I analyze macroeconomic issues from a fundamental perspective, and I analyze market behavior from a technical perspective. Original macroeconomic analysis can be found here and both macro analysis and commentary can be found on my Caps blog. If you like or appreciate my analysis, please add yourself to my Following List

Tuesday, July 19, 2011

Why Using the MACD for Long Term Trend Analysis is Worthless

Again, apologies ahead of time for titling the post so stridently, but I wanted it to capture people's attention, in the same way as I wrote Why Arithmetic Stock Charts Are Worthless. And you will see in this post the exact same reasons exist for this argument as does the logarithmic vs. arithmetic price scale argument.

First, you have to read this post for all of the necessary background (Why Arithmetic Stock Charts Are Worthless). If you don't read that post then don't bother reading this post. I will do a quick summary as to why this difference is important, but I will not rehash everything.

I fully admit that I have been guilty of the sin of using the MACD for long term trend analysis. But I want to show you why I have abandoned it (and replaced it with a more appropriate indicator) for long term analysis.

First consider what the Moving Average Convergence-Divergence (MACD) is and how it is calculated (from Stockcharts)

Standard MACD is the 12-day Exponential Moving Average (EMA) less the 26-day EMA. Closing prices are used for these moving averages. A 9-day EMA of MACD is plotted with the indicator to act as a signal line and identify turns. The MACD-Histogram represents the difference between MACD and its 9-day EMA, the signal line. The histogram is positive when MACD is above its 9-day EMA and negative when MACD is below its 9-day EMA.

As such, the output from the MACD Indicator (both the MACD and the Histogram) is in terms of points.

Consider the following statement:

"The Dow moved 14 points!"

Is that a useful statement? No, it has absolutely no meaning because it has no context.

If the 14 point move happened when the Dow was at 100 (in 1928), then it was a 14% move (pretty significant). If the 14 point move happened when the Dow as at 14000 (in 2007), then it was a 0.1% move (pretty insignificant).

The absolute magnitude of moves (point moves) in the stock market are meaningless values. They are only meaningful when they are related to some reference value (converted into percentage moves).

Which as I said before: ALL GAINS AND LOSSES IN THE STOCK MARKET ARE EXPONENTIAL!! NOT ARITHMETIC!!.

So does that mean we should scrap the MACD altogether?

NO!

The MACD is an extremely useful trend following and momentum indicator, and is very useful in finding momentum divergences for short term moves (where the difference in price over the range of comparison is small).

The problem happens is when one uses the MACD to find divergences in moves where the difference in price over the range of comparison is LARGE. Then the same exponential issue shows up. Fortunately there is another indicator that serves the exact same function as the MACD, can still find momentum and divergences, has the same EMA input, but is instead expressed in percentage terms.

It is the Percentage Price Oscillator (PPO), (from Stockcharts)

While MACD measures the absolute difference between two moving averages, PPO makes this a relative value by dividing difference by the slower moving average (26-day EMA). PPO is simply the MACD value divided by the longer moving average. The result is multiplied by 100 to move the decimal place two spots.

To illustrate why this is a big deal, and why using the MACD instead of the PPO to find divergence in long term trends (where there is a large price change between the two comparison periods) is incorrect, see this example chart:



As another example of looking at indicator behavior in percentage terms, see this long term study post of mine: First Derivative of the S&P 500, Long Term Study

Conclusion

All moves in the stock market are exponential/logarithmic, not arithmetic. Which means that not only do we need to look at price scales in logarithmic terms (so that we can do percentage comparisons not point comparisons), but also make sure our indicators also reflect moves expressed in percentage terms. This is not as critical when the difference in price between the two comparison points is small, but is absolutely critical when the difference in price between the two comparison points is large.
blog comments powered by Disqus