Stock
Selection in the United States
Introduction
& Summary
The estimated forward twelve month earnings yield (FY1
Yield) was the best predictive factor, generating a 34.44% annualized average return in the top
value weighted portfolio and 8.27% return in the bottom (returns are in
sample). This would generate 26.17%
annualized average returns by purchasing the stocks in the best performing
quintile and shorting those in the worst each month.
Methodology
Our screen was
limited to stocks listed on the NYSE and Nasdaq exchanges with a market
capitalization of greater than $100 million.
The in-sample period was from 1988 to 1998, and the out-of-sample period
was from 1999 to 2003. Screening factors
were chosen based on the economic sense of their predictive power of future
returns. FactSet’s Alpha Tester was used
to gather the data.
The methodology
used is similar to that designed by Campbell Harvey using a sorting method for
quantitative stock selection. Our
methodology allowed us to rank individual stocks with respect to a number of
screening factors, placing the stocks into “buckets” based on the factors. We used five quintiles, placing those stocks
with the highest factor in the sample in the top bucket, and the worst in the
bottom. For example, taking dividend
yield as the factor, the stocks with the highest dividend yields were placed in
the top quintile, those with the lowest dividend yields were placed in the
bottom quintile, and the remaining were placed in quintiles two, three and
four. Next we analyzed the returns of
the stocks in each bucket and determined the predictive capabilities of each
factor using various diagnostics.
Each month the
stocks were resorted based on the factor performance for that month. Finally, we evaluated the cumulative annual
return for each portfolio, looking for a stepdown of returns from the top
portfolio to the bottom. In the cases
where there was a consistent positive spread, particularly in recent years,
between the top and bottom portfolios, we determined the factor to be a good
predictor.
Next we reviewed
the predictive strength of all the factors for the in sample period to
determine a subjective scoring system used to weight each factor. These weights were then used to produce a
trading strategy for the out of sample years.
This is called the Subjective Scoring Model.
Additionally, utilizing particular
quintiles from the single-variable models and a mean-variance optimizer, we
were able to determine the “optimal” weights, or scores, for each quintile,
producing the “Optimal” Scoring Model.
Definitions
Diagnostics
:
- Annualized
average return: The annualized geometric average of
post-rank portfolio total returns over all in sample observation
periods.
- Cumulative
return (indexed at 100 – start): The value of $100 if invested at the first
observation date and compounded over intervening periods (in sample time
period only).
- Standard
deviation of returns: The annualized standard deviation of
post-rank portfolio returns overall all observation periods.
- Average
annual excess return - Rm: The annualized geometric average of
post-rank portfolio excess returns above the market portfolio over all
in-sample observation periods.
- Average
annual excess return - Rf: The annualized geometric average of
post-rank portfolio excess returns above the annualized
- Standard
deviation of excess returns – Rm: The annualized standard deviation of
post-rank portfolio excess returns above the market portfolio over all
in-sample observation periods.
- Standard
deviation of excess returns – Rf: The annualized
standard deviation of post-rank portfolio excess returns above the
annualized
- Systematic
risk (Beta): The slope of the regression line
estimated by regressing the average post-rank portfolio returns on the
relevant market portfolio return over all in-sample observation periods.
- Alpha:
The annualized intercept of the
regression line estimation per systematic risk (beta).
- Coefficient
of determination: The R-square of the average post-rank
portfolio returns versus the market portfolio return over all observation
periods.
- Average
market cap: The sum of all constituent market
capitalizations divided by the total number of stocks in the portfolio.
- %
periods > Benchmark: This is the percentage of periods that
outperformed the S&P 500.
- %
periods > Bench Up Market: This calculation is the percentage of
periods in an up market (where the benchmark return is greater than zero)
that the quintile outperforms the benchmark.
- %
periods > Bench Down Market: This calculation is the percentage of
periods in a down market (where the benchmark return is less than zero)
that the quintile outperforms the benchmark.
- Maximum
positive excess return: The highest single post-rank portfolio
excess positive monthly return above the market portfolio.
- Maximum
negative excess return: The lowest single post-rank portfolio
excess positive monthly return above the market portfolio.
- %
periods positive returns to negative: The ratio of
portfolio average returns greater than zero to those less than zero.
- %
periods of negative returns: The percentage of observations that
returns were less than zero over all in-sample observation periods,
indicative of the historical probability of losing money.
- %
turnover: The percentage of stocks that turnover
each month when the stocks are re-ranked.
- Factor
average: The
arithmetic average of the factor of each quintile.
- Factor
median: The median value of the factor of
each quintile.
- Factor
standard deviation: The standard deviation of the factor
of each quintile.
- Cumulative
annual returns: The value of $100 if invested from
the beginning of the year through the end of the year.
- Relative
performance: A ranking of each of the quintiles’
returns for the year. (5= highest
return, ….1=lowest return)
- Max
# of consecutive negative years: The maximum number of consecutive years
where the annual return is less than zero.
- Max
# of consecutive positive years:
The maximum number of consecutive years where the annual return is greater
than zero.
- Cumulative
returns: The value of $100 if invested two
and five years from the end of out of sample time period. (From 12/31/2001
and 12/31/1998 respectively)
Factors
The stocks were
screened by the factors listed below.
Our economic intuition led us to believe these factors posses the
greatest predictive power of stock returns. We conducted univariate screenings,
and each of these factors was looked at independently of the others.
- 1
yr Exp EPS growth: Expected earnings growth rate. Rolling twelve month expected EPS
consensus minus historical trailing EPS divided by historical trailing
EPS.
- FY1
Chg 3mo.: The
percentage growth of projected 1-Year EPS from that stated 3 months
prior. The estimates are based on
IBES consensus forecasts.
- LTM
EPS Yld: Earnings yield. The last twelve months trailing EPS
divided by the closing market price.
- 3 Yr EPS Growth: The consensus analyst
estimate of the average annual growth rate in earnings per share over the
next three years.
- FY1
EPS UvD Ratio: Up versus down EPS estimate
revisions. Sum of the trailing 12
month upwards FY1 estimate revisions minus the downward revisions divided by
the total number of trailing twelve month estimates.
- Div
Yield: Dividend yield. The last twelve months of cash dividends
divided by the closing monthly market price.
- FY1
Yield: Earnings yield. The estimated forward twelve month EPS
divided by closing market price.
See each
factor’s section for detailed definitions and evaluations of each screening
factor.
Summary
of Factor Performace (Value Weighted)
|
Factor |
Annualized
Average Return of Long/Short Strategy |
Average
Rank of Top Quintile |
Average
Rank of Bottom Quintile |
|
|
(in sample returns) |
|
|
|
1yr Exp EPS
Growth |
-0.74% |
2.63 |
2.81 |
|
3 Yr EPS
Growth |
6.47% |
2.81 |
2.69 |
|
Div Yield |
16.92% |
4.13 |
3.00 |
|
FY1 Chg 3mo |
9.55% |
2.94 |
2.5 |
|
FY1 EPS UvD
Ratio |
4.19% |
3.56 |
2.63 |
|
FY1 Yield |
26.17% |
4.88 |
1.75 |
|
LTM EPS Yld |
19.72% |
4.69 |
2.19 |
|
Subjective
Scoring Model |
17.93% |
4.81 |
1.88 |
|
Optimal
Scoring Model |
16.04% |
4.31 |
1.69 |