Historical Financial Statement Information and Future Returns
The eternal search for winning investment strategies underlies the deluge of investment funds, financial experts, and different crafts, otherwise known as valuation frameworks. In a recent seminar, a paper by Joseph Potroski (2002) was discussed with respect to whether historically audited financial information could be utilised to improve future performance. Although the paper does not try to establish an optimal set of predictive ratios, it rationalises that key indicators of profitability, efficiency, leverage and liquidity can be used to eliminate firms with poor prospects from ones portfolio. The author reports that an investment strategy that buys expected winners and shorts expected losers generates a 23% annual return between 1976 and 1996. This strategy is apparently robust across time.
The question that first comes to mind is why should historical information be useful in predicting future returns. If the market is indeed efficient then should that information not already be compounded in market prices. Potroski tries to address this by arguing that the benefits of this analysis generally accrue to small/mid-caps where companies have low share turnover and generally do not share analyst coverage. Potroski is not the first to document this possible market inneficiency. Ou and Penman (1989) and Holthausen and Larcker (1992) both additionally cover an array of financial statement ratios and show that they can accurately predict future changes in earnings and excess returns. Altman's Z, based on 5 scores rather than 9 is also used to model future performance.
So how does one construct the F-score?
1. Profitability
* If ROA (Net Income before abnormals / total assets) > 1 then a company is awarded 1 point, else 0.
* If CFO (cash flow from operations) > 1 then a company is awarded 1 point, else 0.
* If ROA(t) - ROA(t-1) > 0 then a company is awarded 1 point, else 0.
* If CFO > ROA then a company is awarded 1 point, else 0.
2. Leverage and Liquidity
* If LT debt (t) - LT debt(t-1) > 0 then a company is awarded 0, else 1.
* If current ratio (t) - current ratio (t-1) > 0 then a company is awarded 1, else 0.
* If a company did not issue new equity during the year then awarded 1, else 0.
3. Operating Efficiency
* If gross profit margin (t) > gross profit margin (t-1) then a company is awarded 1, else 0.
* If asset turnover ratio(t) > asset turnover ratio(t-1) then a company is awarded 1, else 0.
The maximum (minimum) score that a company can attain is 9 (0). These ratios as mentioned, although chosen arbitrarily are done so with some degree of theoretical (and fundamental) reasoning. For example, key drivers of value include growth and the return on invested capital. These are inherently captured in the profitability and efficiency scores. One could say that this type of study is a substitute for a full-blown fundamental analysis. The fact that we have scores (1 and 0), however, means that we are always going to lose predictive accuracy that we may have attained from a more thorough analysis.
I've undertaken a quick a dirty analysis (click link) of the F-scores between 2000-2010. Where there is missing data (and there is lots of its) or companies did not exist in prior periods they are simply awarded a 0. This analysis isn't perfect by any means and i'll leave it to you to work out whether one can profit from this score. One thing that I think is useful with such an analysis is that it helps us identify companies that may have been overlooked. By doing so we can invest more time on companies that we think are reasonable prospects rather than having to survey the broader index.
The question that first comes to mind is why should historical information be useful in predicting future returns. If the market is indeed efficient then should that information not already be compounded in market prices. Potroski tries to address this by arguing that the benefits of this analysis generally accrue to small/mid-caps where companies have low share turnover and generally do not share analyst coverage. Potroski is not the first to document this possible market inneficiency. Ou and Penman (1989) and Holthausen and Larcker (1992) both additionally cover an array of financial statement ratios and show that they can accurately predict future changes in earnings and excess returns. Altman's Z, based on 5 scores rather than 9 is also used to model future performance.
So how does one construct the F-score?
1. Profitability
* If ROA (Net Income before abnormals / total assets) > 1 then a company is awarded 1 point, else 0.
* If CFO (cash flow from operations) > 1 then a company is awarded 1 point, else 0.
* If ROA(t) - ROA(t-1) > 0 then a company is awarded 1 point, else 0.
* If CFO > ROA then a company is awarded 1 point, else 0.
2. Leverage and Liquidity
* If LT debt (t) - LT debt(t-1) > 0 then a company is awarded 0, else 1.
* If current ratio (t) - current ratio (t-1) > 0 then a company is awarded 1, else 0.
* If a company did not issue new equity during the year then awarded 1, else 0.
3. Operating Efficiency
* If gross profit margin (t) > gross profit margin (t-1) then a company is awarded 1, else 0.
* If asset turnover ratio(t) > asset turnover ratio(t-1) then a company is awarded 1, else 0.
The maximum (minimum) score that a company can attain is 9 (0). These ratios as mentioned, although chosen arbitrarily are done so with some degree of theoretical (and fundamental) reasoning. For example, key drivers of value include growth and the return on invested capital. These are inherently captured in the profitability and efficiency scores. One could say that this type of study is a substitute for a full-blown fundamental analysis. The fact that we have scores (1 and 0), however, means that we are always going to lose predictive accuracy that we may have attained from a more thorough analysis.
I've undertaken a quick a dirty analysis (click link) of the F-scores between 2000-2010. Where there is missing data (and there is lots of its) or companies did not exist in prior periods they are simply awarded a 0. This analysis isn't perfect by any means and i'll leave it to you to work out whether one can profit from this score. One thing that I think is useful with such an analysis is that it helps us identify companies that may have been overlooked. By doing so we can invest more time on companies that we think are reasonable prospects rather than having to survey the broader index.
Comments