Rally-Ho!

Last week we discussed the magnitude and duration of corrections in secular bull and bear markets.  As we discussed, corrections in secular bear markets tend to be bigger than those observed during secular bulls, and tend to last longer.  This week, we’ll turn last week’s analysis on its head and discuss the nature of the rallies, often referred to as “cyclical bull markets”, that occur between the major corrections.  Currently, this side of the bull/bear ledger is more germane considering markets in the US and around the globe have been on a tear higher over the past several months.  Like last week, we’ll review the period from roughly 1950 to the present, picking up with the end of the secular bear market in 1949.  Again, as we discuss differences in performance during secular bull markets and secular bear markets, keep in mind that there have been five secular market cycles since the market peak in 1929: the bear period from 1929 to 1949, encompassing the Great Depression; the secular bull period from 1949 to 1968, which peaked in the wake of the Nifty 50 craze; the bear period from 1968 to 1982, which was marked by political upheaval and “stagflation”; the secular bull from 1982 to 2000, which culminated with the tech craze and the highest valuation levels recorded in modern times; and the current secular bear period, which began in 2000, marked by the tech bust, subdued economic growth, the “Great Recession” and sovereign debt crises.

Over the past 60 or so years, the major corrections (15%+) we discussed last week have been followed without exception by moderate to furious rallies that oftentimes push markets past prior peak levels.  Like market corrections, the nature of the rallies is related to the overall market environment, secular bull or secular bear.  In a mirror image of last week’s data, post-correction cyclical bull market rallies tend to last much longer and achieve much stronger performance outcomes during secular bull markets than the rallies that occur during secular bears.  Let’s look at the S&P 500 data pertaining to the rallies:

* Highlighted Areas represent Secular Bear market periods. 

We’ve witnessed 17 rallies post-correction since 1949, including the present rally.  As you can see above, the median post-15%+-correction rally across all markets has lasted a little more than two and a half years and generated nearly 75% in returns (simple return, not including dividends).  The longest and best performing rally without a 15% intervening decline was the rally from 1990 to 1998, which lasted nearly 8 years and produced a return trough to peak exceeding 300%.  Wow!  The shortest post-correction rally occurred in 1980 as the US grappled with a double dip recession.  Lasting about two-thirds of a year, the rally still managed to produce trough to peak returns of 43%.  This was closely followed in shortness of duration by the short rally between round one of the European sovereign debt crisis in 2010 and round two in 2011.  This relatively short-lived rally also produced the smallest overall return in the data series at 36%.  

Breaking the data down in to secular bull and secular bear periods provides some interesting contrasts.  Median performance for cyclical rallies during secular bulls exceeds performance during secular bears by approximately 30 percentage points, 86% to 58%.  Furthermore, median rally duration is two years longer.  There’s more variance/volatility in the secular bull numbers owing to the fact there are a few significant outliers to the upside, notably the 300%+ return during the ‘90s.  Secular bear market rallies tend to be more consistent statistically in terms of performance and duration.  

Similar to the data last week, there’s no major statistical relationship between starting normalized P/E ratios and subsequent rally performance.  The R-squared, a statistical measure of how well starting P/E levels in this case influence performance, is a mere 0.04.  Correlation is -0.2.  P/E ratios tend to have a much stronger statistical relationship to long-term annualized returns and tend to display little correlation to short-term moves.  Hence, more often than not, it’s not a very good idea to rely on P/E ratios for market timing purposes!  The lowest starting P/E for a post-correction rally was 6.64x in 1982.  This also happened to mark the switch from secular bear market to secular bull market.  The highest starting P/E for a rally was 33.77x in 1998.  Of course, the P/E at the end of the 1998-2000 rally was an astronomical 43x; this happened to mark the end of the 18 year secular bull market.

Where do we stand today?  Since the absolute trough associated with the late 2011 swoon, the S&P 500 has rallied approximately 40%.  The duration of this rally is 1.3 years.  Most market observers believe we’re still enmeshed in the secular bear market that began in 2000.  Using past rallies within secular bears as a guide, this rally is still under the median in terms of duration and overall performance.  There is one thing to consider.  To paraphrase Mark Twain, history doesn’t necessarily repeat, but it certainly rhymes.  During the last secular bear market from 1968 to 1982, the first major post-correction rally from 1970 to 1973 proved to be the longest duration-wise as well as the best performing.  The pattern could be repeating itself here, which bears watching.  While the sample set is small, each secular bear over the past century has ended with normalized valuations in the single digits.  Currently, normalized P/E sits at approximately 23x.  This doesn’t preclude the market from rallying furiously, as seen in some past episodes.  However, there are probably decent odds that investors will witness another round of major market turmoil in coming years.  The good news: time-wise, we are probably in the later innings of the secular bear.  At some point in the not too distant future, investors should see another 15 to 20 year period of outsized global equity returns.

Corrective Action

With markets having rallied significantly over the past several months, both in the US and around the globe, we thought it might be interesting to take a look at the frequency, duration, and depth of market corrections over the past 60 or so years.  We’re not undertaking the exercise in anticipation of major upcoming market turmoil.  Instead, it’s a reminder that investors’ expectations for equity market performance are often romanticized on some level.  Memories can be very short.  Investors often underestimate the frequency of corrective episodes.  In reality, it’s a normal part of the process of investing in equity markets.  Over time, equity market returns have been solid.  Investors just have to endure some bumps and volatility along the way.  Of course, some of the bumps are a lot worse than others.  As we’ve discussed in the past, the severity of corrective action is best viewed within the broader context of “secular bear market” or “secular bull market” action.  These cycles generally have lasted 15 to 20 years over the past century and are generally bookended by valuation extremes; “secular bull markets” end when valuations are severely stretched, while “secular bear markets” have usually ended with long-term valuation indicators at severely compressed levels, such as single-digit P/Es.

For this exercise, we defined a correction as any peak to trough decline exceeding 15%.  Because there is much more data available, we focused exclusively on the S&P 500.  We took the liberty of adding the correction of 1953 in light of the facts that the peak to trough decline was 14.8%, very close to the 15% threshold, and the correction lasted nearly a full year.  On to the data:

*Highlighted Cells represent periods during Secular Bear Markets

From 1950 through the end of 2012, there have been 16 major corrections.  Nine of those 16 corrections have occurred during secular bear markets.  

This brings some interesting observations to the surface, however.  While the frequency of corrections is nearly equal between secular bear and secular bull markets, the nature of those corrections is markedly different.  During secular bull markets, the last of which occurred between approximately 1982 and 2000, the average length between correction troughs and subsequent market peaks is nearly 4.5 years.  The longest stretch between corrections occurred during the mega-1990s bull; nearly eight years passed between the correction in late 1990 and the correction in 1998 associated with the LTCM and Russian debt default debacles.  During secular bear markets (we’ve been in a secular bear since 2000), the length between corrections falls to approximately 2 years on average.  Furthermore, the duration of the corrections during secular bear markets far exceeds the durations generally observed during solid market up periods.  Average correction duration during a secular bear is approximately 1.25 years, with the longest peak to trough decline occurring over approximately two and a half years from 2000 to late 2002.  Secular bull corrections have lasted on average a little less than half a year.  Interestingly, 6 out of the 16 corrections identified here have lasted a quarter of a year or less.  Finally, and probably most important to investors, the magnitude of correction declines (at least those corrections meeting our criteria) during secular bears exceeds that of secular bulls by a decent margin: 33% on average to 23%.  The most devastating peak to trough decline, of course, occurred between late 2007 and early 2009.  That decline of nearly 58% will remain seared in investors’ memories for quite a while.

If, back of the envelope, we total up all the calendar time from 1950 to 2012 during which investors were mired in correction, it comes out to about 14 years of market time.  This represents about 22.5% of the calendar space over that time frame.  Even with the volatility and deep declines investors have experienced over the past 12 years of this secular bear, it’s probably not a stretch to believe that the typical investor would probably underestimate how much time the market spends in corrective stretches (might make for an interesting survey).  

A few final notes pertaining to the data: 10 of the 16 major corrections have been associated with official US recessions as dated by the National Bureau of Economic Research.  As for the other six, those have generally been associated with a gut-wrenching series of negative headlines, such as the near collapse of LTCM in 1998 or the string of crises associated with the European sovereign debt crisis over the past few years.  There is a statistical relationship between the starting Shiller P/E value at the peak and the subsequent magnitude of the correction, but the connection isn’t rock-solid; the r-squared is approximately 0.23 and the correlation is approximately -0.47.  The average starting normalized P/E was approximately 21x, above the long-term average of 16x.  Notably, four major corrections, 1953, 1976-1978, 1980, and 1980-1982 began with below average normalized P/E ratios.  Matter of fact, the early 1980s corrections began with normalized P/E ratios in the high single digits.  

Again, we’re not trying to scare investors here, but just pointing out that to achieve the solid long-term returns of approximately 6% per annum without dividends/9% with dividends in equity markets, you sometimes have to take a few lumps.  It’s always tempting to try to time every single correction to enhance the return profile.  In practice, timing the shallower corrections has proven quite difficult and even counter productive for investors.  Conversely, the deep-dish declines as witnessed in 2008, 2000-2002, and 1973-1974, were generally preceded by a combination of significantly above average valuations, clear signals of impending economic recession via leading economic indicators and other metrics, and clear technical deterioration and signs of distribution in major market indices.  

As the Months Go By...

Listen to enough talking head chatter every week or month, and you’ll hear a number of calendar-based market clichés.  “Sell in May, go away.”  “Santa Claus Rally.”  “January Effect.”  As we enter a new year, we thought it may be interesting to highlight some statistical data on monthly performance over the past 80 or so years to give you an idea which months and portions of the year have been the most promising for investors, and which months have proven discouraging.

First, let’s present a table with the monthly data (1930 through 2012) for the S&P 500. 

A few things stand out.  First, right now we’re in the middle of the “sweet spot” of the bat when it comes to seasonal performance.  December and January have proven to be the best months performance-wise historically.  Even more interesting, this performance has been among the most consistent of the months in terms of performance.  Three quarters of the time, December has been positive, whereas January has witnessed positive performance approximately 64% of the time.  December and January have two of the lowest standard deviation figures among the months and the two of the lowest spreads between the historical maximum return and the historical minimum return.  Hence, there’s definitely something to the notion of a Santa Claus rally, and a general carryover of the good cheer into the New Year.  

The December-January party has typically led to a February hangover of sorts.  February is the third worst performer behind September and June.  Just as December and January are relatively consistent positive performers, February has been consistent in its dourness with the second lowest standard deviation among all the months.  It isn’t surprising to think that there would be some consolidation after two strong months.  Nonetheless, historically, February consolidation has allowed the market to regroup for solid spring performance.  

Is there something to the “Sell in May, go away” maxim?  Definitely.  September and June are the two worst performance months of the year, with May coming in fourth overall (September is the only month that has produced a negative median return historically).  May, June, July, and September are among the worst months in terms of percentage of positive months.  And, the volatility of the return streams from May through October is higher than witnessed during the period between November and April.  Overall, since 1930, the average May to October return is 1.84% vs. 4.76% for the November to April period.  The median return is 2.76% vs. 5.24%.

As with any bit of statistical analysis, nothing is ever set in stone.  There’s always variance making it very dangerous to make big market gambles in any given year based upon factors such as these.  For all we know, performance during the summer of 2013 could prove to be the best of all time. In any case, there’s clearly been a distinct performance advantage historically for the winter and early spring months in contrast to the summer and fall.  And, when it comes to individual months, December has been a consistent, reliable booster for portfolio managers’ performance.  For those with longer time horizons, understanding and incorporating data such as this could help shape certain decisions, such as tweaking asset allocation processes or deciding on the timing of hedges.