Feasible Implied Correlation Matrices from Factor Structures

TL;DR

This paper develops methods to construct feasible implied correlation matrices from factor models, ensuring mathematical and economic validity, with empirical validation on US stock index options.

Contribution

It introduces two approaches—quantitative and economic—to generate feasible implied correlation matrices from factor structures, addressing limitations of existing models.

Findings

The nearest correlation matrix approach effectively estimates feasible implied correlations.

Economic translation of factor-based expected correlations is feasible and consistent.

Empirical tests demonstrate the methods' applicability to real market data.

Abstract

Forward-looking correlations are of interest in different financial applications, including factor-based asset pricing, forecasting stock-price movements or pricing index options. With a focus on non-FX markets, this paper defines necessary conditions for option implied correlation matrices to be mathematically and economically feasible and argues, that existing models are typically not capable of guaranteeing so. To overcome this difficulty, the problem is addressed from the underlying factor structure and introduces two approaches to solve it. Under the quantitative approach, the puzzle is reformulated into a nearest correlation matrix problem which can be used either as a stand-alone estimate or to re-establish positive-semi-definiteness of any other model's estimate. From an economic approach, it is discussed how expected correlations between stocks and risk factors (like CAPM, Fama-French) can be translated into a feasible implied correlation matrix. Empirical experiments are carried out on monthly option data of the S\&P 100 and S\&P 500 index (1996-2020).