Telling Tales – 2019 update

This article provides updated Telltale charts, including returns up to 2019. It analyzes the relative past performance of value and size factors compared to the total US market, as well as studying international and real estate market segments.

Using Telltale charts can be very informative, truly ‘telling the tale’ of what happened over time to portfolio trajectories, illustrating return to the mean properties or lack thereof.

Introduction to Telltale Charts

The following introduction comes from the Bogleheads wiki page about Telltale Charts.
In June 2002, John Bogle published his seminal article titled “The Telltale Chart”. In this article, he began by:

…criticizing the vastly over-simplified but typical way we look at long-term results. We hear, for example, that “small-cap stocks outperform large-cap stocks by almost two percentage points a year,” […] without acknowledging that each and every comparison we see is period-dependent. Whether or not the period has been selected to prove a point, neither the starting date of a comparison nor its concluding date are random.

He was referring to typical growth charts, which are indeed very sensitive to start and end dates, and tend to hide the periods of time that make a big difference to the trajectory. He then suggested the following:

[…] that imposing chart… conceals more than it reveals. I would strongly urge you to not accept that conclusion without transforming it into the telltale chart that is devised simply by dividing the cumulative returns of one data series into another, year after year – in this case dividing the cumulative large-cap stock return into the cumulative small-cap return.

In other words, a Telltale chart shows the aggregate growth of one or several investment assets (e.g. portfolio, fund, individual stock) divided by the aggregate growth of a benchmark asset.

Mathematically speaking, this is a simple way to normalize performance and display it in a relative manner. Telltale charts are very handy because they allow to capture on the same visual the overall cumulative trajectory (from start to end) AND what is happening year over year, therefore allowing one to grasp the big picture as well as the underlying dynamics. It takes a bit of effort to get the hang of it, but it is very informative.

Simba Backtesting Spreadsheet

Simba’s backtesting spreadsheet is a spreadsheet originally developed for the purpose of acting as a reference for historical returns, and analyzing a portfolio based on such historical data.

In the spreadsheet, Telltale charts allow to compare the trajectory of historical returns for various portfolios of interest to a Telltale benchmark (itself a portfolio), in a relative manner. It is interesting to observe that growth charts are actually a special form of telltale charts, and can be generated by the same tool.

The graphs provided in this article have been generated using a customized version of such a spreadsheet, based on revision 19b. Data sources are documented in the spreadsheet and typically anchored on passive index funds from Vanguard and corresponding indices. When index data is not available, annual returns are derived from the Fama-French data library, while applying breakpoint criteria consistent with modern indices (e.g. CRSP), staying as close to real life as possible.

Please note that there is no shortage of variations on value/size breakpoints and related indices. Findings in this article are obviously based on specific data sets. High-level conclusions should apply to a much broader context, but exact numbers (premiums/volatility/etc) will vary.

Value and Size Factors

The following Telltale chart compares the trajectories of portfolios made of a single (US) asset class, namely Large Cap Value (LCV), Mid Cap Value (MCV), Small Cap Value (SCV) and Small Cap Blend (SCB), while the Telltale benchmark is the US Total Market (TSM). Therefore the charts display the growth of an initial investment relative to the growth of the US total market.

The leftmost chart shows the relative trajectory in recent decades (since 1970) of an initial investment of $100. Note that the vertical axis is linear, for better readability. The rightmost chart shows the trajectory since 1927. Note that its vertical axis is logarithmic, to ease the comparison between time periods. Click on the chart for a larger display. A few observations:

  • Only MCV and SCV displayed a clear & sustained premium over TSM, which kept increasing over time thanks to occasional spurts of accelerated growth (early 40s, late 60s, late 70s, early 2000s).
  • LCV didn’t display any meaningful premium, it is the only asset class that seems to ‘return to the mean’ (if the mean is defined as US Total Market).
  • SCB displayed a nice premium, but mostly due to two time periods (early 40s and late 70s). It stayed roughy on par with TSM for two periods of more than 30 years.
  • The MCV trajectory was quite remarkable, displaying a strong premium and less of a roller coaster than SCV.
  • There were long periods of time (decade or more) during which both MCV and SCV underperformed or did not display any advantage, strongly challenging the fortitude of value investors.
  • Size and value effects are clearly linked to the hip, and only a combination of both made a true difference.

Leveraging other capabilities of the Simba spreadsheet, let’s look at a couple of additional charts illustrating the past performance of those asset classes.

The first chart compares standard deviation (volatility) as one possible measure of risk to returns (nominal) for the 1970+ time period. The second chart illustrates commonly used risk ratios for the 1927+ time period (please check the Simba spreadsheet’s glossary for definitions and pointers to more literature). Both charts further illustrate the excellent track record of MCV and SCV.

More about Mid Cap Value and Small Cap Value

Since both MCV and SCV asset classes displayed such impressive performance in the past, let’s dig a little further and look at the corresponding trajectory under a different angle, using more realistic portfolios, rebalanced on an annual basis. Let’s use as a benchmark portfolio the traditional 60/40 portfolio (60% Total-Market, 40% Total-Bonds), US only.

Here are the portfolio nominal drawdowns (based on annual returns, so this doesn’t show intra-year events), comparing a 60/40 portfolio where we replace TSM by either MCV or SCV while keeping 40% in bonds. The SCV drawdowns were a smidge deeper. Note the interesting trajectory during the Internet crisis (2002), where TSM fell hard while MCV and SCV’s fall was much milder. Unfortunately, this didn’t occur again during the financial crisis (2008/09).

Now let’s look at all 30-years cycles which occurred since 1927, plotting the starting year of such cycle as the horizontal axis, and looking at the real (inflation-adjusted) CAGR over the next 30 years. The pink dots (TSM as stocks) were always below the blue dots (MCV or SCV as stocks), without a single exception. It is a pretty amazing chart. In the case of the MCV 60/40 portfolio, the premium was remarkably consistent, between 1% and 2% in most cases. Unsurprisingly, replacing TSM by LCV would not have shown any meaningful improvement.

There was a price to pay though, as illustrated by the following chart, computing the standard deviation (volatility) of annual returns for the same cycles. In the case of MCV, the increased volatility wasn’t that large, but in the case of SCV, it was much more stomach churning for most cycles (much less in recent years though).

Finally, let’s look at the maximum withdrawal rate for the same cycles. This is an interesting metric as it captures absolute returns as well as sequence of returns risk. Unsurprisingly, retirees would have been better off with value-oriented 60/40 portfolios. This would have been difficult to predict without hindsight though, so only retirees using sound adaptive withdrawal methods would have fully benefitted.

Of course, it may seem excessive to fully replace a Total-Market position by a Small Value or Mid Value position. Vanguard TSM’s holdings diversify across more than 3,000 individual securities, while MCV uses ~200 and SCV uses ~850 at the time of writing. One could get rightfully nervous about the low diversification level of MCV (although growth and value companies tend to trade places, but the mid-cap space is still quite limited, ~350 securities).

On the other hand, the securities used by MCV and SCV usually have very little bearing on the TSM trajectory, due to cap-weighting. Tilting one’s portfolio (replacing a portion of TSM equity by MCV and/or SCV) would then seem an attractive approach to further diversify sources of returns and add a little ‘oomph’ to one’s portfolio. Or at least it would have been in the years that we studied. Disclaimer: the author uses this exact approach!

International and Real Estate

The following Telltale chart compares the trajectories of portfolios made of a single asset class, namely International Small Caps, Emerging Markets, Total International, and US Real Estate Investment Trusts (REITs), while the Telltale benchmark remains the US Total Market (TSM).

The chart shows the relative trajectory in recent decades (since 1972) of an initial investment of $100, compared to US Total-Market (TSM). Note that the vertical axis is linear, for better readability. Historical returns aren’t available for Int’l Small and Emerging for the early 70s, so they were approximated to Total International, to allow a consistent display on the chart.

A few observations:

  • Both Total International and Emerging alternated periods of faster growth than TSM with periods of slower (relative) growth. Inferring a ‘return to the mean’ hypothesis might not be unreasonable in this case. In other words, be careful to not fall for recency bias, overreacting to the last decade of underperformance.
    • Even if no sustained premium is visible here, trajectories were quite uncorrelated, hence presenting an interesting diversification opportunity.
  • International Small appeared to display a large relative growth in the late 80s, then more recently fell back under TSM, probably crushing the hopes of corresponding investors.
    • Note that some active funds (e.g. Vanguard VINEX) displayed a better (and sustained) performance than passive funds in this market segment. So far!
    • Please take the Int’l Small numbers in the 80s with a grain of salt, there are some doubts about the reliability of the corresponding data source.
  • As to US REITs, two time periods (late 70s and early 2000s) created a nice premium without too much turbulence on the way. Now if you compare to the MCV and SCV trajectory, you’ll notice that the periods of growth were the same, but the REIT premiums were lower. It appears that REITs often behaved as value investments, but apparently not the ones which displayed the best performance.
    • Vanguard VGSLX (REITs) holds less than 200 individual securities. It is squarely a sector investment. To compare, Vanguard VHCIX (Health Care) holds around 400 securities, Vanguard VGENX (Energy) around 125.

Now let’s look at charts illustrating risks and returns. The first chart compares standard deviation (volatility) as a measure of risk to returns (nominal) for the 1970+ time period. The second chart illustrates commonly used risk ratios. The results of the second chart seem actually quite counterintuitive in this case, illustrating the debatable nature of such ratios in the author’s opinion.

Stocks/Bonds Portfolios and Final Words

To finish, let’s take a quick look to Telltale charts illustrating the trajectory of multiple portfolios mixing US stocks and bonds to various degrees (from 30/70 to 90/10). In this case, the Telltale benchmark is the classic 60/40 portfolio.

The leftmost chart illustrates the relative trajectory for 1970 to 2019 (50 years) of an initial investment of $100. The rightmost chart illustrates another period of nearly 50 years, starting from 1927 and finishing in 1974.

One could infer some tentative conclusions from one of those two charts, and see them invalidated on the other one (give it a try!). Hence a word of caution to finish this article.

Telltale charts are very informative for sure, but the human brain has this tendency to perceive patterns and jump to conclusions a little bit too fast… It is very unfortunate that there is no publicly available data allowing to backtest the value and size factors outside the US, or to backtest international and REIT returns before 1970, in order to provide ‘out of sample’ analysis. In the mean time, one is left to make a judgment call about the possible applicability of such findings in the future. Using a big grain of salt to do so seems in order.

Addendum

Based on feedback received on the Bogleheads forum when the first version of this article was published, a few more charts were created.

More about the size factor

Here is a Telltale chart focusing on the size factor, including Large Cap Blend (LCB), Mid Cap Blend (MCB), Small Cap Blend (SCB), and to put things in perspective, Mid Cap Value (MCV). All relative to US Total-Market (TSM).

Unsurprisingly (given the respective market weights), LCB stayed very close to TSM. The MCB trajectory had a rough start for decades, but did create some nice premium since 1970. The MCV and SCB numbers are the same as previously illustrated, with a nice upward trajectory, although SCB didn’t do much (relative to TSM) in the past three decades. As previously mentioned, combining size and value is what made the biggest difference.

Exchange rates and currency considerations

When looking at return numbers expressed in US dollars and comparing the trajectory of International funds to the US market, we can hypothesize that the strength or weakness of the US dollar compared to international currencies is quite impactful.

The Federal Reserve tracks the “Foreign Exchange Value of the Dollar” where the weights of the currencies of a “large group of major U.S. trading partners” is turned into an index (dubbed H.10). The index weights, which change over time, are derived from U.S. export shares and from U.S. and foreign import shares. Monthly values for the H.10 index are found on the Federal Reserve Web site. Values arbitrarily started at 100 (when the index was created in March 1973; it’s been recently re-aligned to Jan 2006 with a slight change of rules), a high value indicating a ‘strong’ dollar, a low value indicating a ‘weak’ dollar.

The following Telltale chart displays the Total-International performance relative to the US Total-Market (vertical axis on the left), while adding a separate data series (vertical axis on the right) for the H.10 index. We can clearly see negative correlation (-0.4) between the two lines of the graph, consistent with the hypothesis. Also, the H.10 historical average so far is close to 100. Exchange rates definitely have a strong bearing on returns while displaying some rough ‘return to the mean’ properties. This should put in perspective the perceived recent over-performance of US stocks.