Sources of income maintenance

Cancers, cardiovascular conditions, and (to a lesser degree) respiratory disease are the major causes of death during late middle age (NOMESCO, 2015), constituting, for example, more than 70 per cent of the deaths in the Norwegian sample analysed in this study. Some deaths are sudden (e.g. acute myocardial infarction, violent death), but most deaths will occur after a disease history varying from a few months to many years (Reference Kaplan, Baltrus and RaghunathanKaplan, Baltrus, & Raghunathan, 2007).

Many of the individuals analysed here will therefore experience increasing health problems and functional limitations during their last years alive. For those who were retired or pensioners already before the onset of disease, incomes may be relatively unaffected. Among those who were employed before falling ill, income may decline, but this will not always happen. One income maintenance ‘line of defence’ is whether the employed keeps his or her work income in spite of health difficulties. This depends, inter alia, on legal regulations, work tasks, and informal norms. Employment protection legislation may hinder or postpone dismissal of employees who are less productive because of illness (ILO, 2015). Similar constraint may arise from collective agreements between employers and trade unions, or from tradition and established practices. To what extent ill health limits possibilities for managing work tasks, depends on job characteristics, and occupations differ in requirements for physical strength, endurance and necessity of being present. Employers may also choose to keep employees on the payroll even when sick because of valuable competence or personal obligations.

If work income nevertheless declines, reductions in overall income can be curbed by cash benefits from public welfare programmes and social insurance, or from private insurance (Reference ClasenClasen, 2017). Many such programmes exist in Nordic countries: sickness benefits, disability benefits, job-retraining benefits, old-age pensions, preretirement pensions, unemployment benefits, housing benefits, child allowances, social assistance, and some other minor public programmes, as well as some types of employment-based or purely private social insurance (NOSOSCO, 2005). Information on income protection schemes in a number of countries is found in the Social Policy Indicators database (SPIN, 2017) and the Comparative Welfare Entitlements Dataset (Reference Scruggs, Jahn and KuittoScruggs, Jahn, & Kuitto, 2014). Typically, studies of social citizenship rights in Nordic countries paint a comparatively favourable picture (Reference ScruggsScruggs, 2006), as these countries seem to provide wide coverage and relatively generous levels of sickness benefits and disability pensions (Böheim & Leoni, Reference Böheim and Leoni2018; Reference Kangas, Castles, Leibfried, Lewis, Obinger and PiersonKangas, 2010; OECD, 2010).

However, analyses of the legal stipulations and formal rules of these welfare programmes will only give approximate information about how incomes are affected. Social insurance replacement rates, i.e. the ratio between cash benefits when out of work and wages and salaries when working (Reference ScruggsScruggs, 2006), have been characterised as ‘theoretical’ (Reference Peeters, Verschraegen and DebelsPeeters, Verschraegen, & Debels, 2014) and ‘fictious’ (Knoef et al., Reference Knoef, Been, Alessie, Caminada, Goudswaard and Kalwij2016), as they only indicate what certain types of persons with specific sets of characteristics would receive in benefits from specific programmes, if eligible, applying, and rules are followed (see also Reference Smith and ZhangSmith & Zhang, 2011; Reference WhitefordWhiteford, 1995). The impact on the total income situation of those in need will also be difficult to estimate because a recipient may obtain benefits from several welfare programmes at the same time, and it cannot be taken for granted that potential recipients always receive what they are entitled to.

Moreover, taxation will influence how consumption opportunities are maintained. Progressive taxation implies that a reduction in pre-tax income will lead to less reduction in post-tax, i.e. disposable, income. The previous material standard of living may therefore be upheld even when gross income declines, if progression is relatively steep. Tax exemptions and tax deductions for people with particular needs (e.g. deductions for medical expenses) or particular statuses (pensioners, unemployed, disabled) may play a similar role.

Lastly, opportunities for maintaining economic wellbeing when illness and disease occur will be affected by family situation. Individuals living alone may be more likely to experience a deteriorated material standard, since they will often have to rely only on the formal institutions of the welfare state, with less or no access to family resources. Reduced income for the affected person may be countered by spouses’ incomes, and a partner may increase his or her paid employment in order to maintain the material standard of the household.

Thus, the economic welfare prior to death will be influenced by a number of circumstances: health and disease trajectories, whether the affected individual was employed or retired, and the impact of labour protection legislation, employers’ behaviours, welfare programmes, social insurance, taxation policies and family situation. The outcome analysed in this paper – equivalized disposable income – indicates the combined effect of these mechanisms and income maintenance sources.

Previous research

The difficulty in assessing income security among unemployed, sick, disabled and other vulnerable population categories only by analysing welfare programmes and social insurance has been recognised for a long time (Reference WhitefordWhiteford, 1995). Accordingly, previous research has also addressed the actual economic situation of such population categories, using indicators such as income level, poverty or reported financial difficulties. In particular, the situation of the elderly has been investigated. Studies have examined how pensioners’ incomes are influenced by the interaction between life trajectories, family history, and pension systems (Been, Caminada, Goudswaard, & van Vliet, Reference Been, Caminada, Goudswaard and van Vliet2017; Peeters & de Tavernier, Reference Peeters and de Tavernier2015). Findings from Nordic countries suggest that poverty among the elderly is relatively low and has declined in recent decades (Fritzell, Bäckman, & Ritakallio, Reference Fritzell, Bäckman, Ritakallio, Kvist, Fritzell, Hvinden and Kangas2012). Referring to the mid-2000s, OECD found that average disposable income in Nordic countries among those aged 65 and more, relative to the population average, ranged from 72 per cent in Denmark to 82 per cent in Sweden (OECD, 2011).

For the present study, investigations of effects of morbidity and functional limitations on income are particularly relevant. An OECD report showed that in the mid-2000s, average equivalized disposable income among people with disabilities, compared to the average of the working-age population, varied from 88 per cent in Denmark to 94 per cent in Sweden (OECD, 2010, p. 54). A Swedish panel study found that respondents with limiting long-standing illness reported more financial difficulties when interviewed after eight years; this was especially marked among those with lower socioeconomic positions (Reference Lindholm, Burstrom and DiderichsenLindholm, Burstrom, & Diderichsen, 2002). Hospitalisation in Denmark for serious disease, age 43–60 years, tended to be followed by a loss of annual income the next year corresponding to one decile in the income distribution (Reference Rayce, Christensen, Hougaard and DiderichsenRayce, Christensen, Hougaard, & Diderichsen, 2008).

Income effects of disease have in particular been investigated for cancer survivors, with studies from Norway (Reference Benth, Dahl, Luras and DahlBenth, Dahl, Luras, & Dahl, 2014; Hauglann, Benth, Fossa, Tveit, & Dahl, Reference Hauglann, Benth, Fossa, Tveit and Dahl2014; Reference Syse and TonnessenSyse & Tonnessen, 2012), Denmark (Andersen, Kolodziejczyk, Thielen, Heinesen, & Diderichsen, Reference Andersen, Kolodziejczyk, Thielen, Heinesen and Diderichsen2015; Jensen et al., Reference Jensen, Overgaard, Boggild, Garne, Lund, Overvad and Fonager2017) and Sweden (e.g. Eaker et al., Reference Eaker, Wigertz, Lambert, Bergkvist, Ahlgren and Lambe2011). Mostly, average reduction in overall individual income in the first years after diagnosis seems to be moderate (three to seven per cent). Overall incomes declined less than work incomes, due to the compensating effects of welfare benefits. The negative effects on income seemed to be temporary, however. Only small differences in income developments between cancer survivors and analysed control groups could be observed 10–13 years after diagnosis, but income effects were often unfavourable among those with low education (Reference Syse and TonnessenSyse & Tonnessen, 2012). The reason for relatively moderate long-term income effects of a cancer diagnosis could be that those who survive cancer may in many cases regain previous health. In contrast, large negative income effects have been found in studies among patients with multiple sclerosis (Wiberg et al., Reference Wiberg, Friberg, Stenbeck, Alexanderson, Norlund, Hillert and Tinghog2015) and non-affective psychosis (Falk et al., Reference Falk, Burstrom, Dalman, Jorgensen, Bruce and Nylen2016), and among individuals who died of alcohol-attributable causes (Reference Tarkiainen, Rehnberg, Martikainen and FritzellTarkiainen, Rehnberg, Martikainen, & Fritzell, 2018).

Thus, Nordic research has provided considerable information about income among cancer survivors and among disabled working-age people, and a few studies have examined income effects of other severe conditions. Otherwise, knowledge seems scattered, and only a few comparisons between Nordic countries have been undertaken.

Design and research questions

Income security is analysed in this study using a design that differs in several ways from most previous research. The focus is on those who died during late middle age. Such deaths occurred sufficiently often during the mid-2000s to be interesting for income security analyses. Life tables for the year 2000 indicate that among Finnish men who were alive at age 40, 17.9 per cent would die before their 66th birthday (HMD, 2016). Corresponding figures for Danish, Norwegian, and Swedish men were 17.0, 13.2, and 12.0 per cent, respectively. Female death risk was lower, but in Denmark, one in nine women alive at age 40 would have passed away by the age of 66.

Typically, previous studies have selected individuals who were diagnosed with a certain disease or had experienced another significant health event (e.g. hospitalisation), and analysed their incomes during the following years. By using death as the criterion for being included in the study, less known aspects of income security emerge. Since previous studies have primarily focused on survivors, those afflicted by fatal diseases will be less examined, while there will be many such severe disease trajectories in the samples we have analysed. Unfortunately, health information during the 12 observation years was unavailable in our data, but the distribution of causes of death in the relevant age categories (NOMESCO, 2015) indicates that a large proportion in the samples suffered from serious and finally lethal diseases. However, there were certainly many individual variations in the health trajectories prior to their death.

The present study is moreover designed to enhance comparisons between the four analysed countries. Although commonly considered as representing the same Nordic welfare state model (Reference Esping-AndersenEsping-Andersen, 1990; Reference Kautto, Castles, Leibfried, Lewis, Obinger and PiersonKautto, 2010), various differences in welfare programmes exist (NOSOSCO, 2005), for instance as regards benefit levels (e.g. especially high sickness payments in Norway), organisation of social benefits (unemployment insurance was voluntary in Sweden), the relative importance of different types of benefits (housing benefits were more widespread in Sweden than in the other countries), and taxation rules. By using the same inclusion criterion (those who passed away when aged 55–69), a harmonised outcome (equivalized disposable income) and the same length of the observation period (12 years), a comparison of income security in Nordic countries is facilitated.

Three questions are central in the following analyses. First, we examine to what extent the initial level of equivalized disposable income in the samples was maintained during the observation years. We assume that the approaching death and its related circumstances will have had little influence on income at the start of the 12 years study period. Second, we ask whether income developments deviated from what they would have been, hypothetically, if they had avoided their untimely death. Thus, the counterfactual question ‘what would the income have been for those who died, if they had not died?’ is raised (Reference Morgan and WinshipMorgan & Winship, 2007). Third, we address a social equity issue (Reference Kvist, Fritzell, Hvinden, Kangas, Kvist, Fritzell, Hvinden and KangasKvist, Fritzell, Hvinden, & Kangas, 2012) by studying whether income trajectories before death differed by educational level. In the Results section, main findings as regards these three questions are reported. In the subsequent Discussion section, supplementary analyses are used in order to assess possible interpretations.

Overall patterns and the Danish discrepancy

Considerable similarities between the four Nordic countries emerged in the analyses, but there was one striking exception. In Finland, Norway and Sweden, those who died did not only preserve, but even improved, their price-adjusted level of equivalized disposable income during the years preceding death. In Denmark, however, decline was observed after a few years. Since all the four analysed countries are varieties of a Nordic, ‘social-democratic’, welfare model, and all of them experienced substantial overall economic growth from the mid-1990s to the mid-2000s (Nordic Statistics, 2018b), the discrepancy is somewhat surprising. One may also ask why consumption opportunities increased in three of the countries, since it could be expected that work income would decline as death came nearer, and welfare programmes, even when generous, would not replace all work income loss. A detailed report on social protection in Nordic countries in the early 2000s (NOSOSCO, 2005) found that replacement rates net of taxes (‘compensation levels’) for most types of work-related benefits were in the 60–80 per cent range. Thus, some decline in consumption opportunities would not be unlikely – but why did that not happen?

To understand this, the heterogeneity of health trajectories and employment situation in the analysed samples should be considered. Unfortunately, data did not allow for any detailed description of the life courses of the samples, but possible interpretations can be suggested. For some parts of the samples, incomes would hardly be influenced by the approaching deaths, for instance among those who had a ‘normal’ work career up to the time point when death occurred either suddenly or after a very short disease period. Also, among those who had retired already at the start of the observation period, or became pensioners during the study years before experiencing substantial health problems, the coming death and its related circumstances would probably have little income impact. In the Finnish and Norwegian samples, for instance, around 20 per cent were disability pensioners when the observation period started. The risk of income decline would mostly occur among those who were employed before the onset of a prolonged disease which impaired work capacity. They were probably a quite significant proportion of the samples, but exactly how large is uncertain.

Thus, one reason for the improvement in median consumption opportunities (in Finland, Norway and Sweden) was that not a few in the samples had a life situation and income sources which implied that the coming death would have little impact on incomes. Among them, income developments would be more or less parallel to general trends, which in the Nordic macro-economic context during these years meant real growth both in wages, salaries and income-related pensions and benefits. In addition, the particular life phases of the analysed samples did also play a role. At the start of the observation period, when aged 43–57 years, quite a few would have adolescent children living at home. Twelve years later, many children will have left. Both in the Finnish and in the Norwegian sample, average household size was about 2.5 at the start of the observation period, but had decreased to 1.8 in the last year before death. When the household's consumption burden decreased, equivalized incomes could increase even if household incomes were stable or even slightly decreasing.

These general considerations do not explain the particular Danish pattern, however, which was probably related to retirement particularities. The typical income drop on transition to pension has been estimated to be generally larger in Denmark than in the three other countries in 2003 (NOSOSCO, 2005, p. 124, 139). Analyses by OECD, referring to the early 2000s, have estimated that ‘net replacement rates’ for average earners when retiring (voluntary pension schemes excluded) were 54.1 per cent in Denmark, 65.1 per cent in Norway, 68.1 per cent in Sweden and 78.8 per cent in Finland (OECD, 2005, p. 52). In addition, typical age when retiring during the 2000s was lower in Denmark (and in Finland) than in Norway and Sweden (Finnish Centre for Pensions, 2008, pp. 23–24; Statistics Sweden, 2013, p. 29). This suggests that the decreasing median incomes observed in Denmark as death approached (Figure 1) were probably due to relatively many transitions to pensions in combination with more income drops after retirement than in the other three countries.

Figure 4 suggests the likelihood of this interpretation. This figure is a replica of Figure 1, but includes only those who died when aged 55–64. Those who died when aged 65–69 years have been excluded, i.e. the parts of the analysed samples where retirement certainly was particularly widespread. Figure 4 shows, in line with the interpretation proposed above, that in these reduced samples, median income increased also in Denmark, both for men and women, although not as much as in the other countries. It can be added that separate analyses in the Danish sample, including only those who died when aged 55–59 years, indicated an even larger income growth during the 12 observation years (table not showed, available on request).

Figure 4. Median incomes during the last 12 years alive, relative to 12 years before death (= 100 per cent); men and women who died when aged 55–64 years.

Thus, it may be that many underlying income maintenance mechanisms were rather similar in the four countries, but a tendency to earlier transitions to retirement in combination with somewhat larger income decrease when retiring implied that an income decline in the analysed Danish samples occurred. However, typical retirement age was also relatively low in Finland (Finnish Centre for Pensions, 2008, pp. 23–24); the reason that no income decline was observed in the Finnish sample was probably that income drops after retirement were less marked than in Denmark (OECD, 2005, p. 52).

The educational inequalities

Social equity is a proclaimed ideal of the Nordic welfare states (Kvist et al., Reference Kvist, Fritzell, Hvinden, Kangas, Kvist, Fritzell, Hvinden and Kangas2012), and social benefits in the Nordic countries are often considered redistributive since they compensate work income to a higher degree for low-income employees than for higher salaried (Reference Hinrichs, Lynch, Castles, Leibfried, Lewis, Obinger and PiersonHinrichs & Lynch, 2010, p. 358; NOSOSCO, 2005). One might therefore expect that the possible economic burden of dying prematurely would not differ with social status. The analyses (Figure 3) showed, however, that in each of the four countries, the higher educated experienced on average a more advantageous income development during the years prior to death than the low educated.

Better social networks, more family resources and better social insurance could be contributing to this pattern. However, access to work income is also likely to be very important. This is indicated in Figure 5, which shows median work income during the study period for men with lower and higher education (this information was only available for Denmark, Norway and Sweden). Among those with low education, median work income started to drop around 10 years before death and was reduced to zero during the last four years alive. In contrast, those with higher education experienced in general no drop in median work income until about six years before death, and even during the last year alive, median work income was above zero. As those who were able to stay employed would generally have better incomes than those who left employment, differences in access to work income did certainly contribute to the overall better income developments for the higher educated than for the low educated.

Figure 5. Median pre-tax work income prior to death among men who died when aged 55–69 years, low and high educated, per cent of median income the first year of 12 years observation period.

Differences in work participation could be due to different placement in the occupational structure. Employers might be more inclined to keep high educated on the work staff even when afflicted by disease, and those with higher education will typically have occupations where functional limitations more often will be compatible with usual work tasks – such limitations may mean less for a designer or researcher than for a bus driver or a plumber (Reference Lutfey and FreeseLutfey & Freese, 2005; Reference Syse and TonnessenSyse & Tonnessen, 2012).

Since both high and low educated in these samples died, at approximately the same age, it might be that average disease severity during the last years alive did not differ in any major way between educational levels. If so, higher work participation among high educated was not due to less disease severity, but rather to educational differences in possibilities for continued employment even when impaired by disease. Whether disease severity and accompanying functional limitations were similar or varied with educational levels in the analysed samples is uncertain, however. The analysed deaths will differ – many were certainly due to protracted chronic disease, while others would be relatively rapid outcomes of non-external causes (e.g. stroke, aggressive cancer types), and still others would be sudden deaths due to accidents or other external causes. As typical life expectancy is lower among low educated, it may be that disease onset will typically occur at a lower age among them. Data available for the present study did not allow for a closer scrutiny of possible educational differences in health and disease trajectories during the years before death.

When interpreting the educational inequalities in income trajectories prior to death, one should also take into consideration more general trends in income inequality. Figure 6 shows income inequalities between high and low educated during the observation years – not for those who died, as in Figure 3, but in the weighted comparison sample of survivors, i.e. those with similar characteristics, belonging to the same birth cohorts, who lived beyond the last year of the observation period. Figure 6 indicates that also among the survivors, the gap in median income between high and low educated increased considerably during the observation years – by and large in the same way as among those who died. This suggests that the increasing income inequalities between educational levels was not specific for those who died when aged 55–69 years old, but a general feature of developments in income inequalities during the analysed time period.

Figure 6. Median income among low educated in per cent of median income among high educated, in the weighted comparison samples.

Limitations

The present study used high-quality register data in samples where attrition was practically negligible, but information on specific income sources, retirement transitions, and health trajectories during the observation period was mostly unavailable. This limited the possibilities for testing specific explanations. Since the participating researchers were not allowed to export the national individual-level data files to a pooled data file, because of data security regulations, data had to be analysed separately in each country. Thus, a possible role of compositional differences between the country samples for contry differences in income security could not be rigorously examined. The design of this study has advantageous aspects, but an even more ideal design would be to follow samples diagnosed with fatal disease prospectively for a long period of time, with more specific and continuous observations on health, work and income. Whether the results reported in this study are generalisable to other age categories or to other time periods, is of course uncertain. Some welfare programmes have changed in the analysed countries during recent years (NOSOSCO, 2015). This study has focussed on equivalized disposable income, which does not capture how free or heavily subsidised public services contribute to economic welfare. This will hardly mean much for inter-Nordic comparisons, however, since public services were mostly funded in similar ways in these countries during the time period under study.