How these papers were chosen among the week's 103 papers

All 103 papers of the week whose main category is in this field were read, title and start of the abstract, and 25 were kept. They were then ordered by what can be counted: how many research profiles of our readers they come close to, whether a rising topic of the field includes them, and how many categories they were posted in. The abstracts of the first 25 were read in full to choose these, no more than two from one category and one per team.

The 16 numbers of the text were checked against the abstracts. 20 statements on what the papers found were looked for word for word in their abstracts.

Sources: arXiv.

Favoritism with a competitive purpose #

A sympathetic judge might simply appreciate a compatriot’s style. A strategic judge would also consider whom that athlete needs to beat. A new figure-skating preprint finds patterns consistent with the latter: judges award a compatriot premium while penalizing close rivals, and the rival penalty grows with the compatriot’s placement chances before fading when she falls out of contention.

The researchers use judging data and variation in a randomly drawn start order. The compatriot premium declines along that order; before 2022/23, it also increases when close rivals have judges of their own nationality on the panel. A counterfactual analysis suggests these biases can alter close competitions. The broader lesson concerns expert panels everywhere: favoritism may depend not just on personal ties, but on the competitive situation in which experts make their decisions.

Why electricity sometimes costs less than nothing #

Electricity generators can sometimes face negative wholesale prices, effectively paying to supply power. Production subsidies are a familiar explanation: a subsidized generator may still benefit from producing when the market price turns negative. Using U.S. wholesale electricity data for 2000-2023, a new preprint reports causal evidence that the federal production tax credit contributes to this phenomenon. Increasing eligible wind generation by 1 percent raises the probability of a negative energy price by 3.4 percentage points on average.

But the subsidy is far from a complete explanation. At least 18 percent of negative-price hours cannot solely be attributed to the credit, and the estimated effect explains only 10 percent of observed negative prices that could potentially be attributable to it. Effects also differ sharply across markets. The findings caution against treating negative prices as a simple verdict on renewable-energy subsidies; the authors also use their estimate to calculate a plausible value for environmental benefits when prices are negative.

Electric-car incentives are easier to remove than replace #

New Zealand’s Clean Car Discount offers an unusually clear look at both the arrival and departure of a purchase incentive. Researchers examine monthly vehicle registrations from 2014 to 2024, identify breaks around policy dates, and compare New Zealand with Queensland using a difference-in-differences design. After rebates began, electric-vehicle registrations rose strongly, but their average growth relative to Queensland was not statistically significant. A specification allowing the growth trend to change suggests more gradual acceleration.

Repeal looks different: registrations fell by about 72% relative to Queensland, with the decline persisting in 2025. Purchases surged ahead of both policy dates, while average emissions from newly registered vehicles fell during the scheme and rebounded afterward. The preprint suggests that withdrawing incentives abruptly can be disruptive, although road user charges introduced for electric vehicles shortly after repeal complicate interpretation.

Faster trading, thinner markets #

Speed is not necessarily a free improvement in market design. A blockchain-trading preprint develops a model in which faster blocks reduce competition over priority fees, shifting profits from the block producer to arbitrage traders. Liquidity reaches its highest level at an intermediate speed; beyond a threshold, an active liquidity provider can avoid unfavorable trades while passive capital retreats.

Evidence from the Base Flashblocks upgrade, a tenfold increase in block speed, supports those predictions. Priority-fee bids fell, arbitrageurs’ net profits rose, and the active liquidity provider rebalanced faster and earned more. Yet pool depth declined and effective spreads widened, even as prices tracked external markets more closely. The result separates two qualities often bundled together: accurate prices and inexpensive trading need not improve in tandem.

When a new estimator changes the conclusion #

Many policy studies compare outcomes before and after different groups adopt a treatment at different times. A common estimator, two-way fixed effects, can assign negative weights when treatment effects vary. A new preprint asks how much that problem matters in published business research, using language-model agents to help identify eligible analyses and then replicating and re-estimating 355 specifications from 40 papers.

The alternatives do not merely fine-tune the published results. Among comparisons passing a leverage screen, estimates change sign in 25.1 and 18.0 percent of comparisons for the respective modern estimators. Their standard errors are also larger. Including comparisons that fail the screen leaves the overall picture essentially unchanged. For readers outside econometrics, the practical message is that an apparently technical choice of estimator can change both a finding’s direction and its apparent precision.


These preprints share a concern with what lies beneath an observed result. A judge’s score can reflect strategic competition; a subsidy’s effect depends on timing and market conditions; faster infrastructure can redistribute gains rather than improve trading for everyone. Even the statistical machinery used to uncover such effects can change the answer. Understanding incentives therefore requires examining both the rules people respond to and the methods researchers use to measure those responses.

Was this letter useful to you?

Written by AI from the 141 papers of the week in this field. Most are preprints, not yet peer reviewed: the text says what their authors report.

Get one letter by email every week with a free account, all of them with Premium. Create a free account