How the Harvard Loneliness Report Handled Its Own Causal Claim

The title of the Harvard Graduate School of Education’s Making Caring Common report is doing more work than its methodology can support. “Loneliness in America: How the Pandemic Has Deepened an Epidemic of Loneliness,” published in February 2021, reports that 36% of Americans experience serious loneliness, that 61% of young adults aged 18 to 25 do, and that 43% of young adults say their loneliness increased since the pandemic began. Those numbers travelled widely and quickly. What travelled less is how the report itself framed the uncertainty behind them, and where that framing is careful versus where it leans on the reader’s assumptions.

A retrospective self-report, treated as a trend line

The 43% figure is not a before-and-after measurement. It is a single survey asking respondents to recall and estimate a change in their own emotional state over roughly a year. Retrospective self-report of this kind is vulnerable to the respondent’s current mood coloring the recalled past, and to the salience of the pandemic itself supplying a ready explanation for however lonely someone happens to feel at the moment of answering. None of this means the estimate is wrong. It means the report is reporting perceived change, not measured change, and the title’s verb — “deepened” — implies the latter more than the data can bear.

To be fair to the report, this is a known and largely unavoidable limitation of single-wave survey design, and it is not unique to Making Caring Common’s work. But a methods-conscious reading has to hold the causal framing in the title separately from the descriptive statistic in the body, because the two are not making the same claim.

The more interesting number is buried

The finding that does more analytic work, and gets far less attention, is this: about half of lonely young adults said no one had taken more than a few minutes in recent weeks to ask how they were doing in a way that felt genuine. This is not a loneliness score. It is a report about the quality of contact received, filtered through the respondent’s own judgment of what counts as genuine. That judgment is exactly the distinction Cacioppo’s work on loneliness turns on: loneliness is not the absence of interaction but the felt absence of the kind of connection a person needs, and it can coexist with a full calendar of contact that doesn’t register as meaningful.

This matters for how the report handles technology, even though it never states a technology hypothesis directly. Video calls, texts, and social media messages plausibly kept overall contact frequency from collapsing during the period the survey covers. If half of lonely young adults nonetheless reported no genuine check-in, the report’s own data suggest that contact volume and contact quality moved independently — technology-mediated or not. The report gestures at this by pairing the two statistics side by side, but it does not test whether the mode of contact (in person, by phone, by text) predicted the “genuine” judgment. That is a testable question the underlying survey could plausibly answer and the published report does not.

An instrument problem the report doesn’t flag

Making Caring Common’s 36% figure sits awkwardly next to two other widely cited numbers from the same period: the AARP Foundation’s 2018 finding that one in three U.S. adults 45 and older are lonely, and Cigna’s 2020 finding that 61% of U.S. adults report sometimes or always feeling lonely. These are not competing facts about the same underlying reality; they are three different questions. AARP used the 20-item UCLA Loneliness Scale, a validated instrument with decades of use across the epidemiological literature, which is why its estimate is directly comparable to academic work like Holt-Lunstad’s mortality meta-analyses. Cigna’s index and the Making Caring Common survey each use their own constructed measures, calibrated differently, asked of differently sampled populations at different moments. None of the three reports states this plainly enough to stop a reader from treating 33%, 36%, and 61% as three measurements of one thing.

The Harvard report does not claim comparability with AARP or Cigna, and it should not be faulted for a mismatch it never asserted. But it also does not caution against the comparison that its headline number invites journalists and policymakers to make. That omission is the more consequential one, because it is the gap between what a report says and what a report allows to be assumed that determines how its findings get used downstream.

What would resolve the ambiguity

A stronger design would pair a validated, repeated-measures loneliness instrument — administered before and during the disruption in question, on the same panel — with a separate measure of contact mode and contact quality, so that “more time alone” and “less genuine contact” could be disentangled rather than reported in the same breath as if they were one finding. Absent that, the honest reading of the Making Caring Common data is narrower than its title: loneliness was high among young adults in early 2021, contact quality was widely judged as poor, and the two co-occurred. Whether the pandemic caused the deepening, and what role screens played in it, remains an inference the report invites rather than a result it demonstrates.