The Value-Creation Pricing Factor
Master’s Thesis: Improve the Fama-French Five Factor Model ?
The Value-Creation Pricing Factor is my Master’s thesis for concluding the program of Financial Economics at the Erasmus School of Economics, Erasmus University Rotterdam. It’s a supplement to the Fama-French five factor asset pricing model. Replacing Value with Value-Creation-To-Market-Equity provides a significant reduction of alpha (a significant increase in explanatory power) in regressions on the test assets.
Central to the thesis is the topic of ‘value-creation’. Value-Creation, or Economic Value Added (EVA), is defined as Return On Invested Capital (ROIC) in excess of the Cost of Capital (WACC) multiplied by the Invested Capital. The findings show that historically speaking investors tend to assign a significant premium towards firms that create value in the sense of providing a positive return on capital after deduction of the costs of capital. Since 2008 this premium is disappearing, though during the entire sample period (1963-2018) the premium is still very much significant after controlling for the Fama-French risk factors. The findings are revealing investors’ preferences, and recent trends with respect to these preferences are in contrast with the past. For more information about trends in the Financial Markets, I refer to this post.
The description of returns associated with investment improves when substituting value in the Fama-French five-factor model with value-creation to market equity. Additionally, the description of returns associated with sorting on both investment and profitability improves as well. Value-creation is measured by subtracting capital charges from operating income. Sorting stocks on value-creation-to-market produces a pricing anomaly in the US stock markets over the years 1963 to 2018. With an annualized risk premium of 6.29%, stocks with low market equity relative to value-creation, labeled as “cheap” stocks, outperform “expensive” stocks by a magnitude of 7.45 standard errors.
Productivity and expectations
This research suggests that productivity is in fact the result of a firm’s internal investments with raised capital. The view that profitability originates from internal investment is accompanied by the concept of opportunity costs of capital. Long-lasting productivity is not only the result of sound business management, but also a token of a reciprocal relation between a firm and providers of capital. Central to this relation are expectations. A distinction can be made between constructive and destructive productivity. The latter involves productivity short of investors’ expectations. Value-creation is a measure of a firm’s productivity in excess of the cost of capital. Pricing value-creation in units of market equity can give answers to questions such as: When is a productive asset expensive? Or, when is an unproductive asset disproportionally cheap? I call this priced unit of value-creation “the value-creation pricing factor”.
This brings us to the following research question:
“Does replacing the value factor in the Fama-French five-factor model with the value-creation pricing factor improve the description of returns?”
With a redundant value factor, a four-factor model remains as the benchmark. Why is the addition of value-creation-to-market (V/M) potentially an improvement to the four-factor model that excludes the value factor? The empirical tests of the four-factor model show an inability to correctly describe the returns of small stocks whose returns behave like firms that invest a lot, despite low profitability (Fama & French, 2015). The model assigns a large negative statistically significant intercept in regressions on this class of stocks, implying that expected returns are overstated on average. Could the excessive investment of these small stocks be of destructive nature? Will a negative factor loading to value-creation-to-market (V/M) correctly specify the expected returns? These questions, together with theoretical links with the dividend discount model, are the motivation for adding the value-creation pricing factor to the model.
This study replicates and supplements the paper of the Fama-French five-factor model (Fama & French, 2015). I evaluate several asset pricing models by conducting the GRS test by (Gibbons, Ross, & Shanken, 1989). The results are that the addition of the value-creation pricing factor improves the description of returns for portfolios sorted on (1) Size and investment, (2) Size and value-creation-to-market (V/M), (3) Size, profitability, and investment, (4) Size, V/M and investment, and (5) Size, V/M, and profitability. The factor does not improve the description of returns of portfolios sorted on Size and profitability. A pattern in average returns attributed to value-creation-to-market is not well described by the Fama-French four-factor model. The large and significant difference in average returns on portfolios sorted on value-creation-to-market should be added to the list of asset pricing anomalies. With a t-statistic of at least 6.40 up to 7.45, depending on whether one uses factor definitions based on sorts of 2 x 3, 2 x 2, or 2 x 2 x 2 x 2, it is clear that the factor mimicking portfolio “Cheap-Minus-Expensive” (CME) captures some kind of systematic effect. Given its definition, the effect must be related to profitability, the cost of capital, investment, and investors’ expectations for the future – the latter approximated with the market value of equity.
Further, this research shows that in the presence of the value-creation pricing factor (CME), the investment factor (“Conservative-Minus-Aggressive” or CMA), and the size factor (“Small-Minus-Big” or SMB) become redundant when they are formed on 2 x 3 sorts. For the latter, factor spanning regressions produce an insignificant intercept mainly due to high correlation with the profitability factor (“Robust-Minus-Weak” or RMW), while the intercept of investment becomes insignificant due to high correlation with CME. CME by itself always produces a significant intercept in these regressions, despite high correlation with CMA.
For most of the test assets, the models don’t provide a complete description of returns; the null-hypothesis of the GRS test can only be rejected with at least 5% confidence for the 25 Size and profitability sorted portfolios. Regression details reveal many cases of misalignment between univariate firm characteristics and multivariate regression slopes, specifically for RMW and CMA. The models, and factor mimicking portfolios, repeatedly seem to fail to recognize profitability and investment tilts, especially for small stocks and stocks that are part of a different sort than the factor constituents. This defect explains a good amount of the model failures.