Experience Effects

Experience Effects

Experience effects (Malmendier & Nagel, 2011, 2016) are a generalization of extrapolative beliefs that adds cross-sectional heterogeneity by age cohort. The hypothesis: each investor forms expectations about future returns, inflation, and other macro variables as a recency-weighted average of observations during her own lifetime, with younger investors weighting their (shorter) experience more heavily than older investors. Two people facing the same long-run history will therefore form different expectations if they have lived through different parts of it.

Empirical findings. Malmendier & Nagel (2011) — “Depression Babies” — show that individuals who have experienced low stock market returns over their lifetimes report lower willingness to take financial risk, are less likely to participate in the stock market, allocate a smaller fraction of liquid assets to stocks if they do participate, and have more pessimistic future-return expectations. Malmendier & Nagel (2016) extend the same approach to inflation expectations: young people who lived through the high-inflation 1970s have systematically higher long-run inflation expectations than young people who came of age in the low-inflation 1990s. The age-cohort gap is substantial during periods of macroeconomic volatility (the 1970s) and small during stable periods.

Experience effects are mathematically related to extrapolation but motivated by a different psychology — typically the availability heuristic (events you have personally lived through are more cognitively accessible) rather than the representativeness heuristic. Whether experience effects and return extrapolation should be modeled as one mechanism or two remains open.

Intuition

Standard rational-expectations models assume every investor has access to the full historical record and weights it the same way. Experience effects relax this in the most natural possible direction: an investor only knows what she has personally lived through, with more weight on recent events because those are the most cognitively available. A $30$-year-old who has only ever observed bull markets will form different expectations than a $70$-year-old whose memory includes the $1970$s stagflation, the $1987$ crash, and the dot-com bust. The two are extrapolating from the same long-run truth but from different windows of it.

The result is that beliefs are no longer a single object — there is a distribution of beliefs across cohorts, and the dispersion is larger when recent history is more variable. In stable times, all cohorts converge to similar expectations. In volatile times, the youngest cohorts are most reactive (because their lifetime sample is small and recent observations dominate it), while the oldest cohorts are most stable (because each new observation is a small fraction of their lifetime data). This generates testable cross-sectional predictions: differences in stock-market participation by age cohort should track differences in lifetime return experience, and differences in inflation expectations should track differences in lifetime inflation experience.

Properties

(Functional Form) Lifetime weighted average with recency weighting.

The Malmendier-Nagel specification writes individual $i$‘s expectation at time $t$ as
${\mathbb{E}}_{i,t}[r_{t+1}]={\sum }_{k=0}^{{\text{age}}_{it}}{w}_k(\lambda )\cdot r_{t-k},$
where $w_k(\lambda )$ is a weight function with parameter $\lambda$ controlling recency weighting. For $\lambda >0$, recent observations get more weight. The key difference from standard extrapolation is the upper limit of the sum — investor $i$ only sums over data she has actually lived through, so her age determines the length of her sample.

(Cohort Heterogeneity) Younger investors react more to recent events.

A young investor with a short lifetime sample assigns each new observation a larger fractional weight. After a single bad year, a $25$-year-old’s average shifts substantially; a $65$-year-old’s barely moves. The result is testable: cross-sectionally, young cohorts should show more volatile expectations than old cohorts. Malmendier & Nagel (2016) confirm this for inflation expectations using the Michigan Survey — the standard deviation of expectations across young respondents is roughly twice that of old respondents during the high-inflation 1970s and roughly equal during the low-inflation 2000s.

(Depression Babies) Lifetime stock returns predict portfolio risk-taking.

In the canonical Malmendier & Nagel (2011) result, a one-standard-deviation difference in lifetime experienced stock returns predicts an economically meaningful difference in stock market participation and in conditional stock allocation. Individuals who came of age during the Great Depression were less willing to take financial risk decades later, even after controlling for current wealth, income, and demographics. The same pattern holds across many cohorts and many horizons of “lifetime” exposure.

(Inflation Expectations) The 1970s/2000s gap in young vs. old expectations.

The Michigan Survey of Consumer Expectations provides micro-level inflation expectations from $1953$ onward. During the volatile $1970$s, expectations of long-run inflation among respondents under $40$ exceeded those among respondents over $60$ by several percentage points — because the young had spent most of their adult lives in high inflation while the old had a longer sample including the calm $1950$s. During the stable $1990$s and $2000$s, the cross-cohort gap narrowed to nearly zero. The age cohort that experienced the greatest cross-sectional disagreement was always the cohort whose lifetime sample was dominated by the most recent volatile decade.

(Mechanism Open) Source of experience effects is not yet pinned down.

Malmendier & Nagel motivate experience effects via the availability heuristic — events you have personally lived through are more cognitively accessible — rather than via representativeness. But the resulting weighted-average specification is mathematically very close to standard extrapolation; the only difference is that the sum is truncated at the investor’s age. Whether experience effects should be modeled as a separate psychological mechanism or as a special case of extrapolation with a cohort-specific data window is a live question. The empirical patterns are robust either way.

(Application to Asset Prices) Nagel & Xu (2022) embed experience effects in an equilibrium model.

Nagel & Xu (2022) build an asset-pricing model in which earnings growth is i.i.d. but investors form expectations as a long-term weighted average of past earnings growth (with cohort heterogeneity from age). The model fits aggregate stock-market fluctuations and the survey evidence on long-term expectations simultaneously, again from a single extrapolation parameter — confirming that experience effects can do the same equilibrium-asset-pricing work as plain extrapolation when integrated cohort-by-cohort.

Examples

Inflation expectations in the 1970s vs. 2000s

A $30$-year-old in $1980$ has spent the majority of her adult life in the high-inflation $1970$s; her lifetime average of CPI inflation is around $7\%$. A $60$-year-old in $1980$ has $30$ years of additional data including the calm $1950$s, with a lifetime average closer to $4\%$. In the Michigan Survey, the $30$-year-old’s reported expectation of long-run inflation in $1980$ averaged $9\%$; the $60$-year-old’s averaged $6\%$. By $2005$, after two decades of $2$–$3\%$ inflation, the cross-cohort gap had collapsed to less than $0.5$ percentage points — both the young and old had lifetime samples dominated by low inflation, and they agreed.

Depression babies and stock-market participation

Malmendier & Nagel (2011) compute, for each individual in the Survey of Consumer Finances, an “experienced return” — the weighted average of S&P $500$ real returns over that person’s lifetime, with more weight on recent years. They then regress stock-market participation and conditional stock share on this experienced return, controlling for current age, wealth, income, education, and year fixed effects. A one-standard-deviation increase in experienced returns predicts roughly a $7$ percentage-point increase in the probability of stock-market participation and a $4$ percentage-point increase in conditional stock allocation. Strikingly, individuals who had experienced the same historical period at different ages reported different expectations — confirming that it is the individual-level lifetime sample, not aggregate calendar time, that drives beliefs.

See Also

• 8.6. Belief-Based Biases — parent hub.
• 8.6.7. Extrapolative Beliefs — the closely-related framework that experience effects refines with cohort heterogeneity.
• 8.6.5. Representativeness Heuristic — the alternative cognitive root (vs. availability) for the weighted-average specification.