ISR|SRC|Michigan Retirement Research Center|UofM
Mailing List | News | Contact Us
Follow onTwitter
Main photo Promoting research on retirement and Social Security policy

Volume 17, Issue 2 - May 2017

Workshop previews upcoming research

This newsletter features work on two topics of general interest: long-term care (LTC) and cognition’s role in retiree well-being.

Pinpointing cognitive decline’s onset

Johns Hopkins School of Public Health and School of Medicine researcher Lauren Hersch Nicholas presented “Detecting Cognitive Impairment in Financial Data,” National Institute on Aging-funded work done with the Federal Reserve Board of Governors’ Joanne Hsu. Their project hopes to identify the early moments of cognitive impairment before it has been diagnosed and managed.

To do this, Nicholas and Hsu use FRB Consumer Credit Panel data (credit scores, balances, past due accounts, opening of new accounts, etc.) and Medicare indicators of dementia and other chronic/acute conditions, as well as Health and Retirement Study (HRS) Telephone Interview for Cognitive Status (TICS) scores. The study, which is in its early stages, will probabilistically match age, geography, and household consumption to that data.

“Some of what we know, partly clinically and partly anecdotally, is that one of the earliest signs of cognitive decline and dementia is difficulty managing money and paying bills,” said Nicholas. “You may be suddenly making uncharacteristically risky financial decisions: calling your broker and saying your entire portfolio should go into one stock or to be cashed out,” for example, which may leave people open to credit shocks, home foreclosure, and fraud.

“There may be a trail of financial transactions that would let us catch some of these signs of cognitive impairment,” Nicholas said. To be able to pinpoint the onset of decline might help policymakers build safe guards into financial systems and enable caregiver intervention.

“What we would eventually want to be able to do is assess the feasibility of using this type of data to do population monitoring,” to determine whether credit events might be predictive of future dementia diagnoses. Nicholas said, “This would be similar to ways your credit card company is always trying to look at whether someone else is making fraudulent purchases.”

With that goal in mind, the researchers will eventually be working with a machine learning group (a method that automatically builds iterative, analytical data models) to fine tune the predictions.

HRS principal investigator David Weir had thoughts about the study’s machine learning aspect. “Nobody who’s really good at machine learning, algorithms, and data science works in academia,” said Weir. “They all work in the private sector making lots of money…You’re probably not going to be able to do better than they do at crunching credit data, but you might be able to find out what they do and make use of some of that.”

Determining individual-level dementia costs

How much a dementia diagnosis costs an individual may be tough to determine, but it’s important information for a family making financial decisions. “Think about, for example, the problem of whether you want to buy a long-term care insurance product,” said Péter Hudomiet. “Obviously, the benefit for you depends on the chance that you will ever need it, and if needed, how much money you would be able to save.”

Hudomiet presented “Dementia and the Distribution of Lifetime Out-of-pocket Medical Expenditures,” a joint project with fellow RAND researchers Michael Hurd and Susann Rohwedder, that looks at dementia-related, out-of-pocket (OOP) costs, which have relevance for families.  The trio asked three questions to help quantify such expenses:

  1. “What fraction of 65+ year old Americans live with dementia for at least six months before death?
  2. “What is the distribution of lifetime out-of-pocket medical expenditures from age 65 to death?
  3. “How much of the lifetime costs can be attributed to dementia?”

Using the Aging, Demographics, and Memory Study (ADAMS) supplement of the HRS, the researchers modeled dementia probabilities for the entire survey, which aside from ADAMS does not directly observe dementias. “Forty-one percent of the population in the HRS lives with dementia for at least six months before death. This number is actually larger than what we saw in the [previous research] literature,” said Hudomiet.  Demographics play a role in risk, with women, the less-educated, and blacks having a higher incidence.

The mean (average) lifetime OOP spending is $61,000 in 2014 dollars, with a skewed distribution: The median (half spend more, half spend less) is $36,000, but the 95th percentile is $206,000. Hudomiet said, “We are still checking the numbers because they’re higher than we expected.”

Most OOP spending goes toward drugs, which almost everyone buys. The second biggest cost is nursing homes, although only a third of people spend anything on them. The researchers found that dementia increases lifetime spending by $35,000 (2014 dollars) on average, almost exclusively due to nursing home costs.

The trio hopes to turn in future research to informal costs on caregivers, public program spending (Medicaid), and implications for LTC insurance. “The difference between out-of-pocket costs and social costs are enormous,” co-author Hurd pointed out during the comments.

Uncovering LTC insurance’s structural issues

In spite of an aging society with a significant risk of dementia-related nursing home stays, the LTC insurance market isn’t working well. According to Atlanta FRB researcher R. Toni Braun, rejections are high (approximately 36 percent), are not actuarially fair, coverage is incomplete, and it is treated as a toxic asset in the industry, partly because of high administrative costs. This has lead to atrophying numbers of LTC insurance providers. “In 2013, about 60 percent of new policies were written by one of three insurers,” Braun said.

He and co-authors Karen A. Kopecky (FRB Atlanta) and Tatyana Koreshkova (Concordia University?Montreal) built a model and included supply-side frictions (adverse selection, insurer’s market power, fixed and variable overhead costs) and demand-side frictions (Medicaid, means-testing, secondary payers) to try and account for the LTC insurance market’s troubles. The group spent a year calibrating the model to HRS data. It breaks an individual’s life into two periods, young and older than 55, and uses a monopolistic single insurer. The individual has private information on nursing home entry risk.

The model’s results account for high premiums, partial coverage, and low profits. The researchers find that these can be attributed to adverse selection, insurer’s overhead costs, and Medicaid. Even though the model has a single source of private information, that information produces a small or negative correlation between LTC insurance ownership and nursing home entry.

According to the study, the current market helps no one. “The supply-side frictions lead to the insurer, when he sees risks groups with high frailty, he’s going to reject them. The whole pool will not receive an offer of insurance,” said Braun. This produces low take-up rates by the wealthy.

“Medicaid is of first-order importance for the poor because Medicaid is free, and it’s a secondary payer,” Braun said. “If you take Medicaid away, [the insurer] is going to make all his money on the poor and the frail because their demand is biggest. You bring Medicaid in, and they’re gone, so the set of insurable risks becomes a small group in the middle.”

For the middle class, these factors interact in complex ways. “If you take away any of [the frictions], take-up rates go way, way up,” said Braun. For example, “even though I’m in the middle class and there may only be a couple states of nature where I would enjoy Medicaid benefits, that still affects my demand and the outside options of the insurer.”

The authors released this project’s working paper, “Old, Frail, and Uninsured: Accounting for Puzzles in the U.S. Long-Term Care Insurance Market,” in March. It is available on the Atanta FRB’s website.

Nursing home quality and LTC insurance

In past research, RAND investigators Michael Hurd, Italo López, and Susann Rohwedder found that respondents in the RAND American Life Panel (ALP) were willing to pay at least 43 percent more in monthly premiums for policies to avoid a low-quality nursing home.  In this iteration, “Willingness to Pay [WTP] for Long-term Care Insurance: How Much do Individuals Value High Quality Nursing Homes,” the group fielded a new ALP survey to look at whether those additional, willingly paid premiums would equal the lifetime costs of such care.

“The main objective of the project is to learn whether preferences for the quality of nursing homes might help explain the low level of long-term care insurance take up, and whether the data can help us to learn how insurance products could be better designed to meet the needs of respondents,” said López, who presented the latest work. Note that the investigators used subjective, consumer-based quality measures (food, private/shared room) rather than medical ones defined by the Centers for Medicare & Medicaid Services (bed sores, restraints, etc.).

For this survey, the researchers added questions about subjective probabilities of purchase for each policy, as well as randomly assigned variations. One variation described reality-based descriptions of Medicaid and non-Medicaid nursing home services and restrictions. The other offered actual features (daily benefits and length of payout) from currently available LTC insurance policies. They used responses to these variables to determine a respondent’s willingness to purchase a policy.

So far in preliminary work, the team has found that respondents are willing to pay $20 to $40 more in monthly premiums for policies that increase the per diem benefit from $200 to $300 and benefit duration from three years to five. “Then we look at what would be the actual change in the policy premium when you go to [insurance company] websites and take an average, how much would it take to move…to $300 a day, ” said López.

For a daily benefit increase of $100, the average premium for a 60-year-old would increase $127.90 a month. By contrast, the respondents’ WTP averages $37.60 between the study’s two methods of measurement.

In future work, the researchers will compare WTP to actual lifetime costs using actuarial values. “We hope to update these numbers from [estimated] monthly premiums to this more accurate data,” said Lopez.

Stay tuned

While much of the research presented at the workshop is not yet available to the public, many of the projects will have working papers out this fall. MRRC will announce its publications via email, newsletter, and social media.