The Income and Wealth Packages of Older Women in Cross-National Perspective

the income and wealth packages of older women’s (age 65+ years) households and the extent to which low income is paired with low wealth, across a group of six high-income countries. Methods. We use data on income and net worth from the Luxembourg Wealth Study, a new cross-national microdata-base. We deﬁ ne income poverty as having household income less than 50% of the national median and asset poverty as holding ﬁ nancial assets equivalent to less than 6 months of income at the poverty threshold. Results. Older women typically have less income than do members of younger households at the national median, but their wealth holdings are generally much higher than their country’s median wealth holdings. Older women’s households in the United States report the highest net worth across these countries, in part because older American women have comparatively high rates of homeownership. However, American older women are also substantially more likely to be income poor. They also report high levels of asset poverty, as do women across all our comparison countries, with Sweden as a partial exception. Discussion. Further research is needed to identify the most vulnerable subgroups, to integrate analyses of necessary expenditures, and to assess policy implications .

D URING the last 50 years, across the high-income countries, great strides have been made in reducing poverty among older persons. In most rich countries, although older persons still report higher poverty rates than in the total population, elderly poverty has decreased in recent decades nearly everywhere (see http://www.lisproject.org/keyfi gures.htm ). Nevertheless, " older women's " income poverty in particular has not been eradicated, especially in the English-speaking countries, and women's poverty status in old age remains a concern in all rich countries. In fact, due to anticipated demographic shifts, combined with ongoing and expected policy changes, older women's income poverty may rise again in the coming decades ( Smeeding, Estes, & Glasse, 1999 ).
In most rich countries, poverty among younger pensioners (under age 70 years) is no longer a major policy problem, but within this group, older women remain the most vulnerable. Indeed, most elderly poverty is women's poverty, as women typically constitute two-thirds or more of the elderly poor in the rich countries. Previous studies suggest that poverty is especially a problem among women aged 75 years and older who live alone ( Smeeding, 2003 ). One major solution to older women's poverty is private wealth accumulation. However, not all older women are able to save enough to ensure a secure retirement ( Munnell, Webb, & Delorme, 2006 ). Another solution to the problem of elderly poverty may lie in establishing a safety net that helps keep the lowest income and lowest wealth elders out of poverty, through policy interventions that may have little negative impact on the younger and more affl uent elderly -as is accomplished in Canada by means of an income-tested benefi t (the Guaranteed Income Supplement) with a high take-up rate; see Smeeding and Sandstrom (2005) . In order to most effectively design economic security policies for the elderly, it is important to know more about older women's resources, including both income and wealth.
In this article, we extend prior cross-national analyses of older women's economic well-being by assessing income and wealth together and discussing to what extent incorporating wealth in the analysis improves the economic position of poor elderly women. Throughout the article, we invoke the metaphor of the multilegged stool, which is often used to refer to the multiple income streams on which older persons rely. We conceptualize the income stool as having four legs: earnings, capital income (e.g., interest, dividends, rental income, income from savings plans), private transfers (e.g., occupational pensions, alimony), and public transfers (e.g., social insurance, public assistance) and then extend this metaphor to conceptualize a fi fth leg -that is, wealth. We measure wealth as a stock (in what we call wealth packages), although wealth clearly provides an income stream as well.
We are able to consider older women's income and wealth packages together by drawing on the new Luxembourg Wealth Study (LWS) database. The analyses in this article are original partly because the LWS data are new (the database was fi nalized in December 2007) and partly because there are relatively few methodological conventions available for The Income and Wealth Packages of Older Women in Cross-National Perspective measuring and comparing wealth, especially among those with few assets. Although there is an enormous literature on measuring income poverty and income deprivation, there is not yet a substantial counterpart literature on wealth measurement, especially in the cross-national context. Given the importance of assets in supporting consumption and wellbeing in old age, fi lling this lacuna in the research literature on deprivation and old age is our main contribution.
In the next section, we briefl y review the relevant literature on older women's poverty and the growing literature on gender and wealth holdings, highlighting cross-national research. In the following section, we present our empirical analyses, in each case assessing older women in the United States in relation to fi ve comparison countries, Finland, Germany, Italy, Sweden, and the United Kingdom. We chose these six in order to include countries with diverse economic outcomes and widely varying social welfare systems. More specifi cally, these countries span the three major welfare state models that are widely accepted in comparative sociology (see especially Esping-Andersen, 1990 ). The United Kingdom and the United States are exemplars of the so-called liberal model, which is organized, for the most part, to refl ect and preserve consumer and employer markets; most entitlements derive from need based on limited resources. Finland and Sweden represent the Nordic model, generally organized along social democratic lines, with entitlements linked to social rights. And Germany and Italy, like most continental European countries, typically tie social benefi ts to earnings and occupation, and public provisions tend to replicate market-generated distributional outcomes. In the continental countries, social policy is also shaped by the principle of " subsidiarity " that stresses the primacy of the family and community for providing dependent care and other social supports.
Our analyses are mainly descriptive. Drawing on these unique data, we address two core questions: (a) How do older women's income and wealth packages, and their portfolios, vary across countries? (b) To what extent is low income paired with limited wealth, and how does that vary across countries? We close with brief comments on directions for future research.

Background and Literature
Despite major progress in reducing poverty in recent decades, substantial pockets of poverty remain among the elderly, especially among elderly women living alone. The relatively precarious economic position of the elderly in the United States as measured by their incomes is even more evident when we look at cross-national comparative results. Several researchers have used the Luxembourg Income Study (LIS) data to analyze the prevalence and causes of poverty among elderly women ( Doring, Hauser, Rolf, & Tibitanzl, 1994 ;Hutton & Whiteford, 1992 ;Reno, 2007 ;Siegenthaler, 1996 ;Smeeding, 1991Smeeding, , 2003Smeeding & Saunders, 1998 ;Smeeding, Torrey, & Rainwater, 1993 ;Stapf, 1994 ;Wu, 2005 ).
In one of the fi rst studies of elderly women's poverty, Smeeding (1991) found that across seven countries during the mid-1980s, elderly persons in female-headed households were poorer than those in male-headed households, in nearly every age group (55 -59, 60 -64, 65 -74, and 75+ years). Elderly women were especially at risk for poverty in the United States, where 25% or more of elderly persons in female-headed households were poor. Smeeding, Torrey, and Rainwater (1993) , in an eight-country study, further underscored the extreme outcomes seen in the United States. In a more recent LIS study, Smeeding and Sandstrom (2005) compared poverty rates in the United States with those in the United Kingdom, Canada, Germany, Italy, Finland, and Sweden. Their results indicate that American older women have the highest poverty rates among these countries -with poverty defi ned at both 40% and 50% of the national median -in each group they studied: women aged 65 years and over, women aged 65 years and over living alone, and women aged 75 years and over living alone. They fi nd, in particular, that older women's poverty outcomes are markedly better in Canada and in two Nordic countries, Finland and Sweden. Finally, Brady and Kall (2007) explore the feminization of poverty in comparative perspective, reaffi rming the salience of gender as a determinant of old age poverty.
In almost all cross-national research on older women's wellbeing, the main indicator is income. The literature on consumption across countries is more limited (see Sierminska & Garner, 2005 ). Although recent studies suggest that, in the United States, consumption among older women is higher than income and more equally distributed, owing mainly to the fl ow of imputed rent on owned homes, we have no such estimates for other countries on a comparable basis ( Johnson, Smeeding, & Torrey, 2005 ). We conclude that almost all of the researches on older women's deprivation to date has concentrated on income alone or on consumption (much of which may in fact come from sources other than income).
Another relevant strand in the literature draws on panel data to assess economic trajectories and transitions during women's older years, although not necessarily with a focus on poverty. For example, drawing on the Cross-National Equivalent File, Burkhauser, Giles, Lillard, and Schwarze (2005) compare the economic well-being of widows in the United States with those in the United Kingdom, Canada, and Germany. They conclude that, despite diverse social welfare systems, the change in the average woman's economic well-being following widowhood is remarkably similar across these countries.

Women and Wealth
The comparative literature on women and wealth expanded markedly with the April 2006 publication of a Downloaded from https://academic.oup.com/psychsocgerontology/article-abstract/64B/3/402/576761 by guest on 26 July 2018 special issue of Feminist Economics on " Women and Wealth " ( Deere & Doss, 2006 ). Although few of the articles focused specifi cally on older women, several are relevant here -as older women's wealth clearly accumulates throughout their lives. In our review of the gender and wealth literature, we draw heavily on the introductory essay in this issue ( Deere & Doss , 2006) and on the articles that focus on gender and wealth in high-income countries (e.g., Schmidt & Sevak, 2006 ;Yamokoski & Keister, 2006 ). We also draw on another paper produced in association with this special issue but not included in it ( Sedo & Kossoudji, 2004 ).
In their literature review on women and wealth, Deere and Doss (2006) conclude that " although extensive literature exists on women's incomes and the gender wage gap, relatively little work has been done on the gender wealth or asset gap " (p.1). They identify three key reasons for the dearth of research on women and wealth: fi rst, the limited availability of wealth data (relative to income data); second, the near absence of wealth data at the individual level (which forces researchers to study households); and third, conceptual diffi culties encountered in comparing property across household types (especially given the complexity and variation in laws regarding marital property). Schmidt and Sevak (2006) used the Panel Study of Income Dynamics (PSID) to study gender and asset accumulation in the United States. They fi nd evidence of large differences in net wealth between single female-headed households and married couples -differences that exist throughout the wealth distribution. Although some of this gap is explained by differences in observed characteristics (including age, education, family earnings, and portfolio allocation), they conclude that a substantial portion remains unexplained. The wealth holdings of single females are also signifi cantly lower than the wealth holdings of single males. Results from a subsample of young households (with heads aged 25 -39 years) provide no evidence of wealth gaps by gender and family type. Schmidt and Sevak interpret these results to mean that these gender gaps emerge later in life. Yamokoski and Keister (2006) studied the wealth of single females, also in the United States -using the National Longitudinal Survey of Youth. They fi nd that both single mothers and single fathers are disadvantaged in comparison with adults without children and that the greatest gap in wealth accumulation exists between single mothers and single female households without children. They also fi nd that single mothers suffer the most severe economic penalties in household wealth accumulation. Conley and Ryvicker (2005) , in their study of female headship, also fi nd that gender exerts a downward effect on wealth accumulation in the United States. Using the PSID, they fi nd that differential savings rates between women and men account for the gender gap in net worth. Conley and Ryvicker argue that female heads ' savings rates are depressed by fi nancial constraints such as the need to purchase child care or other expenses that are unique to, or associated with, women who head families alone.
A growing body of work focuses on gender and homeownership. In their review of this literature, Sedo and Kossoudji (2004) conclude that homeownership is the main form of middle-class wealth accumulation in most rich countries. Still, these authors note that most studies of homeownership typically ignore gender. Some articles simply omit women from the discussion ( Quercia, McCarthy, & Wachter, 2003 ), whereas others skirt the question by analyzing homeownership patterns only for married couples ( Gyourko & Linneman, 1996 ). Studies of homeownership often include gender through a variable that captures female headship, gender, or marital status as a control variable but not as a point of discussion ( Bostic, Calem, & Wachter, 2004 ;Ioannides & Rosenthal, 1994 ). Sedo and Kossoudji also note that studies of gender-related differences in homeownership from countries other than the United States are very limited. Indeed, they use the Survey of Income and Program Participation to study gender and homeownership in the United States, with a focus on gender gaps. They fi nd that gender gaps are much more pronounced for the probability of homeownership than for home value or home equity. Once households have entered the housing market, differences across gender (and also race and family type) are substantially smaller and sometimes favor those households usually considered to be disadvantaged.
Finally, the use of LIS and LWS microdata for research on older persons ' wealth is just beginning to take hold. Using LIS, Chiuri and Jappelli (2000) and Jappelli and Chiuri (2006) explore ownership status, housing tenure, and, to a more limited extent, the value to the aged of owned homes; they fi nd a mixed picture across countries. Also using LIS, Gornick and Schlesinger (2007) fi nd that, in several countries, older renters are more likely to remain in paid employment than are homeowners, net of other factors. More recently, Smeeding (2006) , Smeeding (2006a , 2006b ), and Gornick, Munzi, Sierminska, and Smeeding (2006) began to examine the incomes and assets of the aged, including both women and men, in comparative perspective. This article builds upon these other LWS studies.

Data
The LWS, a microdatabase, is a newly developed project within the LIS. The LWS database contains harmonized cross-sectional wealth and income data and a range of other demographic and economic characteristics of the households for 10 industrialized countries. Some of these datasets also contain data on expenditures. The primary goal of the LWS project is to assemble and organize existing microdata on household wealth and income into a coherent database, in order to provide a sounder basis for comparative studies on household net worth, portfolio composition, and wealth distributions. The database currently contains either one or two waves of cross-sectional data for each country from the years 1994 -2004. It is the fi rst cross-national comparable wealth database. (For more details, see Sierminska, Brandolini, & Smeeding, 2006a, 2006b , and consult the LIS Web site: http://www.lisproject.org/lwstechdoc.htm .) In this article, we draw on datasets from six countries, collected between 1999 and 2002. These countries include two Anglophone countries, the United States and the United Kingdom; two continental European countries, Italy and Germany; and two Nordic countries, Finland and Sweden.

Income and Wealth Packages: The Aggregate Indicators and Their Components
Our main income variable is household disposable income (DPI), which is defi ned as the sum of total income from earnings (wages and salaries and income from selfemployment activities), capital income (interest and dividends, rental income, income from savings plans, royalties, and other property incomes), private transfers (occupational and other pensions, alimony, regular transfers), and public transfers. Public transfers include public social insurance (public pensions and some universal benefi ts and family allowances) and public social assistance (means-tested cash and near-cash public income transfers). DPI is also net of income taxes and employees ' social security contributions. (Note that DPI includes annuities from life insurance and private pensions, but it does not include imputed rents from owner-occupied housing nor irregular incomes, such as lump sums and capital gains and losses.) Net worth (wealth) consists of fi nancial assets and nonfinancial assets, net of total debt. Financial assets include deposit accounts, stocks, bonds, and mutual funds. Nonfi nancial assets include (owned) principal residence and investment real estate. Finally, total debt refers to all outstanding loans, both home secured and nonhome secured.

Analyzing the Economic Well-being of Older Women: The Unit of Analysis
Analyzing economic well-being among women, or differentials between women and men, is always a challenge because many sources of income and wealth cannot be disaggregated within households. Although wages and pen-sions are usually received by individuals, many public income transfers as well as key wealth components (especially housing) cannot easily be allocated within households to the person level.
In response to the diffi culty, and often the impossibility, of separating income and assets within households, scholars of women's economic well-being (or gender gaps) often conduct their analyses at the household level and compare household types. That is the approach we take in this study. To uphold our central focus on older women, we analyze two types of households. The fi rst type is all households that include older women (aged 65 years and older) as either the head or the spouse, which may or may not contain additional persons. The second type of household -a subset of the fi rst -is composed of one older woman (aged 65 years or older) who lives alone. So, when we refer to the income/ wealth status of older women, we always mean the income/ wealth status of these two types of households: either all older women or older women living alone. The outcomes for these households, of course, pertain to all the members in the household, including nonelderly members. For the population of older women who live alone, person-level and household-level outcomes are the same.
This scheme does not explicitly capture one group of older women -those who are part of extended households and who are neither the head nor the spouse of the head. (In the LWS data, we cannot identify the age or sex of household members who are neither the head of household nor the spouse of the head.) In Appendix Table 1 , we report that 4% -5% of older persons (men and women) live in such households, except in Italy, where the share rises to 11%. Virtually, all these older persons live in households that are not income poor.
Thus, our household-based analyses -like others in this tradition -reveal little about the individual fi nancial wellbeing of women who do not live alone, relative to their own partners or others with whom they share their homes. Although multiple literatures on gender and economics emphasize the importance of understanding intrahousehold inequality ( Sierminska, Frick, & Grabka, 2008 ), we cannot effectively study intrahousehold allocations of income, and especially wealth, with these data at this time.

Equivalizing Income and Wealth and Other Data Adjustments
Following a common practice in income research, we " equivalize " income data by adjusting each household's income to account for household size ( S ) by using an equivalence scale ( S e where e is the equivalence elasticity) -in this case, by dividing the unadjusted income ( I ) by the square root of household size ( e = .5). The use of the square root -meaning an equivalence elasticity of .5 -is the middle point between two theoretical possibilities: no economies of scale and perfect economies of scale. There is much less consensus about how to equivalize wealth ( Sierminska, Brandolini, & Smeeding, 2006b ); we used the same method for adjusting for household size as we used for income. For the most part, among elders, these produce almost the same results as do nonequivalized fi gures ( Sierminska & Smeeding, 2005 ).
To minimize the infl uence of outliers, incomes are bottom coded at 1% of the mean equivalized DPI and top coded at 10 times the unequivalized median. The wealth variables are not bottom coded or top coded; thus, the wealth indicators (net worth in particular) can contain negative and zero values. Because the top ends of these wealth distributions may vary across countries, depending on the quality of the wealth survey and the sampling practices among the richest portions of the population, our analysis relies mainly on medians, not means. The few observations with missing or zero DPI or missing net worth are dropped from the sample. Finally, when we report actual currency amounts, all amounts are expressed as U.S. dollars, adjusted by purchasing power parities (PPPs), using the 2002 Organisation for Economic Co-operation and Development PPP exchange rates. Amounts referring to years prior to 2002 were infl ated to 2002 U.S. dollars using country-specifi c infl ation factors.

Poverty Measurement: Income and Wealth
For purposes of international comparisons, poverty is usually captured in relative terms. (For a discussion of the merits of using relative vs. absolute poverty in crossnational research, see Kenworthy, 2004 ;Smeeding, Rainwater, & Burtless, 2001 .) When analyzing income, most cross-national studies defi ne the poverty threshold as 50% of national median (equivalized) income. In this study, we follow that practice. Note that these income poverty thresholds are higher than the current American thresholds, where the ratio of the offi cial U.S. poverty line (which captures absolute poverty) to median American household cash income was only about 30% -35% in the early 2000 and 2002 ( Smeeding, 2006 ), though the U.S. poverty line corresponded to 50% of the U.S. median when the thresholds were fi rst instituted in 1963.
Although there is a considerable agreement on the appropriate measurement of income poverty in cross-national context, there is no such consensus on wealth povertyeither absolute or relative -because little work exists on this subject in any country and even less in a cross-national context. For this article, we have chosen one particular definition of relative wealth poverty: we classify households as wealth (asset) poor if they hold fi nancial (liquid) assets of less than 25% of median DPI (household DPI), as defi ned above. Our construction of this measure was inspired in part by the work of Haveman and Wolff (2004) , who defi ned " a household with insuffi cient assets to enable it to meet basic needs for a period of time (three months) to be asset poor. " In our analysis, we restrict ourselves to liquid assets, as those are the assets that would be accessible to older persons in times of emergency. We defi ne a household as being " asset poor " if its fi nancial asset holdings, such as deposit accounts, stocks, bonds, and mutual funds, are equivalent to less than six months of income at the poverty threshold of 50% of median income.

Older Women's Income and Wealth Holding Levels
We begin our analysis of women's economic well-being by considering both income and wealth holdings at the median. Using all households within a country as the base, we assess the economic status of households with older women who are heads or spouses, as well as the subset of households that contain only a single older woman who lives alone. To simplify, we refer to these populations as (a) " all older women " or " older women overall " and (b) " single older women " ; the latter group is a subset of the former group.
Median (equivalized) DPI in our two groups of older women's households is reported in Figure 1 along with the median (equivalized) income of all households. The national median household income varies substantially across these six countries, ranging from under $16,000 in Italy to over $21,000 in the United States (in 2002 U.S. dollars). However, at the median, older women overall have less income than typical households. In Finland, Sweden, and the United Kingdom, they have 76% -78% of the income of all households; 84% -86% in Italy and Germany; and 90% in the United States. Single older women fare even more poorly, typically reporting 60% -63% of overall median income.
The net worth (or wealth) picture is starkly different and much more varied (see Figure 2 ). Median net worth of all households also varies substantially across these countries, although the country rankings with respect to wealth are different from those vis-à-vis income. The highest net worth (among all households) is reported in Italy (nearly $78,000) and the lowest in Sweden (about $17,000); the United States falls in the middle of the range among these countries (about $23,000). Although older women's income generally lags relative to all households within their countries, their wealth holdings at the median are, in most cases, well above their country's median. It is not surprising that older households have more assets than the median household, as assets often continue to accumulate up to and beyond retirement. Indeed, this fi nding underlies the main rationale for this article: assets are of crucial importance to older women; yet, little is known about how asset levels vary both across and within countries.
Older women's households (those in which older women are the head or spouse of the head) in the United States report the highest levels of net worth (about $98,000) across these six countries and stand out much more, with respect to their relative position within their own country's distribution. American older women's households report over four times as much net worth as the median American household. Their British, German, and Swedish counterparts report about two to three times the net worth of their country's median household, whereas Finnish older women report net worth of about one-and-a-half times the overall median. Italian older women, in contrast, report net worth equivalent to the median Italian household. The results for single older women are similar but even more varied across countries. Older single women in Italy and especially in Germany have much less (relative) net worth, lagging their nation's median wealth holdings substantially.

Older Women's Income and Wealth Packages
Next, we move beyond income and asset levels to assess the components of older women's income and wealth pack-ages across countries. Table 1 (panels A and B) report older women's disaggregated income packages. One important fi nding is the stark contrast between the income package of older women in the United States and those of their counterparts in other countries. Among older women, the share of income coming from earnings is greatest in the United States: fully 32% for older women overall and 15% for singles (percentages that are statistically signifi cantly different from the corresponding outcomes in the other fi ve countries). Earnings in the other countries constitute far less than that, with an especially marked difference among single women, where earnings are virtually negligible in all the comparison countries. American older women's greater reliance on earnings is consistent with the comparatively high rates of employment among older Americans. As we reported in Smeeding, Gornick, Sierminska, and Leach (2008) , American older women are much more likely to be working for pay than are women in these comparison countries. Fully 19% of American women aged 65 -69 years are  Downloaded from https://academic.oup.com/psychsocgerontology/article-abstract/64B/3/402/576761 by guest on 26 July 2018 employed compared with 8% in Germany, Sweden, and the United Kingdom and 2% in Finland and Italy. Likewise, 12% of American women aged 70 -74 years are employed compared with 4% -5% in Germany and the United Kingdom and 2% or fewer in the other countries. Early and normal pension ages vary across countries, explaining some of these differences in older women's employment rates. Differential employment rates later in life are also driven by differing rates earlier in life; Italian women, for example, report substantially lower employment rates than American women at all ages.
In sharp contrast, the share of income that older women in the United States receive from public transfers (social insurance and public assistance) is dramatically less than in any of the comparison countries. (A table reporting the statistical signifi cance of the cross-country differences, reporting signifi cance levels for each pair of outcomes, is available upon request from the authors. One clear fi nding is that nearly every U.S. outcome is statistically signifi cantly different from the corresponding outcome across all the comparison countries.) Table 1 underscores that the four-legged income stool -comprising earnings, capital income, private transfers, and public transfers -operates differently for older women across these six countries. Whereas the " earnings leg " is especially crucial in the United States, the " public transfer leg " plays a much larger role in the other countries, constituting about 50% -80% of income for older women overall and from nearly 70% to over 90% for single older women.
Older women's wealth packages for the median household are presented in Table 1 (panels C and D). Here, wealth holdings are reported as comprising fi nancial assets, principal residence, and investment real estate. The median household is defi ned as having equivalized total assets between 40% and 60% of the distribution of all households. The most salient fi nding is that wealth packages vary greatly across countries, and three pairs of relatively similar countries emerge. Older women overall hold a relatively small share of their wealth as fi nancial assets (i.e., deposit accounts, stocks, bonds, and mutual funds) -10% to 12% -in Finland and Italy; a moderate share (25% -26%) in the United States and the United Kingdom; and a substantially larger share (44% -54%) in Germany and Sweden (although the shares for Germany and Sweden are signifi cantly different from each other). The cross-national results among single women are quite similar, although the share of wealth held as fi nancial assets is systematically higher than among older women overall. (As reported next, in Table 2 panel A, this results from lower homeownership among older single women than among older women overall.) The common fi nding in these countries is that, with the exception of Sweden and (among single older women) Germany, the primary residence is the largest component of the wealth portfolio.

The Role of Homeownership
As indicated in the fi rst two tables, American older women report comparatively favorable net worth positions. Part of the explanation is their comparatively high rates of homeownership, a form of asset holding that is clearly valuable late in life ( Fisher, Johnson, Marchand, Smeeding, & Torrey, 2007 ) if not readily drawn upon. Although American homeownership rates overall -about 71% -are fairly high, they are not especially high compared with other countries (see Table 2 , panel A). However, in the United States, homeownership is comparatively frequent among older women; fully 82% of American older women's households own their own homes compared with 51% -78% in the other countriesalthough American older women homeowners report the lowest percentage owning their homes outright. Homeownership patterns explain a portion of the single older women's results as well. For example, German single older women report the least median net worth (about $9,500) and the lowest rate of homeownership (33%). In addition, lower homeownership rates for single women result in fi nancial assets being a larger share of women's wealth portfolios (see Table 1 panel D). (We will return to the results presented in Table 2 panel B later, in the discussion titled 'Income and Assets Among Poor Older Women'.)

Income and Asset Poverty Among Older Women
We next look further down the economic distribution to assess the interplay between older women's income poverty and their asset holdings. Policy concerns related to older adults are, not surprisingly, concentrated on adequacy and security in retirement, and assets, in addition to income, constitute an important part of that security. In Figure 3 , we report income poverty and asset poverty among households with older women as head or spouse.
One of the most striking fi ndings in Figure 3 concerns the United States, where older women report very high rates of income poverty. Fully 24% of older women's households (16% plus 8%) in the United States have DPI below the poverty threshold, meaning that American older women are substantially poorer in terms of income -relative to their home country -than are their counterparts in the United Kingdom (16%), in Italy and Germany (11% -12%), and especially in Finland and Sweden, where only 7% are income poor. This fi nding is consistent with the earlier LIS literature cited above.
What about asset poverty? Figure 3 also indicates that an even larger share of American older women are asset poor. Over a third (36%) lack fi nancial assets equivalent to half the income poverty threshold; that is, they do not hold enough fi nancial assets to survive for six months at the poverty level. Yet, in clear contrast to the income poverty results, the prevalence of asset poverty in the United States is not especially high in cross-national terms. Older women report somewhat higher asset poverty rates in the United Kingdom and in Italy (41% -45%), and the rate is even higher in Germany (46%) and, even more remarkable, in Finland (54%). Older women in Sweden are considerably less likely to be asset poor than in any of the other countries In all these countries, of course, there is an overlap between the income poor and the asset poor. When the two types of poverty are considered together, the share of older women's households that are income poor, asset poor, or both is fairly similar in the United States, the United Kingdom, Germany, Italy, and Finland (about 43% -56%). In Sweden, fewer older women -although still one-third -report one or both types of poverty.
The subset of older women who are single and live alone report broadly similar, but more extreme, outcomes -underscoring the economic vulnerability of this group overall. As we report in Figure 4 , in all six countries, single older women's households are more likely to be income poor -with poverty rates about 5 -15 percentage points higher -than are all households containing an older woman as head or spouse. Furthermore, again, across these countries, we fi nd the highest rate of income poverty in the United States, where fully 39% of single older women have incomes that place them below the poverty line. And, again, in all six countries, the prevalence of asset poverty is greater than that of income poverty and also substantially greater among single older women than among older women's households overall. At the same time, even with an asset poverty rate of 45%, American single older women are not especially asset poor vis-à-vis their counterparts across this group of countries; single older women report even higher rates of asset poverty in the other study countries, with the exception of Sweden. Figure 3. Income and asset poverty (assets less than 25% median DPI). Among households (HHs) with older women as head/spouse (older = age 65+ years). Figure 4. Income and asset poverty (assets less than 25% median DPI). Among households (HHs) of single older women (older = age 65+ years).

Income and Assets Among Poor Older Women
Finally, we turn our attention to the income and asset holdings of poor older women, where poverty is defi ned in relation to those who are both income and asset poor (the bottom group in each column in Figure 3 ). As we have reported, homeownership is an important component of wealth among the poor and is a crucial aspect of their economic well-being. Table 2 (panel B) indicates that homeownership among the poor varies dramatically across countries. Homeownership for all income and asset poor households ranges from only 23% in Sweden to a remarkably high 53% in Italy; about one-third (34%) of American income and asset poor households own their homes. Homeownership among poor older women's households in the United States, the United Kingdom, and Italy is quite similar: about 60% -64%. In Sweden and Germany, homeownership is substantially less common (around 43%). The great majority of poor older women householders with owned homes own them outright, although this share is lowest in the United States at 69%.

Summary and Policy Implications
This article has provided the fi rst in-depth descriptive analysis of the joint asset and income position of older American women in cross-national perspective. Although the LIS datasets have long enabled cross-national research on older women's income poverty, there has been virtually no comparative research on older women's wealth holdings. Given the relatively high levels of asset holdings among the elderly, this has been a notable omission. The new LWS database allowed us to begin to investigate asset holdings, in conjunction with income, among older women in six highincome countries -the United States, the United Kingdom, Germany, Italy, Finland, and Sweden.

Summary of Findings and Research Implications
In all six countries, including the United States, older women's households overall typically have less income (adjusted for household size) than do households at the national median. When we disaggregate older women's income packages, we fi nd that American women stand out due to the exceptionally large contribution that comes from the earnings leg and the comparatively small share that comes from the " public income transfers leg. " At the same time, although older women's income lags median national income in all these countries, their wealth holdings are typically much higher than their country's median wealth holdings. Older women's households in the United States report the highest level of median net worth across these six countries. Some of the explanation, cross-nationally, is that older American women have comparatively high rates of homeownership.
The U.S. case has always been most exceptional when we consider older women's income poverty. American older women are substantially more likely to be income poor (see, e.g., Smeeding & Sandstrom, 2005 ). When we consider asset poverty, we see a different picture. Although American older women report high levels of asset poverty, it is not especially high in cross-national context. A partial exception is the Swedish case, where the asset poverty rate is substantially lower than in the other countries, although it is still nearly 30%.
Much remains to be investigated. Future research using the LWS data ought to assess older persons ' well-being more generally -even with the limitations on person-level data as noted by Deere and Doss (2006) . Given that family composition clearly matters with regard to asset accumulation (see, e.g., Conley & Ryvicker, 2005 ;Yamokoski & Keister, 2006 ), households could be further disaggregated according to past marital and parenting status -as well as age, educational level, ethnicity, and immigration status of the household head and/or spouse. Much more could be learned about the interplay between older persons ' employment status and their education, total income, and asset levels, both within and across countries.
It is also crucial that we extend this cross-national picture of income and wealth outcomes to take into account variation in necessary expenditures. The vast majority of older Americans are enrolled in the Medicare program, a public health insurance program. However, Medicare benefi ciaries are liable for substantial out-of-pocket costs, primarily for premiums for medical insurance, as well as for deductibles and an array of coinsurance payments for inpatient and medical services and prescription drugs. Overall, Medicare covers only about half of all elders ' medical expenses, and older Americans, especially those in poor health, often face as much as $4,000 -$7,000 each year in out-of-pocket costs ( Biles, Nicholas, & Guterman, 2006 ). Although data limitations prohibit an accurate comparison of older women's out-of-pocket spending on health care across our study countries at this time, we do know that American households, throughout the age spectrum, pay substantially more out-of-pocket than do their counterparts in these comparison countries ( Smeeding, 2003 ). The average U.S. household now pays more than $800 per year out-of-pocket on health care, which is 1.7 times the amount reported in Italy, 2.5 times the amount in Germany, and 3.5 times the amount spent per household in the United Kingdom ( Organisation for Economic Co-operation and Development, 2006 ). Clearly, American older women's alarmingly high rates of income poverty and their even higher rates of asset poverty (although not high in cross-national terms) must be considered in the context of the large burden they often assume vis-à-vis their health care.
Finally, our results highlight the need for in-depth comparative analyses of policies and institutions that shape older persons ' resources, including not only their income streams but their wealth holdings as well. The portrait that we have sketched here suggests important policy implications for the United States. First, American older women's exceptionally high-income poverty rates highlight the potential benefi ts of strengthening the public income transfers leg of the income stool, including both the social insurance and the public assistance components. Although private income sourcesearnings and to some extent fi nancial assets -are more prevalent in the United States, especially among middleincome elders, and although this self-reliance may be commendable, it is also risky and does little to ensure the economic security of those lower down in the wealth distribution.
Although we recognize the fi scal concerns associated with pay-as-you-go public retirement programs, this public leg is so far more reliable and more effective at protecting elders from the economic uncertainties that characterize all market-based income sources. Our fi ndings also underscore the need to strengthen the public assistance safety net that is so crucial for many older women in the United States. It is well known that low rates of participation are found in the main U.S. income maintenance program aimed at the poor elderly, the Supplemental Security Income program, in part due to strict fi nancial asset tests. As reported in our own earlier work ( Smeeding, Gornick, Sierminska, & Leach, 2008 ), several of our comparison countries have social assistance rules that place no limits on low-income elders ' liquid assets. Governments in other rich countries provide more effective public income safety nets for the elderly, with adequate and well-maintained minimum benefi ts to ameliorate income and asset vulnerability. Indeed, the country in our study with the strongest public income leg, Sweden, seems to perform better both in fi ghting income poverty and in shoring up private assets than does the institutional arrangement now operating in the United States.
Although American older women ' s rates of asset poverty are not exceptionally high in cross-national perspective, they are worrisome nonetheless -especially in conjunction with the prevalence of income poverty. As we reported, over onethird of American older women do not possess fi nancial assets equivalent to even six months of income at the poverty line. Many income-poor older women do own their homes, but the value of those homes is limited and may be diffi cult to access in times of hardship or during a housing slump, and, of course, homeowning itself is not costless. This suggests that policymakers ought to identify better and more reliable methods to strengthen assets among older women, beyond reverse-annuity mortgages or borrowing against the value of their own homes, as the poor elderly will receive little benefi t from such programs. In the longer run, alleviating economic hardship among the elderly could be achieved through securing higher levels of individual savings and more extensive occupational pensions, as long as those were accompanied by safe portfolio options backed by public guarantees. In the short term, however, the surest way to prevent economic hardship among older women, and men, is to provide a fl oor under older households' incomes through government transfers. As our results suggest, self-protection through wealth accumulation alone is not suffi cient.