Patterns in Income Source Expectations for Retirement Among Preretirees

Abstract

Financial planning for retirement is a lifelong process constrained by financial literacy, resources, and competing demands for resources across the life course. Further, social structure shapes the availability of options for funding retirement. The social and economic frameworks surrounding retirement planning are changing, and policy makers and researchers question whether retirement expectations have adapted. To explore this question, this research used k-means cluster analysis of a 2010 survey data set to identify natural groupings of Michigan adult preretirees based on their expectations of income sources for retirement. The cluster analysis identified six distinct groups that hold very different expectations. Most had expectations that are not consistent with projected changes in social structure and resource availability and those that did were more likely to occupy traditionally privileged statuses including being White, male, and married.

Keywords

retirement planning retirement expectations cluster analysis structural inequalities in retirement

Introduction

Financial planning for retirement is a lifelong process constrained by financial literacy, resources, and competing demands for resources across the life course. The life cycle theory of saving explains planning for retirement as an individual’s attempt to maximize utility from lifetime resources by allocating them optimally between current and future consumption (Modigliani, 1986). This theory underscores the importance of an individual’s reasonable knowledge of their expectations about future earning streams during preretirement years and expectations about available retirement funding sources postretirement.

Social structure is critical as well because it frames and shapes the availability of options for funding retirement and what one can reasonably expect over the life cycle. When social structure provides differential access to a society’s options and resources and differences in people’s ability to navigate the options depending on an individual’s social location, there exists structural inequality in that social system (Sorensen, 1996). Also, when social structure changes over time such that future earned income probabilities and retirement funding options change, individuals must be able to react to those changes by adjusting behaviors based on reasonable expectations. Structural change may adjust the nature of inequalities or it may continue to advantage already-privileged groups. Structural inequality theorists have identified that the degree of inequality in a system is dependent on the level of concentration of access to favorable resources, the rigidity of the system, and the ascriptive nature of disadvantaged social locations (Grusky, 2000). Within the United States, inequality in earnings and long-term planning opportunities has existed across race, class, and gender with White, upper-class men occupying the most advantaged positions. Also, in adulthood, social systems have favored the state of lifelong legal marriage between men and women.

This research will identify natural groupings of preretirement individuals based on their expectations of income sources for retirement. A first step to help people prepare for retirement involves determining groups for whom different messages, types of assistance, or financial know-how are needed to improve decision making over the life cycle in the face of structural inequalities and changing social structure. Identifying groupings of individuals based on similarities is a fundamental tool for social research and learning (Jain, 2009). Therefore, through k-means cluster analysis we explore patterns among the expectations that Americans hold for funding their retirements.

The collective nature of planning for retirement for most people adds further complexity. Retirement plans and expectations are formed within a family context, and families take on different forms within and across generations. To account for shared efforts and different levels of dependency, our analysis uses reports of expectations of different self-accrued income sources and also estimates of individuals’ expected reliance on others. We also explore the demographic, behavioral, and social characteristics associated with different reliance groups.

Background

Social and economic frameworks surrounding retirement planning have been changing markedly over recent years. Policy makers and researchers question whether retirement expectations, especially for those most vulnerable in society, have adapted to changes and whether changes will exacerbate the problem of income inequality in retirement as new cohorts reach retirement age.

Momentous social changes have occurred in many facets of the U.S. society during the past 10 years. High unemployment and underemployment have characterized the labor market, and foreclosures and bankruptcies have characterized the housing market. In the workforce, most employers now provide defined contribution (DC) retirement plans rather than defined benefit (DB) plans. The shift from DB to DC plans is effectively a shift of investment risk from employers to employees, thus increasing individual vulnerability (Hacker, 2006). Additionally, globalization has led to a redefined and ever-shifting labor market distinguished by more contingent and contract work, the work least likely to offer retirement plan assistance.

Change has also occurred in the landscape of family life. Families composed of married couples and their children are no longer the most prevalent family structure. More women are delaying childbirth or having children out of wedlock and cohabitation, same-sex couples and single households have significantly increased (e.g., Kennedy & Bumpass, 2008, Fincham & Beach 2010). Single-headed households are disproportionately led by females, and households with married or cohabiting couples are likely to have multiple incomes with women contributing to household earnings (Malone, Stewart, Wilson & Korsching, 2010). All these factors should play a role in individuals’ expectations for retirement income, yet we know little about what that impact is.

Demographic shifts in the age and racial makeup of the U.S. population add to social structural changes. The “baby boom” cohort of about 78 million people born between 1946 and 1964 is now entering retirement. As a result, the retired, elderly population of the United States will grow exponentially in the next two decades. Also, more Americans are living to the age of retirement and can expect an extended period of retirement. The number of additional years individuals can expect to live if they reach age 65 has grown by about 30% since 1960 to 82.2 years for men and 85 years for women (National Center for Health Statistics, 2007). Projections show aging population cohorts becoming increasingly diverse, to the point that, in addition to the doubling of the elderly White population by 2050, the elderly African American population will triple, and the elderly Latino population will increase 11-fold (Kaneda & Adams 2009; Wheeler & Grunta, 2009).

As the retired population has grown, the overall U.S. population fertility rate has steadily declined. Thus, the “old age dependency ratio,” or the number of retired people aged 65 and older divided by the working population aged 20–64, is increasing. In 2010, the dependency ratio was 0.22 and is projected to increase to 0.37 by 2050 (Vincent & Velkoff, 2010). Owing to the growing dependency ratio and more years spent in retirement, the long-term viability of social security in the United States has come into question.

Due to pervasive change, different age cohorts will enter and exit the labor market during different social structural circumstances. Thus, their retirement expectations should be functions of different sets of opportunities and circumstances. Many individuals will need to constantly modify expectations for retirement in the face of new realities in the marketplace and at home.

Past Research

Using the U.S. Census Bureau’s 2009 Annual Social and Economic Supplement, Hayes, Hartmann & Lee (2010) determined the overall average income by source for men and women over 64. The analysis used the following categories: Social Security, Pensions, Earnings, Assets, and Other sources. There were marked differences in the average median incomes (US$20,593 vs. US$38,350) and between the relative source combinations for women compared with men. The gender numbers reveal important vulnerabilities for women. Men have nearly twice the median income as women after age 65 in large part due to long-term gender inequalities in caregiving and labor force participation and earnings. Studies show women access men’s higher level of financial resources, but marital status and life expectancy affect that access (Wong & Hardy, 2009). Divorce leaves women particularly vulnerable when it separates them from resources they previously expected to be available and that factored into their life cycle savings prior to marital dissolution. More than half of women reach age 65 unmarried, and women, on average, outlive men (Hartmann & English, 2009). Thus, our analysis will consider a category for “resources provided by others.” This acknowledges and assesses interdependency.

We are not the first to look for groupings that inform retirement funding behavior. However, past research studies have relied on different sets of variables to generate their schemas. For example, Atchley (1975) defined retirement preparation in terms of three continuous phases, the first two of which, “Remote” and “Near”, address preparation (Atchley, 1975). The phases are described such that preparation is low when retirement is far in the future but becomes more active as retirement nears and one begins to imagine it. This grouping omits consideration of social statuses beyond age and time to retirement but shows potential risks when life cycle savings are not fully informed or aware.

A more recent schema based on citizens of the United Kingdom groups people as nonsavers, undersavers, and adequate savers (Association of British Insurers, 2003). This typology is unidimensional and makes divisions only at the behavioral level. Gough and Sozou (2005) developed a multidimensional typology also using a sample of English who made inquiries about savings and retirement products. To determine groups by propensity to be in a retirement plan, the researchers identified six groups based on a combination of economic, demographic, behavioral, and attitudinal traits (Gough & Sozou, 2005). The factors used in that classification were age, education, gender, income, company provision of retirement plan, routine saving activity, routine carrying of debt, concern with retirement plan when changing jobs, concern over survivor’s benefits, and desire to retire early. Attitudes and behaviors were relevant in discriminating the individuals into groups as were demographics. The six clusters offer ideas to address attitudes among different demographic groups to motivate usage of retirement programs. The study supports the inclusion of thought processes to understand approaches to retirement. However, a sample of only individuals who actively sought information on individual retirement products systematically eliminates groups whose social locations reduce the likelihood of knowledge and use of such products.

Our study builds on past work by exploring the nature of groupings based on individuals’ expectations of what combination of resources they will ultimately utilize to fund their retirements. Expectations reflect beliefs about institutional supports like government, employers and family, and beliefs about what one will be able to achieve alone and in conjunction with institutions and families. Butrica, Murphy, and Zedlewski (2009) found that variations exist across age, gender, race, and marital status, categories traditionally known to relate to social stratification, in the proportion of the group whose retirement resources include components of earnings, capital income, social security, DB plan payments, DC plan distributions, in-kind transfers, public transfers, private transfers, capital gains, and the ability to borrow against assets. In 2004, on average, social security made up 83% of poor adults’ retirement income and 24% of middle- to high-income older adults’ income (Butrica, 2008).

After grouping is identified, we also explore how well behaviors align with expectations and whether there are differences in the groupings by socioeconomic, family, and attitudinal factors previously identified as relevant to retirement planning behavior and vulnerability. Socioeconomic variables like income, full-time labor force participation, enrollment in employer-run retirement plans, and home-ownership have been repeatedly identified as related to financial adequacy in retirement (e.g., Hatcher, 2002; Grinstein-Weiss, Curley, & Charles, 2007; Gustafson, Boldt, & Bird, 2005). Gender, marital status, parental status, health status, and other family and individual characteristics have also been found relevant to financial behavior and financial vulnerability (e.g., Gustafson et al., 2005, Guimaraes, 2007, Hartmann & English, 2009; Whitaker, Bokemeier, & Loveridge, 2013). To fully explore the linkages between demographics and groupings based on expectations, it is necessary to consider how status, attitudinal, and behavioral variables relate to systematic differences in expectations.

We expect to find differences among preretirees in their expectations for reliance on various sources of potential retirement income and on their expectations for other people to connect them to retirement funds. We expect some preretirees to have shifted reliance to newer forms of potential retirement income through DC plans and investments and to be less reliant on traditional forms such as social security, pensions, and home values. We expect others to be entrenched in expectations for traditional retirement income sources or with no viable options aside from continued work. Furthermore, we expect personal, social–economic, and structural variables to differ among these groups. For example, preretirees with higher incomes and education will be less reliant on social security and more so on DCs. Fewer preretirees will expect funds from pensions to support their retirement unless they are union members or live with a union member. Women, particularly homemakers, will be more likely to rely on others for retirement income. Although we anticipate certain findings, this is an exploratory, descriptive study. As such, it will report findings and suggest underlying causes.

Methods: Samples, Measures, and Analyses

Data used for this project were collected through an (Institute for Public Policy and Social Research, Michigan State University) 2010 telephone survey of a representative sample of Michigan adults aged 18 and older. This was one wave of a quarterly State of the State Survey (SOSS) conducted since 1994 that regularly captures a core set of information on statewide attitudes toward key economic and social issues and also addresses one-time topics added to the questionnaire in an omnibus fashion. It was a computer-assisted telephone interview survey of 1,969 noninstitutionalized, English-speaking adults aged 18 years and older in Michigan. The sample was generated using a method reflective of the strategy of the University of Michigan Survey of Consumer Attitudes which combines random digit dialing (RDD) and recontact of previously interviewed households (40% of interviews). Data were gathered from February 2010 through April 2010.

Respondents were randomly assigned to one of the two versions of the interview, and 997 answered Version B that contained questions focused on retirement planning and finances. Our analysis used only responses from individuals who indicated that they were not yet retired and were thus still in the expectation and planning stages of retirement. Therefore, 156 (15.6%) cases dropped from analysis because they were retirees and another 144 (14.6%) dropped due to nonresponse (do not know [DK], no answer [NA]/refused) to one or more variables in the analysis. The final number of interviews included in the analysis was 687.

The majority of cases dropped due to nonresponse (DK = 58, NA = 25) were a result of the single question regarding percentage of retirement assets that will come from one’s own sources (9.9% of eligible cases). The remaining of the cases dropped from nonresponse represents a loss of 7.8% of eligible cases. The typical rate of nonresponse is 20%–40% for income questions and 1%–4% for other survey items (Yan et al., 2010). Given the financial nature of the questions and the high level of specificity about personal finances required by the percent-of-assets question, the levels of nonresponse appear as expected for money-related questions. They likely introduce the same concerns with external validity as would be expected in typical research studies that incorporate survey information on income and investments. An analysis of nonresponders to the percent-of-assets question shows no statistically significant difference between responders and nonresponders on age, race, or marital status. There was a statistically significant difference based on gender, with women more likely to offer DK/NA/refused than men. The exclusion of women who cannot answer (DK) a question about retirement expectations may serve to underestimate the disadvantaged situation of women vis-á-vis understanding their long-term financial situation. The sample was stratified based on regions of the state with oversampling to ensure racial representation. Final results were weighted to account for the dual sample frame design and to Michigan demographic characteristics of race, gender, age, and region to account for oversampling and provide better representation. The demographic composition of the sample compared with the overall United States is shown in Table 1. This article was based on nonretirees only (column 1 in Table 1), but a comparison of sample demographics compared with the U.S. demographics can best be achieved by comparing total sample demographics (column 2 in Table 1) to overall U.S. proportions based on census information (column 3 in Table 1). The Michigan sample profile shows that our sample is more White and slightly more educated than the United States overall. Cluster sizes found in this research should reflect relative proportions within the state of Michigan but deviate somewhat from national proportions.

Table 1.

Sample Characteristics Compared With National Demographics.

	MI nonretirees in analysis n = 687	SOSS MI adults n = 972	Total U.S. adults^a
Gender
Male	49.0	47.1%	49.1
Female	51.0	52.9%	50.9
Age
18–29	24.1	20.7	22.1
30–39	26.1	20.2	17.1
40–49	25.7	21.3	19.0
50–59	16.1	15.5	17.9
60–64	3.6	5.3	7.2
65+	4.5	17.0	17.2
Race
White	82.5	83.2	63.3
Black	16.2	14.5	15.4
Other	1.4	2.3	21.3 (including biracial)
Education
<High school	7.6	6.9	13.2
High school	25.9	27.7	30.0
Some Post-HS	34.1	32.4	28.6
College	20.4	19.8	18.4
College+	12.1	13.2	9.8

Note. HS = high school; MI = Michigan.

^aU.S. Census Bureau (2012a); U.S. Census Bureau (2012b); U.S. Census Bureau (2012c).

Within households containing at least one eligible adult, the respondent was selected randomly using the Trohldal–Carter technique (Troldahl & Carter, 1964). The average interview for survey Version B lasted 14.9 min. The overall completion rate for the survey was 45.7% (38.4% in the new RDD sample and 78.3% in the recontact segment). The refusal rate was 16.1%.

Eight values were utilized in the cluster analysis, seven measuring expected reliance on different resources and one measuring level of dependence on self versus others. Respondents rated their expectation of reliance in retirement on seven different self-accrued income sources: social security income (accrued through respondent’s earnings); pension or DB income (accrued in conjunction with an employer through respondent’s paid labor); 401 K, 403B, individual retirement account, or other DC assets (respondent’s and/or employer’s contributions to a usually employer-managed, retirement-specific investment process); respondent’s own individual savings and investments; respondent’s home value; income from respondent continuing to work; and other income sources (unspecified). Ratings were on a scale from 1 to 5, with 1 meaning complete reliance and 5 meaning no reliance at all. Respondents were also asked the percentage of their retirement income they expected from sources they, themselves, accrued. Responses ranged from 0 to 100%. The measures were all standardized to eliminate scale differences. We utilized a k-means cluster analysis to identify groups of individuals who are more similar to each other and more different than others on the basis of their response patterns on the eight factors. The k-means technique is one of the most widely used grouping processes, and it fits with this moderately sized data set, where all variables are treated as interval level and standardized (Norusis, 2005).

With the k-means technique, it is necessary to determine the optimal number of clusters (k) prior to running the analysis. The question of determining k itself remains an active topic of the study (Sugar & James, 2003), and we have elected to utilize a process whereby you repeat the k-means process experimenting with different numbers of clusters. Using Statistical Package for Social Sciences software, we began with two clusters and then increased the number of clusters until the incremental reduction in mean distance to the cluster centers for the new solution approached zero. At this point, the creation of an additional cluster no longer added classification value.

We determined that six was the optimal number of clusters to describe the data on the eight dimensions. Using the k = 6 solution, we identified cases that fell into each cluster and then analyzed those cases on their characteristics on the eight dimensions of retirement income expectations. Next, we used bivariate analysis to look at demographic characteristics and retirement planning behaviors of each cluster. Finally, to confirm the significance of the overall model and the predictive ability of descriptive variables, we conducted a multinomial logistic regression to predict cluster membership.

Results

Cluster 1 is the largest with nearly a quarter (24%) of the sample. We label Cluster 1 “Expect Shared Reliance with Multiple Sources” because this group reported (1) the highest level of expected reliance on others (37%) for retirement resources and (2) an average level of expected reliance on all cited sources, with home value making the lowest contribution and “other” income sources making the highest contribution. Relative to other clusters, they report the highest expected reliance on “other” income sources (see Table 2).

Table 2.

Cluster Means for Own Source Reliance.^a

	Expect shared reliance with multiple sources (N = 168)	Expect limited, if any, options (N = 159)	Balanced expectation (N = 105)	Expect to be diversified self-reliant worker (N = 80)	Expect investments (N = 60)	Expect formal plan dependence (N = 115)	Total sample (N = 687)
Cluster #	1	2	3	4	5	6
Own SS	3.36 (.78)	2.66 (1.5)	2.48 (.81)	2.23 (.93)	4.37 (.63)	3.05 (1.03)	2.97 (1.19)
Own DB	3.08 (.93)	3.92 (1.2)	4.24 (.89)	1.85 (.80)	4.7 (.64)	2.3 (.82)	3.32 (1.30)
Own DC	3.07 (.97)	4.51 (.85)	2.78 (1.28)	2.51 (1.01)	1.48 (.69)	2.42 (.85)	3.04
Own Save/inv	2.92 (.84)	4.45 (.80)	2.37 (.98)	1.79 (.85)	1.93 (.86)	3.5 (.85)	3.07 (1.26)
Own Home	3.66 (1.13)	4.74 (.66)	2.39 (.78)	2.67 (1.28)	3.3 (1.14)	3.43 (1.19)	3.53 (1.29)
Own Work	3.38 (.94)	3.45 (1.32)	2.62 (.89)	2.09 (.88)	4.01 (.72)	4.25 (.77)	3.33 (1.19)
Own—Other	2.81 (.92)	4.86 (.40)	4.48 (.72)	3.81 (1.12)	4.09 (.76)	4.74 (.49)	4.09 (1.09)
Mean % of self- reliance	63% (29.8)	70% (36.2)	66% (36.5)	96% (12.6)	85% (26.1)	74% (31.4)	72.5% (32.7)
% of sample	24	23	15	12	9	17	100

Note. DB = defined benefit; DC = defined contribution; SD = standard deviation; SS = social security.

^a−1 = full reliance, 5 = no reliance mean (SD). All income sources: p = .000.

Cluster 2 is the next largest with 23% of the sample. We call this cluster “Expect Limited, if any, Options.” Individuals in this cluster report below-average mean ratings on all sources of retirement income except their own social security income. Expectations of social security are still only in the middle range (2.66 with 1 meaning high reliance and 5 meaning no reliance). Those in “Expect Limited if any Options” tend to strongly reject expectation of reliance on DC plans, savings and investments, home value, or “other” income sources. On average, they expect to receive 30% of the resources from someone else, though there is wide variation (standard deviation [SD] = 36.2%).

Fifteen percent of the sample is grouped in Cluster 3, “Balanced Expectations.” This group indicated (1) lower than average (M = 66%, SD = 36.5%) expected reliance on own sources although there is wide variation, (2) medium but above-average expectations for savings and investments, social security, and continued work as sources of retirement income, (3) the highest average rating of all clusters in expected reliance on home value for retirement, and (4) income from DB plans and other income not anticipated. They are similar to “Expect Shared Reliance with Multiple Sources” in that they identify multiple income sources. However, “Balanced Expectations” display more positivity about their expected preparedness as demonstrated by the higher average levels of reliance reported for each source.

Cluster 4, “Expect to be Diversified Self-Reliant Worker,” is a smaller group making up only 12% of the sample. Diversified self-reliant workers expect to generate nearly all (M = 96%) their own income in retirement, and that expectation has relatively low variability compared to other clusters (SD = 12.6%). Compared to other clusters, “Expect to be Diversified Self-Reliant Workers” (1) report more uniform, above-average, expected reliance scores across all income sources, (2) are the most likely to expect to continue working in retirement, and (3) have the highest expected reliance on social security, savings and investments, and DB plans.

“Expect Investments,” the fifth cluster, is the smallest cluster with only 9% of the sample. “Expect Investments” tends to be a relatively self-reliant group (on average expect 85% of their retirement income to come from personal sources) who expect DC plans and savings and investments to make high contributions to their retirement income, and expect social security, DB plans, “other” income sources, and continued work to make little, if any, contribution.

Cluster 6 called “Expect Formal Plan Dependence” accounts for 17% of the sample. These preretirees plan to rely strongly on DB plans and then DC plans as sources for retirement income. “Expect Formal Plan Dependence” reports a low mean expected reliance on savings/investments, home value, continued work, and “other” forms of income. They are about average on their expectations of reliance on personal sources (74%) versus others’ sources.

It is clear that groups hold very different expectations, and some of those expectations are not consistent with projected changes in social structure and resource availability. From here, we explore the differences between clusters in terms of demographic, economic, social, and behavioral variables to see who occupies each cluster. We also consider implications and vulnerabilities for each cluster. Bivariate analyses (t-tests, analysis of variance, and chi-square analysis) reveal significant differences in cluster membership across a wide range of social variables, and the distinctive characteristics of each cluster are described subsequently. See Tables 3– 5. All differences reported in the following section are statistically significant at p < .05.

Table 3.

Within and Between Cluster Differences on Gender, Family, and Community Statuses.

% in category/% in cluster	Reliance w/multi source (N = 168)	Limited, if any, options (N = 159)	Balanced expectation (N = 106)	Diversified self-reliant workers (N = 81)	Investments (N = 60)	Formal plan depend (N = 115)	Total (N = 689)
Sex**
Male (N = 338)	25.1%/50.6%	15.7%/33.3%	19.5%/62.3%	11.5%/48.1%	14.5%/81.7%	13.6%/40%	100%/49.1%
Female (N = 351)	23.6%/49.4%	30.2%/66.7%	11.4%/37.7%	12%/51.9%	3.1%/18.3%	19.7%/60%	100%/50.9%
Kids (N = 686)**
Yes (N = 495)	21%/61.9%	26.1%/81.6%	14.1%/66.7%	7.5%/46.3%	10.5%/86.7%	20.8%/89.6%	100%/72.2%
Marital status**
Married (N = 405)	17%/41.3%	22.2%/56.6%	14.6%/55.7%	8.9%/45%	13.8%/93.3%	23.5%/82.6%	100%/59%
Prev married (N = 82)	22%/10.8%	42.7%/22%	11%/8.5%	8.5%/8.8%	2.4%3.3%	13.4%/9.6%	100%/11.9%
Cohabiting (N = 39)	15.4%/3.5%	15.4%/3.8%	12.8%/4.75%	48.7%/23.8%	2.6%/1.7%	5.1%/1.7%	100%/5.7%
Single (N = 161)	46%/44.3%	17.4%/17.6%	20.5%/31.1%	11.2%/22.5%	.6%/1.7%	4.3%/6.1%	100%23.4%
Employment*
Full time (N = 379)	25.9%/58%	22.4%/53.5%	15.3%/54.7%	9%/42.5%	9.5%/61%	17.9%/59.1%	100%/55.1%
Part time (N = 129)	18.6%/14.2%	20.9%/17%	17.8%/21.7%	15.5%/25%	11.6%/24.4%	15.5%17.4%	100%/18.8%
Homemaker (N = 96)	25%/14.2%	20.8%/12.6%	15.6%/14.2%	14.6%/17.5%	2.1%/3.4%	21.9%/18.3%	100%/14%
Other (n = 84)	27.4%/13.6%	32.1%/17%	11.9%/9.4%	14.3%/15%	7.1%/10.2%	7.1%/5.2%	100%/12.2%
Union (N = 635)**
Yes (N = 92)	27.2%/15.0%	26.1%/15.3%	6.5%/5.7%	7.6%/8.9%	2.2%/3.4%	30.4%/24.8%	100%/14.5%
HH member in union (N = 663)**
Yes (N = 115)	18.3%/14.7%	14.8%/10.7%	4.3%/4.8%	30.4%/43.1%	3.5%/6.7%	28.7%/28.7%	100%/17.3%
Race (N = 687)**
White (N = 549)	22.9%/76.9%	23.8%/84.9%	15.8%/85.7%	10.6%/75%	10.4%/98.3	16.5%/81.7%	100%/82.6%
Black (N = 109)	32.5%/23.1%	20%/15.1%	12.5%/14.3%	16.7%/25%	.8%/1.7%	17.5%/18.3%	100%/17.4%
Community**
Urban (N = 95)	37.9%/21.3%	14.7%/8.8%	8.4%/7.5%	17.9%/21.3%	5.3%/8.3%	15.8%/13%	100%/13.8%
Suburb (N = 207)	21.3%/26%	13%/17%	20.3%/39.6%	15%/38.8%	15%/51.7%	15.5%/27.8%	100%/30%
Small city (N = 206)	21.8%/26.6%	31.6%/40.9%	16.5%/32.1%	6.8%/17.5%	4.4%/15%	18.9%/33.9%	100%/29.9%
Rural (N = 181)	24.3%/26.0%	29.3%/33.3%	12.2%/20.8%	9.9%/22.5%	8.3%/25%	16%/25.2%	100%/26.3%
Total	24.5%	23.1%	15.3%	11.7%	8.7%	16.7%	100%

Note. HH = household.

*p < .05. **p < .001.

Table 4.

Cluster Means in Personal, Demographic, and Economic Behavior.

Mean (standard deviation)	Shared reliance/multiple Sources (N = 168)	Limited, if any, options (N = 159)	Balanced expectation (N = 106)	Diversified self-reliant workers (N = 81)	Investments (N = 60)	Formal plan dependent (N = 115)	Total
Age* (N = 687)	36.63 (13.9)	43.06 (13.2)	38.9 (13.0)	37.4 (13.9)	43.03 (9.2)	44.77 (11.4)	40.47 (13.2)
Age will Retire* (N = 644)	64.05 (7.8)	67.93 (13.1)	69.72 (8.2)	63.2 (6.0)	65.1 (6.4)	63.29 (6.4)	65.71 (9.1)
Income (midpoint)* (N = 686)	76.3 (41.4)	46.3 (28.4)	65.1 (47.8)	71.1 (35.2)	109.5 (53.2)	78.0 (41.3)	70.21 (43.7)
Ability to make monthly payments* 1 = ext difficult, 5 = not at all difficult (N = 681)	3.64 (1.1)	3.62 (1.3)	3.73 (1.3)	3.05 (1.5)	4.35 (.7)	4.18 (1.0)	3.73 (1.24)
Status long-term plans* 1 = very well thought out, 4 = no plans (% with no plans; N = 683)	2.58 (27.4%)	2.82 (41.7%)	2.56 (22.1%)	2.47 (30%)	1.47 (6.8%)	1.93 (7%)	2.41 (24.9%)
Health status* 1 = excellent, 5 = poor (N = 687)	2.73 (1.1)	2.8 (1.1)	2.69 (1.1)	2.14 (.9)	1.91 (.8)	2.14 (1.0)	2.5 (1.1)
Education (in years)*(N = 687)	14.2 (2.5)	13.8 (1.8)	14.4 (1.9)	12.6 (1.3)	15.3 (1.4)	14.5 (2.4)	14.1 (2.1)

*p < .001.

Table 5.

Within and Between Cluster Differences for Economic Variables.

% in category/% in cluster	Shared reliance/multiple sources (N = 168)	Limited, if any, options (N = 159)	Balanced expectation (N = 106)	Diversified self-reliant workers (N = 81)	Investments (N = 60)	Formal plan dependent (N = 115)	Total (n = 689)
Invest in savings Acct (N = 687)*
Yes (N = 393)	23.7%/ 55%	13.2%/ 33.3%	16.5%/ 61.3%	15.0%/ 72.8%	10.2%/ 66.7%	21.4%/ 73%	100%/ 57.2%
Invest in DC plan (N = 684)*
Yes (N = 317)	20.2%/ 38.1%	10.4%/ 21%	16.1%/ 48.6%	10.4%/ 41.3%	17%/ 91.5%	25.9%/ 71.3%	100%/ 46.3%
Invest in stocks/bonds/funds (N = 683)*
Yes (N = 170)	18.2%/ 18.6%	11.8%/ 12.7%	13.5%/ 21.9%	5.9%/ 12.5%	17.6%/ 50.8%	32.9%/ 48.7%	100%/ 24.9%
Home ownership (N = 687)*
Yes (N = 442)	19.7%/ 51.8%	21.5%/ 59.7%	15.4%/ 64.8%	9%/ 49.4%	12.4%/ 91.7%	21.9%/ 85.1%	100%/ 64.3%
Monthly budget (N = 684)*
Yes (N = 472)	18.9%/ 53.3%	25.4%/ 75.5%	18.2%/ 81.9%	15%/ 87.7%	7.6%/ 60%	14.8%/ 62.5%	100%/ 69%
Have health insurance (N = 687)*
Yes (N = 599)	25.7%/ 91.7%	20.5%/ 77.4%	14.4%/ 81.1%	11.9%/ 88.8%	9.5%/ 95%	18%/ 94.7%	100%/ 87.2%
Used Ret. Funds in past 2 years other reason (N = 689)*
Yes (N = 180)	26.7%/ 28.4%	17.8%/ 20.1%	12.2%/ 20.8%	23.9%/ 53.8%	3.9%/ 11.7%	15.6%/ 24.3%	100%/ 26.1%
Total	24.5%	23.1%	15.3%	11.7%	8.7%	16.7%	100%

Note. DC = defined contribution.

*p < .001.

Cluster 1

On average, members of Cluster 1, “Expect Shared Reliance with Multiple Sources,” are equally likely to be male or female, and they have the youngest average age of any of the six clusters (M = 36.6 years, SD = 13.9). This group contains the highest percentage of singles who have never married (44%) and the lowest percentage of currently married (41.3%). Nearly half (46%) of all singles are in this cluster. In addition, they are less likely, on average, to have children, though more than half (62%) report being parents. Compared to other clusters, they are more likely to be African American (23%). Cluster 1s tend to be mid-level earners with average incomes in the mid-US$70,000 range. They are more likely to live in an urban setting than other clusters (21%), though they are fairly evenly distributed across community types. They have lower rates of home ownership (52.8%) than the average respondent, and they are the least likely of all clusters to have a monthly budget (53.3%). This cluster was distinguished by a relatively high level of expectations for reliance on sources of income from others. We found that they have the highest level of expectation that their children will contribute to their retirement (20% of those with children), and they are twice as likely as other clusters to expect parent’s financial resources to help (46%). Just more than half (55%) report having a savings account, but they are less likely than the average preretiree to have a DC plan or to invest in the financial markets.

Comparing the expectations and demographic profile of this group with the current structural environment yields interesting results. This group skews toward being younger, unmarried, and childless, though not all cluster members satisfy those definitions. It also skews toward an expectation for “outside” or unconventional help in retirement (parents, children, “other” forms of resources of their own, resources from others). The middle-range, average expectation ratings for each income source do not demonstrate pessimism in the potential of traditional income sources nor do they demonstrate strong optimism in any sources.

This group may reveal a recent change in normative attitudes among some Americans toward intergenerational economic interdependence. Possibly the shift is a manifestation of what Hansen (2005) suggests is the need for networks of support in the new economy where, as a result of structural changes creating harsher conditions in middle-class and working-class communities, families have increased their reliance on others. Material assistance, however, has been shown to be episodic and primarily responsive to specific needs (Swartz, 2009). Studies also suggest that intergenerational care is more common among those of non-European ethnicity including African Americans, but the results on this question have been mixed (Coleman, Ganong, & Rothrauff, 2006). With a higher than average proportion of urban and African American individuals in Cluster 1, it may be valuable to consider if there is a significant relationship between race/ethnicity and intergenerational support. Also, the disproportionately high number of singles and nonparents in this cluster may result in a higher-than-average perception of the continued, lifelong centrality of one’s family of origin in one’s personal and economic matters. As a society, we have historically viewed nuclear family self-sufficiency as a norm in the middle class (Rapp, 1999), which is the typical class standing of this group’s members based on mean income. Regardless of which explanation holds true—a tendency toward more intergenerational reliance or a false perception regarding the centrality of one’s parents and children in one’s economic life long term—members of this cluster should be encouraged to do more individual planning and saving to ensure their financial status in any eventuality, to share plans for reliance with relevant others, and to share planning with those who will be interdependent. Expectations of this group show an unclear path to retirement security.

Cluster 2

We described Cluster 2 as “Expect Limited, if any, Options,” because they report the expectation of being relatively vulnerable in terms of retirement financial well-being. The socioeconomic descriptors bear this out and show that “Expect Limited, if any, Options” have a lower average income in the less than US$50,000 category and low levels of home ownership, savings, and investments. Two thirds of this vulnerable cluster are women, and 43% of all previously married individuals are in this group. Individuals in this cluster are also more likely to have children (81.6%) and to live in rural areas or small towns (74%). The pessimistic expectations of traditional funding sources for retirement among this group with a high propensity to be from dissolved marriages is reflective of findings that women who raise dependent children outside of marriage are more likely to live in poverty during their elder years than women who were married while raising their children (Johnson & Favreault, 2004).

The “Expect Limited if any Options” group displays characteristics of low-income Americans with regard to health disadvantage as well. Research has shown that financially disadvantaged individuals report poorer health ratings than the more advantaged (Shippee, Wilkinson, & Ferraro, 2012), and Cluster 2 echoes results. Its members report, on average, the poorest levels of personal health and the lowest levels of health insurance coverage (23% with no coverage) of the clusters. Consistent with their more tentative average financial situation compared to other clusters, this cluster’s members are more likely to have monthly budgets but not have long-term retirement plans in place. Only 1 (21%) in 5 has a DC plan and only 1 (33%) in 3 has a savings account.

This low-income cluster is vulnerable under current structural conditions, but the uncertain future of social security puts its members in an even more tenuous situation. People in this group, primarily women, should be encouraged to focus efforts on long-term financial goals. It is unlikely, however, that the group has meaningful options to affect change in its preparation. It is thus incumbent on government, business, and policy makers to retain participation in the function of facilitating retirement well-being rather than passing off most responsibility and risk to individuals. Further, given the vulnerability of so many previously married women, currently married women should be educated on their options and status vis-á-vis shared retirement plans in the event of spousal death or divorce. This cluster, which makes up nearly a quarter of the preretiree population, reveals that struggling individuals recognize their vulnerability.

Cluster 3

Cluster 3, the “Balanced Expectations” group, is average on most variables with the exception that they are more likely to be male (62%), live in suburban areas (39.6%), to be nonunion members (93.8%), and to have no union-affiliated family members in their household (94.3%). About a third (31%) of the cluster is single, a disproportionately high amount compared to the average (23%). In terms of planning, 82% of the “Expect Balanced Expectations” have monthly budgets. Compared to other clusters, they plan to retire at the highest average age (69.7).

This cluster’s expectations and behaviors align well with current structural changes, where responsibility and risk are migrating away from government and business toward the individual and the family The group has balanced expectations, as the name indicates, considering all sources as relevant with the exception of DB plans. Thus, the group is not made vulnerable by our societal transition away from DB plans in most sectors. Reliance on social security, the other potentially declining benefit, is only mid-range for this group, and that reliance is situated within a diversified set of expectations. The group’s near-70 average retirement age expectation and its rating of work beyond retirement show awareness of changing social structure. The group is average on its rates of DC, savings account, and financial market participation.

Cluster 4

“Expect to be Diversified Self-Reliant Workers” are unique in their high expectations of nearly complete reliance on their own resources and entitlements for retirement (96%). They also report the lowest average expected retirement age (M = 63.2, SD = 6.0). This group contains a disproportionately large group of cohabiters (24%), with nearly half (48.7%) of all members of cohabiting couples in this cluster. Over half of “Expect to be Diversified Self-Reliant Workers” (53.7%) do not have children compared to only 28% in the total sample who are nonparents. Of the cluster, 25% is Black compared to only 16% in the total sample. This cluster exhibits potential for vulnerability in terms of retirement well-being. They are less likely to own their homes and have the lowest level of educational attainment of all the clusters. Fewer than half of “Expect to be Diversified Self-Reliant Workers” report working full time. Although the group’s mean income is US$71,000, they report the most difficulty making monthly payments, and over half have used their retirement savings in the past 3 years. They have below-average DC plan participation and have the lowest rate of participation in financial market investments. However, their savings account ownership is among the highest at nearly 3/4 (73%).

Looking only at expectations, this cluster appears cognizant of the structural shifts in retirement planning whereby individuals are being expected to take on more of the risk and responsibility for preparedness. Certainly, they state an intention to be self-reliant, and they highlight all traditional sources of retirement income as relevant. However, this group pays the most homage of any group to DB plans, a declining programmatic form, and their level of financial privilege and acumen do not appear to support their optimistic vision. Their use of savings accounts as a predominant savings vehicle suggests a more cautious risk-averse approach.

As explanation, it is possible that all or part of the group has not yet shifted into statuses perceived as requiring economic interdependence or the start of long-term economic responsibility. The average age of the cluster (M = 39, SD = 14) places the group squarely into adulthood. However, less than half of the cluster is married, less than half has children, less than half works full time, and less than half owns a home. Shanahan (2000) points out that transitions to adulthood have been delayed, and that research suggests the delay derives from increasing ages and variability in when people reach “markers” of adulthood, including leaving school, starting a full-time job, leaving the home of origin, getting married, and becoming a first-time parent. Family development theory also emphasizes changes in role expectations when individuals take on new statuses. For example, both married and parents are normatively expected to have increased accountability for the long-term viability of an economic unit (Chaulk, Johnson, & Bulcroft, 2003). Cluster 4’s lower likelihood of being in roles normatively perceived as triggering economic responsibility may put them at risk of having too few years of active preparation for retirement.

Cluster 5

“Expect Investments” is the smallest cluster composed of only 9% of the sample, but they are the most privileged. “Expect Investments” are 98% White and 82% male. Compared to other clusters, they have the highest average income and educational attainment. Nearly all of them are married (93.3%) and have children (86.7%). Compared to other clusters, they are much more likely to be employed full time, to live in suburban areas, and to own their homes. Their privilege extends to access to health insurance and excellent health, on average. They are the least likely to have used their retirement savings in the past 3 years. On average, they do not have difficulty making monthly payments, and their long-term plans for retirement are very well thought out. They are the most likely to invest in DC plans and financial markets.

“Expect Investments” is distinctive in its expectations and its socioeconomic, attitudinal, and behavioral profile. They emphasize the least structurally vulnerable resources for retirement, and they depict a reasonable likelihood to be willing and able to pursue stated plans. This privileged group makes up fewer than 1 in 10 retirees, and it is composed of the groups traditionally viewed as most advantaged in society—Whites, married, and men. The favorable economic position of this group is likely not transferrable to others simply through literacy campaigns and financial planning, given the centuries of advantage that put them in this position.

Cluster 6

This group has been characterized as “Expect Formal Plan Dependence.” Members of the cluster are more likely to be female (60%), married (82.6%), and to have children (89.6%). “Expect Formal Plan Dependence” members are more likely to be employed full time (59.1%) or to be homemakers (18.3%) than the overall sample. Consistent with their expectations of formal DB and DC plans, they have the highest percentage of union members and the highest percentage with a household member in a union. Like “Expect Investments,” “Expect Formal Plan Dependence” reports a high rate of health coverage and good average health status. Both clusters have above-average income levels and are older, on average, than other clusters. They have fairly well thought-out long-term plans and little difficulty making monthly payments.

“Expect Formal Plan Dependence” is a group that displays expectations apparently cohesive with their socioeconomic position and behaviors. Their older average age may explain the higher union affiliation and DB plan expectations. This more secure cluster, however, may become smaller and smaller over time with changes in retirement risk distribution and formal plan composition. The reduced power and prevalence of unions and the move away from DB plans both put the long-term viability of this retirement strategy at risk. DC plans play a nearly equal role to DB plans in the average set of expectations for this cluster, but the increase in part-time and contract labor in our workforce also threatens to put those plans out of reach of workers.

Multivariate Analysis

The bivariate analyses provide a clearer description of each cluster. They show in a straightforward manner the answer to the question “Who are these people?” However, with so many variables under examination it is necessary to consider if relationships hold when other competing influences are held constant. Thus, we conducted a multinomial logistic regression to predict membership in the six clusters. A multinomial logistic regression was preferred to a discriminant analysis because it does not require that assumptions of normality, linearity, or homoscedasticity be met. The independent variables included in the analysis are gender, age, race, marital status, income, education, employment status, parenthood status, community type, health status, ability to make monthly payments, status of long-term retirement plans, expected age of retirement, any union member in household, use of monthly budget, and home ownership. We used an imputation of the mean for missing values to the variables income and expected retirement age to retain cases for these variables with higher rates of nonresponse. The total sample size used in the multinomial regression was 637, a reduction in 7% of the cases from the cluster analysis. Correlations were examined to check for multicollinearity, and no problems were identified. The resultant model was statistically significant overall and yielded a classification accuracy of 57.4% which is 3.2 times better than the proportional by chance accuracy rate of 18.1%. The overall relationship with cluster membership for all independent variables except race was also statistically significant. The model has a Nagelkerke R ² value of .689.

See Table 6 for a cluster-by-cluster presentation of results. Due to the number of direct comparisons and significant associations, we elected to rely on the descriptive statistics to discuss the cluster characteristics. The regression shows that having a budget, marital status, health status, community type, household union participation, and status of long-term plans were significant predictors in the most dyadic cluster comparisons.

Table 6.

Multinomial Logistic Regression Results, Odds Ratios.

	Reference category: cluster 1, shared reliance/multiple sources (% cases correctly predicted = 51%)					Reference category: cluster 2, limited if any options				Reference category: cluster 3, balanced expectations			Reference category: cluster 4 diversified, self-reliant		Reference category: cluster 5 investments
	vs. Limited options	vs. Balanced expectation	vs. Diversified self-reliant	vs. Investments	vs. Formal plan	vs. Balanced expectation	vs. Diversified self-reliant	vs. Investments	vs. Formal plan	vs. Diversified self-reliant	vs. Investments	vs. Formal plan	vs. investments	vs. Formal plan	vs. Formal plan
Age (in years)	1.04**	1.00	1.05	.99	1.03*	.96**	1.00	.95**	.99	1.04	.99	1.03	.95*	.99	1.04*
Education	.92	1.04	.52***	1.19	.95	1.13	.56***	1.30*	1.03	.50***	1.14	.91	2.30***	1.83***	.80
Health status	1.10	.90	.40***	.46**	.65**	.82	.37***	.42**	.59***	.45***	.51*	.72	1.15	1.63*	1.41
Income	1.0	.98	.98	1.02*	1.0	.98	.98	1.02*	1.0	.99	1.04**	1.01	1.05*	1.02	.98*
Ability make monthly payments	1.28*	.95	.55***	.97	1.29	.72*	.43***	.76	1.00	.57***	1.02	1.35	1.78*	2.36***	1.33
Expected ret age	1.02	1.10***	.96	1.06	.98	1.07**	.94*	1.04	.96	.88***	.97	.90***	1.10*	1.02	.93
Status ret plans	2.19***	1.44	.88	.46*	.87	.66*	.40***	.21***	.40***	.61	.32***	.61*	.52	.99	1.9
Sex (ref: F)	.57	2.84**	1.44	4.32**	.65	4.94***	2.52*	7.52***	1.14	.51	1.52	.23***	2.99	.45	.15***
Kids (ref: No)	1.22	1.25	.17**	.47	2.02	1.03	.14***	.39	1.66	.14**	.38	1.61	2.78	11.86***	4.26
Race (ref: non-W)†	.75	.41	1.02	11.51	.71	.55	1.36	15.3*	.94	2.48	27.88*	1.72	11.25	.69	.06*
Work (ref: No)
FT	1.51	.77	.42	.85	1.02	.51	.28**	.56	.68	.54	1.10	1.33	2.02	2.45	1.21
PT	4.26**	4.14**	3.49*	.96	2.88*	.97	.82	.23	.68	.84	.23	070	.27	.82	3.01
Monthly budget (ref: no)	6.16***	7.08***	15.95***	2.17	2.33**	1.15	2.60	.35*	.38**	2.25	.31*	.33**	.14**	.15***	1.07
Any union in HH (ref: no)	1.00	.32**	2.18	.11***	1.75	.32**	2.18	.11***	1.75	6.87***	.33	5.51***	.05***	.80	16.49***
Own home (ref: no)	2.76*	.45	6.16**	1.46	2.51	.16***	2.23	.53	.91	13.70***	3.25	5.58**	.24	.41	1.72
Marital status (ref: never)
Married	4.60**	1.04	7.56**	55.96***	8.79***	23**	1.64	12.16*	1.91	7.28*	53.93***	8.47**	7.41	1.16	.16
Prev married	3.44*	.54	4.57	1.27	4.96*	.16**	1.33	.37	1.44	8.40*	2.33	9.12**	.28	1.09	3.9
Community type (ref: Urban)
Rural	4.35**	2.90	.50	.55	1.24	.67	.12**	.13*	.28*	.17*	.19	.43	1.10	2.47	2.24
Sm. city/town	7.05***	5.44**	.46	.27	2.54	.77	.07***	.04**	.36	.09**	.05**	.47	.59	5.52	9.30*
Suburb	5.91**	9.38***	1.56	.60	2.15	1.59	.26	.10*	.36	.17*	.06**	.23*	.39	1.38	3.57
% of cases correctly predicted	62%	55%	68%	65%	50%	55%	68%	65%	50%

Note. F = female; FT = full time; HH = household; PT = part time; Prev = previous; Ref cat = reference category; Ret = retirement; Sm. = small; W = White.

^†Variable not significant in overall model.*p < .05. **p < .01, ***p < .001.

Conclusions

As expected, we found differences among preretirees in expectations for reliance on different sources of potential retirement income and in expectations for other people to connect them to retirement funds. Some, but far from all, preretirees have shifted reliance to newer forms of potential retirement income like DC plans and investments and expect to be less reliant on traditional forms such as social security, pensions, and home values. Further, expectations of financial security versus vulnerability in retirement reflect structural inequalities, where individuals who enjoy traditional lines of privilege (gender, class, race, and marital status) continue to be situated most favorably. White, married males with high incomes almost exclusively occupy the cluster that is most prepared in terms of expectations and resources that align with expectations. Specifically, “Expect Investments” expectations are well aligned with the current social trends. This is an upper-income, slightly older, predominantly male group with realistic expectations to achieve solvency in retirement, given socioeconomic status and the structural trends facing the United States. The group is economically privileged and accounts for fewer than 10% of preretirees. Another 15%, “Balanced Expectations,” also reports expectations that are consistent with social structural changes, and this group is also more male and more suburban. They anticipate diverse financial resources in retirement including some reliance on others’ resources, but little reliance on DB plans, one of the declining sources. The group is distinctive in its lack of union affiliation and its older expected age of retirement.

We anticipated some groups would be entrenched in expectations for traditional retirement income sources, and one quarter of preretirees, “Limited, if any, options,” fit those criteria. The “Limited, if any options” group also highlights the relative vulnerability of certain social statuses—previously married versus married and women versus men. It contains almost half of all previously married individuals and two third are women. Its members are highly likely to have children, and they are predominantly from small towns and rural areas. This group has the lowest average income, and members perceive themselves to have few financial options for retirement beyond social security. Upon retirement, “Limited, if any options” show little sign of rising above the cumulative inequality and disadvantages associated with race, class, and gender over the life course. Previously married women also exemplify the significant risks of life cycle savings when personal or structural changes invalidate the late-in-life income expectations their life-long savings behaviors are based upon. The multivariate analysis suggests that marital status is the single most relevant variable for predicting cluster membership. This reinforces the critical role of reliance, family, or lack of family in the retirement planning process.

“Formal Plan Dependence” is an older, middle-income group with a focus on both DB and DC plans. They anticipate both emerging and potentially declining income sources to be salient. They report reasonable health and a high level of retirement planning. This group’s expectations may be in jeopardy as a long-term strategy, given the decline in DB plan availability and the decline in primary labor market jobs that offer any plan at all. The remaining two groups (about 38% of preretirees) reveal high levels of uncertainty and a mismatch between expectations for asset accrual and apparent potential for asset accrual. They reveal that, on average, younger clusters have less focused expectations than older ones. It is unclear, however, whether these differences reflect a true age effect rather than a cohort effect.

The study argues effectively for the importance of financial literacy education and an increase in awareness of changing social structure with regard to retirement options. By improving financial literacy among groups not well positioned for retirement, there is the potential to mediate some of their structural disadvantage. For example, education should extend to developing an awareness among women of the risks they may face upon marital dissolution, so that they may be active in considering their own individual interests. Education and awareness would facilitate more attention and more effective behaviors regarding retirement planning activities over the life cycle. Policy makers, as they actively construct the circumstances of our society, must be aware of the risk of changing structure and of structural inequality on the lives of individuals and on the nation as a whole. Financial literacy initiatives on their own do not have the potential to eradicate problems of inadequate retirement resources for many, even most, people.

Therefore, this study argues for the reinforcement of social security as an institution that is critical to serving the retirement needs of vulnerable groups. The potential decline or deferment of social security and the transition of investment risk from institutions to individuals affect individuals in disadvantaged social locations disproportionately. If social security ceases to exist in a form generally reflective of its current form, our findings reveal that many of our nation’s elderly would be in crisis. For example, the “Limited, if any, options” group does not hold expectations consistent with the direction of social changes nor would their economic reality allow them to comply with the changes. To fend off the crisis that would result from a loss of social security as we know it, we would need an alternative institution to somehow take its place as a safety net. The current direction where individuals bear more and more responsibility on their own is not realistic for all groups. Too many individuals/couples simply cannot take on the full burden given the economic challenges they face during their earning years. Major structural changes in employment, family, community, or government would need to take place to generate a new, replacement “leg” of the retirement stool. Such changes would likely involve wholesale cultural shifts in our collective understanding of our principal social institutions.

Limitations

This study is limited in that the sample is made up only of Michiganians. Also, because it is based on cross-sectional data, it does not speak to how individuals may change their expectations as they evolve through the life cycle, and it cannot reveal whether findings related to age are cohort effects or age effects. Next, steps in the research process would include taking the analysis to a national level to determine whether clusters are consistent with state-level findings and the relative size of clusters nationally. Consistent with the life cycle hypothesis of saving, Wong and Hardy (2009) have found that retirement planning is an iterative process. Thus, a longitudinal examination of individuals’ potential movement between clusters over the life course would build upon the cross-sectional findings here. Finally, developing an instrument to determine cluster membership could lead to the ability to study each group more closely and test potential cluster-specific interventions.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding support provided by a Michigan Applied Public Policy Research Grant.

References

Association of British Insurers. (2003). The state of the nation’s savings. London, England: Author.

Atchley

R. C

. (1975 ). Adjustment to loss of job at retirement. International Journal on Aging and Human Development, 6, 17–27.

Butrica

B. A

. (2008). Older Americans reliance on assets. Ownership and opportunity facts no. 10, Washington, DC: The Urban Institute.

Butrica

B. A

Murphy

D. P.

Zedlewski

S. R

. (2009). How many struggle to get by in retirement? The Gerontologist, 50, 482–494. doi:10.1093/geront/gnp158

Chaulk

Johnson

P. J.

Bulcroft

(2003). Effects of marriage and children on financial risk tolerance: A synthesis of family development and prospect theory. Journal of Family and Economic Issues, 24, 257–279.

Coleman

Ganong

L. H.

Rothrauff

T. C.

(2006) Racial and ethnic similarities and differences in beliefs about intergenerational assistance to older adults after divorce and remarriage. Family Relations, 55, 576–587.

Fincham

Beach

S. R.

(2010). Marriage in the new millennium: A decade in review. Journal of Marriage and the Family, 72, 630–649.

Gough

Sozou

P. D.

(2005). Pensions and retirement savings: Cluster analysis of consumer behavior and attitudes. International Journal of Bank Marketing, 23, 558–570.

Grinstein-Weiss

Curley

Charles

. (2007). Asset building in rural communities: The experience of individual development accounts. Rural Sociology, 72, 25–46.

10.

Grusky

D. B

. (2000) The past, present and future of social inequality. In Grusky

(Ed.), Social stratification: Class, race, and gender in sociological perspective (2nd ed., pp. 3–51). Boulder, CO: Westview Press.

11.

Guimaraes

(2007). Optimal external debt and default (Discussion Paper No. 6035). Washington, DC: Center for Economic Policy Research.

12.

Gustafson

L. V.

Boldt

D. J.

Bird

B. M.

(2005). Determining the saving rate necessary for a comfortable retirement. Journal of Personal Finance, 4, 62–78.

13.

Hacker

J. S.

(2006). Privatization of risk and the growing economic insecurity of Americans. Social Science Research Council: Items and Issues, 5, 16–23.

14.

Hansen

K. H

. (2005). Not-so-nuclear families: Class, gender and networks of care. New Brunswick, NJ: Rutgers University Press.

15.

Hartmann

English

(2009). Older women's retirement security: A primer. Journal of Women, Politics & Policy, 30, 2–3.

16.

Hatcher

C. B

. (2002). Wealth, reservation wealth, and the decision to retire. Journal of Family and Economic Issues, 23, 167–187.

17.

Hayes

Hartmann

Lee

(2010). Social security: Vital to retirement security for 35 million women and men. Washington: Institute for Women's Policy Research.

18.

Jain

A. K

. (2009). Data clustering: 50 years beyond K-means. Pattern Recognition Letters, 31, 651–666. doi:10.1016/j.patrec.2009.09.011

19.

Johnson

R. W.

Favreault

M. F

. (2004). Economic status in later life among women who raised children outside of marriage. Journals of Gerontology: Series B, 59, S315–S323.

20.

Kennedy

Bumpass

. (2008). Cohabitation and children’s living arrangements: New estimates from the United States. Demographic Research, 19, 1663–1692. Doi:10.4054/DemRes.2008.19.47

21.

Malone

Stewart

S. D.

Wilson

Korsching

P. F

. (2010) Perceptions of financial well-being among American women in diverse families. Journal of Family and Economic Issues, 31, 63–81.

22.

Modigliani

(1986 ). Life-cycle, individual thrift and the wealth of nations, American Economic Review, 76, 297–313.

23.

National Center for Health Statistics. (2007). Health, U.S., 2007, with chartbook on trends in the health of Americans. Hyattsville, MD: Centers for Disease Control and Prevention.

24.

Norusis

(2005). IBM SPSS statistics 19 statistical procedures companion (13th ed.). Upper Saddle River, NJ: Prentice Hall.

25.

Rapp

(1999). Family and class in contemporary America: Notes toward an understanding of ideology. In. S. Coontz (Ed.), American families: A multicultural reader (pp. 180–196). New York, NY: Routlegde.

26.

Shanahan

M. J.

(2000). Pathways to adulthood in changing societies: Variability and mechanisms in life course perspective. Annual Review of Sociology, 26, 667–692.

27.

Shippee

T. P.

Wilkinson

L. R.

Ferraro

K. F.

(2012). Accumulated financial stress and women’s health over three decades. Journal of Gerontology: Series B: Psychological and Social, 67, 585–594.

28.

Sorensen

A. B

. (1996). The structural basis of social inequality. The American Journal of Sociology, 101, 1333–1365.

29.

Sugar

C. A.

James

G. M.

(2003). Finding the number of clusters in a dataset: An information-theoretic approach. Journal of the American Statistical Association, 98, 750–763.

30.

Swartz

T. T

. (2009). Intergenerational family relations in adulthood: Patterns, variations and implications in the contemporary United States. Annual Review of Sociology, 35, 191–212.

31.

Troldahl

V. C.

Carter

R. C

. (1964). Random selection of respondents with households in phone surveys. Journal of Marketing Research, 1, 71–76.

32.

U.S. Census Bureau. (2006). Oldest baby boomers turn 60! Facts for Features. Retrieved from November 11, 2009, http://www.census.gov/PressRelease/www/releases/archives/facts_for_features_special_editions/006105.html

33.

U.S. Census Bureau. (2012a). The 2012 Statistical Abstract: Population. Retrieved from January 28, 2013, http://www.census.gov/compendia/statab/cats/population.html

34.

U.S. Census Bureau. (2012b). Educational attainment: Educational attainment in the United States: 2012 - Detailed tables. Retrieved from 28 January 2013 http://www.census.gov/hhes/socdemo/education/data/cps/2012/tables.html

35.

U.S. Census Bureau. (2012c). The 2012 statistical abstract: Income, expenditures, poverty and wealth. Retrieved from 28 January 2013 http://www.census.gov/compendia/statab/cats/income_expenditures_poverty_wealth.html

36.

Vincent

G. K.

Velkoff

V. A

. (2010). The next four decades: The older population in the United States: 2010 to 2050. Current Population Reports, P25–1138. Washington, DC: U.S. Census Bureau.

37.

Wheeler

Grunta

. (2009). Promoting productive aging. Health and Social Work, 34, 237–239.

38.

Whitaker

E. A.

Bokemeier

J. L.

Loveridge

. (2013). Interactional associations of gender on savings behavior: Showing gender’s continued influence on economic action. Journal of Family and Economic Issues, 34, 105–119. doi:10.1007/s10834-012-9307-2

39.

Wong

J. D.

Hardy

M. H

. (2009). Women’s retirement expectations: How stable are they? Journal of Gerontology: Social Sciences, 64B, 77–86.

40.

Yan

Curtin

Jans

. (2010). Trends in income nonresponse over two decades. Journal of Official Statistics, 26, 145–164.