The Limits of Telling People How Long They May Live

The IZA study gives this series a precise answer to its opening question: a life-expectancy statistic may fail to change a retirement plan when it describes similar people without becoming evidence about the recipient’s own future. This final synthesis follows the complete path from population information to personal belief, from personal belief to the retirement horizon, and from that horizon to saving and pension choices. The judgment it supports is not that longevity information is useless. It is that information must become self-relevant, connect with other planning knowledge, and reach a decision before its theoretical value can become a practical effect.

The distinction matters because the study contains two findings that can seem contradictory. Realistic longevity beliefs can matter for consumption, saving, and retirement timing. Yet, among 1,000 working-age adults in Northern Italy, giving people an official estimate tailored to age, gender, and county did not produce a detectable short-run change in own-longevity beliefs or expected retirement age. The contradiction disappears once the planning model and the belief pathway are treated as separate parts of the same system.

The broken chain is the answer

A retirement plan needs a horizon. The life-cycle model treats expected lifespan as the endpoint over which wages, assets, pension benefits, and consumption must be coordinated. A longer expected life can require more resources or a different transition from work to retirement. Under the model’s assumptions, a one-year upward revision in expected longevity would raise the optimal retirement age by an average of 0.43 years, or about five months. That is the model’s answer to what would happen if a person’s own planning horizon changed.

But a population statistic is not yet a changed planning horizon. The information treatment supplied an actuarial estimate for a reference group defined by age, gender, and county. The respondent still had to take the next inferential step: treat that group estimate as evidence about their own expected age at death. Only after that step could the revised horizon reach retirement timing, saving, or a complementary-pension choice.

The IZA results show a break at that connection. Respondents held beliefs about similar people and beliefs about themselves as distinct objects. In the control group, they expected to live to about 85 years and four months themselves, roughly three years longer than the lifespan they assigned to similar people. Twenty-one percent thought similar others would die before age 80, while only 1% expected to die themselves before 80. The overall correlation between the two beliefs was 0.09. The statistic about the group was therefore a weak predictor of the personal horizon that should have guided the plan.

That weak link changes the meaning of the treatment’s null result. The official figure could be accurate as a description of average longevity and still fail as a personal signal. A respondent might accept the population benchmark while giving more weight to private self-information or to an intuition that their own future is different. Tailoring the estimate makes it more specific; it does not automatically make it personally diagnostic.

The same chain contains another weakness: the horizon is not the only input. The study’s framework identifies longevity literacy, pension literacy, and financial literacy as distinct components of a retirement decision. The first concerns how long resources may need to last. The second concerns the public income stream and the rules that produce it. The third concerns compounding, inflation, and diversification. A person can be accurate about one component and still build a poor plan because the components are not connected.

The survey’s evidence makes that concern concrete. More than 32% of respondents underestimated the actuarial lifespan of similar people by over five years, while about 8% overestimated it by over five years. Their average estimate was only about 2.4 years below the official benchmark, but roughly 40% were more than five years away in one direction or the other. The average partly hides the size of the individual planning errors.

Pension and financial knowledge were also incomplete. Only about 32% correctly identified how state benefits were calculated, about 20% correctly identified pay-as-you-go financing, and 6% answered both pension questions correctly.