Abstract: Grandfathering protects incumbents but can impede exchange when protection expires at sale. Beijing’s 2018 school-assignment reform did this: retaining a pre-2019 title preserved eligibility under the prior single-school rule, whereas transfer made the buyer subject to announced multi-school assignment. Using 7,340 completed resales and historical assignments reconstructed from contemporaneous notices, I show that formerly key-school communities recorded roughly six sales per community before and after implementation, while ordinary-school communities rose from roughly three to five. A seasonal design yields a sharp exposed-relative implementation gap, but its magnitude is benchmark-sensitive. There is no detectable pre-deadline acceleration, and a separately benchmarked cumulative comparison detects no two-year differential. Among completed trades, price and discount point estimates move little while marketing-duration estimates rise. A search-and-bargaining model shows how a sale-triggered reset can remove marginal matches, make the selected mean price unsigned, and delay rather than destroy exchange. Comparing title-, property-, household-, and calendar-based protection under stated partial-equilibrium conditions makes a time-limited household entitlement one concrete benchmark to evaluate; the data do not rank transition rules in welfare terms.

Abstract: Variance treats favorable and adverse earnings realizations as the same risk. I use public CPS histories and strictly later SIPP portfolios to separate lower-tail distance, upper-tail distance, total tail span, and orientation. Among 27,752 annual transitions beginning with no reported stock-and-mutual-fund holdings, a one-cross-cell-standard-deviation wider lower tail predicts a 2.71-percentage-point lower probability of reporting positive holdings one year later; a one-standard-deviation wider upper tail predicts a 1.92-point higher probability. The two distances remain jointly significant after 200 CPS resamples rebuild the first stage (p = 0.0022), while the total span is imprecisely estimated. A scale-free incremental Kelley contrast predicts a 1.05-point lower transition relative to an 8.79% rate. Measurement is consequential: zero-inclusive Kelley reliability is 0.74 in CPS but falls to 0.37 beyond zero incidence, mean, and dispersion. The estimates are descriptive: historical group distributions are neither exogenous nor elicited beliefs.

Abstract: Using transaction-level data on 189,530 Chinese retail investors over the 2013–2016 boom and bust, we measure experience as counts of large realized gains and losses and find that it reshapes the disposition effect, the tendency to sell winners and hold losers, in an asymmetric way. A two-standard-deviation increase in accumulated large losses raises the propensity to sell winners rather than losers by about 23% of the baseline effect, while an equivalent increase in gains lowers it by about 21%. Composition-matched simulations show the asymmetry is not a mechanical artifact of building experience from past trades. To explain both the baseline effect and this asymmetry, we build a memory-based recall model in which similarity-weighted retrieval of past outcomes guides selling. A single retrieval channel reproduces all three patterns without shifting preferences.

Abstract: Does repeated elicitation improve measurement or mainly expose how it works? Using 56,050 intertemporal-choice elicitations from 6,808 Japanese households, I freeze 2010–2015 responses and predict 2022–2024 wealth for 3,099 households conditional on 2016 wealth. The displayed-rate mean is negatively associated with future wealth rank and split-block ORIV is steeper. But capping the 40% endpoint at 20% makes the negative slope imprecise; saturated first-response and repeated-share models reject displayed-rate linearity. Post-first responses add virtually no out-of-sample prediction beyond the first category. Repetition improves diagnosis—revealing persistence, endpoint use, and coding sensitivity—more than prediction. Estimates are predictive, not causal.

Abstract: I ask whether elevated retirement non-rollover around job loss is specific to involuntary displacement or also accompanies job separation more generally. I use two regimes of the Survey of Income and Program Participation to measure pension and retirement lump-sum distributions that respondents neither rolled over nor planned to roll over. Permanent layoff is associated with a 1.74-percentage-point increase in the annual non-rollover probability, relative to a 0.79-percent no-layoff mean. On an exact common sample, the permanent-layoff contrast is 1.55 points, whereas the involuntary–voluntary separation contrast is 0.40 points; their difference is 1.15 points (p = .021). I then restrict attention to prior workers who reported a 401(k) or thrift account before the outcome year. Relative to continuing workers, voluntary and involuntary separators have increases of 1.95 and 2.78 points. The incremental involuntary–voluntary difference is 0.83 points, with a 95-percent confidence interval from −0.47 to 2.13. The evidence documents a separation-associated elevation that recurs in both survey regimes, but it does not causally distinguish distribution access from financial distress because the survey neither links a distribution to the job that ended nor observes employer-plan rules.

Abstract: Places industrialized by administrative allocation carry a distinctive demographic profile. In China’s 1982 census, the never-married contrast between ages 20–24 and 30–34 steepens by 4.23 percentage points per standard deviation of industrial composition inside Third Front provinces relative to elsewhere; predetermined siting controls reduce it to 2.24 points. We reconstruct pseudo-1982 profiles from the 2000 census marriage histories using birth-year and completed-age indices. Their differentials are 3.95 and 3.86 points; on identical respondents, the completed-minus-birth-year index difference is −0.08 points (wild p = 0.341). The birth-year-indexed pseudo-minus-actual estimate is −0.28 points, with a wild 95-percent interval of [−2.45, 2.20]. The point estimates align across differently selected retrospective and resident-stock estimands, although the interval permits meaningful differences and neither current-residence design identifies a program effect. In an exact point decomposition, education shares account for 0.59 points, or 15.5 percent of the comparable collapsed gap; the 3.21-point within-education component is imprecise under wild inference (p = 0.118). A negative but imprecise sex-ratio estimate provides little support for a positive male skew. Risk-window-aligned cohort comparisons likewise do not deliver a robust Third-Front-specific campaign shift. The result is a reproducible demographic regularity and an estimand lesson, not an average treatment effect of the Third Front.

Abstract: Do labor-market conditions at school leaving predict what households save for later in life? I link 4,120 Japanese women to job-openings ratios at predicted graduation. Among component-complete observations when women are married, worse entry conditions are followed by more saving with zeros retained. A direct contrast places most of the estimated response in family-common, child, and other rather than spouse-designated flows. Among positive-saving observations, spouse shares fall, while the differential wife–husband response is imprecise. Estimates use 39 national cohorts and are sensitive to persistent-process inference; within-year evidence is mixed. The results extend labor-market scarring to the purposes of household saving.

Preliminary abstract: Does the same investor treat a loss differently when it is hers rather than an algorithm’s? We study 100,000 investors on Alipay’s Ant Fortune platform (September 2020 to September 2024) who simultaneously hold a self-directed mutual-fund portfolio and a robo-advisory account (BangNiTou). Comparing the same investor across two accounts that differ only in who chose the holdings, we find that the response to losses does not weaken under delegation—it reverses sign. On her own account she holds and adds to losers (averaging down); on the delegated account she sells losers and chases winners. A single loss-aversion parameter cannot produce this sign flip. It follows, we argue, from delegation changing what a return means: on her own account a loss is processed through realization utility, while on the delegated account it is both a signal about the algorithm and a source of dissonance she resolves by exiting. We estimate the model structurally and quantify the return investors forgo through self-directed loss aversion. Full empirical results remain in progress.