Thursday, April 21, 2011

The Ivy Portfolio Review and Tests

The Ivy Portfolio by Mebane Faber is a great read and is a staple reference in my library.

Take a look at this chart:
 I'd say that's a pretty decent return on a strategy since 2008. + 36% return vs. -3.6% on the S&P 500.  The above is a simple back test of a strategy from the Ivy Portfolio that I did on etfreplay. It uses the 19 asset Ivy Portfolio which performed better on the moving average model than the portfolios with less asset classes. I want to point out  that this portfolio goes back only to 2008 where market timing clearly had a big advantage. Longer term numbers are decent but not like the above. You can expect to beat buy and hold with less draw down by about a percent per year but not destroy it like the above test may lead you to believe. Probably charts like the above have sparked interest in using moving averages to reduce risk in portfolios. Do they work on lengthier horizons? According to the Ivy Portfolio book the answer is yes. According to my own back testing results the answer is also yes.

There are two main investing ideas in the book - Diversity through asset allocation, and simple market timing to avoid bear markets. Faber goes into detail on how the endowments invest in various asset classes which helps bring more stable returns with less draw downs. (2008 was a bad year for the endowments because essentially all asset classes failed except treasuries. Asset allocation still helped but was much less effective)

The simplest allocation is us stocks, international stocks, real estate, and commodities. (Invest in etfs obviously not rip off mutual funds) The more complex portfolios essentially sub-divide those asset classes into finer cross sections. The discussion on these ideas is very good but of even more interest to me is the timing model Faber promotes in the book.

The timing model is incredibly simple but performs well.  When the asset class is above the 280 day moving average you buy that asset class. If it is below the 280 day moving average you sell that asset class and move into cash. Just the percentage of the portfolio dedicated to that asset class would be in cash. Larry Connors uses the 200 day moving average in virtually all of his swing trading strategies. He's found having the long term trend at your back helps reduce risk. I do want to point out though, that Connors is not a fan of moving cross over strategies. He questions the longer term results Link

Robert Colby's work in his book the Encyclopedia of Technical Market Indicators shows his back testing results on simple moving average systems like this with very positive results going very far back in time vs buy and hold.

Using these moving average systems seems to reduce risk but can be prone to whipsaw in sideways market. There appears to be a fairly simple solution to this problem.

In Faber's models he updates the portfolio once per month. Buys and sells are only done once per month regardless of any signals during the month. At first I thought this was just to make it easy to manage. After testing specifically if it made a difference I found it actually improved performance significantly. The reason is simple. You don't get whipsawed buying then selling the position around the moving average as it moves back and forth. I was surprised that simply buying on the same day each month improved performance so much. There might be even better ways to avoid whipsaw like waiting x amount of days after a cross over to make a position change to make sure the trend change is real but I have not tinkered with that. The monthly management method works well and is far easier to work with.

 Here are some of my own testing results below:

The first chart takes all signals on any day of the month on the $spx index. If the price closes below the 280 day moving average it is sold, and if it crosses above the 280 day moving average the market is bought.

You can't buy shares of the pure index so the dollar amounts are more about comparing the two results than actual returns.
 You can see some evidence of whipsaw eating into profits on the chart. The back test goes too far back to get it all in on this chart though.
 Here is the once a months management method. Performance is much better.

Less whipsaw and fake outs.

Overall I think this is a good system to be using. I'm using it in conjunction with the relative strength models I often post about. You can actually combine the two models but I run them as separate portfolios. 

I want to point out that I've done tests using a 120 day exponential moving average and got good results. It might be worth looking at that time frame as well. 

If you don't have the book, here is link to Faber's paper on market timing relative strength model. link


  1. My own returns bear your test out. From 1/1/08 to yesterday I get S&P -7.75% and me +13.76% (I keep 20% in cash and do other things different from Ivy).

  2. I'm glad you got good results. I think all investors need to have some kind of exit plan from the market given the way government sponsored bubbles keep happening. Let's lock in profits on the ride up.