Screening - Fuqua School of Business



The Fama & Freedom 3 ½ Factor Model

[pic] [pic]

Submitted by:

Scott Gavlick

Todd Hoskin

Edward Kim

EXECUTIVE SUMMARY

In deference to our name’s sake, Fama and Freedom, we have chosen to approach stock selection from a value-oriented, technical basis – the FAF 3 ½ Factor Model. Four factors were analyzed, two fundamental and two technical in an attempt to capture differing characteristics and potentially differing market signals. First the factors were analyzed independently and then on a combined and scored basis. These factors included Operating Income/Enterprise Value, Operating Income/Net Assets, Momentum (in terms of the 1-month lagged return versus the 13-month lagged return) and Volume Moving Average. For all portfolios a long-short strategy was assessed as well as the simple long portfolio investment strategy. In conclusion we determined that a simple long position in the Operating Income/Enterprise Value Portfolio generated the strongest performance from several difference measurements including highest average return, Sharpe Ratio and holding period return. The Scored Portfolio performed nearly as well while the other independent factors produced results that were materially lower. In conclusion we recommend the position in the long-only Operating Income/Enterprise Value Portfolio.

Screening

We have chosen to limit our universe to small market capitalization US stocks (between $50MM and $700MM in market value). We have also excluded American Depository Receipts so as to only deal with stocks of US companies and thus avoid any currency considerations. We believe that the market for small market capitalization stocks is less efficient than the market for large capitalization stocks. This is due to the lack of analyst and institutional coverage on these entities and limited public information. These companies may also have the potential to grow into large capitalization stocks. If we invest while the companies are small, we may be able to benefit from the subsequent growth of the company into a large capitalization stock.

Factors

Value Play

We felt that value stocks offered better investment returns than growth stocks when investing over a large time horizon. There will be many periods along the way where this relationship will reverse, but in the long run we took that position that the returns from value investing should beat the returns of growth investing. Thus, we used a longer time horizon in our calculation of returns. We use a time series from July, 1986 to December, 2005. This time series has benefit of two major stock market events: the market crash of 1987 and the internet bubble of 1999. From this time series, we can see if our model can survive multiple major market events.

The two measures of value that we used were:

Operating Income / Enterprise Value

Operating Income / Enterprise Value was used because we felt it would be a better measure of a firm’s ability to generate ongoing profits based on the level of capital required than the more readily available earnings-to-price ratio. Operating income also avoids items included in earnings such as extraordinary items and gains from discontinued operations.

Operating Income / Net Assets

We used this ratio as opposed to return on equity to avoid accounting items in operating income discussed above. This ratio has a weakness of extreme results when both operating income and net assets are negative and when net assets is negative but operating income is positive.

Momentum

We calculated momentum as the total returns over one year, with a one month lag (i.e. the period between 13 months ago and 1 month ago from today). The idea is that a stock that has performed well recently will continue to perform well.

Volume

We created a volume moving average crossover by using the average daily volume of the last five trading days of the month and subtracting the 52 week volume moving average. We then divided this number by the 52 week moving average. The idea behind this factor is that unusually high or low volume would translate into positive or negative returns over the next month for the stocks falling into these fractiles.

Approach – The Scored Portfolio

We decided to create equally weighted quintiles based upon our scoring system. We chose to stay with an equally weighted portfolio because we have already set an upper and lower bound on market capitalization for our universe. Stocks in our universe must have a market capitalization of at least $50 million and no more than $700 million. We do not take into account any transaction costs as they can vary from investor to investor and would seriously complicate our analysis. We rebalance each month and remove or add stocks based upon their new rankings in our quintiles.

Our approach was to subjectively score our four factors based upon the results of each individual factor calculated by FactSet. We then created a heat map for each factor looking at returns, standard deviations, and location on the heat map. Based upon this analysis, we came up with the following subjective factor scoring.

See Heat-Maps in the Factor Analysis Sections Below.

[pic]

#1 Scored Portfolio

Our model (long position in the Scored Portfolio) beat our benchmark, the S&P 500, in returns over our time series. From July, 1986 to December, 2005, the S&P 500 returned an annualized average return of 11.6% while our model returned an annualized average return of 25.6%. The $100 invested in our model in December 01, 1986 turned into $7,599 by November 30, 2005. $100 invested in the S&P 500 over this same time period turned into $803. As can be seen in the chart, the return of our top performing quintile also beat the return of the long-short position of top and bottom quintile over this period.

[pic]

[pic]

Scored Portfolio – December 1986 through 2005

[pic]

Scored Portfolio – 1998 through 2005

One of the more interesting time periods over which to analyze the data is 1998 through 2005. As shown in the total return chart below the long-only position in Tier 5 still outperforms the other scored portfolios over the long-term (1998-2005), but over the critical 1998 through 2000 period actually generates a return below that of Tiers 1 and 3 as well as the S&P500. Being able to forecast when the value strategy is appropriate would be extremely profitable – but beyond the scope of our analysis.

[pic]

#2 Operating Income/Enterprise Value

The returns for the quintiles are monotonic with the returns increasing as we reach the top quintile. Our best portfolio, the fifth quintile, pretty consistently performs beyond that of the bottom quintile, implying a consistent return from a long-short strategy. Again, the model does not do well in periods when the market as a whole becomes more focused on growth and value oriented companies underperform (especially 1999 and 2003). As indicated earlier – over the period of 1986 through 2005 the Operating Income/Enterprise Value Portfolio actually performs better than that of the Scored Porfolio.

[pic]

[pic]

[pic]

#3 Operating Income/Net Assets

Again, the returns for the quintiles on Operating Income/Net Assets are monotonic. Our key finding with this factor is that the results – as implied by the heat map – are consistent with those generated by the Operating Income/Enterprise Value assessment. However, while similar they are not as strong. As such less weight was given to the top and bottom quintiles when applied to the Scored Portfolio.

[pic]

[pic]

[pic]

Momentum

Momentum as a stand-alone factor proved to be valuable on a less consistent basis then the first two factors. Nevertheless, on a stand-alone basis performed extremely well when the portfolios consisting of the more fundamental factors (Operating Income/Enterprise Value and Operating Income/Net Assets) performed poorly. This can be seen clearly if December 1998 through December 2000 is isolated and returns compared across factors. But one cannot conclude that this relationship implies predictability.

[pic]

[pic]

[pic]

Volume Moving Average

The volume moving-average factor was the only factor of the four that generated quintiles that were not monotonic. Over the time-period assessed quintile 4 generated stronger performance than quintile 5. The number of years in which this factor was able to predict superior results in quintile 5 was insignificant. However, this factor was very good in predicting those stocks that would perform the worst, even in the critical 1999 and 2003 years.

[pic]

[pic]

[pic]

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download