Yesterday's post presented an analysis of floor and ceiling levels for the percent of stocks to be held in a rebalanced portfolio. Today's post presents a similar analysis for a portfolio that takes a different strategy. Instead of rebalancing toward a target percentage, it rotates strongly to stocks or to bonds, depending on which asset has recently had better returns. How strongly? Until this post, the working assumption has been to rotate the entire porfolio, going 100% to stocks or 100% to bonds. Today, the topic is whether it's best to rotate stocks to a floor higher than 0% or a ceiling lower than 100%. All these analyses are part of an on-going series that began on June 3, 2019, comparing traditional rebalancing with the alternative strategy of rotation.
The chart below summarizes 48 combinations of floor and ceiling values for rotating the portfolio between mostly or entirely stocks and mostly or entirely bonds. (For specifics about the data, see "Some Technical Details" at the end of this post.)
How do similar combinations of floor and ceiling values effect a portfolio's 25-year returns, adjusted for inflation? The next chart illustrates the main results, annualized.
The average in the chart, representing a ceiling near 70%, paid handsome returns of about 6% above inflation when the floor was 10%. That result was quite close to the performance of rotating from a ceiling of 100% to a floor of 40%. As illustrated by these two examples of 70-10 and 100-40, different combinations of floor and ceiling values may produce similar long-term accumulations. They will differ, however, in potential drawdowns. That necessitates an analysis of relative returns, adjusted for drawdown risks.
The story gets a bit more complicated at this point. The historical evidence indicated that three factors had to be considered together: drawdown risks, potential for real returns, and the window-size for comparing whether recent outcomes had been better for stocks or for bonds. To illustrate these dimensions, the chart below shows examples of relative returns (risk-adjusted) when large-cap stocks and 10-year Treasuries were compared over three-month and twelve-month windows.
Several points in the chart deserve emphasis:
- The relative returns for rotation portfolios were consistently high, generally above 1%. In contrast, previously reported results for rebalancing ranged from 0.3% to 0.8%. Thus, overall, rotation portfolios in these data-sets produced better risk-adjusted returns than rebalanced portfolios. (As will be noted in future posts, exceptions to this conclusion are rare.)
- Relative returns tended to be higher when the floor percent was lower. On reflection, this result makes sense. The trigger for rotating out of stocks and into Treasuries is that Treasuries, based on recent performance, are giving better returns; furthermore, their drawdown-risk is low. Therefore, reducing stock-exposure by a larger amount at such times should mean a greater reduction of potential losses. And the historical evidence says that it did.
- The best results came from portfolios with a ceiling at 50% in stocks, with a floor of 0% or 10% and a reference window at or near 3 months. Although not shown in the chart, a ceiling below 50% had somewhat lower relative returns. Very high ceilings, such as the 100% example in the chart, stayed close to the average outcome. The sweet spot may be about 50% to 60%.
The lessons to be drawn from these findings depend on your goals as an investor:
- If your main concern is to limit drawdowns, your lower boundary for a rotation portfolio should be 0% to 15%. This finding begs the question: "Which is better? Holding a rebalanced portfolio with a floor at 10%, or rotating a portfolio around low percentages, such as a floor of 0% and a ceiling of 15% to 25%?" My next post will examine exactly this question.
- If your main concern is to maximize real returns, then rotate your portfolio from high exposure to stocks (70% or more) to low exposure (0% to 10%).
- If you prefer a compromise on boosting returns while also limiting drawdowns, your best strategy is to rotate between about 50% to 60% in equities when they are strong, and 0% to 10% when they are weak, with the remainder in bonds. Another question is begged: "How do rotations like 50-to-0 or 60-to-10 compare to rebalancing at, say, 40-60 or 25-75?" As long-term portfolios, historically, all of these offered good risk-adjusted performance. But is any one of them distinctively better than the others for an investor who wants to optimize relative returns? This, too, will be covered in upcoming posts.
These analyses aggregated statistics and data from a large number of observations, which could be categorized in several ways:
- Four countries: the U.S., Japan, Germany, and Australia. For all these countries, public monthly data was obtainable for (a) the total return, including reinvested dividends, for an index of large-cap stocks going back at least 25 years in local currency; (b) at least 25 years of dividend history for country-specific, constant-maturity, 10-year Treasury bonds in local currency; and (c) at least 25 years of inflation history as measured by a country-specific index of consumer prices.
- Seven 25-year periods. Four quarter-centuries were available for the U.S., between April 1919 and March 2019, and one each in Japan, Germany, and Australia from April 1994 to March 2019.
- Different reference windows. Generally, the analyses presented here are averages from comparing bond-returns and stock-returns over windows of three months or twelve months. Unless otherwise noted, the results were the same for these two windows. Other analyses, not presented here, also considered windows varying from one to twelve months, plus combinations of multiple window-sizes.
- Floor-ceiling settings. The results in this post are from 48 pairs with floors of 0%, 10%, 15%, 20%, 25%, 30% or 40% in stocks and ceilings of 40%, 50%, 60%, 70%, 80%, 90% or 100% in stocks. These seven floors and seven ceilings generated 49 possible combinations. However, one of them, 40% floor and 40% ceiling, is really a rebalancing strategy, not rotation, and was therefore excluded.
All together, the seven sets of 25-year periods for different countries, the two reference windows, and the 48 floor-ceiling settings generated 672 observation points, each of which had multiple statistics such as annualized nominal and real returns, maximum 25-year downdowns, worst-month drawdowns, and more, both for the large-cap indexes and for country-specific Treasuries.
Primary data-sources were the St. Louis Federal Reserve for Treasuries, inflation, and some stock indexes; Yahoo Finance for the historical total returns of other stock indexes; and older historical data for the U.S. as published by Robert Shiller. Data published directly by index-providers, such as the Deutsche Borse and MSCI, was sometimes used for verification or to fill gaps.