Dot-Com Redux
How to invest after valuation risk strikes
By: Nick Schmitz
Leading up to the dot-com bubble burst, small value stocks dramatically underperformed large growth stocks for years.
Below are the indexed factor premia for the five years up to year-end 1999 for buying small stocks instead of large stocks and value stocks instead of growth stocks.
Figure 1: Indexed Global Factor Premia (Through Year-End 1999)
Source: Ken French Library
After five years of market exposure, small caps and value stocks compounded dramatically less than large caps and growth stocks across all three developed markets.
If that looks extreme, it's eerily similar to what has happened again over the last few years.
Figure 2: Indexed Global Factor Premia (Last Five Years)
Source: Ken French Library
2022 broke the trend, and some have advocated “buying the dip” in large growth. Most of the FANMAGs have experienced almost bankruptcy-level drawdowns of late, so why not wade back into the growth waters now?
But if you’d “partied like it’s 1999” as a growth investor, how long would the hangover have lasted?
We wrote about this in late 2020 in our piece on the "blunt ratio" (critiquing the Sharpe ratio). There we noted how “stable” growth was at staying down for decades at a time (after the Nifty Fifty and dot-com bubbles burst), thereby producing a pretty good Sharpe ratio compared to deep value, with it’s much more volatile bounce backs.
Here, however, we look at the upside small value. We show the same metrics as the charts above for the 22 years following 1999 (even though this period ends in another large growth bubble).
Figure 3: Indexed Global Factor Premia (Since 1999)
Source: Ken French Library
The hangover for large growth following 1999 now looks more like chronic liver failure to us. Indulging in the valuation risk euphoria of the late '90s was a “moment on the lips and a lifetime on the hips” in the grand scheme of investing.
The pain was just getting started in year one of the extremely compressed x-axis above. Past large growth bubbles took about 10-15 years to fully unwind, leaving about 70-90% of aggregate wealth generation potential on the table. This duration and depth would also include the Nifty Fifty bubble and, of course, Japan’s 1989 asset bubble, when valuation dynamics were reversed across the Pacific.
But this is a mere observation about the distributional history of returns. In financial markets we can guess at the cause of these symptoms. There were specifics to each bubble, of course, but what unified all of them were stretched valuations and margin mean reversion while long-term forecasts remained as deeply uncertain as they always are.
Virtually everywhere we’ve looked across assets, geographies, and market history we’ve found that the risk of catastrophic long-term capital impairment has been correlated with stretched valuations and stretched growth expectations. Conversely, long-term premiums have been correlated with compressed valuations and compressed earnings margins, in our research.
Before this reversal started, we focused somewhat obsessively on these topics across IPOs, across geographies here and here, and within the cross-section of stock returns here, here, here, here, here and here.
So after writing about it non-stop from 2019 through 2021, where are relative and absolute valuations currently?
For value versus growth, the state of relative valuations was well summarized in AQR’s year-end note: “value’s returns in 2022 were extremely strong, and the spread only moved from roughly the 100th percentile [of market history] to the 94th, which makes us very optimistic about the prospects of continued normalization in 2023 and beyond.” For small caps versus large caps, we recently wrote a piece on the topic showing just how much room we believe there is to run here in figure 4.
And 2022’s down market produced more attractive absolute valuations, in our opinion. But despite a 20% down year for global markets, not everywhere looks attractive as a result of the discounts. 2022 saw the FANMAGs “dip” a massive 50-60% from their peaks. However, even after that, at an average price tag that’s still 5x revenue, the FANMAGs are still about 500% more expensive than Japan’s TOPIX index at 1.1x revenue—even though the FANMAGs saw a a 10% drop in earnings while Japan’s operating income was up over 5% last year.
We think there’s ample room for continued normalization of multiples that should continue to support small-cap and value stock returns globally in the decade ahead.
Many might think this is less relevant to them if they don’t actively target growth and just buy the whole index. However, the question is perhaps more salient for many investors than they know. Most all passive retail ETF products are capitalization weighted. Larger stocks take up a larger proportion of the invested money compared to smaller stocks. Popular “total market” indexed funds typically invest ~70% of every dollar in large cap stocks, not in a balanced exposure to the total market of stocks. Cap weighting works very well during a large-cap growth rally. But it concentrates more and more of investors' exposure in just a few active bets as the valuations of the biggest stocks go up, just as they have in recent history.
But this can be catastrophic, according to our past research, when valuation spreads mean revert as they did after the last tech bubble. In 2020, we ran a simulation that showed popular total market indexes with capitalization weighting would have lost 9.3% per year when valuations normalized fast (1999-2004) compared to equal-weighted versions of the same index constituents (this is without microcaps in either index).
And once again, this appears to have begun to occur in 2022. The S&P 500 index was down 20%, but the equal-weighted version of the S&P 500 was only down 13% (SPX vs Invesco: RSP).
If simply avoiding valuation risk sounds like an overly simplified investment philosophy that skips over much of the intricate, case-specific nuance of industries, eras, and mega-trends... it admittedly is. It’s also differentiated from the standard forecasting approach in bubble markets, in that it would have worked decade after decade, historically, in our research.