Using PMI Data For Tactical Asset Allocation

The 200 day moving average is perhaps one of the most well-known tactical asset allocation filters and many analysts suggest that you should be long the stock market if the Index is greater than the 200 day MA, and flat the stock market if the Index is less than the 200 day MA.

For example, the following chart plots the buy and hold performance of the SPY (the black Line), and the performance of trader who is only long if the SPY is above it’s own 200 day moving average (the blue line):

The bottom pane of the above chart also plots the amount of available equity allocated to cash. Note that if the SPY is beneath the 200 day moving average, the trader using the 200 day moving average filter will have a 100% cash position.

It’s worth pointing out that although the annual return of the 200 day MA strategy lags the annual return of buy and hold, on a risk adjusted basis (using the CAR/MDD as our metric) we can see that the 200 day MA strategy was superior during this period:

Buy and Hold. CAR/MDD =  0.17

200 Day MA Strategy. CAR/MDD =  0.26

The next test shows what happens if rather than move to cash when the SPY is beneath the 200 day moving average, we instead allocate 100% of available equity to the short-term U.S. treasury Bond ETF “SHY”.

Note that during the sample period if we allocated 100% of available equity to SHY when the SPY was below the 200 day moving average, the CAR/MDD produced by the strategy was: 0.28

Of course, we can also optimise the length of the moving average that we use.

The following graph plots the CAR/MDD if using a variety of X period moving averages for our tactical decision:

It can be seen from the above graph that during the sample period the better risk-adjusted returns were produced if using a a moving average length of between 80 – 180 days or 220 – 340 days.

However, many people are uncomfortable with the idea of using a simple moving average of price to determine when to enter and exit the market.

PMI Data For Tactical Asset Allocation

So, for the next tests our filter will instead use the data of one the most popular leading economic indicators, the ISM Purchasing Managers Index (PMI).

Without getting into the finer details, the PMI is a closely watched economic indicator because it tells us how business purchasing managers view the current economic climate.

Given that purchasing managers have key insights into sales, employment, inventory and pricing, it stands to reason that if they are optimistic then we should invest in SPY.

If on the other hand the PMI suggests pessimism, it seems logical to invest in SHY.

The precise strategy will be to rotate between SPY and SHY. However, instead of using the 200 day moving average to determine which ETF to hold, we will instead use the PMI data.

We’ll buy SPY if the PMI is greater than 50 (which indicates economic optimism) and we’ll buy SHY if the PMI is less than 50 (which indicates economic pessimism).

For example, we can see the PMI values plotted beneath the following SPY chart. I have also included the buy and sell arrows for the SPY if using the PMI > or < 50 rule:

It can be seen in the following performance chart that during the sample period, the PMI 50 rule produced better results (both absolute and risk adjusted) than either buy and hold or the 200 day moving average of price.

Note that the CAR/MDD produced by the PMI 50 strategy was: 0.52

The above tests suggest that if you’re uncomfortable using a simple moving average of price to make your tactical asset allocation decisions, economic data can provide a more logical and just as effective substitute.


  • Kevin at SignalPlot

    Reply Reply February 12, 2017

    Interesting work with the PMI. Can you tell me where you got the PMI data from? And did the PMI strategy take into account the actual release date of the PMI data since the data is released with a one month lag?

    • Llewelyn James

      Reply Reply February 12, 2017

      Hi Kevin,

      I used the #NAPMI ticker provided by Norgate Premium data.

      Upon manual inspection of the official values published by ISM, I noticed that there were occasionally some small discrepancies between the Norgate values and the official ISM values.

      So I also cross-checked the strategy results using the data provided by

      As you correctly point out, the one month lag was accounted for. For example, the August 2016 PMI was 49.4 and the Sept 2016 PMI value was 51.5.

      This meant that the strategy moved from SHY to SPY, but not until Oct 2016 when the Sept PMI value was released.

      I hope that helps.

      Kind regards,


  • Sergey Vedernikov

    Reply Reply February 12, 2017

    Hi James,

    Thanks for the interesting post! I keep the list of macro indicators that are probably useful for timing S&P, and was thinking about adding PMI to it after reading your post.

    But upon more thorough testing it turned out that PMI really helps to time last two bear markets, and outperforms B&H by roughly 40% since 2000. However, since 1950 it badly lags B&H, having final equity of about 30% of what could be achieved by now by simple B&H.

    Another suspicious thing that hints that this research is probably overfitted is that changing threshold a bit, i.e. exiting the market when PMI < 49, 48, …, produced very different results – PMI *did not* time last two drawdowns, giving only small improvement over B&H, or lagging B&H / not giving any improvement at all.

    So, my honest opinion about this indicator: rubbish. Try unemployment-related stuff – some of it really works robustly since 1950.

    • Llewelyn James

      Reply Reply February 13, 2017

      Hi Sergey,

      You’re right that the PMI was not effective in the 50’s, 60’s or 70’s. Since the 1980’s the PMI filter has worked O.K.

      I suppose it could be hypothesised that purchasing managers were not as well informed before the adoption of computerised spreadsheets.

      As for fitting, a PMI value of 45 has tended to work the best during the past 40 some years. I was simply testing the value of 50 because it is the logical value between expansion and contraction.

      I agree with you about unemployment data. For example, a simple strategy that’s worked well for some time is to only be long equities if the unemployment rate is below it’s 12 month moving average or if the price is greater than it’s 12 month average. Although I’m not always satisfied that the historical unemployment data isn’t a little forward looking with respect to revisions.

      Thanks for sharing your opinion.

      All the best,


Leave A Response

* Denotes Required Field