Gyroscopic Investing

Paul,

Sorry to reach you on the forum instead of contacting you directly--I couldn't find an email or form on your mad money site.  I was wondering if you could run the "max drawdown" over a larger sample, perhaps as long as 1972-present.  I thought it was a great addition for assessing risk in various portfolios.  Although 2001-2009 was an excellent start, running the same analysis over a 456 pt sample would be even better for comparison purposes. 

Would that be possible?

Thanks,
Wonk

Wonk wrote: Paul,

Sorry to reach you on the forum instead of contacting you directly--I couldn't find an email or form on your mad money site.  I was wondering if you could run the "max drawdown" over a larger sample, perhaps as long as 1972-present.  I thought it was a great addition for assessing risk in various portfolios.  Although 2001-2009 was an excellent start, running the same analysis over a 456 pt sample would be even better for comparison purposes. 

You can reach me here just fine. BTW: my email address is in the top banner of the web site. Feedback at MadMoneyMachine dot com.

I talked about Max Drawdown on the most recent episode of Mad Money Machine (http://madmoneymachine.com/2010/11/05/m ... -bums-out/). I pointed out that in order to get a more accurate Max Drawdown stat, we need higher-resolution data than the annual returns brought into Simba's spreadsheet. So I created a modified version of the spreadsheet to bring in monthly data from Yahoo! Finance's adjusted historical quotes. Unfortunately, that data source does not go all the way back to 1972. For VTSMX, for example, it only goes back to 1996. Of course, many funds in Simba's spreadsheet use "simulated" returns in order to go back to 1972. I stopped at 12/31/2002. (Oh, and one "glitch" in my spreadsheet is that because of the way the spreadsheet computes a portfolio's returns, the portfolios are rebalanced monthly - hardly something that would be done in an actual portfolio.

To compute Max Drawdown more accurately, we need the higher-resolution monthly data because if we sample at only year end we may not see big drops that occurred and rebounded within the year. The HBPP itself is a good example of this.  In 2008 I'm showing a 1.8% drawdown for it when using only year-end data. But when using monthly data, I compute a drawdown of 12.8%. Of course if we used daily quotes it could only be higher still.

One of the interesting results of looking at Max Drawdown side-by-side with Standard Deviation is that the resulting scatterplots of various lazy portfolios look almost identical. (See http://madmoneymachine.com/2010/10/22/m ... 2001-2009/) And the Standard Deviation plots used only annual data. This gives me confidence that the Standard Deviation plots in Simba's spreadsheet provide a good comparison of risk between the portfolios going back to 1972 even if we cannot easily compute the Max Drawdown for the entire duration.

I can tell you this: I have also modified Simba's spreadsheet to compute Max Drawdown using only the annual data. And from the looks of it, most portfolios had their Max Drawdown in 2008. And it looks like 1974 was a close second with very similar max drawdown numbers to 2008 for most portfolios. Interestingly, using annual data for HBPP its Max Drawdown occurred in 1981 (5.2%) and for PBPP it was in 1994 (3.7%).  Of course this is wildly different than if Max Drawdown were computed monthly. For HBPP and PBPP it was 2008 with 12.8% and 15.8%, respectively (with data only going back to 2002).

Because 2008 was the worst year for most portfolios using annual data, I think the numbers for Max Drawdown using monthly data going back to 2002 probably represent a really good proxy for the values if we could compute them back to 1972. I think the numbers from 2008 are "bad enough" to use when evaluating the risk of the lazy portfolios. Therefore, trying to compute more precise measurements probably add little gain for lots of pain.

As a measurement of risk, I like Max Drawdown over Standard Deviation because it more easily translates into something tangible for the investor. However, it is much more difficult for me to compute.

Paul,

Thanks for the info.  I agree max drawdown is a more useful metric--especially for most investors.  I would think that most investors can stomach a year or two of -5% returns on their portfolio, but most will panic in the midst of a -45% monthly drawdown.

Paul,

I also think expressing risk in terms of drawdown is a great way to illustrate the issue. Browne would often speak of asset risk in terms of the kinds of losses they've had and total portfolio losses. Your description is a very clear way of discussing the issue in realistic terms for investors. 

There was also a max drawdown that was calculated from my 'Total Real Return' data, published here.

The specific max drawdown dates in 2008 are from Mar 18, 2008 - Nov 19, 2008

But, there are a few things to consider when interpreting that (or any other) data:

1) That was just one example of a Permanent Portfolio. Everyone's portfolio would be slightly different depending on when you last rebalanced your portfolio and how you reinvested your dividends. The data set I used simply extended Harry Browne's original data set (1970-2003), which had reinvested its S&P 500 dividends back into the S&P 500. Most PP holders likely would have reinvested their dividends into Cash. (Also, after 2003, I used TLT and SHY in lieu of individual T-Bonds).

2) In that monthly data, I purposefully used the values of the best day in March '08 (Mar 18th) and the worst day in November '08 (Nov 19th) to fully show the max drawdown in the data and chart — so as not to smooth that drawdown over. So, the max drawdown for that particular PP was fairly well demonstrated from a daily perspective.

3) If I run a brand new Permanent Portfolio (starting with 25% GLD, 25% SPY, 25% SHY, 25% TLT) started at the close of March 18, 2008, using Google Finance, with all dividends re-invested to CASH I am seeing a nominal Max Drawdown of -14.88% in 2008. (I don't have time to adjust that for monthly inflation). Of course, using VTI instead of SPY would have shown slightly different results.

The point is.... everyone's Permanent Portfolio is slightly different.

I'll just point out that Harry Browne probably would have described Mar 18, 2008 - Nov 19, 2008 as an extremely unusual situation. There was a run-up before Mar 18, 2008 and a quick rebound after Nov 19, 2008. The only PP holders who would have noticed a 13% or 15% drawdown were those who monitored their PPs daily with with detailed intraday historical charts — something that was never recommended by Harry Browne. Harry Browne recommended only checking the PP a few times a year -- or after cataclysmic events.

I would also mention that back when the PP was started, in the early 70s, there was no easy way to monitor the portfolio so closely. A typical lazy PP investor may have checked their portfolio on November 19th, 2008 and simply seen it was time to rebalance, not realizing that roughly 13%-15% had been lost since the exact moment of 4PM EST on March 18 2008.

EDIT: For those who are interested, the S&P 500 lost 39% of its value from Mar 18, 2008 - Nov 19, 2008. Here was the news report for Nov 19, 2008