Monday, July 11, 2016

Investing: Humans or Machines?

I ran a test to try and compare human money manager strategy against machine money management strategy. It wasn't scientific and it had a handful of holes, but, I looked at 14-months of returns (from 02/2015 to 06/2016) for both the professional (1'ish % fee-level) human managers and Wealthfront's machines (on level 7 in terms of aggression). The comparison was of public equity portfolios only, with roughly the same "international"/US "mix." I calibrated the humans and the machine with roughly the same risk tolerance levels.

The humans handily beat (nearly double the machines' return percentage) the machines most (65%) of the time. When the machine's did win, they ok; 50% better than the humans.

I did not compare tax loss harvesting which is important. My gut tells me though, based on just looking at the numbers peripherally, that the machines win in a big way here. Harvesting algorithms are pretty good if you don't involve human discretion, _but_ they can get into nasty trouble in down markets, and this is where another interesting bit of data came out.

When you look at the overall market nastiness that occurred over the past year, the humans destroyed (sometimes as much as 5x better) the machines in particularly ugly months. The implication here is right in line with my hypothesis, as well as the marketing material human money managers spew, that humans are smarter when markets start to tank. Put another way, when markets tanked in a given month, the losses the machines took were much larger than those the humans took.

I ignored fees. You can't argue with the fee savings incurred when machines manage money. Machine money management companies like Wealthfront charge nearly a quarter on average of what you pay humans to do.

Result: you're paying humans more to protect your money in down market scenarios. If the machines can get smarter in down scenarios (seems really hard to solve this), there's no need for humans to manage money anymore.