In response to our previous post and chart on this topic, several people asked for a statistical analysis to accompany the graph.  Hopefully we’ve enhanced the explanation and can issue a proper warning, which, after thorough historical analysis, seems all the more warranted.  We will be developing and expanding on this post in the weeks to come, but submit the following background and preliminary findings:

Background and Summary Findings:

• M2 is the broadest form of money supply currently reported by the Federal Reserve and we have found that large changes in it (what we call M2 Volatility) coincide with stock market volatility and significant price corrections in a critical part of the year–the third week of July to the third week of October.
• M0, the monetary base, though more commonly touted by economists, does not appear to be a good predictor of stock market volatility or of potential stock market corrections (data to be presented in the future).
• Specifically, we use M2 non-seasonally adjusted (NSA) because we have found that the very seasonality (volatility) that the Federal Reserve attempts to remove from the data is what is most correlative with stock market volatility and corrections.*  The Fed publishes M2 NSA along side M2 SA for all to see, so there is no conspiracy–they are simply a bit short-sighted in stressing M2 SA.
• As M0 and seasonally adjusted data are the norm, we believe this analysis is important because it addresses a potential hazard to stock market prices heretofore not widely addressed*.
• Based on an analysis of M2 Volatility in the three month period commencing the end of the third (approximate) week of July and into the third week of October for the past 28 years, we believe that this year, 2009, (1) there is a significant chance of a large (average ~17%) eventual correction as measured from the close of July 17 09, (2) there is a significant chance of increased volatility as measured in percent price range over the period (average ~20%), and (3) that, based on M2 Volatility, equities are in a more precarious position than last year at this time, which experienced a 33.38% loss in the corresponding three month period.

Explanation of graph:

• At the top of the above graph, in the first subgraph is the S&P 500 cash index going back to late 1981 with M2 plotted in cyan, with its 13 week (one calendar quarter) moving average. Weekly data for M2 goes back to January 1981, though monthly data (not usable for our purposes because of low resolution) goes back to 1959.
• The yellow indicator in the second subgraph (M2 Volatility) is the percent change of M2 from its value 13 weeks prior.  The yellow trendlines were hand drawn and show secular trends in the broadening and  contracting of M2 Volatility. Note the percent change is not annualized.
• In the third subgraph are the 12 month (bright red) and 6 month (dark red) ranges of basis point (bp) swings (divided by 100 for presentation purposes) in 13 Week % Change M2.  Basically, they represent the 6 and 12 month range of M2 Volatility.
• In the bottom subgraph is the 10 year US T-Note yield minus the 13 week US T-Bill yield (the 10 Yr – 13 Wk spread), which is one of the simplest representations of bank lending profitability.  The greater the spread, the more profitable it is for banks to borrow on the short cheap end and lend at the more expensive long end.

A closeup of the above graph of the last five years only is here:

Preliminary observations:

• It is immediately apparent that M2 13 Week % Change (M2 Volatility) has been broadening since mid-2006 with higher highs and lower lows.
• As of last week’s M2 figure (reported July 23 09 at 4:30 pm), the 12 month spread (bright red / third down) reached its highest level (674 bp / 6.74 percentage points) going back to 1981.
• There is an approximately quarterly periodicity to extremes in M2 Volatility (yellow).  This lends credence to the theory that money supply tends to exert its influence in three month increments.

Looking further at the quarterly periodicity of M2 Volatility (yellow), most years tend to display the same temporal pattern of peaks and troughs as the previous years.  Eventually, however, the pattern breaks, the relative importance of which will be researched in the future.  In general, the spikes tend to occur in the last week of December/first week of January, the third week of April and the third week of July, with April usually up and July usually down.  There is a notable absence of any regular high or low in M2 Volatility in October, as might be expected at this quarterly mark.

The periodic spike up in the third week of April appears to be particularly important because if M2 contracts too much over the summer into the third week of July, there is a tendency to have a correction in the S&P 500 going into the third week of October. Because we are now closing the third week of July and because most major market corrections are underway into the third week of October**, we will be looking at this three month period exclusively in this report.

Data:  The above graph is 1981 to present, as before, but marked with vertical lines to show the third week of July spike low (red) and the third week of October (gray).  Data from the graph was exported to a spreadsheet for further analysis and to identify key thresholds.   

Analysis:  In the 28 year sample, at the beginning of the three month period commencing at the end of the (usually) third week of July, a (1) 13 Week % Change in M2 (M2 Volatility) (4th column) less than 0.2% along with (2) a 12 month M2 Volatility basis point (bp) spread (6th column) of 320 or more , has coincided nine times with declines of 9% or more measured from the close of the weekly bar at the beginning of the period to the low set in the following 13 weeks, with the average decline being -16.69% [stdev 10.39%, n=10] versus an average decline for other years of -4.12% [stdev 3.79%, n=18].

Two failures were 1) a false positive when the three month period in 2006 met these parameters but had no low that was lower than the close of the beginning week, and 2) a false negative when for the -15.72% decline in 1981, which had a beginning 13 week M2 Volatility of 1.08% (materially greater than 0.2%), but did fit the second criterion, having a max 12 month bp spread of 358.

For the three month periods that did not fit the two criteria, there were average gains of 7.07% [stdev 6.61%, n=18] versus average gains of 4.76% [stdev 4.79%, n=10] for those that did.

The 12 month bp spread of M2 Volatility set for the week ending July 17 09 is an all time high and the beginning M2 Volatility was -1.04%, the second lowest after the July 21 2000 -1.44% reading.  The one saving grace of this period is perhaps the 10 year-13 week Treasury spread of 3.49%.  The largest loss during the three month period with a Treasury spread greater than 3.00% was -8.74% in 1982 (which also had a 24.89% gain), with an average loss of -4.69% [stdev 3.08%, n=6] and average gain of 9.48% [stdev 10.98%, n=6].

When the beginning 12 month bp spread in M2 Volatility was greater than or equal to 320, the average percent range of the S&P 500, or what could be characterized as price volatility, from high to low in the three month period (range / price at beginning of period) was 19.25% [stdev 10.05%, n=13] as opposed to 11.05% [8.13%, n=15] for other years.

Conclusions: Two caveats with respect to the data are: (1) the sample size is small and the average gains/losses are often with the standard deviations, and (2) the boundaries chosen in the two parameters to coincide with 9 of the worst declines are a bit arbitrary, so measurements at about the boundary areas could probably go either way.  What is clear is that we are in no way close to the boundaries for the current three month period in 2009.  And, while the data presented here do not imply causation by themselves, they do confirm correlation of M2 Volatility with stock market prices. 

What we might expect is not only a large eventual correction (ranging in the 9.33% to 33.38% [avg:  16.69%] experienced in previous years), but also a wide price range over the next three months into the third week of October (avg: 19.25%) .  So, while prices can certainly climb higher, there is great danger that all but the most nimble traders or long term investors (with no stops) can be eventually shaken out.  Much of the retail money that left the equities markets last year was permanent, and a second year of thrashing poses an even greater threat to the long term health of the markets.

Further, we noted the possible saving grace that Treasury spreads are high now, which can induce banks to lend and support equity prices.  However, the Fed’s new ability granted by Congress to pay interest on excess reserves held by banks at the Fed can easily erase this potential benefit, and further studies on actual lending activity are warranted.

______________________________________________________

* Thanks to Robert Wenzel, editor of EconomicPolicyJournal for suggesting the use of M2 non-seasonally adjusted as opposed to seasonally adjusted, and for noting the similarity of contracting money supply in 2009 versus 2008, which was the impetus for this study.

** Robert Wenzel points out this is because, “it is the period of great consumer intensity away from savings and toward consumption. New school clothes, winter clothes and preparation for Thanksgiving and Christmas. From mid-August to December, October is at the vortex of a shift toward consumer buying, away from savings–which includes stock liquidation.”