Update: how to use the Macro Index for trading and investing


Now that I’ve completed the first version of the Macro Index, I am thinking of ways to use it. I first explained how to potentially build a trading model around the Macro Index here.
[wpdatachart id=79]
Garrett’s comment reflected what I had in mind when I started building the Macro Index in November.

I am a little bit concerned about the data-mining component of the Medium-Long Term Model, but I am not significantly concerned.
Ray Dalio in his book Principles said

Investment systems built on machine learning that is not accompanied by deep understanding are dangerous. This is data mining.

The Medium-Long Term Model’s components are not data mined, because they inherently make sense. They are accompanied by “deep understanding” (e.g. yield curve inversion). HOWEVER, it’s the binary component part that I am slightly concerned about. E.g. in the past, every bear market happened after the 10 year – 3 month yield curve inverted. What happens if this bull market ends when the 10 year – 3 month yield curve is at 0.1%? You can see how having a hard binary rule (all-or-nothing) can be dangerous, because it doesn’t take into account risk:reward. Gradients are safer.
This is why Ray Dalio prefers to scale in. For example, let’s assume that your target is to “sell and shift to 100% cash when the 10 year – 3 month yield curve inverts”. Instead of staying 100% long until the 10 year – 3 month yield curve exactly inverts, why not:

  1. Shift to 80% long when the yield curve is at 0.4%
  2. 60% long when the yield curve is at 0.3%
  3. 40% long when the yield curve is at 0.2%
  4. 20% long when the yield curve is at 0.1%
  5. 0% long when the yield curve is at 0%

This example is just for illustration purposes, but you can see how scaling in and scaling out makes more sense. It’s not an “all or nothing” trade.
So with that being said, I am going to split the Macro Index into 3 different models. Here’s what our models are going to look like in the future.

  1. The Macro Investing Model is suitable for peoples’ retirement accounts if you’re holding non-leveraged SPY. This is for accounts that make a position change infrequently, and merely want to avoid the bulk of bear markets (which occur in the context of economic deterioration).
  2. The Macro Trading Model is for traders who want to make a position change once every few months. There are 2 versions: a long-only version and a long-and-short version. The long-and-short version includes short positions during bear markets, whereas the long-only shifts to 100% cash during bear markets. While returns will be higher for the long-and-short version, I will use the long-only version. In bear markets, the name of the game is capital preservation. The real money is made from bull markets.

*It can be very dangerous if you short too early. Much more money has been lost on the short side than on the long side. Never doubt the market’s ability to be irrationally bullish in the medium term.
Personally, I will use a combination of the Medium-Long Term Model and the Macro Trading Model: Long-only in the next bull market.
The current bull market is so aged that even though I still think the stock market will go higher, the long term risk:reward isn’t worth it. Hence why I am not using these models to trade right now.
In order to trade properly, you need to use a strategy from the beginning to end of a bull market and bear market cycle. Starting with a strategy near the top of a bull market isn’t great for long term risk:reward.
As I mentioned, the simplest version of the Macro Investing Model yields an average of 10.2% a year when traded with $SPY.
*When running backtests, you need to adjust for beta and leverage. That’s why we run all backtests on $SPY first before progressing to $SSO and $UPRO.

Previous version of the Macro Investing Model

This was the initial version of the Macro Investing Model, as explained here.

Sell your $SPY when the Macro Investing Index falls to <= 0.5 (go from 100% long $SPY to 100% cash)
Buy back your $SPY when the S&P 500 makes a “golden cross”, whereby its 50 day CLOSE moving average rises above its 200 day moving average (go from 100% cash to 100% long $SPY)
Once a “golden cross” occurs, wait for the Macro Investing Index to rise to >=0.8 before you start monitoring the Macro Investing Index again for a dip <=0.5

From May 1968 – October 2018:

  1. Buy and hold $SPY yields an average of 6.8% per year
  2. This version of the Macro Investing Model yields an average of 10.2% per year with $SPY


The new and improved version of the Macro Investing Model is based on the idea that we don’t want to use binary indicators. Let’s add several “scale in and scale out” components to reduce risk.
I have created 3 Tiers for the Macro Index:

  1. Tier 1: when the Macro Index falls to <= 0.5
  2. Tier 2: when the Macro Index is between >0.5 and <= 0.6
  3. Tier 3: when the Macro Index is between >0.6 and <= 0.7
  4. Tier 4: when the Macro Index is >0.7

Remember what I said in this post:

0.5 for the Macro Index = the long term bearish sign.
0.6 for the Macro Index = the long term warning sign.

The new version of the Macro Investing Model incorporates the 0.6 warning sign.

When the Macro Index is in Tier 3 or Tier 4, be 100% long the S&P. If not…
When the Macro Index is in Tier 2 AND the 6 month moving average of the Macro Index is going down, shift to 100% cash. This is a warning sign. If the Macro Index rises back to Tier 3 or 4, switch back to 100% long the S&P 500. This was a FALSE ALARM. If not…
Remain in 100% cash when the Macro Investing Index falls to Tier 1. Buy back your S&P 500 when the S&P 500 makes a “golden cross”, whereby its 50 day CLOSE moving average rises above its 200 day moving average (go from 100% cash to 100% long $SPY). Once a “golden cross” occurs, wait for the Macro Investing Index to rise to >=0.8 before you start monitoring the Macro Investing Index again for a dip <=0.6

From December 1968 – present, this yields an average of 10.05% per year for the S&P 500.

  1. You can see that the returns deteriorated by only a little bit (from 10.2% to 10.05%)
  2. However, you bought extra protection, in case the future mega-crash happens when the Macro Index is at 0.6 instead of when the Macro Index is at 0.5

The slight underperformance comes from the FALSE ALARMS being slightly bigger than the advantage of selling earlier in the start of a bear market.
I tested multiple scenarios:

  1. If you go to 50% long when the Macro Index falls to Tier 2, your average annual return is 10.14%
  2. If you go to 25% long when the Macro Index falls to Tier 2, your average annual return is 10.1%
  3. If you go to 0% long when the Macro Index falls to Tier 2 (which is what I espouse), your average annual return is 10.05%

*As of December 2018, based on this strategy, you would be 0% long the S&P 500

Macro Trading Model

I am only going to talk about the long-only version of this model for now, because I have not completed the long-and-short version. But here’s a brief explanation regarding how to use the Macro Index when shorting in a bear market.
The Macro Trading Model is like the Macro Investing Model, except it takes the concept of “scaling in and scaling out” one step further.

When the Macro Index is in Tier 4, be 100% long the S&P. If not…
When the Macro Index is in Tier 3, switch to 25% long the S&P. If the Macro Index rises to Tier 4, switch back to 100% long the S&P. If not….
When the Macro Index is in Tier 2 AND the 6 month moving average of the Macro Index is going down, shift to 25% long. This is a warning sign. If the Macro Index rises back to Tier 3 or 4, switch back to 100% long the S&P 500. This was a FALSE ALARM. If not…
Remain in 100% cash when the Macro Investing Index falls to Tier 1. Buy back your S&P 500 when the S&P 500 makes a “golden cross”, whereby its 50 day CLOSE moving average rises above its 200 day moving average (go from 100% cash to 100% long $SPY). Once a “golden cross” occurs, wait for the Macro Investing Index to rise to >=0.8 before you start monitoring the Macro Investing Index again for a dip <=0.7

You can see that the only difference here is that there is another “scale in and out” between Tier 3 and 4. Doing the backtest:

  1. When the Macro Index enters Tier 3 and you switch to 75% long: this yields an average of 9.96% a year
  2. When the Macro Index enters Tier 3 and you switch to 50% long: this yields an average of 9.85% a year
  3. When the Macro Index enters Tier 3 and you switch to 25% long: this yields an average of 9.75% a year
  4. When the Macro Index enters Tier 3 and you switch to 0% long: this yields an average of 9.64% a year

These charts demonstrate the returns and drawdowns. While you can see that option number 4 has the lowest return, it also has smaller drawdowns.


My personal preferance is #3: when the Macro Index enters Tier 3 and you switch to 25% long. This yields an average of 9.75% a year
As you can see, we have sacrificed some returns from 10.2% to 9.75%. But we have bought extra protection for ourselves instead of fitting the data.

Future versions of the Macro Trading Model (both long-only and long-and-short)

  1. The Medium-Long Term Model yields an average annual return of 14.7% with $SPY
  2. So far, the Macro Trading Model (my favorite version) yields an average of 9.75% per year

While The Macro Trading Model looks like a massive downgrade, don’t forget that the Medium-Long Term Model also includes a lot of “big correction” indicators. That’s where most of its outperformance comes from.
In the future, I will incorporate some of these “big correction” indicators into the Macro Trading Model, which will increase the Macro Trading Model’s returns.
In the end, what I want to see is the Macro Trading Model (long-only version) yielding an average of 11%-12% a year with $SPY, while being less “curve fitted” than the Medium-Long Term Model.
In other words, the performance will only be slightly worse, but it be much less curve fitted.

Any questions/comments? Comment below!

8 comments add yours

  1. Hey Troy!
    Thanks for this amazing work… Apologies for my ignorance here, but can this index be used to identify big drops (above 15%) such as in 2011 or 2015?
    Regards,
    Remi

    • Hi Remi,
      At the moment it cannot. But it will, once I add some of the Medium-Long Term Model’s components into this model.

    • Hi Ray,
      The rules are simple. But in order for me to describe it effectively (i.e. to excel), i need to have those parameters.
      Basically, just think that as Macro gets progressively worse, you shrink your long position size

  2. Troy, on the Drawdpwn chart, I presume that your favoured Option #3 is the colour Green one (although there actually seem to be more colours than Options, so I am not too sure about that). Am I right in seeing the max DD for that is ~-11.5% against the #1 Option maxDD of ~-17%?
    As an aside, Ned Davis Research has a model that measures various fundamental metrics, and scales out progressively from 100% to 80% to 40% to 0% as the metrics deteriorate.

Leave a Comment