Investment concentration and crowding
- are ‘FAANGs’ a unique market phenomenon?
June 2019 | Share this article
Kristian Rung Weeke
Head of RFP and Client Intelligence, Sparinvest
In this 'Insights' paper we take a closer look at the concept of investment concentration and crowding. Whether it occurs in single stocks like the FAANGs, in sectors or in factors, the tendency for investors to flock to what is popular poses the risk of inflating prices to potential bubble proportions. We examine the uniqueness of the ‘FAANG’ concept from an historic perspective, the previous performance of this ‘big and pricey’ segment and the risks inherent in such concentrations of capital.
- The current cumulated index weight of the most expensive ultra-large caps (e.g. FAANGs) is not at a historical high when compared to previously occurring ‘top 5’ stock concentrations
- The valuations of the biggest stocks are elevated and the inter-stock correlation of the ‘big and pricey’ segment is high compared to previously
- Thus while investor crowding today - measured by simple factors like market caps - has not exceeded levels seen in previous bull markets, key difference lies in the types of companies behind today’s crowded stocks. These are fundamentally different to previous ‘market darlings’ in a number of ways: (1) some of the companies are relatively new, (2) there are fewer industrials amongst them and (3) they are potentially more homogeneous (tech-skewed). This poses a risk, with some of the risk indicators showing up in valuations and inter-stock correlations
- The important lesson from history is that these risk indicators can be connected to higher risks of large future drawdowns
- Empirical evidence suggests that ‘big and pricey’ stocks underperform the market, and factor crowding increases risk of large drawdowns
Setting the scene
As per 31 May 2019, sector allocation to Information Technology in the MSCI US Index was almost 22%. Together with Financials, these two sectors - out of 11 - account for roughly 35% of the index. Looking at the top 10 index stocks, we find that Apple, Amazon, Facebook, Google and Microsoft (roughly speaking the FAANGs, but with Microsoft instead of Netflix) account for a nearly 15% of the total index of more than 600 constituents. This is an obvious case of single name concentration and investor crowding in the sense of huge amounts of capital chasing the same securities or sector. Both history and financial academics have pointed to an increased risk of subsequent large draw-downs following large concentrations in factors, sectors or single stocks.
The very grouping of the FAANGs back in mid-2017 and their abbreviation into a single word signals something unique. As companies, they might very well be unique. But as a group of investments, they are essentially the result of: 1) the tendency for investors to ‘crowd’ and 2) a market definition, where market cap weighted indexes – being the basis of the market – by definition bet on past or recent winners. Today these colossal five are an unavoidable part of any well-diversified portfolio and, whether you as a portfolio manager like them or not, you have to consider them as they make up a considerable part of the market that you are referencing. If you like them, you must still consider the risk - at least from an absolute return perspective - that these tech-skewed, monopoly-like companies could run into some interconnected obstacle, get disrupted themselves or simply prove unable to keep up accelerating growth rates. If you don’t like them, you still also need to consider them, as the risk of not owning them leaves you with the risk of being left behind if they continue to rise.
In this paper, we measure crowding by various measures, taking a step-by-step approach. We begin by analyzing the degree of crowding using simple measures: relative size of the largest stocks and costliness (measured by P/E) of these stocks historically. This gives us an intuitive measure of concentration and crowdedness in these stocks. Then we look at the correlations between the largest stocks. If stocks are heavily followed by investors, they will tend to move more together causing the inter-stock correlations to rise. Finally we discuss more complex measures of crowdedness, based on multi-metric models.
Concentration and valuations
Although the FAANG phenomenon today seems unique it is not unprecedented. The formation of large market concentrations in only a few stocks is far from new. Back in the seventies, the ‘Nifty Fifty’ stocks ruled the day. These 50 stocks were a group of premier growth stocks, such as Xerox, IBM, Polaroid, and Coca-Cola, that became institutional darlings. All of them had proven growth records, continual increases in dividends, and high market capitalizations. They were often called ‘one-decision stocks’ (buy and never sell), because their prospects were so bright. They peaked in the early seventies priced at an average P/E ratio of 42 in 1972. After the 1973–74 bear market slashed the value of most of the Nifty Fifty, many investors vowed never again to pay over 30 times earnings for a stock.
Fast forward to today, where we are looking at the FAANGs priced at an average P/E of close to 60, compared to 22 for the S&P 500. No doubt they are pricey – but what about the degree of concentration? Exhibit 1 below shows the aggregated market cap of the top-5 stocks as a % of the total US market (with the top-5 names highlighted below the graph). It is obvious that although the current top five share of 12% is a large concentration, it is in no respect extraordinary in historical perspective. The concentration in 1999-2000 was higher – topping at 14%. Going back to the eighties, the top five constituted more than 20% of the market. The difference between then and now relates more to the high number of relatively new, tech-related stocks, the low number of industrial companies and the relative high pricing of the elite five.
Exhibit 1: Top 5 market cap stocks as % of all US and median P/E
Source: Sparinvest DEFCON database. Median P/E are year-end.
So what about the historic long term performance of the ‘big and pricey’ stocks? Does this justify the recurrent flocking of investors into these stocks? This is examined in Exhibit 2 below, showing the inferior performance of this segment compared to the market in terms both of lower returns and higher volatility.
Exhibit 2: Top 5 market cap stocks as % of all US and median P/E
||‘Big and pricey’ stock segment TR
||S&P 500 TR
Note: CAGR is the compound annual growth rate. The defined ‘Big and pricey’ segment performance is calculated by taking the median performance of the most expensive half of the 10% largest stocks, measured on market cap, in S&P 500 at any time
Source: Fama/French Database and Sparinvest calculations
So why do investors keep targeting this segment? Apart from the obvious “Flock mentality” of the market, one important (and potentially dangerous) reason is the very way in which market indices are constructed. Market cap weighting encourages benchmark-aware investors to allocate to these stocks, effectively boosting the concentration even more. And this spiraling effect gets even more pronounced with investors pouring more and more money into index funds. During 2017 alone, more than $692 billion flowed into index funds across all asset classes, compared to $7 billion in outflows from actively managed funds1. The result of this ‘passive flocking’ might be higher degrees of crowding in the future, followed by higher risk of drawdowns and more volatility in the market.
Getting back to the question of current crowdedness, we now look at correlations. As mentioned, these can serve as a measure of crowdedness because if stocks are heavily followed by investors, they will tend to move together more, causing correlations to rise. Perhaps even more importantly, if correlations of crowded stocks rise, the risk of inter-correlated drawdowns gets higher.
Exhibit 3: Average pairwise correlations of top-10 shares in MSCI US
Source FACTSET and Sparinvest calculations. Data are until end 2018. The graph shows the average pairwise correlations of the 12-months trailing returns of the rolling top-10 shares in the indices. The end-2018 top-10 in the index were: Alphabet/Google, Apple, Microsoft, Amazon, Facebook, Johnson & Johnson, JP Morgan, Exxon, Bank of America and Visa. Google is only included once in this calculation and not (as in the indices with both the (A) and (C) shares). The correlations are based on equal weights of the 10 constituents.
Exhibit 3 above shows the average pairwise correlations of returns of top 10 shares in MSCI US in yellow. The graph shows that we are in fact at very high – and volatile – levels of correlations compared to previously. This tendency is even more pronounced when compared to the average US equity market correlation using the Cboe S&P 500 Implied Correlation Index2. This has been falling steadily since 2011. Although the two measures are not directly comparable (one being historical and the other being implied forward-looking) it does give us an indication of the current correlation of the top 10 stocks being at a relatively high level compared to the average market correlation.
Multi-metric crowding models
Above we have been looking at crowding based on relatively simple measures. We now turn to more complex multi-metric models based on the approach presented by Nicolas Rabener of FactorResearch.com in a August 2018 paper. This model uses a five parameter setup to estimate an integrated normalized score (z-score) of crowdedness in factors based on how extreme the value of the underlying parameters are. The model uses valuation, dispersion, correlation, momentum and volatility as input factors to explain crowdedness.
The model in the paper estimates a large degree of crowdedness in the Technology sector and a significant risk of a 15% or larger drawdown over the next 12 months connected to this crowdedness3. The results are shown in Exhibit 4 below.
The top graph shows an extreme degree of crowding both in the tech sector and in the momentum factor (with values above one being crowded). A large part of this crowding comes from extreme valuations (dark blue in the graph). The bottom graph estimates a considerable (50%) risk of a 15% or larger drawdown over the next 12 months following this crowdedness. Although this risk has been mitigated by sell-offs since the publication of this blog post, a considerable degree of tech crowding and risk must be expected to remain.
Exhibit 4: Multi-metric model of Tech sector crowding
Crowding Model: Score Dashboard
Tech Sector Crowding: Probability of a Drawdown of 15% or More
Source: Factor Research
2 The Cboe S&P 500 Implied Correlation Indexes is a market-based estimate of the average correlation of the stocks that comprise the S&P 500 Index (SPX). Source: www.cboe.com/ImpliedCorrelation.
This material does not constitute individual investment advice and cannot form the basis for a decision to buy or sell (or an omission thereof) of securities. The material has been prepared for information purposes only and investors are encouraged to seek necessary professional advice before buying or selling securities. Sparinvest does not undertake any responsibility for the advice given and actions taken or not taken in respect of this material. There are always risks involved when investing and it is stressed that past performance or past return cannot be considered as a guarantee for future performance or return. Investors may not get back the full amount invested. Sparinvest makes reservations for possible typing errors, calculation errors and any other errors in the material.