Document Sample
CS Powered By Docstoc
					Quantitative Stock Selection

    James F. Page III, CFA
    May 2005
Project Summary

1.   Why Quant Selection is Attractive
2.   Methodology
3.   Historical Back Testing
4.   Model Results
5.   Dynamic Weights / Regime Change
6.   Benchmarks
7.   Next Generation Models
8.   Concluding Thoughts
I. Quantitative Stock Selection
 Quant Stock Selection

(1) In aggregate, certain fundamental, expectational, and
    macro variables may contain valuable information in
    predicting stock returns
(2) Not unlike traditional fundamental analysis, just more
Quant Stock Selection
 Anecdotal evidence suggests ~80% of stock picking is done „by
  hand‟ (individuals making calls on fundamentals)
   Relies heavily on talent (or luck) of individual analyst

   Individuals can only process so much information (sector focus)

   Human nature suggests cognitive biases likely

 Market structure may perpetuate mis-pricings (Street incentives,
  value weighted benchmarks, short sale restrictions)
 Little academic research on subject (trade rather than publish)

 Evidence suggests that investors systematically over pay for
 Quantitative selection is scaleable
Quant Stock Selection
 Black box nature of model

   Explain approach without revealing too much information

   Attribution analysis – must be able to explain performance

 Protecting against common modeling errors

 Credibility of simulated results

 Adapting to individual client restraints
Quant Stock Selection
Market Neutral

   Generate returns from both undervalued and overvalued stocks
   At present, high market valuation = low future returns
   Market exposure is commodity but good stock selection is valued
    (higher management fees)
   Low return expectations combined with geo-political environment
    suggests absolute return approach prudent
II. Methodology
1.   Hypothesize
        Develop candidate list of potential factors that may assist in
         predicting stock returns (valuation, growth, etc.)
        „Priors‟ reduce data mining
2.   Back Test
        Decide on “universe” for testing (capitalization, index, sector, etc)
        Use sorting or regressions to test individual candidate variables
        FactSet‟s AlphaTester currently available to Duke students
3.   Rebalance
        Periodically rebalance portfolios (monthly, annually, etc.)
4.   Analyze Results
        Consider factor performance and consistency (both long and
         short candidates) in predicting returns balanced against
        Select most promising factors for inclusion in the model
5.   Weight
        Once individual factors selected must decide on weights for
         final model by either:
          a)   „Eye balling‟ best factors and assigning weights for a scoring
          b)   Pushing individual factor portfolios into a mean-variance
III. Historical Back Testing
    Historical Back Testing

   Access to reasonably accurate historical data is costly
   FactSet‟s AlphaTester is currently available to Duke students
   Two approaches common in practice
       Regression of factors on security returns (Panel, etc.)
       Sorting universe into fractiles based on factor characteristics (AlphaTester)
   Must protect against common modeling errors
       Survivorship bias
       Information / reporting lags
       Data mining
       Inaccuracies in data
   Credibility of simulated returns is critical
    Historical Back Testing
Term 3 Model Discredited
 Errors in Historical Returns
       Scrub Example.xls
   Survivorship Bias
     Difficult to rule out unless you spend a lot of time examining results

   Fractile Misspecification
     MSFT grouped in F1 Div Yield for 85-04 because of Special
   Betas not believable
     Subject to similar errors as returns information

     Makes market neutral simulation difficult

   Combing factors into comprehensive model increases complexity
 Historical Back Testing
To Mitigate Potential Errors:
 Universe Selection is critical component

    Market Cap weighted

       Adds to turnover (98-00)

       Unstable sector allocations

       Less undervalued firms to buy

    Revenue weighted

       Sector bias

       Less overvalued firms to sell

    Actual Indices (Preferred method)

       Limit universe to actual benchmark

       Limit survivorship bias

       Historical indices available (but option not turned on for Duke)

       Greatly enhance credibility – look to acquire for next year‟s class
 Historical Back Testing
To Mitigate Potential Errors:
 Factor Syntax
    If you do not get this right – data is worthless (lots of opportunities
      to get it wrong)
    Consider consolidating our “approved” syntax for future students as
      starting point
    Expectational (instead of accounting/fundamental) produced
      significantly fewer errors
 Survivorship Bias
    Selecting “Research Companies” does not protect without:

       Appropriate Syntax on Factors
       Correct specification of Universe
           Sanity checks on early period companies
           # of NA companies can be signal
 Errors
    You must clean historical data
    Consider median returns as back of envelope option
 Historical Back Testing
 Use historical indices as universe

   S&P 500

   Barra 1000

 Start with “approved” list of factor syntax

 Clean historical results (particularly returns)

 Do not rely on betas to construct market neutral portfolio

 Research ways to limit reliance on AlphaTester

   Look for other data providers – ask managers what they use

   Interface with CompuStat/IBES directly?

 Once comfortable with model, begin sorting real time ASAP
IV. Model Results
Model Results
Desired Universe: S&P 500

 „Considered‟ to be highly efficient

 Value weighted index suggests low hanging fruit

 Historical data for testing is plentiful and reasonably accurate

 Highly liquid (market impact costs and borrow)

 Very scaleable because of market capitalizations

Actual Universe:
 First choose US Companies with highest sales (~ 500)

 Had to switch to Market Cap because of data limitations
Model Results
Universe Comments:
 Unstable during bubble period (1998-2000)
 Less undervalued firms to buy (but more overvalued firms to sell)
 Sector allocations float with market sentiment

 Rebalanced “official” results annually due to time consuming nature
  of “cleaning returns”
 Equal number of companies in each bucket
 Equal weight returns
 Did not impose sector constraints
 Included two groups of Factors – Fundamental and Expectational
 Actively looking for “Quality” factor to add to the model
 Assume “beta” exposure is equal is both portfolios – probably

Results seem “too good” – further „cleaning‟ necessary
Model Results
Individual Factor Performance
Monthly Statistics 1989 – 2004
Long Factors correlated with Value and visa versa

                           View Portfolios
Model Results
          Fixed Weighting Scheme
Model Results
        Scoring Model Heat Map
Model Results
          Summary Statistics
V. Dynamic Weights /
     Regime Change
Dynamic Weights / Regime Change
   A factor‟s effectiveness may vary in different states of
    nature (PE ratios impacted by inflation)
   Certain market / macro conditions may favor growth or
    value (value was dog in late 1990s)
   Dynamic factor weights allow model to capitalize on
    conditional information
   Few managers currently employ dynamic weighting
   This area “is the Holy Grail” of Quant Strategies
Dynamic Weights / Regime Change
Forecasting Regime Change
 Inflection point for style (growth or value) relative performance

 Used S&P 500 Barra Value and Growth Indices as Proxies

 Examined macro economic variables that might assist in
  forecasting these inflection points
 Two variables demonstrated “promise” in forecasting style relative
  performances over the following year
Dynamic Weights / Regime Change
Regime Change – Factor 1
Dynamic Weights / Regime Change
Regime Change – Factor 2
Dynamic Weights / Regime Change
       The Same Can Be Applied to View Portfolios
   Expectational Factor #2 and Regime Change Factor #1:
           Prediction of Long outperforming Short
Dynamic Weights / Regime Change
       The Same Can Be Applied to View Portfolios
   Expectational Factor #2 and Regime Change Factor #2:
          Prediction of Long Outperforming Short
VI. Benchmarks
Value or Equal Weight?
 Since 1990, EWI has outperformed by 177 basis points

 Turnover for EWI is 6x which begs the question …

   Can we separate turnover between model signals and weighting
Value or Equal Weight?
 Significant Implications for Sector Weights / Tracking Error
Value or Equal Weight?
 Correlations drift through time – implications for tracking error
Value or Equal Weight?
 EWI had positive loading on the size premium

 EWI has significant exposure to the value premium

                     Fama-French Risk Factor Exposures

Value or Equal Weight?
 EWI has 82% correlation with 500 / Barra Growth

 EWI has 96% correlation with 500 / Barra Value

 Further proof of value tilt
Value or Equal Weight?
 Obvious Pros and Cons to both

 EWI benchmark will make returns look less impressive, but
  help explain turnover
 EWI may be a better match for style

 Provide more stable weighting for sector allocations

 Equal weight is newer idea – historical data is limited

 If possible, choice should match weighting scheme of
VII. Next Generation Models
Next Generation Models
   Refining Dynamic Factor Weights
     Preferably done outside of FactSet

   Migration Tracking
     May contain information to enhance returns or limit turnover
Next Generation Models
Modified Versions of S&P 500 Model
 Separate Models for Sector and Stock Selection
 More Conservative
   More positions
   Limited tracking error

 More Aggressive
   Directional

   Less positions

   Leverage

Other Domestic Models
 S&P Mid-Cap 400 / Russell 2000

International Models
 Developed / Emerging markets
VIII. Concluding Thoughts
Concluding Thoughts
 How long will excess returns exist

 How to stay „ahead of the curve‟

 Cost of data

 Credibility of simulation

 Returns during first 12 – 24 months

 Balance between turnover and model signals
Concluding Thoughts
 Quantitative Stock Selection Appealing

 Outperformance Seems Possible

 Long/Short Consistent with Absolute Return Approach
James F. Page III
Jimmy became interested in quantitative stock selection during
   Campbell Harvey‟s Global Asset Allocation and Stock Selection
   class and a follow-up course dedicated to quantitative stock
   selection. He received his Bachelor of Science degree from the
   University of Florida and will receive his MBA from Duke University‟s
   Fuqua School of Business in May 2005. Prior to enrolling at Duke,
   he spent four years in the Equity Research Department of Raymond
   James & Associates in St. Petersburg, FL. He is also a CFA charter