Predicting Earnings
motivated by e-mail from James J

There are financial rituals that require estimating company earnings ... next year (or next quarter).

>Such as?
Well, there's Gordon' equation ... associated with the Dividend Discount Model for determining a "fair" price for a stock.

It involves estimating the dividend (or earnings) growth rate and ...

>So you're going to estimate a growth rate?
For earnings, yes. In fact, if we know the Earnings Per Share (EPS) over the last umpteen years (or quarters) and they've increased at a fairly constant rate, then we can fit some exponential curve to the earnings and use that to extrapolate.

For example, if we know the (annual) EPS from year 1994 to year 2003 as in Figure 1, we can fit an exponential curve and use it to predict the EPS for 2004.

>Is it any good ... this prediction?
I have no idea ...

>Then why talk about it?
I was about to say that I have no idea where I can find historical Earnings.
However, if you can find ten years of earnings you can download a .ZIPd spreadsheet:

Just RIGHT-click here and Save target.

Note:
In the spreadsheet you enter ten years worth of EPS (or prices or any other collection of ten numbers) together with the associated Year (example: years 1994 to 2003) and some "future" year (example: 2004, as in Figure 2).

The spreadsheet uses the Excel GROWTH function to get a "best fit" exponential and predicts a future value (as in Figure 1).
(And you'll need POSITIVE numbers in column C, else you're in trouble.)

>Don't you have a real example?
If we find in the annual returns for the S&P 500 from 1991 to 2000 and assume we start with a $1.00 portfolio (in 1991) and stick in the portfolio values (in Column C), then the spreadsheet predicts a portfolio worth $5.23 at the end of year 2001 (as in Figure 3).

>And?
Yes, well ... the actual portfolio would have been $3.48

However, you'll need a set of ten numbers which actually have a near-exponential growth in order to have much faith in the extrapolation.

Besides, the 1990s were GREAT years ... but 2000 and 2001 were lousy.


Figure 1


Figure 2


Figure 3