0
$\begingroup$

I am comparing the performance of 2 investment portfolios.

To simplify the numbers:

After day 1, portfolio A is up 1.5% and portfolio B is up 1.7%.

After day 2, A is up a total of 2% over the 2-day period, and B is up 2.4%.

After day 3, A is up a total of 2.2% over the 3-day period, and portfolio B is up 2.7%.

And so on.

Now I calculate the percentage-point difference in their returns. So after day 1 it is 1.7 - 1.5 = 0.2. After day 2 it is 2.4 - 2.0 = 0.4. After day 3, 2.7 - 2.2 = 0.5.

Then I chart that %-point difference on an Excel scatter chart. After over a year of this, there is a clear trend. So I use Excel to add a trendline.

The chart’s X axis is a simple series of integers representing each day that an entry is added. I update the chart 5 days a week. So the data are complete (not a sample).

The Y axis is the thing I am interested in: the %-point difference in performance of the 2 portfolios over time.

The first entry for the chart is (1,0.0). (The 0.0 is the first number in the table column, and represents the initial zero difference in performance between each portfolio.) The most recent entry is (426,9.9). The Y column has gone up more or less steadily from 0.0 to 9.9.

Is this one of those rare (I think) cases where I should force the trendline's y intercept to 0? If I do that the line would be somewhat different than it is currently.

$\endgroup$

1 Answer 1

1
$\begingroup$

It's a bad practice (and also wrong theoretically) to force the intercept of a trendline to be zero. You want to find the 'best' fitting line to your data points, why would you insert a bias to the fitting?

It has nothing to do with the fact, that in reality, the intercept should be zero. That is an information you know, and not an information you want to get from your trendline. You would like to use your trendline for e.g. forecasting the difference. Forecasting for the starting day is not a valid goal, you know that difference without any trendline.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.