1 Introduction

In comparative welfare research Nordic countries often stand out in terms of low degrees of income inequality and poverty. In fact, these characteristics are often considered to be some of the key elements of the so-called Nordic welfare model (e.g. Nygård 2013; Kautto 2001). Also when it comes to among older Scandinavians, the relative poverty rate plunged drastically since the 1960s due to the introduction of modern pension schemes (Fritzell and Ritakallio 2010; Gustafsson, Johansson and Palmer 2009; Gustafsson and Pedersen 1996; Gustafsson and Uusitalo 1990). In countries like Finland and Sweden, which are in focus here, income poverty among older adults has for several decades been on a relatively low level in international and European comparison (Eurostat 2014; Ahonen 2011), but it has not, however, disappeared altogether as a social problem related to older people. Not only has the overall level of income inequality started to climb since the mid-1990s (e.g. Immervoll and Richardson 2011), it has also been shown that older females often face a higher risk of poverty than do older men (Zaidi 2010; Arber 2006; Price and Ginn 2003). In addition, many measurements for estimating poverty focus on household economy. This means a risk of underestimating the economic situation of economically dependent women and older adults in households with an overall stable economy [European Anti-Poverty Network (EAPN) 2014].

In the literature on old-age poverty women's higher poverty incidence is often called the’ feminization of poverty’ (cf. Brady and Kall 2008; Pearce 1978), and it has been suggested that this problem is connected to the overall weaker labour-market position of women, lower income levels and their larger shares of the responsibility for domestic care [Tuominen, Nyman and Lampi 2011; Statens offentliga utredningar (SOU) 2003]. Accordingly, women’s careers are characterised by a higher incidence of career brakes due to family leaves, part-time employment, and so-called a-typical contracts that make them more exposed to lower incomes and thus lower pensions in old age. Moreover, although there are some differences in female labour-market participation between Finnish and Swedish women, with Swedish women adhering to the housewife model for a longer time in history than Finnish women (e.g. Hiilamo 2002), pre-Second World War cohorts of women in these countries display a systematically lower participation rate than do later cohorts of women (e.g. Thelin 2013; Julkunen 1994; Jallinoja 1991). One important outcome of this is that older females in both Finland and Sweden, respectively, face a higher risk of income poverty that do older males (Jungerstam and Wentjärvi 2012: 112)—something that has been highlighted both in the Finnish and Swedish discussion about ageing and age policies (e.g. Government of Sweden 2013; Sosiaali- ja terveysministeriö 2009).

The aim of this article is to contribute to the understanding of the abovementioned problem by examining differences in objective and subjective income poverty between older women and men in Finland and Sweden. The article is based on survey data collected in 2010 among 65-, 70-, 75- and 80-year-olds living in the West-Finnish region of Österbotten/Pohjanmaa and the North-Western region of Västerbotten in Sweden (the GERDA survey)Footnote 1 (N = 3260). In contrast to national surveys of retirees’ income levels using some kind of sampling technique (e.g. Palomäki 2011), the GERDA survey can be considered as a population study in terms of people in these four age categories. It is crucial to investigate both objective and subjective aspects of income poverty, since these two reflect different dimensions of a person’s livelihood that do not necessarily coincide (cf. Kautto et al. 2009; Kangas and Ritakallio 2008). Whereas subjective income poverty refer to the subjective experiences of economic scarcity in an everyday-life context (e.g. Dirven and Berghman 1995; Saunders, Halleröd and Matheson 1994), objective income poverty, comes into question when a person’s (disposable) incomes fall below a certain line, for example a politically negotiated poverty line (EAPN 2014; Bridges and Gesumaria 2013; Ravallion 1998; Dirven and Berghman 1995).

The overall aim of the article can be divided into two research questions. First, it examines whether there are differences in objective and subjective income poverty among older females and males in these two Scandinavian regions. Second, it analyses whether such differences are explained mainly by gender or if other possible predictors, such as educational level or civil status, also play a role here. The article thereby contributes to the literature on poverty in old age, and to the discussion about income equality between older females and males in Scandinavia. If older females systematically face higher poverty risks than men, this would indicate a gendered injustice problem that is at least partially sustained by the Nordic welfare state, its pension systems and labour markets. If that is the case, it is something that does not go well together with the notion of the women-friendly Nordic welfare model (cf. Julkunen 2003; Goul Andersen and Hoff 2001). Finland and Sweden can be seen as an interesting pair of countries to compare in this sense, since they share very similar welfare-institutional configurations and have a rather homogenous cultural and political characteristics (ibid.). Therefore we can also expect the ‘feminisation of poverty’ to take rather identical forms. It should be noted, however, that the level of poverty among older adults are relatively low in these two countries if we compare to other countries in the world, and that the level of wellbeing of elders surpass that of many other countries.

The rest of the article is structured in the following way. The next section presents the theoretical framework alongside a brief review of previous research in this field. The third section discusses the data and methods used, the penultimate section presents the findings, and the final section summarises the results.

2 Theory and Earlier Research

Poverty is a notoriously difficult social phenomenon to define, especially when it comes to questions about poverty in a global scale (cf. Spicker et al. 2006). But also in a European context poverty is often an elusive concept that is complicated to assess (e.g. European Commission 2015: 140–141). The difficulty lies not only in finding suitable criteria or methods for measuring poverty, or if we should regard poverty as some kind of material deprivation or as a reflection of income inequality (e.g. Brady 2005), it has also to do with the question as to whether poverty should be seen as something absolute or as something relative (Kangas and Ritakallio 2008). If we restrict the discussion to income poverty, absolute income poverty, put simply, asserts that there is some kind of (universal) income threshold under which the sustainment of human life and physical functional capacity becomes impossible, such as the UN definition “two dollar a day”, whereas the relative definition uses the conception of ‘normal’ living, and the possible exclusion from it, as its starting point. Consequently, relative income poverty not only means the absence of resources necessary for physical capacity, it also implicates an exclusion from ‘normal’ living in relation to shared social values and standards (EAPN 2014; Spicker et al. 2006).

An example of this kind of relative measure is the EU income poverty threshold that sets the (relative) risk-of-poverty line at 60 % of the median income in each country and year, respectively (EAPN n.d.; Fritzell et al. 2011). This indicator, which reflects the overall distribution of income, asserts ‘that households with disposable incomes below this threshold should be treated as being at risk of poverty rather than being in poverty’ (Fritzell et al. 2011: 7). Although ‘at-risk-of-poverty’ rates are often used and fairly easy to comprehend, they are often problematical since they are calculated on the basis of disposable household incomes, which assumes that all members in a given household have access to an equal share of the disposable income. Moreover, they often fail to draw light on the depth or persistency of poverty, since they only calculate the share of households beneath a given poverty threshold (Khandker and Haughton 2009).

Another key distinction in this respect is the one between the income poverty approach (Jäntti and Danziger 2000) and other measures of resources necessary for a normal living, such as subjective assessments of how one gets by in everyday life (e.g. Kangas and Ritakallio 2008). Although the income poverty approach has a central position in both European and global poverty research (Alcock 2006), subjective methods have gained ground in recent years since they not only measure the lack of resources in a theoretical way but also give an assessment of the practical impacts that such a lack may have on people’s everyday lives (Alcock 2006; Ravallion 1998). It is important, however, to remember that relative income poverty measurements used in most European countries do not represent a “lack of resources”, rather they reflect the degree of inequality in the distribution of income. The fact that an individual or a household earns less than 50/60 % of the median income does not necessarily mean that the individual/household is going to lack resources or suffer individual deprivation.

In addition to being a difficult phenomenon to capture, the incidence of poverty has been found to be concentrated to certain age groups. Already in the early-1900s, Seebohm Rowntree (1901) found that poverty is closely associated with childhood, early middle age and old age. Although the emergence of the welfare state has mitigated these poverty risks to some extent, old age is still likely to be a period in life when persons become exposed to poverty due to retirement (e.g. Scharf and Keating 2012; Smeeding and Sandström 2005; Middleton 2002). It should however be noted that the relation between poverty and old age is not necessary a consequence of chronological age in itself, but rather a reflection of the different socioeconomic situation and conditions following retirement (Alcock 2006).

When discussing old age, it is important to remember that any given age categorization entails a certain amount of arbitrariness, and that old age has different meanings in different cultures and times (Wilson 2000). Here we define older adults as persons aged 65 or more, since it largely corresponds to the official retirement ages in Finland and Sweden (Nordic Council of Ministers 2010). Poverty in old age is to a large extent a gendered problem. A majority of poor people living in unacceptable conditions in developing countries are women (UN Women Watch 2016). Also in a Nordic context, the ‘feminization of poverty’ can be traced in income data for older adults. For example, in countries like Finland and Sweden, the average level of pensions has been shown to be systematically higher for men than women (e.g. Flood 2014; Sjögren Lindquist and Wadensjö 2012; Ahonen and Bach-Othman 2009).

In other words, the determinants as well as the consequences of poverty seem to be strongly related to a person’s gender. Not only do women face a higher risk of experiencing income poverty in old age due to more fragmented working careers and greater domestic responsibilities, they often face another structural inequality in terms of lower pensions generated by income-related pension schemes (Arber 2006; Price and Ginn 2003). Besides getting lower pensions than men, women’s subjective experiences of poverty are likely to differ from those of men, but also to be less investigated by the scholarly community, which may lead to a risk of underestimating the true consequences of female poverty such as stigmatization or structural dependencies (Lister 2004; Kabeer 2003; Walker 1981).

The concept’ feminization of poverty’ (cf. Brady and Kall 2008) gained ground in the 1970s as a reaction to the fact that the income poverty risk for women had increased in relation to that of men (Lister 2004; Bianchi 1999). Still today this remains a problem, both among people in working age and older people (Brady and Kall 2008), and it is often considered as a structural disadvantage inherent in the contemporary welfare states of Western democracies (e.g. Walker and Maltby 2014). This is especially problematical for the Nordic countries, where the relatively high levels of gender equality have long been used as a trade mark for Scandinavian women-friendliness (e.g. Goul Andersen and Hoff 2002).

The opinions as to what the explanations to the gender differences in income poverty risks may be are partly divided between scholars. Among the most influential explanations we can find variables that relate to the number of years in working life and life-long earnings, but also socio-demographic variables such as gender and civil status are important (Thelin 2013; Arber 2006; Price and Ginn 2003). If we look at female labour participation on a general level, during the period between 1970 and 1990 women in Western democracies acquired roughly the same level of education as men, which means that the gender wage gap cannot be accounted for by differences in human capital (Arber 2006; Lister 2004). Instead we need to look for answers in the fact that women generally choose employment in different sectors of the labour market than men, and that women tend to stay employed in the same position for a longer time than men (Orloff 2009; Middleton 2002; Gough 2001). Furthermore, female careers are often disrupted by child births and domestic responsibilities, which also account for a large share of the wage gaps, as well as differences in poverty risks in old age (ibid.). Given the two-tier structure of pension systems in Finland and Sweden, where the replacement rates of work-related pensions are dependent upon one’s income, gender differences in income accumulated during one’s working life is later mirrored in terms of pension gaps between older women and men (Flood 2014; Sjögren Lindquist and Wadensjö 2012; Kautto 2011).

This does not mean however that there is a direct link between women’s weaker labour market affiliation and a higher income poverty risk in old age. If the magnitude of labour market participation during one’s active life had a direct influence on the poverty risk in old age, the gender differences in poverty risks could be expected to be larger in countries where female labour market participation is on a much lower level than male participation (Kautto 2011). This is not the case. Instead we find that gender differences in income poverty are much higher in countries where women tend to participate very actively in paid labour, such as the Nordic countries (Ahonen and Bach-Othman 2009; Ahonen 2006). By contrast, the size and type of household seems to have a somewhat stronger bearing for the risk of facing income poverty in old age. In countries where older females often live in large households, such as Southern European countries, the gender differences in old-age income poverty risks are less accentuated than in countries where it is more common for older women to live alone (Ahonen and Bach-Othman 2009; Ahonen 2006).

Previous research also suggests that although the general income poverty rate among retired persons in many Western countries has decreased during the last decades or so, the poverty rates among the oldest old seem to climb, regardless which country we look at (cf. Scharf and Keating 2012). Moreover it seems that it is predominantly older women, and the single ones in particular, that are likely to face income poverty when they grow old (Ahonen and Bach-Othman 2009; Smeeding and Sandström 2005; Shaw and Lee 2005). The few countries that do not follow this general pattern, such as Iceland and New Zeeland, also tend to display low levels of general poverty among retirees (OECD 2011).

When it comes to gender differences in incomes and poverty in Scandinavian countries like Finland and Sweden, there is considerably less research available than for the working-age population in general (Ahonen and Bach-Othman 2009). One exception from this rule is a longitudinal income study that was made in Finland between 1966 and 1990 (Jäntti et al. 1996). According to the study, the average income poverty rate among older adults started to plunge drastically in Finland during the 1970s, although the poverty risk was found to be higher among older men living alone than among older women living alone. Interestingly enough, this pattern was found only for the period prior to the early-1980s. After that the groups shifted places so that older women living alone started to face higher income poverty risks than single men. One plausible explanation to this result is the expansion of the Finnish income-related pension system since the early-1960s, and the fact that the abovementioned wage gaps between sexes also tend to explain why older men started to get higher income-related pensions than older women from the late-1980s on. Moreover, the cutbacks that were made in the flat-rate people’s pension system during the 1990s crisis, affected older women more severely than men, since a larger share of older women rely partly or solely this pension benefit (cf. Jäntti et al. 1996).

A somewhat similar picture can be drawn for Sweden, where female labour market participation, historically speaking, has been lower than in Finland (Sjögren et al. 2012; Hiilamo 2002) and female working careers tend to have been characterised by disruption due to family responsibilities and short-term contracts (Gustafsson et al. 2009; Gustafsson and Pedersen 1996). According to a Swedish study (Thelin 2013), the subjective experiences of income poverty in old age differ a lot between persons and also between sexes. Not only were the experiences of poverty firmly related to the persons’ life-specific trajectories, such as birth cohorts and working-life experiences, but also living conditions, civil status and gender were related to experiences of poverty.

In a study of Finnish retirees’ economic situation (Rantala and Suoniemi 2007), it was shown that the relative income situation of retirees started to improve during the early 1990s, mainly as a result of the fact that the economic depression had a curbing effect on the overall development of income. During the period of economic recovery and rising incomes in the early 2000s, pensions started to lag behind in relation to the general income development, which also meant an increase of the income poverty risks in certain groups. Most severely hit were women aged 75 or older with an overall poverty rate of 29.3 % in 2004. The corresponding rate for men in the same age category was only 12.2 %. Also civil status was found to have bearing for the poverty rate; older adults (both men and women) living together with a partner faced lower income poverty risks than single persons (Rantala and Suoniemi 2007).

According to a similar study made by Statistics Finland in 2009 (cited in Rantala 2011) the income poverty rate among single people aged 75 or more was considerably higher than among married or cohabiting couples, but it was also shown that the prevalence of poverty seem to increase with age in a more accentuated way for women than for men (Rantala 2011). A plausible explanation to this finding is that a higher proportion of older women live alone, and that the mortality of low-income males is somewhat higher, which would also explain why men that live longer often are relatively more well-off (cf. Ahonen 2006). Furthermore, as noted above, another plausible explanation can be found in the interconnection between the Finnish labour market and its pension system.

On the basis of this, we can expect to find gender differences in objective and subjective experiences of income poverty among older adults in the two regions under scrutiny, and gender to have a strong predicting role for the odds of income poverty.

3 Data and Methods

3.1 Data and Study Population

As noted above, this article seeks to explore gender differences in experiences of objective and subjective income poverty among older adults in Finland and Sweden, and to test what role gender plays as a predictor of poverty when simultaneously controlling for other variables. The data was obtained from a cross-sectional survey that was conducted 2010 as part of an inter-regional research project, the GERDA Bothnia project (see GERDA 2012). The overall aim of this multidisciplinary project was to map living and health conditions of older adults living in the Bothnia region, i.e. on both sides of the Gulf of Bothnia, in Västerbotten in Sweden (15 municipalities/211,884 inhabitants) and Österbotten/Pohjanmaa in Finland (17 municipalities/178,000 inhabitants).

As part of the project a questionnaire was sent to all persons aged 65 (born 1945), 70 (born 1940), 75 (born 1935) and 80 (born 1930) in rural municipalities and to every second person in the most populous town in Österbotten and every third person residing in the two most populous towns in Västerbotten. The urban residents were systematically sampled on the basis of the official population register in Finland and the official tax authority in Sweden. This technique selected every second or every third person in the abovementioned age groups and towns, respectively (Herberts 2011). A total of 10 696 questionnaires, each and every one according to the mother tongue of its receiver, were sent out in September–October 2010.Footnote 2 After reminders a total response rate of 64 % (n = 6838) was obtained. The questionnaires were answered by 3779 persons in Sweden and 3059 in Finland, resulting in a total response rate of 70.7 and 57.2 %, respectively. The response rate was higher among the two younger age groups (approx. 66 %) than those aged 75 and 80 (62 and 59 %, respectively). Furthermore the survey generated a lower response rate among the Finnish-speaking Finns (52.9 %) than among Swedish-speaking Finns (61.5 %).

3.2 Outcome Variables

The 2010 GERDA survey contains one item on older persons’ gross income as well as two items assessing their subjective financial situation. On the basis of the review of previous research in this field (e.g. Sjögren et al. 2012; Kautto 2011), we decided to use a dual approach for assessing income poverty among older adults, since we believe that such an approach captures the poverty problem in a fuller sense than a one-dimensional approach. This means that we in this article focus on income poverty, experienced both in an objective and a subjective way. While objective income poverty (henceforth ‘objective poverty’) pertains to the actual income disposable to a household or a person, subjective income poverty (henceforth ‘subjective poverty’) relates to subjective evaluations of the economic situation. Not only does this procedure reflect poverty in terms of a given poverty line, which is often a ‘top-down’ construction created by experts and politicians, it also give voice to ordinary persons’ evaluations of their economic resources (e.g. Kangas and Ritakallio 2008). Accordingly, the two outcome variables used in the analysis were objective poverty and subjective poverty (cf. Table 1). The first variable, objective poverty, is a dichotomized variable constructed on the basis of an item assessing one’s gross income per month.Footnote 3 Also the second outcome variable, subjective poverty, is a dichotomised variable based on older person’s subjective assessments of their economic situation.Footnote 4

Table 1 Descriptive variable statistics by country and background variables (N = 6838), valid percentages
Full size table

As to the first outcome variable, persons with a gross income under or equal to 1000 €/10 000 SEK are categorised as experiencing objective poverty, whereas persons with an income over this line is categorised as non-poor.

In general, this categorisation corresponds to the general poverty line used in Finland and Sweden at the time of the survey, although the match is not perfect. According to official income statistics from Statistics Finland (2008) the relative income poverty line, defined as incomes under 60 % of the disposable median income per consumption unit, was set at 13,800 € per year, which gives a monthly poverty line of 1150 € per month. In 2010, the corresponding poverty line in Sweden was set at 119,460 SEK per year and consumption unit, which gives a monthly poverty line of 9955 SEK (Statistiska centralbyrån 2012). The categorisation used in this study, however, differs from the official categorisations used by Statistics Finland and the Swedish Statistical Bureau, since the GERDA survey data presents gross incomes (not disposable incomes), gives individual accounts of incomes (not household incomes) and uses a categorical variable (not a continuous variable) for the assessment of income. The ordinary approach in poverty research is to depart from household incomes, which are equivalised (adjusted to the number of family members) in order to calculate the theoretical distribution of disposable incomes within the household (Rantala 2011; Palomäki 2011; Kautto 2011). This has not been possible here, since the GERDA survey data assesses personal incomes, not household incomes, something which may have consequences for the comparability with findings from studies of household incomes. It may also be problematic, since older persons can easily make errors in their evaluations of gross incomes, whereas it is probably easier to remember their disposable incomes. The use of a categorical measure, in turn, takes away some of the discriminatory effect of the variable at the same time as it complicates the calculation of poverty rates. Consequently, although the poverty line used in our study differs from the one used in official statistics, it has been considered the ‘next best’ thing. Furthermore, by setting the poverty line a bit lower, we lower the risk of exaggerating the proportion of poor persons.

As to the dichotomisation of the second outcome variable, subjective poverty, persons stating that they have ‘some difficulties’, or that it is ‘difficult’ or ‘very difficult’ to make ends meet are categorised as subjectively poor. Also this procedure is not straightforward and needs discussion. For example, it could be argued that also those having ‘some difficulties’ making ends meet, could have been classified as non-poor, since almost all people face some kind of economic difficulties sometime during their lives. The argument for coding the original response categories ‘some difficulties’, ‘difficult’ or ‘very difficult’ as expressions of subjective poverty is twofold. First, we wanted to make clear distinction between those who did not experience any difficulty at all in making ends meet, and those who experienced some, or many, difficulties. Second, after having analysed an alternative coding of the variable (with those experiencing ‘some difficulties’ belonging to the non-poor group), we found that the number of persons reporting experiences of poverty was so small that it would not have been possible to conduct regression analyses.

3.3 Independent Variables

Six independent variables were used including a country dummy (see Table 1). The first variable is gender, which obviously plays a central role in the analysis since the aim is to assess gender differences in incidence of objective and subjective poverty. The other variables were age (65, 70, 75, 80 years), civil status (single, partnership), self-rated health (poor, good), and education (less than 10 years of schooling, 10 years or more).Footnote 5 Self-rated health was dichotomised so that the response categories excellent, very good and good were put together into a new category (good) whereas the two other categories formed the other new category (poor). As to the variable education the cut-off point was set at 10 years of schooling so that less than 10 years represented low education whereas 10 years of schooling or more represented high education. The choice of independent variables was dictated by the number of available measures in the GERDA data as well as previous research on older persons’ incomes or poverty (e.g. Ahonen and Bach-Othman 2009; Rantala and Suoniemi 2007). Finally, in order to be able to control for any possible association with country, a country dummy variable was also used. Although Finland and Sweden admittedly share many similarities as to their societal and cultural configurations that would lead us to expect similar patterns of old-age poverty, it is possible that there are also specific country characteristics (such as the pension system) that may reflect upon the results. In order to control for such effects, the country dummy was introduced in the analysis.

3.4 Analyses

The analysis was conducted in two steps. In the first phase the prevalence of objective and subjective poverty among women and men in each country was assessed with the help of ordinary contingency tables. In the second phase, we analysed the explanations to objective and subjective poverty by using logistic regression models with country dummies. The first model (M0) tested bivariate associations between each of the independent variables and the two outcome variables. The second model (M1) is a multivariate model that assesses the explanative power of socio-demographic variables according to the following equation:

  • Logit (objective/subjective poor) = b0 + b1(gender) + b2 (age) + b3 (civil status)

In the next model (M2) self-rated health and education was added to the analysis according to the following equation:

  • Logit (objective/subjective poor) = b0 + b1(gender) + b2 (age) + b3 (civil status) + b4 (health) + b5 (education)

In the fourth and final model (M3) also the country dummy was introduced. Consequently the final model is constructed as:

  • Logit (objective/subjective poor) = b0 + b1(gender) + b2 (age) + b3 (civil status) + b4 (health) + b5 (education) + b6(country)

where b0 is a constant and b1, b2,… b6 are estimates for the parameters, β1, β2,… β6.

4 Findings

As shown in Table 2 the prevalence of objective poverty is related to gender, age, civil status, health status and education. As to gender differences, we see that women display a higher prevalence of objective poverty compared to men and that this pattern is visible in both countries, although the differences tend to be more significant in Sweden than in Finland. In both countries and gender groups the prevalence of objective poverty is higher in older age groups, so that 80-year-olds report far higher prevalence of poverty than other age groups. In Finland 54 % of 80-year-old women reported objective poverty while the corresponding rate among men was approximately 42 %. In Sweden the corresponding percentages are approximately 53 and 16 %, respectively.

Table 2 The share of older adults experiencing objective poverty by country, gender and other background variables (% of valid cases within each group)
Full size table

In both countries the relevance of civil status for objective poverty follows a somewhat different pattern for women and men, so that the poverty prevalence tends to be higher for women living with a partner compared to single women, whereas single men report somewhat higher poverty prevalence than men living with a partner. However, when it comes to the relation between self-reported health and objective poverty we see that persons with poor health generally report higher poverty prevalence, irrespective of country and gender. The same pattern goes also for education, where those with ten or more years of education report lower poverty than those with fewer years of education.

If we turn our attention to the prevalence of subjective poverty (Table 3), we see a similar pattern as in the case of objective poverty. In both countries a larger share of women report subjective poverty than do men, even if the differences are much smaller than in the case of objective poverty. As to the relevance of age and civil status, however, we find a slightly different pattern to that of objective poverty. In both countries the prevalence of subjective poverty is more evenly distributed among age groups than in the case of objective poverty, and single persons tend to report higher poverty prevalence than those living with a partner. However, in the case of health and education the pattern is similar to that of objective poverty, so that those with ‘poor’ health and shorter education have a higher prevalence of subjective poverty.

Table 3 The share of older adults experiencing subjective poverty by country, gender and other background variables (% of valid cases within each group)
Full size table

Thus far the analysis has focused on bivariate association within each country and gender category. What picture arises when we conduct multivariate analyses for the predictors of objective and subjective poverty, respectively? In Table 4 the odds ratios of objective poverty, with 95 % confidence intervals, are presented.

Table 4 The odds of experiencing objective poverty of older adults in Finland and Sweden. Odd ratios (OR) and 95 % confidence intervals (CI), N = 6838
Full size table

Model 0, which reports bivariate associations, shows significant associations for gender, age, health, education and country, but not for civil status. The odds ratio for reporting objective poverty is considerably smaller for men and those with good health and more school years, but higher for older age groups as well as those living in Finland. In model 1, which calculates odds ratios for gender while simultaneously controlling for socio-demographic variables (age and civil status), we see that men still have a significantly smaller odds than women of being poor, while the odds is higher in older age groups and among persons living with a partner. This pattern is also visible in Model 2, which controls for health and education, and we can see that the odds for objective poverty is significantly smaller for persons reporting good health and longer education. In Model 3, finally, which also controls for country, we can find similar associations as in the other models, but we also see that the odds of objective poverty is significantly higher among older adults living in Finland than in Sweden. When checking for interaction effects we found a significant interaction between the variables country, gender and education (not reported in the table), but this effect did not alter the odds ratios reported in Model 3.

In Table 5, the odds ratios for subjective poverty are reported. Model 0, which reports bivariate associations, shows that males, persons living with a partner, persons with good health, better educated persons as well as older Finns report smaller odds of being subjectively poor than their respective reference groups. However, no significant association could be detected between subjective poverty and age group. In Model 1, which controls for socio-demographic variables, older males retain smaller odds than women of being subjective poor, and we can also see that 80-year-olds display significantly smaller odds of being poor than 65-year-olds. The same also goes for those living with a partner.

Table 5 The odds of experiencing subjective poverty of older adults in Finland and Sweden Odds ratios (OR) and 95 % confidence intervals (CI), N=6838
Full size table

Model 2, which also controls for health and education, shows a similar picture but also reveals that persons with good health and more school years have significantly smaller odds of being subjectively poor compared to their reference groups. This pattern is true also for Model 3, which includes a country dummy. As can be seen, older Finns report significantly smaller odds of being subjectively poor than those living in Sweden.

5 Discussion

The aim of this article was to investigate differences in the prevalence of subjective and objective income poverty between older women and men in Finland and Sweden, and to test the assumption that poverty constitutes a problem of gendered injustice, or a feminization of old-age poverty, by analysing the explanative power of the gender variable while simultaneously controlling for other variables, such as education. On the basis of our findings a number of conclusions can be drawn.

First and foremost, the results indicate that a systematic gender difference in experiences of both objective and subjective poverty indeed seems to be the case, since older males in both countries report lower poverty rates than women, even after controlling for age, civil status, health, education and country. As could be seen in Tables 4 and 5, the odds of objective poverty is 75 % lower, and in the case of subjective poverty 27 % lower, for men than for women. Secondly, when analysing the association between gender and the two forms of poverty while simultaneously controlling for other variables, such as health status or education, we found that the control variables played a lesser, albeit significant, role in relation to that of gender. This lends support to the assumption that gender plays an important role for the prevalence of both objective and subjective poverty, and that the control variables have a slight mediating effect on this association. In the case of objective poverty we see that the gender difference in odds ratios increased when inserting the socio-demographic (M1) as well as the health and education variables (M2), while the insertion of the country dummy did not have any visible impact. In the case of subjective poverty, on the other hand, we found that insertion of control variables reduced gender differences in poverty odds (M1 and M2), while the insertion of the country dummy had only a slight increasing impact.

It is important to note, however, that the amount of unexplained variance is considerable here, both for the modelling of objective poverty (−2 Log likelihood: 6,350,841 Cox and Snell R Square: 0.159 and Nagelkerke R: 0.229 for the full model) and for the modelling of subjective poverty (−2 Log likelihood: 8,046,364 Cox and Snell R Square: 0.047 and Nagelkerke R: 0.064 for the full model). This suggests that some other characteristic, such as previous incomes for example, may play a role for the prevalence of objective and subjective poverty. Due to the limitedness of the GERDA survey data, i.e. since it is a cross-sectional survey and there is no data on previous incomes, it is not possible to test this assumption.

It is possible, though, to detect a clear association between objective poverty and the control variables used, suggesting that this kind of poverty has a tendency to be higher among older age groups, persons living with a partner as well as Finns, while it is significantly lower among better-educated persons and those reporting good health. To some extend these findings are in accordance with those found in earlier research, for instance Rantala (2011), Rantala and Suoniemi (2007), Smeeding and Sandström (2005) and Ahonen and Bach-Othman (2009) that have found a higher prevalence of poverty among older women than men. These previous studies have also shown that the prevalence of poverty tends to increase with age, but to be negatively associated with health and education. In previous research the higher prevalence of poverty among older women has been explained by pointing at an overall weaker labour-market position of women, lower income levels during their active working years, and their larger shares of the responsibility for domestic care (Tuominen et al. 2011; SOU 2003). Accordingly, women’s careers are characterised by a higher incidence of career brakes due family leaves, part-time employment, and so-called a-typical contracts that make them more exposed to lower incomes and thus lower pensions in old age.

What is somewhat surprising in this study is the effect of civil status and country on objective poverty. Interestingly enough, older persons living with a partner and Finns were more exposed to objective poverty than their reference groups, whereas the opposite was true for subjective poverty. One plausible explanation as to why persons living in partnerships face a higher risk of objective poverty could be that they have estimated their real incomes (incomes before taxes) in relation to their partner and that the response reflect some kind of household-internal distribution of real incomes that is impossible to control for on the basis of the data at hand. The country difference in objective poverty, in turn, can to some extent be explained by differences in pension rights, suggesting that the Swedish pension system is more efficient than the Finnish when it comes to counteracting old-age poverty. What is interesting though is that older Finns reported lower subjective poverty than Swedes, which does not go neatly hand in hand with such an interpretation, and that may suggest that some cultural factor could be at play here. For instance, given that income levels have been historically lower in Finland and that poverty has been a greater historical ‘social evil’ in Finland (e.g. Hiilamo 2002), it is plausible that such historical conditions may be reflected in subjective assessment of poverty in form of less modest income expectations and thus a lower inclination to report subjective poverty.

A third conclusion that can be made is that the systematic, and gendered, difference in objective and subjective poverty undermines the picture of these Nordic countries as bastions of a gender-equal society for older people. Although it is true that Nordic countries display a high degree of gender equality, for example by fostering high levels of female employment and safeguarding a high level of political rights for women (e.g. Nygård 2013; Goul Andersen and Hoff 2001), systematic gender differences in income among the older populations undermines this picture and suggests that the encompassing pension systems in these countries fail to counteract such injustices. The problem of old-age poverty, or the ‘feminization of poverty’ (cf. Brady and Kall 2008; Pearce 1978), does indeed seem to be a problem facing today’s older populations, and this is something that needs to be addressed and discussed in the public sphere, since it may have implications for the overall wellbeing of older people.

Finally we need to address some of the limitations of this study. The first limitation relates to the focus of the study. On the basis of the data used here, no inference as to the whole older population in Finland or Sweden can be made, only to the older population living in the regions of Österbotten in Finland and Västerbotten in Sweden. A second limitation relates to the way that incomes are measured in the survey. It is possible that persons in general and older persons in particular, could face difficulties in estimating their gross incomes, instead of estimating their disposable income. The GERDA survey asks for ‘incomes before taxes’, which is a somewhat extraordinary praxis compared to other surveys, where respondents are asked about their disposable incomes. A third limitation is the limitedness of the survey in terms of its cross-sectional character and the lack of data that could help to improve the model fit of the regression models, for example data on previous incomes. Notwithstanding these limitations, the findings show that there are systematic gender differences in the prevalence of both objective and subjective poverty, and they may also be said to offer at least tentative support for the assumption that gender plays an important role for explaining poverty prevalence among older Finnish and Swedish adults. In order to substantiate this interpretation, though, further research is warranted.