26 October 2009

Indicator of the Week: Monthly MACD By Rocky White, Senior Quantitative Analyst

Hat Tip qqq.

Foreword: The market has rallied quite strongly during the past several months. A couple of popular long-term technical indicators, the Stochastics and the RSI, signaled a buy at the end of April. We just had another major technical indicator give a buy signal. I'm talking about the Moving Average Convergence Divergence, better known as the MACD.

The MACD can be looked at over different time frames (for example, hourly, daily, weekly or monthly). I'm taking a long-term perspective here and looking at it monthly. In the chart below, the MACD is calculated by subtracting a 26-month moving average from a 12-month moving average of the Dow Jones Industrial Average (DJIA). I plot the difference of those moving averages (red line below), and then calculate the nine-month moving average of that line (dotted line). This dotted line is the signal line. When the signal line crosses above the MACD, it's a buy signal. The most recent buy signals are marked by the red dots (I only consider one signal every 12 months).

Historical Analysis: Going all the way back to 1900 on the Dow, this is just the 25th MACD buy signal. Below is a table showing annualized returns for the Dow following these buy signals. The second table below is for comparison, and shows typical Dow returns since 1900. In the short term, you will see the signal shows no outperformance to the market. In fact, the three-month returns following a signal have tended to underperform the market at other times. But once you get out to six months, especially at the one year mark, you see quite a bit of outperformance. The market has averaged a gain of about 14.5% in the year after a MACD buy signal. This is way better than the typical market return of less than 7%. Furthermore, the market was positive one year later 83% of the time after a signal. Compare this to 64%, which is the percentage of time the market has usually been positive a year later.


No comments: