Conformal Predictive Portfolio Selection
JAFEE 2024 Winter Conference.
- portfolio selection
- conformal prediction
- uncertainty quantification
- finance
In one sentence. Wrap any return predictor in conformal prediction intervals, then choose the portfolio from the intervals rather than from point forecasts — a model-agnostic way to let prediction uncertainty enter the allocation.
Prediction uncertainty in portfolio choice
Mean-variance selection balances the mean and variance of returns; quantile-based selection targets tail risk. Both rely on distributional quantities estimated from historical data by some predictive model, and each such model carries uncertainty of its own that the allocation step normally ignores.
The CPPS framework
Conformal Predictive Portfolio Selection forecasts future portfolio returns, computes conformal prediction intervals for those forecasts, and selects the portfolio on the basis of the intervals. Conformal prediction supplies interval coverage without distributional assumptions, so the framework accommodates a wide range of predictive models — autoregressive models, random forests, neural networks — without changing the selection step.
The framework is instantiated with an AR model and validated empirically, where it delivers superior returns compared with simpler strategies.