There is a moment familiar to anyone who has done this work: the 4 a.m. moment, when you are awake because someone else is awake, standing in a kitchen you did not expect to be standing in, doing something that has no name and no invoice. No one bills for it. No one pays for it. And yet, when economists at the University of Southern California sat down to count what dementia costs the United States this year, that unbilled hour turned out to be one of the largest line items in the country's ledger. In a study published online in Alzheimer's & Dementia on 24 June 2026, Johanna Thunell and colleagues put the 2026 total at $818 billion. Of that, $237 billion is the value of care that families give away.
The machinery behind the number matters, because it is not a receipt; it is a simulation. The team built the US Cost of Dementia Model, a dynamic microsimulation that ages a synthetic but nationally representative population forward in two-year steps, drawing on the Health and Retirement Study from 1998 to 2020, Medicare and Medicaid administrative data, and Census projections. It estimates that 5.7 million US adults aged 51 and older are living with dementia in 2026 (95% confidence interval 5.6 to 6.0 million), 5.1 million of them aged 65 and older, supported by roughly 5.2 million care partners. Those care partners will provide 6.8 billion hours of unpaid care this year. To turn hours into dollars, the model uses a replacement rate: $34.92 for every hour, the national average as of July 2025 of what it would cost to hire someone to do the same work. Multiply, and you get $237 billion. A companion methods paper by Bryan Tysinger and colleagues, published the same day, spells out the alternative approach, valuing the same hours at what the caregiver could have earned instead, which produces a different figure entirely. The choice of method is not a technicality; it is most of the difference between the low and high estimates that have circulated in this field for fifteen years.
Now shrink the aggregate to household scale. Dividing the study's own totals, 6.8 billion hours across 5.2 million care partners averages about 1,300 hours each per year, which is roughly 25 hours a week, every week, with no holiday. That is a second job, and the model says these people are frequently in their prime working years. The consequences show up elsewhere in the same ledger: $23 billion in forgone earnings across people with dementia and their care partners combined, and $15 billion assigned to the care partners' own lost quality of life, the emotional and physical wear that a spouse or a daughter absorbs quietly. The largest item of all, $320 billion, is the quality of life lost by people living with dementia themselves. It is the most contested number in the paper and the least like money, and it belongs to the person, not to the system.
Set against all that, formal medical and long-term care costs $222 billion this year. Read that again beside the $237 billion of unpaid hours: families supply more value in kind than the entire health and long-term care system spends in cash. Of the $222 billion that does move, Medicare and Medicaid cover about 70 percent, roughly $154 billion; patients and families pay about 20 percent, roughly $46 billion, straight out of pocket. So the household is hit twice, once for what it buys and once for what it substitutes. The study also reports that the distribution is uneven: non-Hispanic White individuals account for about 60 percent of the dementia population but about 66 percent of total costs, while Black and Hispanic individuals account for 21 and 19 percent of the population respectively and roughly 17 percent of costs each. Lower measured cost is not good news. It often means care was never purchased in the first place, and someone at home did it instead.
What the imputation really represents, then, is a transfer that the national accounts do not record. Every hour priced at $34.92 is an hour withdrawn from a paid shift, from sleep, from a marriage, from a teenager who needed help with homework. The dollar sign is a translation device, a way of making visible labour that the market ignores precisely because it is done out of love and obligation. Julie Zissimopoulos, who leads the US Cost of Dementia Project, framed the point in terms of policy, saying the work "can help guide decisions about how to allocate resources" as the dementia population grows. Families reading the same number will hear something blunter: the system's books balance because yours do not.