9 articles · sorted by date
When a model is misspecified, its best-fitting parameter can still point you in the wrong direction. That is the problem with pseudo-true values: they minimize the model’s own objective, but they may not match the quantity you actually care about. This paper studies that gap in a linear population minimum distance problem, where Bayesian decision-makers use priors motivated by the same objective. The authors characterize which sequences of priors make posteriors concentrate on the pseudo-true value, and they find that this convergence is fragile: small changes in the prior can break it. That means pseudo-true values are useful for decision-making only in special cases. The paper also gives a constructive result. It derives simple confidence intervals that guarantee correct average coverage for the true parameter across every prior in the class studied, even when there is no bound on how large the misspecification is. In other words, the paper draws a line between a misleading target and a dependable way to report uncertainty.
A resume keyword can look impressive and still fail in the real world. In one Fortune 500 insurance carrier, data from 10,765 hired agents, 2022–2025, linked three separate systems: an applicant tracking system (ATS, which stores candidate profiles), a human resource information system (HRIS, which stores performance outcomes), and a behavioral assessment. The result was stark. Out of 8,181 unique skills pulled from ATS profiles, 3,597 were testable, and not a single keyword predicted production after correction for multiple tests. Thirty keywords were actually anti-predictive, and the median keyword was tied to 25% lower odds of production. Requiring insurance experience alone would have rejected 2,863 agents who later produced $17.7 million in annual premium credit. By contrast, the behavioral assessment reached AUC 0.647 on its own and 0.735 when combined with the other data. The paper also found that speed-to-production followed a measurable economic constant of $54 per day per agent, or $35 per day after controlling for source channel and tenure.
Balance tests, pre-trend checks, and instrument-validity checks are usually treated like roadblocks before the real analysis begins. This paper argues they should become part of the estimate itself. The authors propose residualizing a baseline estimator against the vector of diagnostic check statistics, which strips away the part of sampling noise that those checks explain. In their framework, that single move can do three things at once: reduce distortions from selective reporting after looking at checks, lower variance without changing the estimand when the original model is correctly specified, and minimize worst-case bias under bounded local misspecification within the class of linear adjustments. They apply the method to the randomized controlled trial in Kaur et al. (2024) and find that, even though all balance checks pass comfortably, residualization makes the baseline point estimate larger and its standard error smaller. In that application, the gain is equivalent to about a 10% increase in sample size. The core message is simple: the information in a diagnostic check does not have to sit on the sidelines; it can help make estimates sharper and more stable.
When prices or debt start climbing faster and faster, the real question is whether the series is just noisy or genuinely self-reinforcing. This paper argues that the answer should come from the observed path itself, not only from a hidden data-generating process. It introduces a descriptive framework for “path-explosive behaviour,” using four diagnostic layers: level geometry, growth rate dynamics, normalised curvature, and log-space behaviour. Those layers are designed to separate true multiplicative growth from I(2) dynamics, and they do so without distributional assumptions or asymptotic critical values. The method also adds two absolute gate thresholds before assigning a composite intensity score. For pairs of series, it checks co-explosive behaviour with a Jaccard co-occurrence index and non-parametric intensity concordance measures. The paper tests the framework in simulation across four DGP regimes and applies it to real house prices, commodity prices, public debt, and Spanish tourism destinations. Its broader claim is simple: in settings where discrete institutional decisions shape growth, a realization-centred description can be more informative than a traditional unit-root test.
When market turbulence hits, the size of price swings can rise together across assets, and that joint motion matters for risk and portfolio decisions. This paper introduces the realized copula of volatility, a nonparametric way to measure how the hidden ups and downs of volatility move together, using high-frequency returns and local volatility estimates. The authors show their estimator is consistent, whether the sample period is fixed or gets longer, and they also derive a functional central limit theorem for the measurement error in the time-invariant marginal copula of volatility. In simulations, the method tracks both the empirical and marginal copula of volatility well, even with only a moderate amount of high-frequency data over a relatively short sample. The accompanying goodness-of-fit test also shows strong size control and excellent power. Applied to high-frequency transaction data from futures contracts tied to the U.S. equity and treasury bond markets, the framework points to a Gumbel copula as a near-perfect fit for the realized variance processes.
When a theory is only partly right, the hard part is deciding how much to trust it. This paper tackles that problem by putting competing restrictions on a model on a price tag: a shadow price that measures how costly each restriction is to the data fit. The authors build a unified framework for weak, noisy, or approximate restrictions, alongside nuisance control covariates, and let the data choose a tolerance level with a Stein-type risk criterion. They also use a debiasing step based on Karush–Kuhn–Tucker conditions, and introduce individual shadow prices to judge the empirical relevance of different restrictions one by one. A plateau rule is proposed to help separate signal from noise. The paper establishes consistency and asymptotic normality for the estimators, and characterizes the individual shadow prices. Simulations and an application to a Solow growth model show how the approach can work in practice when model uncertainty makes simple yes-or-no model selection too brittle.
A business is not worth the same thing if tomorrow’s workers own tomorrow’s profits. That is the core claim of this paper: standard ‘fair market’ formulas price a firm as if current owners will keep collecting future residual profits. The paper says those formulas are usually built from two pieces — the firm’s current net asset value and the present value of expected future profits. But that logic fails for partnerships and employee-owned firms such as a 100% ESOP, an Employee Stock Ownership Plan, or a worker cooperative. In those cases, the future worker-members or partners are the residual claimants at those later times, so the future residuals do not accrue to today’s shareholder-residual-claimants. The result is simple but important: any fair market valuation that assumes those future profits belong to current owners is inappropriate. The paper frames this as a distinction between property rights and personal rights, and argues that the usual valuation formulas do not fit these ownership forms.
Asset prices are not always explained by one market force alone. This paper adds another layer: spatial interactions, a way to let high-dimensional assets influence one another inside pricing models. The authors define a new quantity called spatial rho, a counterpart to market beta, and build a Spatial Capital Asset Pricing Model before extending it to a broader Spatial Arbitrage Pricing Theory with multiple factors. For observable factors, they propose a generalized shrinkage Yule-Walker estimator that uses ridge regression, a stabilizing tool for cases with many variables. When the factors are hidden, they first use autocovariance-based eigenanalysis to extract them, then apply the same estimation strategy. The paper also establishes asymptotic properties for these estimators when both the dimension and sample size grow. Simulated and real-data examples are used to show the method’s effectiveness and usefulness.
If you hedge a stock against the market, you may be blocking the wrong thing. The paper argues that the familiar CAPM beta, which many traders read as “market moves cause stock moves,” does not fit realized equity returns cleanly. The problem is the “aggregator contradiction”: the market return is built from the same stocks it is supposed to explain, unless it is lagged or measured with the stock left out. To sort that out, the author treats CAPM as a structural causal model, a map of cause-and-effect links. In the most plausible setup, an external driver Z pushes both the market return and the stock return, so the market is a proxy rather than a mechanism. In that reading, ordinary least squares beta is an attenuated signal of how well the market captures Z. The paper also shows that “beta-neutral” portfolios can still be exposed to macro or sector shocks, and hedging on the market can import index-specific noise. Using stylized models and large-cap U.S. equity data, the paper finds that contemporaneous betas behave like proxies. Any real market-to-stock effect, if it exists, shows up only with a lag and is economically modest. The practical takeaway is simple: CAPM should be read as associational, not causal.