In a recent article on the Accumulation/Distribution Index and the Money Flow Index (MFI) developed by Marc Chaikin, I noted that the MFI is generally combined with a momentum indicator such as the Stochastics or RSI. In this article, I would like to show you how this might work by combining the MFI with the RSIosc.
The RSIosc is a modification of the Relative Strength Index, developed by Welles Wilder, and has been covered in previous articles. This oscillator is calculated by taking the difference between a 9-period simple moving average of a 14-period RSI (RSI(14)) and a 45-period exponentially weighted moving average of the RSI(14). When the 9-SMA is above the 45-EMA, the oscillator is above zero, and viewed as “positive”; when the 9-SMA is below the 45-EMA, the oscillator is below zero and is negative. Marc Chaikin developed the Money Flow Index in an effort to improve the signals from the Accumulation/Distribution Index. The MFI is a 21-period summation of the Accumulation/Distribution Index values divided by a 21-period summation of the volume. These are the default values for this indicator, and while some find a shorter period to be more useful for weekly data, I have found that using 21 weeks provides a good picture of the intermediate-term trend.
Objectively combining the signals from the RSIosc and the MFI would require that both agree before a buy or sell signal is generated (i.e. they are either above the zero line or below it). However, I favor a more subjective style of analysis in which one looks for positive or negative divergences for early warnings of a trend change and an eventual violation of the zero line. The first chart I would like to look at is a weekly chart of Amgen (AMGN) from the latter part of 2002 through the end of 2005. As is often the case, I will first look at the weekly analysis where I find the signals to be more clearly defined, which goes along with my belief that the analysis of the intermediate-term trend is most important.
Figure 1
On October 4, 2002, the MFI, which was already above its rising 21-period WMA, also moved above the zero line (line a). The RSIosc was already in positive territory and had been for four weeks. AMGN opened the following week at $43.30 and closed the week strong at $48.09. Both indicators confirmed the new highs in April 2003, but as AMGN closed at $70.77 on July 18, 2003, the MFI (line 1) and the RSIosc (line 2) both formed negative divergences, suggesting that neither volume nor momentum was confirming the price action. Both the MFI and RSIosc continued to deteriorate over the next two months, and on September 19th dropped below the zero line as noted by vertical line b. AMGN continued to correct for the next year, as it retraced 50% of the rally from $30.57 to $70.45.
The first signs that the correction was over appeared in July and August of 2004 as the RSIosc started to form a positive (or bullish) divergence (line 3) and then moved back above the zero line on August 20th. At this point, the MFI was above its WMA, which had also started to turn higher; however, the MFI did not move back into positive territory until October 1, 2004 (line c). The following week AMGN opened at $58.33 and continued to decline over the following few weeks eventually reaching a low of $52.00. The new correction lows were not confirmed by either of the indicators. If you had bought the Monday after the weekly MFI turned positive, this would have been a drawdown of 10.8%. As noted on the chart (see circle), the MFI dropped just barely below the zero line as the 50% support level at $52 was hit, but AMGN quickly rebounded and closed the week strong—a positive sign. I have often noticed similar formations in a number of indicators, which can allow further fine-tuning of entry and exit points.
Both indicators turned negative in the spring of 2005 and stayed negative until the latter part of June 2005. However, they did turn positive in time to catch the sharp rally to the September highs at $86.92. As is the case with many of the studies we have discussed, trend line analysis can be very useful. For example, you will note that the MFI dropped back below its WMA in the latter part of October and failed to move back above its WMA on the following rally. This set the stage for a break of the uptrend (line 4) that warned of a deeper correction. AMGN dropped back to the $65 level the next year.
Figure 2
Further trend line analysis is illustrated in Figure 2, which shows the ETF-based on the European MSCI EAFE Index. The first period I would like to focus on is the latter part of 2004, as both the RSIosc and MFI confirmed the December 2004 highs at $53.50 (point 1). After a two-week setback, EAF again turned higher and reached the $55 level, but both the MFI and RSIosc failed to confirm these highs as noted by the points labeled 2. Three weeks after the highs, the MFI dropped below its WMA, and by April 15th, the RSI had violated the zero line. By early July 2005, one was able to draw a well-defined downtrend in the MFI (line a). By the middle of July, EAF had once again turned higher, and on August 12th, 2005, the RSIosc moved back into positive territory. The MFI held above the zero line throughout the correction and broke its downtrend (line a) on September 9th, providing further evidence that the correction was over.
EAF corrected for three weeks in October 2005 allowing a new uptrend to be drawn in the MFI (line b). The MFI confirmed the January 2006 highs (point 3), but then turned lower even as EAF continued to move higher. By April 13th (line B), when the MFI’s uptrend (line b) was broken, the MFI had formed multiple negative divergences. In early May, as EAF pushed above the $72 level (point 4), the MFI was significantly lower and had formed a well-defined downtrend (line c). The signals from the RSIosc were less clear, as while it was higher in May than it was in January, it matched but did not exceed the September highs (line d).
Figure 3
By first looking at the weekly analysis, one can then move to the daily data, which will generally improve the entry and exit points. The third example takes a look at Goldman Sachs (GS), a financial stock that in the summer of 2007 got quite a bit of attention. The first thing to note from the weekly chart is that the MFI peaked in March 2006 and has since been in a steady downtrend (line B). This divergence from the price action (line A) indicated declining volume as GS moved higher. In September 2006, the MFI moved back above its WMA as GS rallied from $167 to over $212 in February 2007. The violation of the weekly uptrend in the MFI, line C, came two weeks before the late February plunge that eventually dropped both the MFI and RSIosc into negative territory. The RSIosc, it should be noted, was very strong during the rally but also diverged at the February highs.
The daily MFI formed a negative divergence at the February highs, line D, and was well below its WMA as GS peaked on February 22nd. Both the weekly and daily MFI stayed negative until late April, with the daily MFI managing to form a short-term uptrend, line E. As GS rallied to new all time highs above $230, the weekly MFI was well below its downtrend (line B) and the previous high, while the RSIosc barely made it above the zero line (point 1). By the end of June both of the weekly studies were below the zero line. The daily analysis was actually weaker as the MFI, despite significantly higher prices in April and May, failed to move above its downtrend, line D. This was confirmed on May 15th as the uptrend in the daily MFI, line E, was broken. The RSIosc turned negative a week later (point 2). On June 18th, the WMA of the MFI also violated the zero line (point 3). Over the next two months, GS was very weak, dropping below the $160 level.
So, what is the best way to use the MFI and RSIosc combination? Though I have not often seen such long-term divergences as those observed in GS, they do occur periodically and avoiding issues where they are observed would be a mistake. Though I do pay attention to 12- and 18-month divergences, I find that the divergences that develop over eight to 12 weeks generally give the best signals. Secondly, I have always found trend lines on the indicators to be very useful, as the breaking of the up- or downtrends will confirm the divergence. This is more valuable to me than the crossings of the zero line. Generally, I start with the weekly or even monthly data and work my way down to the shorter time frames. Then, as always, the key ingredient is managing the risk and being sure that you never take a position where there is enough risk to damage your portfolio. If the signals to buy or sell are valid but the loss if your stop is hit would be too high, then you are better off standing aside. In the next article, I will look at other ways that volume can be used to better analyze supply and demand.
If you have a question about this article or would like to have a trading method discussed in more detail, I can be contacted at tomaspray@intershow.com.
Tom Aspray began doing computer analysis of the financial markets in the early 1980s. He helped to introduce many of the technical tools that are now very popular though his writings and lectures around the world. He now works as a private consultant and educator.
