mas financial tools

What does it do?
Perform several three factor regressions (Market, SMB, HML) for a fund, over different time-frames. Graphically display the results.
See: The Cross-Section of Expected Stock Returns
What does it mean (simplified)?
Sample results for 'Vanguard Small Cap Value Index': regression view and style view
There is also a new Gallery that lists the results of other users.
  • alpha: how much 'extra' return did the fund have that could not be accounted for by the model. this is almost never large or statistically significant.
  • B(MKT) = market factor: most 100% stock funds have a market factor near 1.0. higher values may indicate leverage. lower values may indicate cash and or bonds.
  • B(SMB) = size factor (small minus big): positive values indicate the fund's average holding is smaller than the market
  • B(HML) = value factor (high minus low): positive values indicate the fund's average holding is more 'value' oriented than the market (based on book to market ratio)
  • R2: measures how well the fund returns match the model (values close to 1.0 indicate a good statistical fit)
  • +/- value for each factor: a 95% statistical confidence interval
  • p value for each factor: ask a statistician! roughly speaking, lower values indicate that the factor coefficient is more statistically significant.
  • timeseries chart for each factor:
    • red line is based on regression results for rolling sub-periods, centered at a given date.
    • blue line is regression result for entire period.
    • red and blue areas represent confidence intervals.
  • histogram for each factor:
    • red areas represent frequency, based on on regression results for rolling sub-periods.
    • blue line is regression result for entire period.
  • style chart:
    • red is each sub-period's HML and SMB factors (with confidence rectangle).
    • blue is result for entire period (with confidence rectangle - often too small to see).
    • This is oriented to appear similar to Morningstar's style box, but the meaning and scale are somewhat different.
  • Pre-defined FF_LV, FF_LC, FF_LG, FF_SV, FF_SC, FF_SG portfolios are based on "6 Portfolios Formed on Size and Book-to-Market (2 x 3)"
Why would someone use it?
Curiosity mostly. I developed it to satisfy my own curiosity.
Any utility that you derive from it is merely coincidence.
I would not recommend it for any real use, see disclaimer below.
But happy analysis nevertheless.
Are there any known limitations?
Of course, the following tend to not fit the model well (low R2 values):
  • non US stock funds
  • active funds holding significant amounts of non US stocks
  • single stocks and undiversified portfolios (i.e. sector funds)
  • funds holding cash and/or bonds (sorry I do not have TERM and DEFAULT data for the 5 factor model)
  • commodities, futures, and more exotic assets
  • illiquid ETFs that may contain stale end of day prices
Others:
  • the code may be buggy
  • the data may be incomplete or inaccurate
  • my application of the concepts may be inaccurate
  • This hosting arrangement may be slow, bandwidth limited, and may change in the future.
How does it work?
Java Webstart (JRE 1.6, most features will work on 1.5). Download
Warning: depending on the data and parameters chosen, this can be quite memory intensive.

Data from:
Ken French's data library
Yahoo Finance

My own ugly code (hopefully no one is looking here...):
mas.ff3

Additional Java libraries:
JAMA : A Java Matrix Package (public domain)
JSci - A science API for Java (LGPL)
JFreeChart (LGPL)
javastatsoft (GPL)


???
More to come? Maybe one day.


Disclaimer: Do not trust anything here to be accurate or useful. Yadda, yadda, no liability, etc...
I am not a statistician, economist, financial planner, or anything else that would lend credibility to this tool.
But I did stay in a Holiday Inn Express last night!
mas stats