Extending Profitability to 1927-1953

In this paper we conduct an out-of-sample test on gross profitability and investment in the same manner as Davis, Fama, and French (2000) for the pre-Compustat period (1926-1955). We hand-collect financial statement data from Moodys Industrial Manuals using the company PERMNO list first created by DFF. In total, we collect data from 1,291 firms, largely industrial firms but with some utilities. We then run Fama-Macbeth (1973) regressions using gross profit, scaled operating profit, scaled net income, and investment along with existing variables like book-to-market, market equity, one-month reversal, and one-year momentum. We find that the premiums on gross profitability and investment are not significant for any part of our sample period. For the overall sample period as well as the first half (before the 1933 Securities Act), our accounting data is often missing or cross-sectionally inconsistent. Despite the better-quality data in the period after 1935, however, neither gross profitability not investment have significant Fama-Macbeth slopes. We believe this is caused by inconsistent and incomplete accounting data, chiefly the number of firms that combine SG&A and COGS data into one "cost" number and the inclusion of investment-like costs, like R&D, in COGS or SG&A. This causes gross profitability to not reflect direct economic profitability as closely as in prior research. However, net income has significantly positive coefficients during this period and is not subsumed by gross profitability; this contradicts prior research for the post-1962 period. More data cleaning and analysis is needed in order to form firm conclusions on the gross profitability, net income, and investment premiums during this period.