What the Information Experiment Actually Changed
The IZA experiment provides the cleanest test of the question at the center of this series: does an official life-expectancy figure move the personal belief and retirement plan that should, in theory, depend on it? The important task here is to separate the causal contrast from the interpretation. We need to know exactly what the treated respondents received, which measured outcomes moved or stayed still, and why a near-zero result is evidence about this intervention rather than a universal verdict on information.
The experiment was designed around a chain. A person could receive a more accurate estimate for similar people, revise their own expected age at death, and then revise an expected retirement age or a complementary-pension choice. The chain can fail at any link. The results are most informative when each link is named rather than compressed into the sentence that “retirement planning did not change.”
The treatment and its contrasts
The survey included 1,000 people aged 25 to 60 in Trentino–South Tyrol. Respondents first supplied the knowledge measures and an estimate of the average lifespan of people with their relevant age, gender, and county characteristics. The random intervention then occurred; own-longevity and planning outcomes were elicited afterward.
A randomly selected half received a numerical actuarial estimate for a person matching their age, gender, and county. The other half—the control group—was told that the national statistical office calculates life-expectancy estimates by those characteristics but was not given the numerical figure. Both groups therefore encountered the idea of official longevity statistics. Only the treatment group received a number that could directly correct their estimate of the reference group.
This is a stronger comparison than simply asking whether people who know more make different plans. People who seek financial information may already differ in income, education, health, patience, or confidence. Random assignment instead creates a treatment and control group within the same survey, so the difference in outcomes can be attributed to receiving the numerical actuarial information, subject to the limits of the design.
The outcome sequence followed the proposed mechanism. Respondents reported their own expected longevity, their expected retirement age, and their satisfaction with retirement planning. They also reported choices related to complementary pension schemes, including willingness to participate and whether they would annuitize private pension funds. The study could therefore look for an effect on the belief most directly connected to the statistic, an effect on the planned work transition, and effects on related pension decisions.
The central result is a set of null findings. The tailored actuarial figure did not produce a statistically detectable change in own-longevity beliefs. It did not change expected retirement age. It also did not change willingness to join a complementary pension scheme or the stated choice to annuitize a complementary pension fund. The message did not travel from a population estimate into the principal personal and planning outcomes that the life-cycle model made relevant.