Center for Translational Research on Aging Projects
The Center for Health and Wellbeing is home to a National Institute of Aging Roybal Center, called the Center for Translational Research on Aging. The center has eight active pilot projects, two of which were carried over from a prior grant.
Interactions between Economic Status, Psychological Well-being, and Age: Analysis of an Unconditional Cash Transfer Trial in Kenya
PIs: Johannes Haushofer and Jeremy Shapiro, Princeton University
Within the aid industry and the research practitioners and organizations that support it there is a great deal of emphasis on “sustainability.” This term tends to mean one of two things within the industry: a) organizational sustainability or b) individual sustainability. Organizational sustainability refers to an organization being able to sustain its operations on a long-term basis without ongoing receipt of donor funds. Individual sustainability refers to a single intervention or development program being able to lift an individual or individual family out of poverty on a long-term basis. This focus on sustainability has fueled support for interventions such as microfinance or agricultural support services, which have the potential to allow poor recipients to earn returns and increase their future income. Critically, however, these interventions may have substantively different impacts for older recipients, and the research that has evaluated these interventions has not adequately considered potentially heterogeneous effects for the elderly.
The purpose of this project is to fill this gap in the literature. In particular, we intend to conduct secondary analysis on a dataset that was created during a randomized control trial impact evaluation of an unconditional cash transfer program. The initial study evaluated the impacts of cash transfers on economic outcomes (such as expenditure and income) as well as measures of psychological well-being. In this secondary analysis, we will explore heterogeneous effects of cash transfers on the elderly, and widows in particular. The purpose of this analysis will be to assess whether cash transfers “work” for the elderly in significantly different ways than they do for younger, and potentially more economically active, recipients. We will further link economic impacts to effects on subjective and objective (measured through cortisol levels, a stress hormone) indicators of psychological wellbeing.
Does being surveyed affect subsequent reports of subjective well-being?
PIs: Johannes Haushofer and Jeremy Shapiro, Princeton University
The central challenge of measuring well-being in social science is that “true” well-being is unobserved and usually has to be estimated through self-reports. This fact creates a number of concerns for the interpretation of such reports of subjective well-being (SWB). Apart from the obvious question how survey measures of SWB relate to the latent target variable, a further important concern is that eliciting self-reports of SWB may itself affect SWB, or reports of SWB.
The literature has described a number of survey effects. “Question-behavior” effects refer to a change in behavior or SWB due to being asked prospective questions about one’s own future actions or outcomes. Thus, subsequent SWB might be affected by being asked about one’s own projections for it. In contrast, the well-known Hawthorne refers to a change in behavior as a result of being observed in an experiment; for instance, participants in a randomized controlled trial which aims to improve productivity using an intervention can affect productivity simply because participants know about the goal of the experiment and respond in terms of productivity (demand effect). We propose to study a third type of survey effect: simply being surveyed about SWB may affect subsequent self-reports of SWB. This effect differs from a “question-behavior” effect because respondents are asked about current, not future SWB; and it differs from a Hawthorne effect because respondents are not exposed to an intervention – they survey is the intervention.
We propose to study this question in the context of an ongoing field study in rural Western Kenya. We have previously conducted a randomized field experiment on the Unconditional Cash Transfer program by the NGO GiveDirectly in this region. In this study, a treatment group received unconditional cash transfers, and their survey responses were compared to two control groups: a spillover group, i.e. control households in treatment villages, and pure control households, i.e. control households in control villages. Importantly, the treatment villages were surveyed both at baseline (wave 0) and endline (wave 1), while the pure control villages were surveyed only at endline (wave 1). This poses a threat to the integrity of the study, and more broadly raises the question whether this double surveying affected responses.
Age and the correlates of wellbeing: moving from correlation to policy
PI: Angus Deaton, Princeton University
The goal of this project is to determine whether, under what circumstances, and exactly how the large literature on the correlates of wellbeing should be used in setting policy. “Happiness” regressions find, for example, that people who are older, better educated, better-off, more religious, or who live in Wyoming or Hawaii, have higher life-evaluation and/or better emotional lives. How might such regressions be useful for policy? One possibility is that they could be used to inform people about what makes lives better: for example, people may be happier in middle age if they realize that being elderly is on average associated with high life evaluation and emotional wellbeing. Another possibility is to use the regressions to calculate “shadow prices” for circumstances, by converting each circumstance’s effect on wellbeing into its equivalent money value using the effect of income as a base. These shadow prices measure the monetary value of outcomes and can be used in benefit-cost analysis to compare with the costs of interventions. Of particular interest here is the extent to which shadow prices change with the age of the individual. For example, it seems that additional income has its largest effect on life evaluation in middle age, and that the effect declines after age 50. Knowing this would be useful for information purposes—telling people about what is likely to matter for them, which they often do not know—as well as in the design of policy, such as social security policy. In spite of the superficial attractiveness of such ideas, there remain unresolved the fundamental questions of just when the numbers that come out of these happiness regressions can be legitimately used in this way, whether they mean what they seem to mean.
Wellbeing, self-reported health, and suicide
PI: Anne Case and Angus Deaton, Princeton University
The goal of this project is to document and better understand the relationship between measures of self-reported wellbeing (SWB), self-reported health status (SRHS), and suicide. The pilot will look at the changing patterns of suicide by age, both in the US and around the world. In recent years, the suicide rate for the elderly in the US has fallen below the suicide rate for those in middle-age, and this pilot will investigate a possible link between the reversal of the age pattern of suicide and a matching deterioration in SRHS in middle age—one not observed among the elderly. The pilot will examine whether changing levels of pain experienced by different age groups can explain both outcomes. The pilot will match vital statistics information on causes of death from the CDC and the WHO to data on SWB from Gallup’s World and US national polls, to data on SRHS and health conditions from the National Health Interview Survey and the National Health and Nutrition Examination Survey (NHANES), and to data on risk factors and SWB from the Behavioral Risk Factor Surveillance System (BRFSS).
Setting limits and their relation to wellbeing in end of life care
PIs: Jeremy Shapiro, Princeton University; Geoffrey Rees, University of Chicago
For the past forty years and more, abundant research has focused on patient preferences for end-of-life care. The increasing reliance on advance directives in health care reflects a concern to enhance the evaluative well-being of patients and families when negative experiential well-being intensifies. Since passage of the Patient Self-Determination Act (1991), it has become a standard of care to attempt to match the care patients receive at the end-of-life with their known preferences. Underlying these efforts is the conviction that self-determination enables people to live more comfortably in the present and die with greater dignity in the future.
Equally important in the promotion and use of advance directives is the hope that they will reduce the provision of expensive and medically inappropriate care at the end of life. Rather than set explicit limits on care, the reliance on advance directives seeks to guide patients and their families to set their own limits, following the conviction that only patients and families can evaluate for themselves the contribution of care to their well-being. At the same time, expanding from the universal default status of cardiopulmonary resuscitation, more and more additional life-saving and intensive procedures are becoming defaults of care, adding to the expense of dying and the concern that too often end-of-life care is a complex negotiation to end treatments that arguably should never have started. Even as the costs of end-of-life care continue to increase, becoming particularly intense in the last six months of life, little research so far has focused specifically on public attitudes about this spending, and about how people relate policy on spending to evaluations of their own well-being. This gap in research is particularly notable when considered in light of the widely acknowledged discrepancy between spending and quality in medical care at the end of life, that greater spending does not equal greater well-being, and that they are often inversely correlated.
Our research aims to fill this gap and thereby contribute to a growing national conversation about how to improve the quality of dying in the United States, by asking about how people assess spending on end-of-life care in relation to their evaluative well-being. Working through MTurk we will conduct a three-part survey of a broad sample of the US population. In the first part we will capture on a Likert scale participants’ responses to a series of statements about setting limits on spending on end of life care. In this section, we will also include a variety of life satisfaction measures, permitting us to explore how attitudes towards end-of-life care relate to evaluative well-being. In the second part of the survey, we will present participants a variety of scenarios of care at the end of life, and ask them to make decisions about spending, including options to reallocate funding to support of general health care for the population. In the third part of the survey we will ask respondents to report their well-being through use of the Cantrill ladder, their experience in caring for anyone at the end of life, and experiential information about pain and health problems. In this final section, we will specifically explore how advance directives and other methods of allocating end-of-life care relate to individuals anticipated evaluative well-being, and their perceptions of how such methods might impact the evaluative well-being of others.
Together these parts will enable us to explore how people think about intensive care at the end of life as a public good that supports individual well-being. And how assessments of thriving relate to determinations of appropriate spending. The results will enable us to develop grounded answers to the questions: How do people relate their individual evaluations of well-being and end-of-life care to their expectations for responsible public policy on spending for such care? To what extent do people support paying for others when they don’t associate the same level of care with well-being for themselves? To what extent and under what circumstances do people consider that medical care at the end of life is wasteful if it doesn’t match their own evaluations of well-being?
Working at the intersection of psychology, behavioral economics, and ethics, our research is particularly important because it promises to generate insight into the challenge of setting limits on spending on end-of-life care while also supporting the evaluative well-being of patients and families. Just as ethics contributes a framework for the study of the meaningfulness and value of human action, so also behavioral economics contributes a framework for the study of the psychological and economic background of human action. Taken together they promise to generate powerful insights into the ways that people value medical care at the end of life as a source on the one hand of individual welfare, and a source on the other hand of common good. Through carefully developed survey questions and scenarios, the research will consequently provide understanding not only of what people think about spending on end-of-life care, but how they relate what they evaluate as supportive of their own well-being to what they think is fitting to authorize in spending on others.
This research will shed light on several important topics that are currently not well understood:
- All persons are encouraged to complete advance directives and living wills, but little is known about how completion of them contributes to a sense of present evaluative well-being, and how awareness or lack of awareness of the costs of end-of-life care contributes to the decisions people make about the care they wish for themselves and others, for example about how people evaluate their future well-being in the event of terminal illness.
- More and more aging adults are asked to make specific decisions about care for themselves and for family members, again with little insight so far into how people understand and relate the perceived economic value of the care about which they are making decisions to their evaluative well-being.
- A central concern for aging adults is to remain independent, yet little insight is available into how people think about the claim of aging adults to expensive care as means of sustaining their independently evaluated well-being, and when and how such expenditures count as burdensome to the general population.
- Likewise little research so far explores the difference between how medical professionals assess appropriate spending as a proxy for well-being compared to how patients and families think about what is appropriate or not, and especially how professional expertise and other social determinants contribute to the sense of well-being – or lack thereof—that people associate with greater expenses on medical care.
Our research consequently promises clinical application by generating insight into the decision making process of patients and families, opening opportunities to improve shared-decision making through empirically grounded understanding of how patients and families value medical care in their present as well as their future evaluations of well-being.
Finally, as the population ages and the pool of Medicare recipients expands, so also the more general population of insured is also expanding, placing a strain overall on available medical resources and prompting greater public awareness of the potential need to make more explicit decisions about how to allocate spending. Doing so will require deeper understanding of the ways in which the aging population values spending on end-of-life care in relation to their well-being and their willingness to make trade-offs that affect the well-being of the broader community.
Toward new evaluative well-being questions
PIs: Marc Fleurbaey, Princeton University; Dan Benjamin, Jakina Debnam, and Ori Heffetz, Cornell University
This proposal relates to the following points from the call for applications: (3) Explore the magnitude of survey mode effects, contextual effects, and other biases in existing measures of both ExSWB and EvalWB and methods for mitigating them; (4) investigate the extent to which people’s perceptions of their SWB adapts to changes in life circumstances (e.g. longterm unemployment, chronic disease, and widowhood) and how those relationships differ over the lifecycle.
The overall goal of this project is to come up with recommendations regarding specific wordings for SWB questions to be asked on large scale surveys. Such recommendations would hopefully be useful to governments, national statistical offices, and others who conduct large scale social surveys. Indeed, such recommendations would be very timely: there is growing effort, following Stiglitz-Sen-Fitoussi (2009) and others, on the part of national agencies to start collecting SWB data. But which SWB data to collect, and how to phrase the questions?
We focus on evaluative questions (such as ladder-of-life or life satisfaction) and examine them from three perspectives: a) due to imprecise wording, is there heterogeneity in how people perceive the scope of the questions (such as the relevant time period, whether it is about the respondent’s strictly personal situation or includes others)? b) do different respondents use the scales (typically 0-10) in different ways due to self-anchoring to personal goals, reference groups, past experience, so that answers are not comparable? c) how do the answers relate to the respondents “ideal” (i.e., informed, pondered) preferences?
In a first wave of the survey (questionnaire enclosed), we have examined the first two issues for eight different questions (three standard questions about satisfaction and happiness, and five new questions inspired by previous work by members of the team), using introspection after first asking the evaluative SWB question. Preliminary results show that there is indeed variation in the perceived time and person scope of various questions, and that self-anchoring is pervasive.
In the second wave, which is the object of this application for funding, we would like: 1) to test new questions, which will be constructed with the goal of combining as little ambiguity as possible about the meaning and scope of the question, as little self-anchoring as possible, as much deliberation as possible, without constraining expression of values and preferences and without imposing cognitive burden; 2) to reformulate some of the introspective questions, in order to make it clearer to respondents that we are asking about how they understood the question rather than their personal preferences (for instance, in the first wave we asked about what they thought about when answering the question, and this combines their understanding of the scope and their own preferences).
This project will therefore, with respect to existing evaluative questions, contribute to a better knowledge of the problems with their formulation (point 3 of the call) as well as a better understanding of self-anchoring, which includes the adaptation phenomenon (point 4 of the call). It will also contribute to constructing new, hopefully more effective, SWB questions, with the goal of reducing noise and getting responses that are more comparable across the population and across time, and are as faithful as possible to the respondents “ideal” preferences.
The project can be viewed as pursuing efforts at probing SWB questions made by Krueger and Schkade (2008), Steffel and Oppenheimer (2009) and Ralph et al. (2012). It also builds on previous work by members of the team (references below), that has shown that various SWB questions do not equally relate to the respondents’ preferences and choices, and has explored the ethical and empirical conditions for constructing interpersonally comparable well-being indexes.
Survey Methods For Measuring Affect
PI: Dylan Smith, Stony Brook University
There is a growing emphasis on capturing subjective appraisals of emotional experiences, as a supplement to traditional “objective” economic performance measures such as income, or epidemiological measures of health such as longevity. This pilot study tests the validity of several recall-based methods for capturing affect in daily life. Specifically, it examines and compares assessments taken on the same day as the observation day (the “end of day” approach), versus assessments taken the next day. The next day assessment will be either with or without “instantiation” of the previous day, as operationalized in the Day Reconstruction Method (DRM). DRM is a structured diary methodology designed to enhance accurate recall via instantiation of the previous day, which is done by dividing the day into discrete “episodes.” These three approaches differ substantially in terms of the level of burden placed on respondents, and in terms of cost of implementation. The “next day” approach, which relies on memories of the previous day, is the least burdensome as it involves a (potentially) brief survey that can be administered by phone, in written form, or by web survey at any time of day, and at the respondent’s convenience. The “end of day” approach is also brief, but as its name suggests must be completed at the end of the day, which represents a significant constraint. Finally, a full version of the DRM requires around 45 minutes to complete, and thus is not amenable to a brief phone survey (although abbreviated versions have been successfully scaled up for national surveys). To explore the validity of these three approaches, this study will compare the different recall-based methods to real-time assessments taken during the observation day (ecological momentary assessments [EMA]). It will also employ a mood manipulation on the observation day, to examine how well it is captured by differing recall methods, and how it may affect recollection of other parts of the day.
Investigating The Relation Between DRM And Experienced Yesterday Measures
PI: Arie Kapteyn, University of Southern California
The Day Reconstruction Method (DRM) elicits a full-day diary of episodes accompanied by an inventory of emotions/moods during each of the episodes. The DRM proves to be successful in reinstating feelings of the previous day (Kahneman et al., 2004) and has become a widely adopted method for eliciting experienced well-being. The DRM however takes considerable response time and hence is hard to incorporate in on-going major population surveys. Several simplifications have been proposed, where rather than asking about the whole previous day, one selects a small number of activities and asks for emotions during the performance of these activities.
A different approach to simplification is taken by the Hedonic Well-Being 12 measure (HWB12; Smith and Stone, 2011), which asks respondents for an overall day assessment on 12 dimensions (happy, enthusiastic, content, angry, frustrated, tired, sad, stressed, lonely, worried, bored, pain) on a 5-point scale. Similarly, the Gallup Experienced Well-Being Index includes 10 items (anger, depression, enjoyment, happiness, sadness, stress, worry, learning or doing something interesting, smiling or laughing a lot, being treated with respect) on a binary scale, in addition to whether the respondent would like to have more days like yesterday.
This project aims to investigate the relation between aggregate measures like the HWB12 and DRM-like approaches. To that end we will conduct a time diary study for one day (“yesterday”) among 2000 respondents in the Understanding America Study. We will moreover elicit emotions for a number of periods. By also asking the HWB12 for “yesterday” we can study the relation between the HWB12 and the DRM. Proper randomizations will take care of possible anchoring effects.