Skip to main content
Track Stochastic oscillator
17 / 42
Search the archiveLibrary
Library

2001issue C121-6

Two tests of a rate-adjusted earnings-yield gap

First show why a drifting stock-versus-bond yield gap is a weak raw forecast. Then rank the same rate-adjusted spread inside a 36-month lookback, so the test is whether the current gap is extreme versus its own recent range, not versus a once-and-for-all cutoff.

  • A rate-adjusted spread subtracts the contemporaneous 10-year Treasury yield from the earnings yield and is only a crude stock-versus-bond comparison.
  • A fixed yield-gap cutoff can stay crossed after the series drifts, and quartile splits of the raw spread were not a consistent next-year filter in the historical sample.
  • A 36-month stochastic oscillator ranks the latest spread inside its own lookback window, printing 100 at a three-year high and 0 at a three-year low.
  • Editorial reading: the fairer test asks whether the current gap is extreme versus its recent range, then checks later nonoverlapping 12-month index changes at those overbought-oversold prints.
Entries in this reading1 entry
Current trackStochastic oscillator170 readings

Two questions, one series

The earnings yield is trailing twelve-month index earnings divided by the index level, expressed as a percentage. Subtracting the contemporaneous 10-year Treasury yield produces a rate-adjusted spread, used as a crude stock-versus-bond comparison.

That series can be evaluated in two steps. The first step asks whether the raw gap itself is a usable forecast. The second step ranks the latest gap inside a fixed lookback window and asks only whether it is extreme versus its own recent range.

Why the raw spread is a weak forecast

A fixed dividend-yield-versus-bond-yield cutoff that held before 1959 stayed crossed afterward. An absolute earnings-yield exit in the early 1990s would have left the later advance outside the rule.

From April 1953 through March 2001 the earnings-yield minus 10-year Treasury spread ranged from -4.2% to 7.8% with a median of -0.37%. The earnings yield had not been above the 10-year yield since September 1980.

Four equal 140-month quartiles of the raw spread produced next-12-month total returns of 14.4%, 16.1%, 6.0%, and 20.2%. The un-normalized difference was not a consistent forward filter in that sample.

A 36-month ranking of the same series

A 36-month stochastic oscillator ranks the latest earnings-yield minus 10-year Treasury spread inside its own three-year lookback window. It prints 100 at a three-year high of that spread and 0 at a three-year low. Those prints are the overbought-oversold marks for this input.

The oscillator is a 0 to 100 ranking of the latest observation inside a fixed lookback. Here the input is the rate-adjusted earnings-yield spread rather than price.

From May 1956 through March 2001 the 36-month stochastic averaged 41 with a median of 33.2. The next-12-month total return was 3.9% in the lowest decile and 5.3% when the reading was below 10.6.

Next-year S&P 500 return after a 36-month rank of the yield gap

Ranking the S&P 500 earnings yield minus the 10-year Treasury yield inside a 36-month window is the fairer test of the same series: a floor reading (min–10th percentile, stochastic stuck at 0) was followed by only a 3.9 percent average 12-month total return, while ranks above the 25th percentile averaged 13.5 to 17.8 percent. Bars are the article’s tabulated bucket averages over 535 months, not a redraw of the oscillator plot.
Ranking the S&P 500 earnings yield minus the 10-year Treasury yield inside a 36-month window is the fairer test of the same series: a floor reading (min–10th percentile, stochastic stuck at 0) was followed by only a 3.9 percent average 12-month total return, while ranks above the 25th percentile averaged 13.5 to 17.8 percent. Bars are the article’s tabulated bucket averages over 535 months, not a redraw of the oscillator plot.S&P 500 · 36-month stochastic; next 12-month hold · 1956-05-01T00:00:00.000Z to 2001-03-31T00:00:00.000Z

Lookback is 36 months on trailing-12-month S&P 500 earnings yield minus the 10-year Treasury yield. Subsequent returns include dividends. The min–10 and 10–25 bands both start at a stochastic of 0 because many months sat on the three-year floor.

Extreme prints and nonoverlapping holds

In nine nonoverlapping cases since 1953 when the stochastic printed 100, the following 12-month index change had a median of 18.4% and a mean of 12.4%, and the index was higher in seven of the nine cases. A nonoverlapping hold is a later 12-month outcome window that does not share months with the next extreme-signal test.

In 16 months when the stochastic printed 0, the next 12-month index change had a median of -2.7% and a mean of 1.5%, with a decline in eight of those 16 years. The last three late-1990s zeros were not reliable inside a single 12-month window.

A January 2000 reading of 100 was followed by a -2.0% 12-month index change, and at the March 2000 peak the oscillator stood at 7 with a 22.6% decline over the next 12 months.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
17 of 42 in the Stochastic oscillator track
20021-2 pp.Next on Stochastic oscillatorConstructing a two-line stochastic from a range-normalized closePercent K locates the latest close inside the lookback high-low span and scales that location by 100.
All readings on this track · 42 readings
  1. 1987Stochastic fast and slow construction as a rebuildable stack
  2. 1989Building the stochastic oscillator from close location
  3. 1990Monthly stochastics as a multi-year bond regime filter
  4. 1990Walk-forward screen for yen indicator rules
  5. 1990Slow stochastic construction for index pullback entries
  6. 1991Random Walk Index construction with an adaptive lookback
  7. 1991Building a two-stage stochastic oscillator from close location
  8. 1992Constructing fast and slow stochastic oscillator lines
  9. 1992Constructing nested stochastic lookbacks
  10. 1994Construct the four-state price-volume rank before filtering it
  11. 1996Crowded stochastics, false breakouts, and hidden stops
  12. 1997Fade and follow entries from stochastic extremes
  13. 1998Oversold confirmation as a staged rule-based-entry case
  14. 1999Constructing regular and slow stochastic oscillators
  15. 2001Construct a variable-interval simple moving average from stacked extremes
  16. 2001Threshold RSI and stochastic setups with next-bar stops
  17. 2001Two tests of a rate-adjusted earnings-yield gap
  18. 2002Constructing a two-line stochastic from a range-normalized close
  19. 2002Inspect mechanical stochastic daytrade rules on one bar
  20. 2003Constructing an adaptive stochastic RSI
  21. 2003Four parameters that construct a stochastic oscillator
  22. 2004Volume breakout as signal, pullback as entry
  23. 2004A first currency-market checklist with two averages and a slow stochastic
  24. 2005Shared-scale cycle indexes with companion oscillators
  25. 2005Current-bar versus prior-bar range construction for the stochastic oscillator
  26. 2005Two-session moving-average pullback short
  27. 2006Market condition as a permission layer for moving averages and oscillators
  28. 2008Lock the stop at support before sizing a stochastic entry
  29. 2010Construct a center-line volume oscillator and read it with a stochastic oscillator
  30. 2010Sharpened RSI turns with rainbow averages and a slow stochastic
  31. 2011Build a Spearman rank oscillator from ordered closes
  32. 2012Gold as a regime-dependent hedge in the euro-area crisis
  33. 2012Pairing moving averages with variable-length stochastics
  34. 2014Two-leg stochastic stress oscillator as a rebuild drill
  35. 2014Ingress dates as price bases for relative strength, stochastics, and moving averages
  36. 2017Constructing a dual EMA stochastic from range normalization
  37. 2018Constructing a two-stage stochastic RSI for comparable price-oscillator divergences
  38. 2018Combining a weekly stochastic, a long moving average, and two-day resistance
  39. 2018Weekly and daily stochastic readings with a long moving average and support
  40. 2018A confirming workflow for rotating from discretionary to staples
  41. 2019Stochastic scan thresholds, averages, and formula syntax
  42. 2020Constructing Slow %K as a two-stage helper
By fieldIndicators and filters682 readingsBy year200176 readings
All 170 readings tagged Stochastic oscillator
Also on Stochastic oscillator5 readings
1982Three gates on a 1982 pork-belly short1984Constructing the slow stochastic from a five-session range1985Daily ARIMA range as a filter for intraday stochastic divergence1986Auditing stochastic crossovers with moving-average baselines1987Stochastic fast and slow construction as a rebuildable stack