Thursday, May 13, 2010

More on the math ... start vs. end-of-day treatment

An observant reader identified an error in my math in yesterday's post, which I should have caught immediately. Long ago I realized that outflows should be treated as end-of-day events. Let's consider yesterday's post a bit more.

We started the day with $327,000; there was an outflow of $337,000 (i.e., $10,000 more than what we started with) and we had residual cash of $940. How could this be anything but an end-of-day event? If it was a start-of-day flow, we'd immediately put the account into a deficit, right?

The proper way to think about this is that we began with an investment, it grew in value because of the market, we sold out to capture the gain, and then we sent the client a distribution. The distribution, in essence, has nothing to do with performance and, in keeping with time-weighted returns, its impact should be eliminated or reduced. By ignoring it completely and simply treating it as an end-of-day event, we focus on the growth of the portfolio during the day, which yields a return of 3.35%.

I can't think of a more perfect example to demonstrate why end-of-day makes sense for outflows. More and more firms and software vendors are adopting this approach (and start-of-day for inflows) because they've seen the benefit it has.


Back to yesterday's original post for a moment.  The start-of-day return I came up with was wrong, but because it was one basis point away from the end-of-day return, I ignored it. It should have been suspect because of past problems I've seen, but I was lulled into thinking it was okay. That happens a lot, doesn't it? We tend to look at the outliers for errors, not considering that returns that look right may, in fact, be wrong! The real start-of-day return is -108.87%, which is totally absurd (losing more than we have?). That would have gotten my immediate attention, and I would have recalled the problem with such treatment. Goes to show that just because something looks right doesn't make it so.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.