Summary
Introduces and documents "last-place aversion" (LPA): individuals exhibit especially strong disutility from occupying last place in a distribution, beyond what standard risk aversion, prospect theory, or inequality-aversion models predict. Using laboratory experiments and survey data, the paper shows that LPA affects risk-taking behavior, dictator-game transfers, minimum wage preferences, and support for redistribution in the General Social Survey (GSS). Low-income individuals near the bottom — but not at the very bottom — may oppose redistribution because it threatens their ability to distinguish themselves from an even-lower group.
Key Claims
- Lottery experiment (LPA in risk taking): In 6-player groups with randomly assigned wealth ranks, last-place players choose a risky lottery (equal expected value, but chance to move out of last place) 13.4 percentage points (pp) more than players in ranks 1–5 (23% above the baseline rate of 59.1%). The effect is flat across ranks 1–5 and nonlinear at last place. Survives controls for: being below median, being below previous round, expected rank change, and Fehr-Schmidt disadvantageous/advantageous inequality. Holds when balances accumulate across rounds (with the effect extending to 5th place as well, consistent with LPA-under-uncertainty for the second-to-last player).
- Modified dictator game (LPA in transfers): Subjects given $2 must give it to the person directly above or below them (each separated by $1, so giving downward costs a rank). Fehr-Schmidt inequality aversion predicts always giving downward. Observed: second-to-last-place subjects give upward 25–50% of the time — directly falsifying Fehr-Schmidt for subjects near the bottom. Effect is specific to second-to-last place; subjects at other ranks almost always give downward.
- Minimum wage survey: Original survey of low-wage workers. Overall support for a minimum wage increase is
80%, but significantly lower (10–15 pp) among workers making $7.26–$8.25 — just above the 2014 minimum of $7.25. LPA predicts these workers fear losing the status distinction of being above the last-place wage group. (Reported in the Kuziemko et al. (2014) file but the wage range refers to a survey conducted close to the paper's submission.)
- GSS redistribution analysis: In nationally representative GSS data, individuals in the 6th through 8th income deciles — above poverty but below the median, the rough analogue of "second-to-last place" — support redistribution significantly less than their income decile and background characteristics predict. The LPA-group coefficient (−0.096) is as large as the coefficient on moving up one income decile (−0.087). Holds across racial subgroups; racial resentment is not the primary explanation (LPA effect for whites is strongest in "whitest" regions, and nonwhites in the LPA group show similar patterns).
- Distinguishing LPA from competing models: LPA is empirically separable from (a) standard diminishing absolute risk aversion (predicts last-place player least likely to gamble), (b) prospect theory with mean/median reference point (controlled for; last-place effect remains), (c) being below previous round (controlled for; effect remains), (d) expected rank change (controlled for; effect remains), (e) Fehr-Schmidt inequality aversion (predicts always transfer downward; LPA breaks this for second-to-last place), (f) Bolton-Ockenfels ERC (equity, reciprocity, and competition), (g) Charness-Rabin distributional preferences, and (h) a linear rank effect.
- Behavioral mechanism: LPA formalizes utility as g(r)=1(r>1) — a step function giving a bonus to all but the last-place individual. Under income uncertainty, the second-to-last-place player also faces nontrivial probability of falling into last place, producing low-rank aversion that is concentrated at the bottom of the distribution.
- Broader applications: LPA potentially explains consumers' tendency to buy the second-cheapest wine (avoiding association with the last-place product), resistance to letting others leapfrog in queues (Oberholzer-Gee 2006), and evidence that Moving to Opportunity (MTO) boys who moved to better neighborhoods showed increased criminal activity — consistent with low academic rank in the new school environment encouraging effort substitution to other domains (Kling, Ludwig, and Katz 2005).
Concepts Introduced or Extended
Entities Mentioned
Quotes
"Last-place aversion suggests that low-income individuals might oppose redistribution because they fear it might differentially help a last-place group to whom they can currently feel superior."
"The differential decrease in support for redistribution associated with being in the LPA group (–0.0958) is larger than the decrease associated with moving up an income decile (–0.0870)."
My Take
The laboratory identification is exceptionally clean — LPA is separated from a comprehensive list of competing models in a single paper, which is unusual for a behavioral economics contribution. The dictator-game result is the most striking: it directly falsifies Fehr-Schmidt for subjects near the bottom without requiring auxiliary assumptions. The minimum wage survey is suggestive but small and targeted (low-wage workers in one survey); the GSS evidence is correlational. The key limitation for policy is the gap between the lab's precise last-place operationalization and the real world, where "last place" is undefined; the paper is honest about this. The mechanism is compelling but the paper cannot fully close the alternative that the LPA group simply has different political preferences for other unobserved reasons. The connection to MTO criminal behavior is speculative but thought-provoking — LPA may have implications for education and neighborhood effects beyond redistribution.