We present proposals for three specific Pilot Projects. The first, a two-year interdisciplinary project, falls under the thematic heading """"""""Age-Related Changes in Everyday Context."""""""" The second and third are one-year projects, proposed for years 1 and 2 respectively, which fall under the thematic heading """"""""Demographic Change, Late-life Well-being and Public Policy."""""""" Project One (years 1 and 2): """"""""The daily lives of older adults: Developing measures of daily cognitive function and behaviors for population-based surveys"""""""" (Martin Sliwinski and Janet Wilmoth, project leaders). Age-related declines in cognitive function are associated with significant health outcomes, such as mortality (Sliwinski et al., 2006) and dementia (Small et al. 2000). Cognitive aging also potentially has profound consequences for daily behaviors in older adults. However, relatively little is known about this relationship because the available measures in population-based surveys do not adequately represent daily life as it is experienced by the individual. Most population-based surveys include measures of cognitive functioning based on adapted versions of the mini-mental status exam (MMSE) or the telephone interview cognitive screen (TICS). But these measures only capture relatively substantial cognitive impairment and therefore are unable to characterize the wide range of subtle cognitive changes that occur with age (Lipton et al. 2003) or the daily variation in cognitive functioning that emerges in response to daily stress (Sliwinski et al. 2006). Similarly, population-based surveys typically contain various general measures of depressive mood, social activities as well as measures of physical and instrumental activities of daily living that assess the individual's ability to perform basic self-care tasks. Yet, the relatively wide time frame of reference for these measures, often the past two weeks or longer, relies heavily on accurate recollection, which may be impaired in older adults, overiooks short-term intraindividual variability, and is susceptible to reporting biases that threaten their validity (Carstensen &Hartell 2006). Therefore, the purpose of this two-year pilot project is to develop valid measures of daily cognitive functioning and behaviors that can be used in population-based surveys. Data will be gathered from 75 older (age 60 and over) partnered households (total N=150) using a survey that employs a daily diary design. One-half of the sample will be interviewed once in-person followed by daily CAPI phone interviews for six consecutive days. The other half of the sample will complete a computer-based survey via the internet for seven consecutive days. The survey will include measures of cognitive functioning based on objective cognitive tests as well as self-reports of daily cognitive failures. In addition, it will gather detailed information on daily behaviors that indicate social engagement, self-care and maintenance, daily stress, physical symptoms, health-related activity constraints and mood. Both partners in the household will be surveyed to allow comparisons of self and partner reports. This project has several aims, the first of which is to validate the use of objective cognitive assessment procedures administered over the internet to older adults. We have developed a battery of cognitive tests (Mogle et al., in press), based on measures collected from various aging samples with a range of demographic and socioeconomic characteristics, that are amenable to repeated testing and remote administration over the internet (e.g., intuitive interface, clear instructions). These tests include assessments of secondary and primary memory, working memory, processing speed and executive function. This pilot project would involve adapting these procedures for use with older adults and establishing their convergent and divergent validity. A related aim is to validate daily self-reports of cognitive failures by comparison with objective cognitive testing. Daily cognitive failures (e.g., forgetting to take medicine, inability to complete a complex task) may represent real world consequences of cognitive aging as well as the results of daily challenges (i.e., stressors) competing for limited attentional resources. A problem with current self-reports of cognitive problems is that they require recollection of events over long intervals, often two weeks or longer, which assumes intact and accurate retrospective memoiy. Assessing daily experiences and.?vents vyijl minimize the heavy reliance on^recollectipn and related biases and potentially provide more validindicatcTrs of real-worid cognitive functioning.-"""""""" A second aim will be to develop and validate measures of daily behaviors that indicate social engagement and demonstrate self-care and maintenance. Daily social engagement includes activities such as such in-person, telephone, internet contact with family and friends, and participation in specific social activities including religious rituals, sporting events, paid/unpaid work, educational courses, and recreational programs. Daily selfcare and maintenance involves activities such as exercise, shopping, fixing meals, dispensing/administering medications, and bathing/grooming. Similar to self-reports of cognitive problems, self-reports of behaviors such as these are subject to recall error. In addition, they also are often not broad enough in scope to capture the wide range of older adults'daily activities. Assessing daily behaviors will potentially provide more valid measures of the routine activities in which older adults regulariy participate.
The third aim will be to examine the relationships among daily cognitive performance, cognitive failures, and behaviors. Recent work has drawn attention to the in the importance of reactivity and resilience to daily stress for understanding well-being (Almeida 2005). By linking sensitive indices of cognitive functioning to how individuals respond to daily challenges, this project will provide unique insight on the interplay between cognitive aging and maturational changes in quality of life. Finally, a key aim is to investigate the feasibility of incorporating these computer-administered measures in a large-scale, geographically-dispersed survey sample obtained via random sampling. This pilot represents the first phase in this feasibility assessment. After developing and validating the measures of daily cognition and behaviors, we will pursue a second phase which will focus on alternative methods for implementing the computer-based survey. Previous efforts to incorporate an internet survey into the Health and Retirement Study (Couper et al 2007) indicated rates of internet usage are low among older adults and those who do have internet access are different than those who do not on a range of demographic, financial, and health-related variables. This suggests that samples of older internet users are not likely to be representative of the older population. Therefore, in the second phase of this project we will compare internet-based administration to methods such as loaning low-cost computers to participants or adapting hand-held devises (such as the Nintendo DS) for participant use. Both of these methods of administration would require computer programming and technical support that is beyond the scope of this particular pilot project. The potential long-term scientific returns to this pilot project are substantial. Adaptation to the challenges of aging occurs within individuals in real time. Examining responses to daily stress would provide a window into the ebb and flow of day-to-day health limitations, frustrations and problem solving that may be missed by research limited to the study of major life events and transitions (Almeida, 2005). In addition to providing insight into the daily lives of older adults, including measures such as these on population-based surveys will enable gerontologists to map daily life onto various developmental trajectories and transitions. For example, adding a daily diary component to a survey such as the Health and Retirement Study, or the recently-funded National Survey of Disability Trends and Dynamics (for which a first wave of data collection is planned in 2010), would allow for comparisons in daily cognitive failures and health-related activity constraints across racial/ethnic groups. It would also enable researchers to identify how daily cognitive failures and health constraints on daily social interactions predict changes in living arrangements or reliance on others for ADL assistance. In addition, the impact serious health conditions and disability on daily behaviors could be ascertained. Project Two (year 1): """"""""Health, Disability Insurance, and Family Structure;"""""""" Perry Singleton, project leader. The decision to form and dissolve marriages has important economic and behavioral implications. Ostensibly, individuals form unions to capture the numerous pecuniary and non-pecuniary benefits that marriage provides. Many of these benefits result from investment in marriage-specific capital, including raising children, specializing in market and nonmarket work, and certain types of saving. However, as Becker, Landes, and Michael (1977) suggest, the potential gains from marriage is determined in part by the likelihood of divorce since individuals invest less in marriage-specific capital if the marriage is likely to dissolve. Thus, the analysis of the divorce decision has broad reaching implications for fertility, labor market behavior, and savings. Recent studies examine specifically the link between disability onset and marital dissolution. The most plausible mechanism of an effect of disability onset on earnings is straightforward: several empirical studies demonstrate an adverse effect of disability onset on earnings (Bound 1991;Dwyer and Mitchell 2001), so, if a marriage was predicated on the expectation that the disabled spouse would have earned income in the absence of the disability, then a disability and its associated loss in earnings may precipitate divorce. To be sure, there are other mechanisms by which disability affects divorce, but the negative effect through earnings seems first-order. In recent work, Singleton (2008) builds a conceptual framework for disability and divorce based on previous theoretical work on marital instability. In a seminal paper, Becker, Landes, and Michael (1977) accounted for divorce by introducing a degree of uncertainty in the marital decision, allowing the perceived benefit of mamage at the time of union to differ from what is actually received. In theory, the probability of divorce increases the further realized outcomes differ (for better or for worse) from expected incomes at the time of marriage. The basic theoretical premise of uncertainty and marriage has two implications when extended to disability and divorce. First, the difference between the expected and realized values of marriage increases as the ex ante incidence of disability onset declines;intuitively, the likelihood of divorce is greater if the disability onset is relatively unanticipated. And second, the impact of disability on divorce increases the more adverse the effect of disability on the value of marriage. Singleton (2008) provides empirical evidence to support these theoretical implications regarding disability and marital dissolution. In particular, he examines the association between disability and divorce by age and education which, in theory, should be greatest among the young and more educated - a natural conjecture since disability onset would cause a greater deviation between expected and realized potential among the young and educated who, ex ante, are less likely to be become disabled. To test this conjecture, he compares divorce rates (shortly before and after becoming disabled) among the recently disabled to the nondisabled population using retrospective data from the Survey of Income and Program Participation. The data confirm the theoretical prediction: the association between disability and divorce is greatest among young and educated males compared to their older, less educated, and female counterparts. There is no apparent association between disability and divorce among older, more longstanding marriages. The evident association between disability and divorce found in Singleton (2008) contrasts with earlier work by Kofi-Charies and Stephens (2004). In that study, the authors examined disability onset and marital dissolution using longitudinal data from the Panel Study of Income Dynamics, and concluded that there is little to no association between disability onset and divorce. But in contrast to Singleton (2008), Kofi-Charies and Stephens (2004) did not take into account the potentially heterogeous effect of disability on divorce - in the case of Singleton (2008), by age and education - limiting the scope to which disability could have an effect. The proposed research will examine the underlying mechanism for the linkage between disability onset and marital dissolution found in Singleton (2008). The most obvious mechanism, mentioned above, is the loss in earnings due to disability onset. But many factors may associated with disability onset may lead to marital dissolution, including unobservable characteristics such as impatience or carelessness. To test for the proposed mechanism, the research will exploit a particular policy, implemented in 1987, that dramatically increased disability benefits under the Canadian Pension Plan, which serves all Canadian citizens except those residing in Quebec. If the loss in earnings is the mechanism by which disability and divorce are correlated, then an increase in disability insurance generosity, which replaces earnings in the event of a disability, should attenuate the association between disability and divorce. For several reasons, the 1987 legislation to the Canadian Pension Plan is an ideal natural experiment to examine the effect of disability insurance on divorce. First, there were no substantive changes to the Quebec Pension Plan, which serves Quebec citizens, so residents of Quebec serve as an obvious comparison group. Second, in contrast to other structural reforms to disability insurance programs, the 1987 legislation was discrete and significant, which implies that the effect of disability on divorce can be credibly identified. Third, the policy affected a large nurpber of people, so there_are adequate data sources available to construct a suitably-sized sample of the population before and^aftertlrc legislation was implemented. Aril ^finally, and perhaps most importantiy, the policy relaxed the eligibility criteria (fewer years of previous work would be required) and increased the base benefit amount (so benefits increased regardless of previous earnings history). These two policy features would have a particular effect on the young: ages when the association between disability and divorce is greatest. The main data source for the analysis is the Survey of Consumer Finances, from Statistics Canada. These micro-level data cover the time period prior to and after the 1987 legislation and contain the relevant variables of interest: age, education, region of residence, work disability status, marital status, and disability pension receipt. Although the data are available for public use, they must be purchased directly from Statistics Canada. The effect of the reform on marriage rates is estimated as the relative change in divorce rates between the disabled and nondisabled before and after the 1987 legislation between non-Quebec citizens and residents of Quebec. The proposed analysis, based on Singleton (2008), is intended to serve as the starting point for a broader research agenda on health and family structure. For example, the effects of eariy-life disability on work and marriage may have cumulative effects over the life course. Thus, it is important to consider the economic wellbeing and social networks in old-age of individuals who became disabled early in life. In addition to marital dissolution, future work would examine the role of disability and health in the marriage and remarriage decisions. An important question to consider is whether marital sorting occurs by both observable and latent health, and how this affects the marriage market as a whole. And in addition to the disability insurance, the proposed analysis will examine the impact of medical and technological advances on the decision to form marriages, dissolve marriages, and to invest in marital-specific capital. Similar to disability insurance, medical and technological advances may attenuate the strain that adverse health shocks put on social networks. Developing a broader understanding of these decisions, how factors such as insurance and medical and technological advances affect these decisions, and the cumulative effects of these decisions over the life course have direct implications for family dynamics, optimal social insurance, private insurance mechanisms, and ultimately the well-being of the aged and disabled. Project Three (year 2): """"""""Medicaid, Long-term Care Insurance, and the Housing Behavior of the Elderiy"""""""" (Gary Engelhardt, project leader). This project is motivated by two facts about the elderiy. First, they face substantial risk of long-term care expenses. These are exactly the type of risks that insurance theory predicts there should be a demand for market-based insurance. However, ownership of private LTC insurance is low among the elderiy. Second, contrary to the orthodox life-cycle model, the elderiy do not spend down their wealth fast enough at the end of life, especially their housing wealth (Venti and Wise, 1989, 1990, 1991, 2001, 2004). In fact based on tabulations from HRS """"""""exit"""""""" interviews, 68% of elderiy homeowners at age 70 died as homeowners. The analysis in this project centers on Medicaid as a key explanation for both these seeming contradictions to established economic theory. The idea that Medicaid may explain the low demand for LTC insurance is not new. In particular, Pauly (1990) and Brown and Finkelstein (2007, 2008) suggest that the low level of LTC insurance purchases by the elderiy may be rational, because Medicaid may crowd out LTC insurance demand. However, the impact of Medicaid on homeownership has been relatively little explored. A key feature of Medicaid is its income and asset tests. Under the asset test, non-housing wealth must fall below a threshold in order for the individual to be eligible for Medicaid payment for long-term care. Importantiy, housing wealth is exempt from the asset test. The elderiy may self-insure against uncertain end-of-life expenses by holding a buffer stock of assets as a precaution (Palumbo, 1999). In addition, if the individual has a bequest motive, the stock of assets can also serve as a buffer and a potential bequest if it goes unused (Dynan, Skinner, and Zeldes, 2004). Given the preferential treatment of owner-occupied housing under the Medicaid asset-test, housing appears to be an ideal asset to hold for both precautionary and bequest motives, possibly rationalizing why the elderiy hold onto their housing so long. This will be examined using a stochastic dynamic programming model of elderiy Medicaid, long-term care, and housing behavior. The riiodel'oJfBfewn and Finkelslfeiri (2008) serves as a point of departure. They examined the impact of Medicaid on the demand for LTC insurance under the assumption of a single source of wealth and no bequests. To jointly explain the housing and wealth decumulation puzzles, their model will be expanded to include a bequest motive, a distinction between housing and non-housing assets, and the exemption of housing assets from Medicaid's means test. The dynamic programming model also incorporate the ability of the elderiy to liquidate housing assets by changing from home owning to renting. The core of the theoretical framework and much of the notation is based on Brown and Finkelstein (2008). Formally, utility, U, is derived from consumption, C, and long-term care services, F, and exhibits constant relative risk aversion. The individual faces a per period probability of survival of p, with period L being the known maximum length of life. With probability 1 - p , the consumer dies and receives the terminal payoff O, the utility of bequests. When alive, the individual may be in one of four care states, indexed by s: at home, receiving no care {s = 1 );at home, receiving home health care (s = 2);residing in an assisted living facility (s = 3);or, residing in a nursing home (s = 4). Economic resources come in three forms. First total wealth,W[T], is the sum of housing wealth, Wh , and non-housing wealth, Wn, i.e., WT =W'N +WH. Both earn a riskless pre-tax return, r. Second, the individual receives annuity income. A, from Social Security. Finally, if the individual has private long-term care insurance, it will pay the lesser of actual long-term care expenditures, X, and a benefit cap, B. Let P denote premia paid for LTC insurance.
Showing the most recent 10 out of 19 publications
