Earlier this week I received two requests for a different approach to my monthly update on the long term market trend, one that plots a regression through the fourteen decades of US market data. My intention is illustrate that the stock indexes do not move in a straight lines but are subject to long-term ("secular") trends. A linear regression through the complete series essentially finds a straight line that best 'fits' the monthly averages of daily closes, splitting the data points above and below the line.
Because the range of price data is so extreme, I use a log-scale vertical axis to ensure that advances and declines are the same relative size regardless of the price (e.g., a 33% decline will be the same vertical distance regardless of the price.
My standard regression-to-trend chart uses "real" (inflation-adjusted) prices chained in today's dollars. Here's the latest update using a base 10 log scale for the vertical axis and an exponential regression through the data:
The two requests I received were for a regression through the nominal index data -- in other words, with no inflation adjustment. My personal view is that a nominal chart of 14 decades is dangerously misleading. Consider: A dollar in 1871 has about the purchasing power of a nickel today, and the slide in value has been incredibly erratic, especially in the decades before FDR abandoned the gold standard.
Here again is the real (inflation-adjusted) S&P Composite, this time using a base 2 log scale vertical axis with an exponential regression. With a base 2 log scale, Excel is able to plot more numbers on the vertical axis. I've included four callouts on the lower series, which illustrates the percent above and below the regression. The four numbers in red are the percent variance from the trend at the market peak in 1929, the trough in 1932, the Internet Bubble peak and the latest value as through January 2013:
Here is the nominal version -- no inflation adjustment -- with the same four callouts on the lower series.
My view is that the current market price is significantly above trend, which suggests excess valuation. The nominal version of the chart puts that deviation from trend at a level that is clearly a nonsensical exaggeration of the overvaluation, hence my preference for the "real" data.
I'll close with an overlay of the nominal and real index, the latter chained in today's dollar value. And I've included the annualized rate of inflation (deflation) in the background.
In a follow-up commentary, I'll explore the pros and cons of using two regressions, one for the earlier decades and another for the more recent decades, say since the introduction of the 401(k) and IRAs in the early 1980s.
Note : For readers unfamiliar with the S&P Composite index, see this article for some background information.