公众“过度自信”、退休收入体系认知与退休规划行为研究报告

Overconfidence, Knowledge of the Retirement IncomeSystem, and Retirement Planning* Philippe d’AstousHEC MontréalFranca GlenzerHEC Montréal January 13, 2026 Abstract Previous research shows that the level of confidence in one’s financial ability is importantfor decision-making, especially in the realm of retirement planning. We expand on this liter-ature by using survey responses to objective and subjective measures of financial literacy andretirement knowledge. We find that even though overconfident individuals are more likely tostate that they have a retirement plan, they are less likely to have registered retirement savings,and when they do, they hold lower balances. Our findings highlight a potential mechanism inwhich overconfidence in one’s knowledge of the retirement system raises expected income re-placement rates, which—consistent with a standard consumption–saving model—reduces pri-vate saving. Overconfident individuals also have biased inflation perceptions but take fewerprotective actions to mitigate the effect of inflation. Finally, we find that overconfident individ-uals decrease their scores with repeated participation in different waves of the survey. Theseresults suggest that calibrating confidence about one’s knowledge of the retirement system andof macroeconomic factors may be important for improving private retirement saving. JEL classification:D14, G53, J26.Keywords:Overconfidence, Financial literacy, Retirement, Inflation. 1Introduction As public pensions and defined-benefit plans decline, individuals increasingly bear the burdento finance their consumption in retirement (Lusardi et al., 2011, Behrman et al., 2012).At thesame time, consumers face ever more complex financial choices in preparing for retirement (Choiet al., 2004), often including multiple layers of income sources such as public pensions, employer-sponsored plans, and registered savings accounts. Previous research has shown that financial lit-eracy matters for many financial outcomes, including more effective retirement planning (Lusardiand Mitchelli, 2007, van Rooij et al., 2011, Behrman et al., 2012, among others). Recent research has highlighted important nuances regarding what factors facilitate soundretirement planning. First, beyond objective knowledge, self-assessed knowledge – often termedconfidence – also matters for financial decisions, and in some contexts may matter even morethan actual knowledge (Parker et al., 2012, Anderson et al., 2017, Angrisani and Casanova, 2021,Lawrence et al., 2024). Second, specific knowledge about the retirement system appears to affectretirement planning above and beyond general financial literacy (Debets et al., 2022, Boisclair et al.,2023, Castagno et al., 2025).To date however, the role of individuals’ confidence in retirementliteracy has not yet been addressed. An important obstacle to studying the effects of objective andsubjective retirement knowledge on financial choices is that very few surveys provide an objectiveset of questions on the retirement income system.For this reason, researchers often resort tomeasuring overconfidence as a difference between subjective knowledge of the retirement incomesystem and objective financial literacy (e.g. Angrisani and Casanova, 2021). In this paper, we leverage data from a large-scale survey on the Canadian retirement incomesystem.The survey collects standard measures of financial literacy, self-perceived financial lit-eracy, and socio-economic characteristics such as age, education, and income. Crucially, it alsoincludes a detailed questionnaire that captures objective knowledge of all pillars of the retirementsystem, alongside self-assessed knowledge of the system more generally. In addition, respondentsreport whether they have a retirement plan—a common proxy for retirement preparedness in theliterature—their expected income replacement rate in retirement, and specific actions they havetaken to prepare, such as enrollment in defined-benefit (DB) or defined-contribution (DC) plansand saving in tax-advantaged individual savings accounts. Finally, the survey measures respon- dents’ perception of the previous year’s inflation and the adjustments they made in response. We define overconfidence as the extent to which individuals’ self-assessed knowledge, rel-ative to the sample average, exceeds or falls short of their demonstrated knowledge on objec-tive tests.Formally, we measure overconfidence using the distance between standardized self-assessed knowledge scores and standardized test-based knowledge scores, relative to the averageperformance of survey respondents. This effectively benchmarks each individual’s self-perceptionagainst how they and others in the sample actually perform on the knowledge tests. Individualswith scores near zero are well-calibrated, recognizing either that they know or that they do notknow. Negative scores indicate underconfidence, where individuals know more than they thinkthey do, while positive sco