Colin Priest | Searcharxiv

arXiv2026

Portfolio-aware scenario transfer of reserve movements: evidence from Lloyd's syndicates

A claims-development movement on one portfolio may not transfer directly as a stress to another size. We estimate how to transfer such scenarios between portfolio sizes, using 632 syndicate-years from 117 Lloyd's syndicates over 2015-2024. The measured movement is the change in gross claims estimates on selected older underwriting cohorts, divided by opening gross claims outstanding plus reserves taken on in the year whose development it includes; it is not a pure measure of opening-reserve deterioration. Its dispersion falls materially with reserve size, and a size-based scale predicts held-out syndicates far better than a single market scale. But the transferred upper tail understates the smallest books: in the smallest reserve-size quartile, six of 158 held-out observations exceeded the transferred 99.5th percentile, against 0.79 expected, from three syndicates. Square-root scaling has descriptively lower held-out pinball loss at 99% and 99.5%, though more misses among the largest books. For a GBP 500m portfolio the transferred empirical 99.5th percentile is 0.34 of reserves (95% credible interval, conditional on the fitted model, 0.23-0.52), against 0.67 in the raw market pool. Both are Syndicate 5820's 2016 movement. The contribution is inspectable scenario comparability, not a calibrated capital quantile or a next-year forecast.

q-fin.RM↗

arXiv2026

Same Facts, Different Remedy: Linguistic Profiling and Arbitrariness in LLM and System One Decisions

Large language models are increasingly used to resolve consumer complaints, and the way complainants write varies with age, education and socioeconomic status (SES). We audit three remedy-decision systems using sociolinguistic rewrites of US consumer financial complaints that are instructed to preserve every fact: two generative LLMs that return a single label (GPT-4o-mini and GPT-5-mini) and a probabilistic "System One" model that returns a distribution over remedies (Jev). Using pre-registered hypotheses tested once on held-out complaints, we separate two phenomena that single-label audits conflate. At the group scale, older, college-educated and higher-SES styles receive slightly more generous remedies; these effects replicate but are small, with every monetary-remedy gap within 0.5 percentage points. At the individual scale, two style versions of the same complaint receive different monetary decisions 2.6-6.3% of the time. In a variance-component analysis of the probabilistic model, persona components shared across the two rewrite models account for about 1.6% of within-complaint variation. The persona gaps are associated mainly with delivery (vocabulary, grammar, politeness) rather than content. Positive controls show that all three systems respond 49-387 times more strongly to remedy-relevant facts than to style. Rewriting every complaint into one standard style reduced GPT-4o-mini's education and SES gaps by 94% and 87% while retaining fact sensitivity, but reversed two of the probabilistic model's gaps and did not make individual decisions more stable. Independent LLM auditors flagged material fact changes in 3.3% and 11.7% of rewrites, and instability persisted among unflagged rewrites. Fairness audits of LLM decision systems should report group disparities and individual instability as separate outcomes.

cs.CY↗