No. of Recommendations: 5

Lee:
Thanks for the illustrative example.
To make the intent clearer, I will re-write the BI formula as:
Downside = 10Yr_Avg_Low * (2004_BV/10Yr_avg_BV) * (2004_ROE/10Yr_Avg_ROE)
In other words, the downside value is the 10 year average low,
normalized by BV and ROE.
What I found particularly striking about your example, however,
is that the BV for PFE has been rising very rapidly. So, the BI
formula would lead to a corresponding rapid increase in the downside.
This is perhaps as it should be.
However, I asked myself -- what if I smoothed the values somewhat,
either because I don't think that the spike in BV is real (is that
possible), or just because I wanted to be conservative and assumed that
the stock price would rise more gradually to catch up to intrinsic
value as defined by the BI formula?
So, I tried a little experiment ...
I computed a moving average of sorts, where the 10-year average is the
average of 1993, a weight times 1994, weight_squared times 1995, and so
on -- in essence, a lagged average where the oldest data is emphasized,
and the newest data is de-emphasized. I hasten to add that this is
counter-intuitive, and driven strictly by my desire to de-emphasize the
recent spike.
I would have preferred to "test" this by predicting last couple of
years' numbers, but since I did not have 1993 and 1992 data, I
approximated them using 9-year and 8-year lagged averages.
Here are the results using straight averages:
2002 2003 2004
Actual_High 42.50 36.90
Upside 69.98 70.04 73.18
Downside 43.11 42.88 45.99
Actual_Low 25.10 27.90
Notice that 2004 prediction for downside is $45.99, consistent with
your calculation. Downside for 2003 was below actual low, but above
actual low for 2002.
Here are the results using lagged averages:
2002 2003 2004
Actual_High 42.50 36.90
Upside 47.58 46.25 48.45
Downside 29.01 28.10 30.00
Actual_Low 25.10 27.90
This would predict a downside of $30 for 2004! Downside for 2002 and
2003 is close to actual lows.
My sense is that this is a classic example of curve-fitting :-)
At the very least, I would guess that the weighting factor would
have to be adjusted for each stock individually.