewed at wileyonlinelibrary.com] 2, . . .N and t = 1, 2, . . .T(2) with β −3 , β −2 , β −1 being the coefficients associated with the insurance status 3 years or more, 2 years, 1 year before becoming insured, respectively; I it−3 , I it−2 , I it−1 dummy variables coded as 1 if an observation represents the status 3 years or more, 2 years, 1 year before taking out life insurance, respectively, and 0 otherwise; symmetrically, β 1 , β 2 ,…, β 25 represent the coefficients associated with the status 1 year, 2 years,…, 25 years or more after having become insured; I it1 , I it2 ,…, I it25 dummy variables coded as 1 if an observation represents the status 1 year, 2 years,…, 25 years with the same indicator for different categories of assets, insurance assets are slightly more democratically distributed F I G U R E 3 Proportion of individuals with life insurance at the household level over time F I G U R E 4 Proportion of individuals with life insurance at the household level by age