Poverty of widows in Europe

Transcription

1 Poverty of widows in Europe Anikó Bíró Central European University, The University of Edinburgh October 7, 2011 Abstract In this paper I investigate the relationship between widowhood and poverty among women aged 50 and above in Europe. I analyze the e ect of widowhood on income, wealth, health, and labor force status. I argue that not only the lack of the deceased husband s income, but also the worse health condition and earlier retirement of the widows contribute to the unfavorable nancial conditions. I also analyze the role survivors pensions have in mitigating the adverse e ects of widowhood. The results of the paper are based on the Survey of Health, Ageing and Retirement in Europe, and its retrospective third wave (SHARELIFE). Using retrospective data I can analyze the dynamics of the adverse e ects of widowhood. 1 Introduction Old-age poverty is strongly associated with poverty in widowhood. Widowed individuals, and especially widowed women are more likely to be in poverty than the rest of the population. In this paper I analyze the poverty of widows in Europe. I provide some descriptive evidence on the nancial di culties of the widows. Based on the Survey of Health, Ageing and Retirement in Europe, about half of the widows who are aged 50 and above report nancial di culties, which is around ten percentage points higher than the similar ratio among the married and single women at the same age category. Such di erence cannot be seen among the men. Apart from the lack of the deceased husband s income, other factors can also contribute to the poverty of widows. I investigate the short and long run e ects of widowhood on the nancial circumstances, health, and employment status. Health problems and exiting the labor market can exacerbate the deprivation in widowhood. I also analyze the cross country di erences in the determinants of poverty in widowhood. This analysis can provide some insight on the e ciency of the various social security systems in preventing widows poverty. The results of this paper are based on the Survey of Health, Ageing and Retirement in Europe (SHARE). I use the second wave of the survey, which covers individuals aged 50 and above, and I use also the third wave, which is a retrospective survey. The cross sectional analysis based on the second wave of the survey makes it possible to analyze the e ect of widowhood on the living conditions at older ages, whereas based on the retrospective data I analyze the immediate e ects of widowhood. In the empirical analysis I take into account that widows are on average in worse socioeconomic status even before widowhood than the rest of the female population. I am grateful for comments received at the 3rd SHARE User Conferenc in Tallinn. 1

2 The estimation results indicate that the death of the husband has immediate adverse e ect on the nancial, health, and labor market status of the widow. These e ects are on average also long lasting, but are less severe for those widows who are still working and who have post secondary or tertiary education. There are cross country di erences in these e ects, however the variation in the negative e ect of widowhood on the nancial status cannot be explained by the di erences in the generosity of survivors pension. The rest of the paper is structured as follows. In section 2 I provide a brief review of the related literature and discuss my contributions. In section 3 I describe the data, and in section 4 I present the estimation results. I relate the estimation results to the survivors pension systems of the analyzed countries in section 5. Section 6 concludes. 2 Related literature The rst part of the related literature provides evidence on the relative poverty of widows. The second part of the related literature analyzes to some extent the factors leading to the poverty of widows. The main contributions of this paper are to provide an international comparison on the poverty of widows in Europe, and to investigate the in uencing role of employment, health, and social security systems on the living conditions of widows. As an additional novelty, I can also utilize the retrospective nature of the third wave of SHARE, based on which I analyze the dynamics of the adverse e ects of widowhood. Using the European Community Household Panel data, Ahn (2005) documents that among the 65+ population around 31% are widowed, and for women this ratio is 45%. Widowers have on average higher income than widows, and the income di erences across countries and between genders are larger for those living alone. Due to widowhood the income of women decreases more than of men. According to Ahn, the ratio of household income after and before widowhood ranges between 50 90%, thus there are substantial variations across the countries in Europe. Using the SHARE data, Tinios et al. (2010) provide evidence that widowhood signi cantly increases the probability of persistent poverty, but they do not investigate how this in uencing mechanism works. Siegenthaler (1996) compares ve European countries and the U.S. in preventing poverty among widows. Based on this comparison he concludes that those old-age security systems are the most e cient in preventing poverty which provide a minimum income to all. Among the analyzed countries, the Dutch and Swedish systems are found to be the best in preventing poverty of widows. Monticone et al. (2008) provide an overview of survivors pension in the EU. All countries provide some type of survivors pension, although the conditions for eligibility are becoming increasingly strict. Smith and Zick (1996) analyze the increased mortality risk and worse health status of the surviving spouse. Potential reasons can be the bereavement, stress, and role changes associated with widowhood. Smith and Zick base their empirical analysis on the U.S. based Panel Study of Income Dynamics and death certi cate information, and they di erentiate the mortality of the surviving spouse by the mode of death of the spouse. They nd elevated mortality risk for widowers, especially if the death of the spouse was sudden. On the other hand, for widows they do not nd signi cant increase in mortality, which they explain by that the lack of need for role changes and by emotional preparedness. In this paper I investigate the e ect of widowhood not on mortality, but on morbidity. Hurd and Wise (1987) analyze the circumstances that lead to the disproportionate poverty of widows in the U.S. They nd that one explanatory factor of poverty is the low accumulation of wealth prior to the death of the husband. Hurd and Wise also nd that poor widows loose a higher percentage of the household 2

3 wealth with the death of the husband than those who are better-o, partly because of the absence of life insurance. In a related study, Sevak et al. (2003/2004) claim that the poverty of widows in the U.S. is not only due to the lost income of the husband, but also to selection. Based on the Health and Retirement Study, they provide evidence that poor women are more likely to become widow, and a substantial number of widows in poverty were poor also during marriage. They also nd positive relationship between poverty and the duration of widowhood. Based on the SHARE data I also provide some evidence on the selectivity into widowhood in section 4.1. Burkhauser et al. (2005) compare the economic e ects of widowhood in Canada, Germany, Great Britain, and the U.S. Using individual level data from the Cross-National Equivalent File they focus on the size and variation of private and public income sources that can o set the loss of the husband s income. The SHARE data is not suitable to determine individual level replacement rate of the survivors pension, therefore in this paper I use country speci c indicators of the generosity of the survivors pension. On the other hand, I can extend the analysis of economic well-being with additional focus on health and labor force participation. 3 Descriptive statistics In this paper I use data from the SHARE and SHARELIFE. The Survey of Health, Ageing and Retirement in Europe (SHARE) is a multidisciplinary and cross-national panel database. It covers individuals aged 50 and above and their spouses. The rst wave of the SHARE data was collected in 2004, and the second wave in SHARELIFE is the third wave of data collection for SHARE, which focuses on the SHARE respondents life histories. The data on detailed retrospective life-histories were collected in Based on the second wave of the SHARE, 7:3% of the men and 25:1% of the women in the sample are widowed. Table 1 provides some comparison on the age, economic background, and health status of widows and widowers, and the control groups. The binary indicator of nancial di culty is set to one if the household reports to make ends meet with great di culty or with some di culty. The wealth and income indicators are ppp adjusted, discounted to year 2005, and are divided by the square root of household size. I use the square root of household size in order to account for the returns to scale within households. The mean of the ve SHARE-imputations is used here. 2 The self-reported health is on the 1-5 scale, with 1 corresponding to excellent, whereas 5 corresponding to poor health status. The objective health indicators (chronic health conditions, ADL di culties, symptoms) are generated based on the reported health problems of the respondents. The presented statistics indicate that the widowed are older, and in worse health status than those who live with a spouse or who have not been married. However, there are considerable di erences between the two genders in terms of the economic indicators. As for males, the reported (per capita) wealth and income of widowers is smaller than of the rest of the males in the sample, but these di erences are of moderate magnitude, and for the two wealth indicators statistically insigni cant. The subjective indicator of nancial di culties is approximately equal for the two subgroups of males, reporting nancial di culties is slightly less prevalent among the widowers. On the other hand, widowed females report considerably lower 1 This paper uses data from SHARELIFE release 1, as of November 24th 2010 and from SHARE release 2.4.0, as of March 17th The SHARE data collection has been primarily funded by the European Commission through the 5th framework programme (project QLK6-CT in the thematic programme Quality of Life), through the 6th framework programme (projects SHARE-I3, RII-CT , COMPARE, CIT5-CT , and SHARELIFE, CIT4-CT ) and through the 7th framework programme (SHARE-PREP, and SHARE-LEAP, ). Additional funding from the U.S. National Institute on Aging (U01 AG S2, P01 AG005842, P01 AG08291, P30 AG12815, Y1-AG and OGHA , IAG BSR06-11, R21 AG025169) as well as from various national sources is gratefully acknowledged (see for a full list of funding institutions). 2 The SHARE provides ve imputed datasets for the nancial variables, such as income and wealth. 3

4 income, wealth, and also more severe nancial problems than those who are not widowed. These ndings indicate that poverty in widowhood a ects the females more than the males in terms of these observed indicators. According to the t-test of the equality of means, the subgroup di erences are signi cant at the 1% signi cance level, with the exception of the reported nancial di culties and wealth indicators for males (the reported p-values refer to two-tailed t-tests, and equal variances across the subgroups are not assumed). At the bottom part of Table 1 I report two measures of poverty by widowhood and gender. The rst one is a measure of absolute poverty, where I de ne a respondent to be poor if her ppp adjusted annual income is below 3,650 Euro (corresponding to 10 Euro per day poverty line). The second one is a measure of relative poverty, where I categorize a respondent to be poor if her annual income is less than 60% of the country speci c median income. These indicators also show that widowhood is more likely to imply poverty among women than among men. males females not p-value of the not p-value of the widowed widowed equality test widowed widowed equality test age nancial di culty net worth (1000 EUR) nancial wealth (1000 EUR) income (1000 EUR) self-reported health number of illness number of ADL di culties number of symptoms poor, absolute poor, relative Table 1: Descriptive statistics (sample means) based on SHARE wave 2 In Figure 1 I present the ratio of female respondents reporting nancial di culties by marital status and by country. The graph shows that widows are more likely to report nancial di culties than the control group in each country, with the largest di erence in Sweden. However, there are large di erences across countries in the prevalence of reported di culties. These di erences can be partly due to the di erent nancial circumstances, but also partly due to the di erent reporting behaviors. As the indicator of nancial di culties is based on a subjective measure, this indicator can be strongly in uenced by the response patterns of the respondents. Nevertheless, this empirical problem does not invalidate the observed di erences between the widows and the rest of the female respondents in each of the countries. I structure the retrospective SHARELIFE data into a panel structure with observations ranging from 1930 to Among all these observation points, in case of males 1:2% is widowed, whereas for females this ratio is 5:3%. If the sample is restricted to those aged 50 and above then these ratios are 4:0% and 17:3%, respectively. In Table 2 I present some descriptive statistics for those who are widowed and who are not. These statistics are based on the pooled subsample of individuals aged 50 and above in the referred year. Those who are widowed are older on average by about 10 years. The binary indicator of nancial hardship equals one if the respondent reports that there was a distinct period in her life of nancial hardship, and the starting year of that period equals the referred year. Widowed women are more likely to report such di culties than those who are single or married, but similar di erences are not present for men, among 4

5 Figure 1: Ratio of female respondents reporting nancial di culties based on SHARE wave 2, black bars corresponding to widows, white bars to the control group whom the widowers are slightly less likely to report nancial hardships. These ndings are in line with the statistics related to nancial di culties based on the second wave of SHARE (see Table 1). The indicator of poor health equals one if the respondent reports a distinct period in her life of poor health, and the starting year of that period equals the referred year. The illness indicator equals one if in the given year a period of serious illness started, which lasted for more than a year. These two health indicators show that the widows are on average in worse health condition than the rest of the female respondents. However, based on these statistics it is not possible to di erentiate the e ect of widowhood on health from the e ect of older age. Again, for males these di erences are less straightforward: only the subjective indicator of poor health shows that the widowers are in worse health condition. I also check if widows on average leave a job or retire in a given year with di erent probability than those who are not widowed, and the statistics indicate that these probabilities of exiting the workforce is lower for the widowed respondents. On the one hand, this can be due to the higher motivation of the widows to stay employed, on the other hand, if the widows are already outside of the labor force by age 50 then that can also contribute to this nding. Figure 2 shows by age categories the mean of the indicators of nancial hardship, poor health, and leaving job for females. It can be seen that nancial hardship is more prevalent in the younger age categories (age 50-69), and this higher prevalence is present stronger for widows than for the rest of the female population. However, widows are generally in worse health status according to the indicator of poor health, irrespective of the age category. Only in the oldest age group there are no clear di erences with this respect between those who are widowed and who are not. As for leaving a job, the only clear di erences can be observed in the age group 60-69: the widows are less likely to leave a job in this age category. This di erence is due to the lower proportion of retiring, whereas the widows aged are slightly more likely to retire than those 5

6 males females not widowed widowed not widowed widowed age nancial hardship 0.29% 0.24% 0.28% 0.37% poor health 1.29% 1.53% 1.22% 1.49% ill 0.76% 0.69% 0.78% 0.86% leave job 4.35% 2.40% 2.86% 1.46% retire 3.86% 2.23% 2.46% 1.39% Table 2: Descriptive statistics based on SHARELIFE, aged 50 and above in the control group not widowed widowed financial hardship poor health leave job Figure 2: SHARELIFE ratios for females, by age categories 4 Estimation results 4.1 Determinants of widowhood Before investigating the e ect of widowhood on poverty, I analyze which individual characteristics increase the likelihood of being widowed. This analysis also sheds light on the di erences between the widowed and the control group prior to widowhood. I estimate three linear probability models of widowhood. First, using the second wave of SHARE, I regress the binary indicator of widowhood on the age, years of schooling, and country of the female respondents. Next, using the SHARELIFE data, I estimate a random e ects linear model, where the dependent variable equals one if the respondent becomes widowed in the next year. The sample is restricted to those who are not widowed in the given year. The explanatory variables in this model are the age, the years of schooling, and the country of the female respondents. Finally, I restrict the 6

7 SHARELIFE data to those who are aged 50 or over in the given year. Because of the time-invariant nature of the regressors, FE models cannot be estimated. SHARE SHARELIFE SHARELIFE, 50+ age age [66.93] [47.66] [36.50] schooling schooling [3.51] [13.46] [3.64] Constant Constant [30.50] [1.08] [10.95] Observations 15,489 Observations 817, ,700 Number of individuals 13,107 12,216 Robust t statistics in brackets signi cant at 10%; signi cant at 5%; signi cant at 1% Table 3: Determinants of widowhood, country dummies not reported The results presented in Table 3 con rm that the probability of widowhood increases with age, which is a natural consequence of the similar age of the spouses. In addition, these results show that those with higher level of education are less likely to become widowed. The cross-sectional result indicates that one additional year of schooling decreases the likelihood of being widowed by 0:3 percentage point. The panel result shows that at ages 50 and above, one additional year of schooling decreases the probability of becoming widowed by 0:07 percentage point. The small magnitude of the coe cients under the SHARELIFE estimates is due to the di erent outcome variable: out of the total observations of non widowed females, only for around 0:4% is the indicator of becoming widowed equal to one (1:2% among those aged 50 or above). The years of schooling is considered to be predetermined in these models, since widowhood is typically an old-age phenomenon, whereas education takes place at younger ages. Thus reverse causality is not likely, which would not be true if instead of education, income or wealth were included as a regressor. On average, higher education level is associated with better socioeconomic background and with better nancial status. As a consequence, those who are widowed are likely to be in worse nancial status than the control group even before the widowhood, because they have on average lower education level. This nding is in line with the related results in the literature (see e.g. Sevak et al. (2003/2004)). 4.2 Cross sectional results In this section I analyze the e ects of widowhood on the economic status using the second wave of the SHARE data. I estimate linear regressions of the logarithm of nancial wealth, income, and of the binary indicator of nancial di culties. It is described in Section 3 how I generate these variables. First I estimate only a single coe cient of widowhood, then I extend the model with interactions with receiving survivors pension bene ts, having received bequest of 5; 000 EUR or more from the deceased husband, and the years spent in widowhood. In each model I also control for the age, health status, employment status of the respondent, and I also include binary indicators of post-secondary or tertiary education, having children, and country dummies. The estimated coe cients of the widowhood indicators are reported in Table 4. In Appendix A I report the estimated coe cients of the rst set of models, which do not include interaction terms. The estimation results indicate that widowed women have on average signi cantly lower nancial wealth and income (per capita), and by around 10 percentage points higher probability of having nancial di culties than the control group, ceteris paribus. Receiving survivors pension is estimated to signi cantly mitigate 7

8 ln( n. wealth) ln(income) n. di culty # illness employed widow [13.30] [12.99] [10.96] [2.73] [8.81] R-squared Observations 15,688 15,688 15,327 15,694 15,688 widow [14.70] [14.01] [11.35] [0.23] [7.16] survivors pension widow [6.70] [11.20] [3.76] [4.13] [0.77] widow [13.80] [13.05] [10.37] [2.87] [9.28] bequest widow [4.11] [4.21] [0.17] [1.27] [2.88] widow [5.90] [7.37] [5.86] [2.76] [1.31] years of widowhood [3.75] [1.30] [0.03] [1.17] [8.05] Robust t statistics in brackets signi cant at 10%; signi cant at 5%; signi cant at 1% Table 4: Estimated e ects of widowhood based on SHARE wave 2 these adverse e ects of widowhood on the poverty indicators. If the widow received substantial inheritance from the deceased husband then that also has a positive e ect on the widow s economic status. However, this positive e ect can be seen only on the objective measures of wealth and income. The length of widowhood is estimated to exacerbate the negative e ect of widowhood on the nancial wealth holdings. This is a reasonable nding as the lack of the second income has a cumulative e ect over time. Again, this e ect cannot be seen on the subjective measure of nancial di culties. Further estimations also reveal that the adverse e ects of widowhood on wealth and income are stronger for the older respondents, which can re ect the more severe consequences of long lasting widowhood. On the other hand, the estimated adverse e ects are weaker for those who have post secondary or tertiary education, and who are still working. These results are reasonable as those with higher education and who are still earning depend less on the husband s income. In the last two columns of Table 4 I report the estimated e ect of widowhood on the number of reported illnesses, and on the probability of being employed. In these models the health indicators and the employment indicator are omitted from the control variables. The results indicate that widows are ceteris paribus in worse health condition, but they are more likely to be employed than the control group. The estimated adverse e ect on health is estimated to be stronger for those who receive survivors pension bene ts. This result is puzzling. However, the signi cant coe cients of receiving bequest and years of widowhood on the probability on employment are more reasonable: a longer period of widowhood ceteris paribus can induce women to postpone retirement so as to compensate for the lost income of the husband, and receiving substantial bequest might decrease the motivation for staying in the labor market. I also estimate the models of nancial outcomes with allowing country speci c e ect of widowhood, and of receiving survivors pension bene ts. These coe cients are reported in Table 5. These results show that widowhood has negative e ect on the nancial wealth holdings in all countries. This negative e ect is the strongest in the two post-socialist countries and Greece, whereas in Switzerland it is close to zero 8

9 and statistically insigni cant. The second part of the table shows that this negative e ect is signi cantly mitigated by the survivors pension in 8 out of the analyzed 13 countries, using 10% signi cance level. The estimated average e ect of widowhood on income is also negative for all countries. Receiving survivors pension has unanimously positive e ect on the widow s income, although this e ect is not signi cant in all of the countries. The results indicate that survivors pension has the strongest compensating e ect in those countries where widowhood is the most detrimental on income if the widow does not receive survivors pension. Finally, widowhood is estimated to exacerbate nancial di culties in all of the countries. Survivors pension can mitigate the likelihood of reporting such di culties, but this favorable e ect is signi cant only in Denmark, Switzerland, the Czech Republic, and Poland. I return to these results in section 5, where I discuss the various survivors pension systems in the analyzed European countries. widow widow survivor pension widow ln( n. w.) ln(inc.) n. di. ln( n. w.) ln(inc.) n. di. ln( n. w.) ln(inc.) n. di. AT [4.23] [4.35] [3.64] [4.29] [4.26] [3.21] [1.89] [1.67] [0.81] DE [2.86] [3.13] [3.32] [2.57] [4.55] [2.32] [0.92] [3.98] [0.16] SE [2.19] [7.13] [5.96] [2.60] [5.65] [5.57] [3.97] [2.45] [0.47] NL [2.56] [3.18] [3.60] [3.15] [3.60] [2.99] [2.24] [0.34] [0.36] ES [1.37] [1.90] [2.84] [0.61] [3.56] [1.62] [0.27] [3.69] [0.12] IT [4.51] [0.91] [1.90] [4.35] [2.76] [2.47] [2.19] [2.98] [1.55] FR [1.42] [4.84] [2.85] [1.23] [3.98] [1.42] [0.51] [1.73] [0.66] DK [5.82] [10.54] [3.56] [5.84] [9.98] [4.02] [1.56] [4.75] [3.72] GR [5.59] [2.07] [3.29] [7.18] [4.04] [3.64] [3.87] [6.04] [1.46] CH [0.26] [2.96] [2.31] [0.27] [3.71] [2.88] [1.06] [1.39] [2.28] BE [5.30] [4.39] [3.33] [5.36] [4.12] [2.60] [2.13] [2.95] [0.10] CZ [8.99] [7.78] [5.12] [9.09] [7.38] [6.33] [3.08] [3.06] [3.29] PL [5.31] [7.74] [3.42] [6.15] [6.91] [4.39] [2.48] [1.38] [2.33] Robust t statistics in brackets signi cant at 10%; signi cant at 5%; signi cant at 1% Table 5: Country speci c estimated e ects of widowhood based on SHARE wave SHARELIFE results Using the retrospective SHARELIFE data, I estimate linear probability xed e ects models on the e ects of widowhood. I analyze ve dependent variables: nancial hardship, poor health, serious illness, leaving job, and retiring. The de nitions of these variables are provided in Section 3. The sample is restricted to those aged 50 and above. The estimated results are comparable to the average marginal e ects based on random 9

10 e ects probit models. The linear probability xed e ects model is preferred since that allows the indicator of becoming widowed to be correlated with the unobserved individual characteristics, such as unobserved living conditions. 3 Apart from the widowhood indicator, I also include the squared age in the xed e ects model. The estimated coe cient of the squared age is signi cant at 1% signi cance level, but close to zero for all ve outcome variables. In Table 6 I report the estimated coe cients of becoming widowed in the given year, and one or two years before. The coe cients of the lagged indicators can reveal if the adverse e ect typically begins only one or two years after the husband s death. In case of leaving job and retiring I also include in the models the indicator of becoming widowed in the next year. The reason for this speci cation is that if the husband needs personal care near the end of his life then the wife might choose to exit the labor market so as to provide care. The results indicate that if a woman leaves her job due to (or prior to) widowhood then that happens predominantly via retiring. The models of all ve outcome variables indicate that the adverse e ects of the death of a husband appear within the same year, and these e ects are weaker one or two years before or after. However, this nding can be due to measurement error, which follows from the survey method. Since it is a retrospective survey, the widows might relate the negative life experiences to the event of the husband s death. For example, it is likely that if the period of nancial hardship started one year after the death of the husband then that is still reported as starting in the same year. Becoming widowed increases the likelihood of reporting nancial hardship or poor health in the same year by around 3 percentage points, ceteris paribus, and these e ects are signi cant. The likelihood of becoming seriously ill also increases due to widowhood, but this is less prevalent than reporting poor health status. Women are also more likely to leave their job in the year of the death of the husband, and this happens predominantly via retiring. An explanation for this nding can be that after retirement the wives can devote more time to the needs of the husband. This explanation is supported by the positive coe cient of becoming widowed in the next year: the bad health of the husband might induce the wife to retire. However, this one year ahead e ect is weaker than the e ect estimated for the same year. If I interact the indicator of the husband s death in the current year with age then these extended estimations indicate that the e ect of widowhood on nancial hardship and leaving job decreases with age. Thus the immediate detrimental e ects on economic status of the widow are estimated to be more severe if the widow is younger. Finally, I reestimate the model of nancial hardship with including the indicators of reporting poor health and leaving job as explanatory variables. These estimation results are reported under the last column of Table 6. Under this speci cation the estimated e ect of becoming widow is close to the results of the rst speci cation, although both poor health and leaving job increases the probability of reporting nancial hardship. These results imply that widowhood has direct e ect on the nancial circumstances, and the indirect e ects through health and changing employment status are relatively minor. The raw data also indicate that the adverse e ects of widowhood appear already in the year of the husband s death. In Figure 3 I depict the reported nancial hardship, poor health, illness, and retirement indicators as a function of years in widowhood. The analyzed sample is restricted to widows aged 50 and above. The graph clearly indicates the jump in these indicators around the husband s death. The propensity to retire starts to increase already before widowhood. Although these statistics cannot lter out the age e ect, there is no clear long-run increasing patterns in the indicators of health problems or 3 The Hausman test rejects at 10% signi cance level the consistency of the random e ects linear probability models of nancial hardship, poor health, and leaving job. When running the Hausman test then I do not use robust standard errors. 10

11 n. hardship poor health ill leave job retire n. hardship become widow the current year [8.67] [7.64] [2.83] [4.03] [4.11] [8.34] become widow the previous year [1.80] [1.74] [0.08] [0.03] [0.40] [1.71] become widow years before [0.66] [0.54] [0.32] [1.68] [1.91] [0.63] become widow the next year [1.90] [2.02] poor health [3.95] leave job [7.37] Observations 226, , , , , ,806 Individuals 13,243 13,243 13,243 13,091 13,091 13,243 Robust t statistics in brackets signi cant at 10%; signi cant at 5%; signi cant at 1% Table 6: Estimated e ects of becoming widowed based on SHARELIFE, females aged 50+ nancial di culties. However, these statistics on themselves overestimate the immediate adverse e ects of widowhood, as compared to the xed e ects results. The SHARELIFE survey asks not only the beginning year of nancial hardship, poor health and period of illness, but also the ending year of that. Based on these details the length of these adverse periods can be calculated. If the period of nancial hardship or bad health is still ongoing at the time of the interview, I de ne the ending year as The statistics presented in Table 7 indicate that if a period of nancial hardship starts at the year of widowhood then that lasts on average for 6.6 years, with a median of 5 years. The length of reported poor health is of similar magnitude, whereas the mean and median length of serious illness is longer by 1:5 2:5 years. Thus the estimated adverse e ects of widowhood not transient, but typically last for several years. Estimating zero-in ated negative binomial regression on the pooled SHARELIFE data indicates that the death of the husband increases the expected length of nancial hardship, poor health and illness at the average by around 2, 3 and 1 months, respectively. These estimated e ects refer to di culties starting at the year of the husband s death. 4 mean median std. dev. n. hardship poor health illness Table 7: Length of pediods of di culties starting at the year of widowhood, in years, based on SHARELIFE, females aged 50+ I re-estimate the xed e ects models of nancial hardship, poor health, illness, and retirement with allowing the immediate, lagged, and forwarded e ects of widowhood to vary across countries. Since the 4 Apart from the indicators of the husband s death and aged squared I also control for age in these pooled estimations. In the nancial hardship model I also include the binary indicators of poor health and leaving job as control variables. The estimated e ects are not conditional on reporting a period of nancial hardship, poor health or illness, thus if someone does not report such condition then for her the estimated e ect is zero. 11

12 5.0% 4.5% 4.0% 3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% Years in widowhood financial hardhsip poor health illness retire Figure 3: Dynamics of reporting nancial hardship, health problems, and retirement before and after the husband s death, based on SHARELIFE strongest e ects appear at the same year as becoming widowed, I report the coe cients only of the concurrent indicator of widowhood. These results are reported in Table 8. In case of the e ect of widowhood on illness and retiring, the country speci c e ects are signi cant only for few of the countries. The average e ect is small on these outcome variables, and due to the smaller country speci c sample sizes the standard errors of the country speci c coe cients are relatively large. As for the indicators of nancial hardship and poor health, the di erences in the e ect of widowhood across countries are clearer. Sweden, Denmark, and Switzerland are estimated to fare the best in this respect: for these countries the nancial hardship and poor health status do not become signi cantly more likely in the year of becoming widowed. The most severe estimated immediate e ect of widowhood is in Italy, France, and Poland. In these countries the death of the husband is estimated to immediately increase the probability both of reporting nancial di culties and poor health status, these e ects are strongly signi cant, and are around 5 percentage points in magnitude. 5 Survivors pension 5.1 Institutions Survivors pension can mitigate the poverty of widows, as it can partly compensate the loss of the husband s income. By the middle of the 20th century, some type of survivors pension was provided in all of the analyzed countries. 5 Such pension bene ts are needed since women might not be entitled to such level of pensions which would ensure su cient income. However, survivors pension can also imply e ciency losses through disincentives to work, and can also lead to perverse redistributions. These issues are discussed in 5 The International Labour Organization provides information on the date of the rst law related to survivors pension. The source of this information is 12

13 n. hardship poor health ill retire AT [2.02] [1.56] [2.37] [0.63] DE [1.37] [2.10] [0.92] [1.01] SE [0.87] [1.15] [1.50] [1.35] NL [2.22] [1.90] [1.32] [1.40] ES [2.16] [2.76] [1.30] [2.07] IT [3.46] [2.70] [0.18] [0.83] FR [2.76] [3.39] [0.45] [1.15] DK [1.29] [0.36] [0.89] [1.40] GR [4.48] [1.67] [0.91] [3.31] CH [1.33] [1.60] [1.79] [1.17] BE [2.50] [3.49] [1.05] [1.98] CZ [3.27] [0.81] [1.50] [1.11] PL [3.39] [3.06] [0.81] [0.71] Robust t statistics in brackets signi cant at 10%; signi cant at 5%; signi cant at 1% Table 8: Country speci c estimated immediate e ects of becoming widowed based on SHARELIFE, females aged 50+ more details by Estelle (2009). There are di erences across the analyzed European countries with respect to the entitlement rules and bene t levels of survivors pension. In the following I summarize these country speci c characteristics, based on Monticone et al. (2008) and MISSOC (2010). Austria Widows are entitled to the pension if they were married to the deceased spouse or were divorced but were entitled to maintenance payments. The deceased husband must have had 60 insurance months within the last 120 calendar months. The amount ranges between 0% and 60% of the old-age pension to which the deceased husband would have been entitled to. This percentage depends on the ratio of the husband s and the wife s income, and also on the total sum of the survivor s pension and the wife s own income. 40% of the old-age pension is provided if the husband s and wife s income were equal. The maximum monthly amount is around 2.3 thousand EUR. Belgium The main conditions of receiving survivors pension are that the deceased husband was insured, and the marriage lasted at least one year. The amount of the pension is 80% of the deceased husband s actual or hypothetical retirement pension. For divorced spouses this amount is lower. The maximum pension for a surviving spouse equals the maximum old-age pension for a single person (22.5 thousand EUR per year in 2010). 13

14 Czech Republic In order to be entitled to survivors pension the deceased husband must have been entitled to old-age pension. The widow receives the survivors pension for one year, after that she receives this payment only if she has reached the retirement age, or if she su ers from invalidity, or cares for a child or parent. Divorced spouses are not entitled to the bene ts. The bene ts consist of two parts: the at-rate basic bene t (84 EUR per month), and 50% of the amount to which the husband would have been entitled to. There is no maximum amount. Denmark The system of the Danish survivors pension has been modi ed several times in the past two decades. If the death of the husband occurred after 2002 then he must have paid at least two years of contributions so that the spouse becomes entitled to the survivors pension (previously it was 10 years). From 1992 an additional condition is that the deceased person was over age 67. Divorced spouses and cohabiting partners are also entitled to the survivors pension, provided that the cohabitation lasted at least two years. Before 1992 the amount of the survivors pension was 50% of the pension of the deceased spouse. From 1992 the surviving spouses receive a lump sum amount. This amount is 50% of the capitalized value of the pension entitlements of the deceased husband. If the deceased has contributed since 2002 then the surviving spouse also receives a lump sum payment of around 6.7 thousand EUR, which is gradually decreasing with the age of the deceased spouse. France Widows receive survivors pension from the general retirement scheme if they are aged at least 55, and if the husband was a bene ciary or met certain conditions. The age limit is gradually decreased. Widows can also receive bene ts from the complementary pension schemes if the husband was a liated, and if the widow is above the de ned age limit (55 and 60 years, depending on the pension scheme). Divorced widows are also entitled. In the general scheme the widows receive 54% of the old-age pension of the deceased husband. In the complementary schemes this ratio is 60%. The maximum amount of survivors pension in the general scheme is around 9 thousand EUR per year. Germany The basic conditions of entitlement are that the deceased person must have had at least 5 years of insurance, and the widow was married to the deceased husband at least for one year. Widows who are aged 45 or above and with reduced capacities to work or with a dependent child are entitled to the "major" widows pension, otherwise they are entitled to the "minor" widows pension. The "major" pension amounts to 60% of the pension to which the husband would have been eligible. This ratio is 25% in case of the "minor" pension. If the husband dies before age 63 then the widows pension is reduced. In addition, earnings of the widow are credited against the survivors pension. There is no statutory maximum amount of the payable pension bene ts. For marriages concluded after 2001 and for couples aged 40 or below on 1 January 2002 the "major" replacement rate is 55%, and the survivors pension is payable for maximum 2 years. Greece A condition of being entitled to survivors pension is that the deceased spouse must have paid contributions at least for 4500 days, or for 1500 days out of which 300 accrued during the 5 years before the date of death. From 1999 the widows are entitled to the survivors pension irrespective of age. The normal amount of the survivors pension is 50% of the pension of the deceased spouse (before 1993 it was 70%). If the widow works or receives a pension then she is entitled only to the half of the normal survivors pension. However, from age 65 the full survivors pension is received. The maximum amount of the bene ts since 1993 is around 1.3 thousand EUR per month. 14

15 Italy Widows are entitled to survivors pension if the husband was already receiving a pension or if the husband had at least 5 years of contribution out of which 3 years were during the last 5 years. If the couple was divorced then the widow might still receive survivors pension, but only by decision of a judge. The normal amount of the bene ts is 60% of the deceased husband s old-age pension. However, this ratio is reduced if the widow s total income exceeds 3 times the minimum pension. There is no statutory maximum pension. If the conditions for the survivors pension concerning the deceased are not met, then the widow can be entitled to a one-o allowance, which is proportional to the amount of contributions paid by the husband. The Netherlands The conditions of entitlement are that the deceased spouse was insured at the time of death, and that the widow was born before 1950 (valid in 2010). At age 65 the survivors pension is replaced by the general old-age pension. The divorced spouse might receive survivors pension if the deceased husband was obliged to pay alimony. The basic monthly amount of the bene t is around 1.1 thousand EUR. However, this amount is income dependent: work related income is deduced, but 50% of the minimum wage plus one third of the income in excess of this amount is disregarded. The survivors bene t may not exceed 70% of the statutory net minimum wage. Poland Widows (including divorced widows who are entitled to alimony) are eligible for survivors bene t above age 50, or in case of total incapability for work. The deceased husband must have received a pension at the time of death or must have paid contributions su cient for old-age pension. The children of the deceased who are aged under 16 are also eligible to survivors bene ts. If there is only one recipient then the amount of the survivors bene t is 85% of the old-age pension of the deceased. If there are two or three recipients then the total amount of bene ts is 90 or 95% of the pension of the deceased, and this amount is divided equally among the recipients. There is no statutory maximum pension level. Spain The deceased spouse must have been retired and receiving old-age bene ts, or must have been an active contributor with at least 500 days of contributing days in the 5 years before the death. If the deceased was neither a pensioner nor an active contributor then the widow might still be entitled to survivors pension if the husband contributed for at least 15 years. Divorced spouses are also entitled to the survivors pension. The amount of the pension is in general 52% of the base pension, and 70% if there are dependent family members. The monthly base pension is the sum of the contribution bases for a continuous period of two years divided by 28. The statutory maximum monthly amount is around 2.5 thousand EUR. Sweden The surviving spouse is entitled to survivors pension if she is aged under 65 and has lived together with the deceased uninterruptedly at least for 5 years or lives with a child aged under 18. A person living together permanently with the deceased is also considered as spouse, provided that she had or was expecting a child at the time of the death. However, divorced spouses are not entitled to the bene ts. The survivors pension is paid for 12 months, and the payment is continued if the widow lives with a dependent child under age 12. The annual amount of the guaranteed pension is 2.13 price base amounts (i.e. around 9 thousand EUR), and the income related pension is 55% of the deceased husband s pension base. There is no statutory maximum amount. Switzerland Widows can receive survivors pension both from the rst and second pillar of the Swiss pension system. The rst pillar is the state pension, and the second is the occupational pension. The 15

16 conditions of entitlement are that the deceased spouse contributed for one year ( rst pillar) or was insured or received old-age pension at the time of death (second pillar). The widow must be at least 45 years old and married for 5 years. Divorced spouses are also entitled under some conditions. The surviving non-married partner is also entitled provided that she has one or more dependent children. The amount of the bene ts in the rst pillar is 80% of the old-age pension corresponding to the determining average income. In the second pillar the bene ts are 60% of the disability pension that the deceased would have received. The maximum amount in the rst pillar is around 1.2 thousand EUR per month. In the second pillar there is no statutory maximum. It is clear from this summary that in all countries the entitlement to survivors pension is conditioned. The typical conditions are related to the insurance status of the deceased spouse, and to the age of the widow. The bene ts are typically de ned as a percentage of the pension bene ts to which the deceased spouse would have been entitled to. The Netherlands is an exception with this respect, where the survivors bene t is basically a at-rate amount. The survivors pension is typically an annuity type bene t, but in the Netherlands and in Sweden the widows cannot receive this type of bene t after reaching retirement age, and in Denmark the bene t is a lump sum payment. In Table 9 I summarize the basic magnitude of the survivors pensions. % of the deceased Other Survivors pension/ Survivors pension/ spouse s pension bene ts GDP (%), 2007 old-age pension (%), 2007 AT 0-60% BE 80% CZ 50% + 84 EUR/month DK 50% of capitalized value ,700 EUR lump sum FR 54% 60% in the complementary schemes DE 25% GR 50% IT 60% NL Basic monthly bene t: 1,100 EUR PL 85% ES 52% SE 55% CH 80% Table 9: Basic magnitude of the survivors bene ts, as described in the text, and expenditures on survivors pension (source: OECD Statistics) 5.2 Relation to the adverse e ects of widowhood There is no clear relationship between the estimated adverse e ects of widowhood on income and wealth, and the generosity of the survivors pension bene ts. The results rather indicate that survivors pension has important role in ghting widows poverty in those countries where widowhood would otherwise have strongly negative e ect on the nancial circumstances. Based on the rst part of Table 5, widowhood has the most detrimental e ect on the objective indicators of economic status in Denmark, Sweden, the Czech Republic, and Poland. These countries have various policies of survivors pension, and in Poland these bene ts are among the most generous ones. The second 16