QUAFI Research

The research behind QUAFI

QUAFI Research publishes working papers on finance, behavioral finance and financial econometrics. Each one studies a question that shapes how QUAFI analyzes markets and builds portfolios. All papers are free to read, download and cite.

6 working papers
Latest 26 Aug 2026
Published in English

QUAFI Working Paper series

All papers

LatestWP 2026-0626 Aug 2026

When Is One Asset a Substitute for Another?

By Agustín A. Shehadi C.

Suppose an investor holds an optimized portfolio and wishes to replace one of its constituents with an asset that plays the same role, whether because of a mandate, an access restriction or a preference. Which asset should be offered? The intuitive answer is the one whose inclusion leaves the covariance matrix least disturbed, as measured by a matrix norm. We show that this answer is wrong in three specific and correctable ways. The sharpest is that a matrix norm is sign-blind. It penalises a candidate that strictly improves the portfolio exactly as much as one that strictly damages it. We derive the correct criterion from the envelope theorem, as the first-order change in the value of the optimization problem evaluated at unchanged optimal weights, and we show that it costs no more to compute than the naive one. We then address a limitation that the covariance formulation cannot see at all. A covariance of one-period returns is a single-horizon object, and two assets can co-move at high frequency while their cumulative returns diverge permanently. We propose the h-period tracking variance and its associated spread variance ratio as the horizon-aware criterion, prove that its boundedness in h is equivalent to cointegration of the log prices with a unit cointegrating vector, and argue that this continuous statistic is preferable to hypothesis testing when thousands of candidates are screened. We formalise the selection bias induced by minimising a noisy criterion over a large universe and show that it grows like √2 log M, so that restricting the candidate set helps only logarithmically, a fact that has been used to justify more than it can support. A simulation study confirms that the first-order criterion is correct to O(ε²), that it fails badly at discrete substitutions near a binding constraint, and that it nonetheless ranks well enough for a two-stage procedure to recover the exact optimum.

Financial economicsEconometricsFinance

Read the paperPDF

Where it shows in QUAFI

Derives the criterion behind QUAFI's alternatives screen. A substitute is ranked by its first-order effect on the optimized portfolio's welfare and by its horizon-matched tracking variance, and the shortlist is then re-optimized exactly, so that a matrix distance is never used to decide.

WP 2026-0515 Aug 2026

The Nature of Volatility Across Financial Instruments

By Agustín A. Shehadi C.

We study the distribution of returns for 4,334 instruments in six asset classes, namely US equities, exchange-traded funds, US and Argentine fixed income, Argentine equities and Argentine depositary receipts over foreign shares, using 9.2 million daily observations over ten years. Three findings organise the paper. First, the estimated tail index lies between 2.3 and 3.3 for every class. The variance therefore exists everywhere and the fourth moment nowhere, which renders the F-test, the χ² variance test and the Jarque–Bera test inadmissible on these data. We report them alongside their robust counterparts to show the difference. Second, tail heaviness and volatility are distinct attributes that order the classes differently. The two fixed-income classes have the lowest volatilities and the heaviest relative tails, while the most volatile class in the sample has the lightest. Argentine hard-dollar sovereigns dominate US equity on mean, volatility and Sharpe ratio at the same time while carrying the panel's lowest tail index, so that the mean–variance ranking and the tail ranking point in opposite directions. Third, aggregation thins the tails monotonically without delivering normality by one month, and the mean reversion that would justify horizon-dependent allocation is modest where it exists, with six-month variance ratios of 0.71 to 0.77 for equity-like classes, and absent in Argentine sovereigns. Accounting for the observed distribution rather than a Gaussian match reduces the optimal US equity allocation by 23% for a log-utility investor, and the welfare cost of the Gaussian assumption rises with risk aversion. We separate inference at the pooled and instrument levels throughout and show that the two can disagree sharply. The comparison of US equity against Argentine fixed income yields p ≈ 10⁻⁹⁰ pooled and p = 0.42 across instruments. We argue that the second is the defensible number, and we draw the implications for how risk and horizon profiles should be constructed.

Financial economicsEconometricsFinance

PDF

WP 2026-041 Jun 2026

Beyond Expected Utility: Decision Theory at the Frontier of Financial Modeling

By Agustín A. Shehadi C.

Through the no-arbitrage Euler equation, every asset price is a statement about preferences. The price of a payoff is its expectation under a stochastic discount factor built from the marginal utility of the investor who holds it. This survey maps how the field has moved beyond the expected-utility core of von Neumann, Morgenstern and Savage in response to the systematic ways in which real markets and real investors depart from it. The departures are organized into four breaks. The first concerns risk attitudes. The Allais paradox falsifies the independence axiom and motivates rank-dependent, prospect-theoretic, disappointment-averse and reference-dependent utilities, with consequences for the equity premium and for the pricing of skewness. The second concerns ambiguity. The Ellsberg paradox falsifies the existence of a unique subjective prior and motivates Choquet, maxmin, variational and smooth ambiguity preferences, together with robust control. The third concerns time and custom. Discounted utility fails dynamic consistency, which motivates recursive utility and habit formation, the latter being the formalization of custom as a moving standard of consumption. The fourth concerns frictions. Liquidity and asymmetric information break the representative, fully informed agent and dissolve the notion of a single marginal utility. Throughout, the unifying object is the pricing kernel. Each behavioral feature or friction is a precise modification of it, and each is judged by the asset-pricing puzzles it does or does not resolve, above all the equity-premium puzzle and the Hansen–Jagannathan volatility bound. The paper closes with the open problem that ties the frontier together, which is the identification of preferences from data.

Financial economicsBehavioral financeFinance

PDF

WP 2026-031 May 2026

The Statistics of Return Forecasting: Heavy Tails, Estimation, and the Limits of Predictive Inference

By Agustín A. Shehadi C.

Forecasting asset returns is hard for reasons that are, at bottom, statistical. This paper gives a self-contained and mathematically rigorous account of why, and of the estimation theory that the difficulty demands. It formalizes the forecasting problem for a stationary, ergodic return process, decomposes predictive risk, and shows that the predictable component of returns is small relative to irreducible noise, so that estimation error governs out-of-sample performance and the finite-sample bias of predictive regression is of first order. It then develops the two technical obstacles that distinguish financial data from textbook settings. The first is serial dependence. Valid inference about forecasts requires limit theory for dependent sequences, variance estimators that are consistent under heteroskedasticity and autocorrelation, and the modern apparatus of predictive-ability testing. The second, and the central theme of the paper, is heavy tails. Returns are regularly varying with a finite tail index, so high moments may not exist, classical central-limit arguments and sample-moment estimators become fragile, and one must turn to extreme-value theory and to tail-index estimation. Particular attention is given to empirical confidence and prediction intervals. When asymptotic normality fails, validity must be recovered by resampling that respects both dependence and heavy tails, and by distribution-free conformal prediction with finite-sample coverage, extended to the non-exchangeable and non-stationary regime that finance inhabits. The paper closes with a workflow and a set of open problems. Its unifying message is that predictive inference on returns is a problem of estimation under dependence and near-infinite-variance tails, and that its solution is a disciplined combination of robust estimation, resampling and distribution-free coverage.

Financial econometricsFinancial economicsFinance

PDF

WP 2026-021 Apr 2026

The Behavioral Foundations of Asset Prices: Investor Psychology, Limits to Arbitrage and Market Anomalies

By Agustín A. Shehadi C.

Classical finance assumes that investors are rational optimizers and that competition drives prices to fundamental values. Decades of evidence, from the excess volatility of stock prices and the size of the equity premium to long-horizon overreaction, strain that assumption. Behavioral finance responds by grounding asset pricing in two ideas with empirical support. Real investors are subject to systematic psychological biases, and the arbitrage that is supposed to correct the resulting mispricings is itself limited. This survey, written for a numerate reader with no specialist training, develops both ideas from first principles and shows how they combine into testable theories of market behavior. It reviews the core biases, namely heuristics, loss aversion in the sense of prospect theory, mental accounting, overconfidence and the disposition effect, and traces how each maps to a market phenomenon. It then presents the three canonical behavioral models that reconcile short-run momentum with long-run reversal, and the sentiment-based view of the cross-section of returns. The efficient-markets rebuttal is given its full weight. Anomalies split roughly evenly between over- and underreaction and are fragile to methodology, a caution reinforced by the multiple-testing and replication evidence of the past decade. The paper closes with the adaptive-markets synthesis and an agenda for research, including the practical question of how a disciplined quantitative process can help ordinary investors act against their own biases.

Behavioral financeFinancial economicsFinance

PDF

WP 2026-011 Mar 2026

Predicting Alpha: A Survey of Return-Forecasting Signals and Their Evaluation

By Agustín A. Shehadi C.

An alpha signal is a quantity that can be computed from information available today and that helps predict the part of an asset's future return which is not compensation for bearing known risks. The search for such signals is the central activity of quantitative investing, and the same statistical machinery that uncovers a genuine signal will, used without care, manufacture many false ones. This paper surveys what the empirical asset-pricing literature has learned about predicting alpha, for a reader with quantitative training but no specialist background in the field. It builds the definition of an alpha signal from first principles and separates a return predictor from the risk-adjusted edge it may contain. It maps the four areas in which predictors have been documented, namely the cross-section of firm characteristics, the time series of the market premium, the microstructure of trading, and the dynamics of volatility. It then sets out how a candidate signal should be evaluated. The requirements are out-of-sample accuracy, economic significance after trading costs, corrections for the number of signals tried, and calibrated statements of uncertainty. The paper closes with the open problems, among them non-stationarity, the decay of signals after publication, the low signal-to-noise ratio at daily horizons, and the integration of machine learning with multiplicity-aware evaluation. The main lesson of the literature is that measurement, and not discovery, is the scarce skill.

FinanceFinancial economicsBehavioral finance

PDF

See the research at work

The findings of these papers shape QUAFI’s optimizer, its risk profiles and its alternatives screen. The app is free to use during the launch.

Try QUAFI Visit quafi.io