Distributional Reference Class Forecasting of Corporate Sales Growth With Multiple Reference Variables

TL;DR

This paper presents a distributional forecasting method for corporate sales growth using multiple reference variables and rank-based algorithms, improving forecast accuracy and detecting biases in expert predictions.

Contribution

It introduces a novel reference class selection approach with dimension reduction techniques for distributional sales growth forecasting, applicable to large datasets.

Findings

Dimension-reduced variables improve forecast performance

Forecasts closely match actual sales growth distributions

Method detects biases in expert and model forecasts

Abstract

This paper introduces an approach to reference class selection in distributional forecasting with an application to corporate sales growth rates using several co-variates as reference variables, that are implicit predictors. The method can be used to detect expert or model-based forecasts exposed to (behavioral) bias or to forecast distributions with reference classes. These are sets of similar entities, here firms, and rank based algorithms for their selection are proposed, including an optional preprocessing data dimension reduction via principal components analysis. Forecasts are optimal if they match the underlying distribution as closely as possible. Probability integral transform values rank the forecast capability of different reference variable sets and algorithms in a backtest on a data set of 21,808 US firms over the time period 1950 - 2019. In particular, algorithms on dimension reduced variables perform well using contemporaneous balance sheet and financial market parameters along with past sales growth rates and past operating margins changes. Comparisions of actual analysts' estimates to distributional forecasts and of historic distributional forecasts to realized sales growth illustrate the practical use of the method.