Analysis · Finance
A Winning Stock Model Can Still Lose on Your Benchmark
By AWEI · AI-compiled · Published · Analysis prepared · 1 source · doi.org
Adding wage information improved a stock return model for 22 portfolios but weakened its adjusted fit for 10 others. That split, obtained by counting entries in a 2023 Pakistan study’s comparison table, is more revealing than its headline conclusion that the six-factor model outperformed its competitors.
Eleftherios Thalassinos and colleagues studied 173 non-financial firms from June 2010 to June 2020. They formed 32 portfolios using company size, book-to-market ratio, profitability, investment and a labor measure. Regressions compared how different combinations of factors accounted for monthly portfolio returns above a three-month Treasury bill benchmark.
The crucial comparison adds one factor to an existing five-factor model covering market exposure, size, value, profitability and investment. The extra factor compares returns of portfolios classified by low versus high labor income growth, using salaries and wages payable from annual reports. It is a constructed return spread, not a direct measurement of employee skill or productivity.
The authors ranked models using adjusted R-squared, a measure of historical fit that penalizes additional explanatory variables. In their comparison table, adding the labor factor raised one portfolio’s score from 33.11% to 45.25%, while another fell from 39.15% to 38.62%. The uneven results mean that a model favored overall need not be the strongest benchmark for every portfolio.
These are historical associations. They do not establish that changing wages causes stock returns to change, and the paper reports no separate forecast test on unseen data. Its scope is also narrow: one country, non-financial firms, exclusion of negative book-to-market ratios, and a sample the authors considered insufficient for conventional portfolio constructions.
The practical lesson is to match a model comparison to the portfolio being assessed. Here, the useful question is how much the added factor improves that portfolio’s historical benchmark. Whether it improves future forecasts remains unresolved.
Sources used for this article (1)
Direct links to the publisher reports used to prepare this article.
- Source 1
- A Comparison of Competing Asset Pricing Models: Empirical Evidence from Pakistan Eleftherios Thalassinos, Naveed Khan, Shakeel Ahmed …
