There have been a number of studies/ blogs published recently on the use of moving averages as a trend following tool.
In short the general conclusion is that if you're looking for good risk adjusted returns, they still hold up pretty well.
Many of these blogs/ studies looked at US markets so I thought I'd add my two pennies worth with some analysis of the FTSE 100 using various moving average crossovers.
In addition, a couple of posts ago, I shared some of my research on the use of Mean Reversion strategies in different market environments. I found that MR strategies had the bulk of their returns at points that the 5 month return on the FTSE 100 was negative.
I wondered if you could improve on this finding by finding a better FTSE filter than 5 month rate of return. So this is a step back before I step forward again with another post looking at trend following vs Mean Reversion strategies.
Without further ado here's how various MA X over systems have fared on the FTSE 100 since 1990. Long only by if the short MA is above the longer MA, if not go to cash. No dividends included. Return on cash a conservative 1.5% per year for the duration of the study. No transaction costs.
Not really any need for the numbers as the charts speak for themselves. Moving average cross overs lag the market, but they do allow you to get out quick enough and get back in quick enough.
Is there any way to improve these results?
As I blogged about here, instead of going to cash, you could trade Mean Reversion strategies. You could also switch to bonds as Mebane Faber suggests here, but I don't have the data for that to test. I'm sure you would have been able to find better than 1.5% APR on your cash pre 2008 as well.
In terms of tweaking various moving average combinations, no doubt, but this will only be possible with a huge dollop of curve fitting so I doubt it would be robust going forward.
However I have spotted something else that works.
Momentum + Value
One way to modulate moving average use is to combine it with a value score like the P/E 10. I've already blogged about this here, have a good read for details of the P/E 10. Effectively you only pay attention to a moving average cross under (i.e. short crossing below long) if the market is over valued. Otherwise you stick to buy and hold on the theory that the market is under valued. Like the housing market, the idea is that things can remain 'expensive' for many years before correcting.
Here I used a P/E 10 value of greater than 20 as a value representing 'expensive'. Shiller's P/E ratio is based on the S&P 500 and US companies, but as you can see, it's value levels are equally applicable to the FTSE 100.
The 10 Month MA
The best results come from using the 10 Month Moving Average in conjunction with value as I found in my original post on the P/E. It must be noted though that one bad month missed can make all the difference here. So I don't think the 10 month MA is particularly magical, but it has done well.
In my original post on the P/E i managed to track a similar approach on the S&P 500 going all the way back to 1881 with ridiculously good results thanks to compounding.
Here you can see the raw 10 month MA vs the use of a value filter with the 10 MA. The data goes back to 1984 this time.
For comparison, here is a similar 5 month return (5ML) filter. This acts like a moving average whereby you leave the market if the return from the last 5 months is negative. Once again the value filter increase returns.