This post is available as a PDF download here.

Summary­­

  • Traditional value strategies simply sort the investment universe based on one or more valuation metrics (e.g. book-to-market, price-to-earnings, etc.) and purchase the securities that look the cheapest.
  • However, this process is often prone to structural sector bets, which are uncompensated sources of risk within a strategy.
  • By comparing the value of stocks within each sector along with the value of stocks among sectors, we can move beyond the simple value methodologies that are always over- or underweight certain sectors.
  • This allows us to develop value strategies that either maintain sector neutrality or tactically rotate among sectors, both of which can be more useful in targeting specific portfolio objectives than the more traditional value approach.

 

While the family tree of value investing can trace its roots back to Graham and Dodd, it was not until after modern portfolio theory arrived on the scene that the notion of the “value factor” appeared in the literature.  This timeline is somewhat tautological, as violations of CAPM (i.e. factors) could only be discovered after CAPM itself was proposed in the 1960s.

For value, it was Basu (1977) who found a significant positive relationship between value-related variables – specifically earnings yield ratios – and average returns within U.S. stocks.  Later, book-to-price would be established as well.  Eventually, in 1992, Eugene Fama and Kenneth French would propose their Fama-French 3-Factor Model to incorporate the Size (large minus small capitalization stocks) and Value (low minus high book-to-market stocks) factors alongside the market factor.

In the same year, Morningstar introduced the style box: a nine-square grid, used to capture an investment’s style across two dimensions (large-to-small and value-to-growth). Originally, the value dimension was based upon price-to-earnings and price-to-book, which meant that “growth” was merely a euphemism for “expensive.”[1]

Under rational pricing models, growth and expensive may actually be synonymous.  For example, using Gordon’s growth model, we know that dividend yield (D/P) is equal to the expected (required) rate of return (k) minus growth (g).

Assuming a rational market, the expected return should be driven by sensitivity (beta) to risk premia (F), i.e.

Since expected return is driven by a stock’s risk exposure, a relatively expensive stock must either be one that is less risky (as measured by whatever factor model we are using) or growing very fast.  So, we’ll forgive the decades of authors who have used “expensive” and “growth” interchangeably.

In 2002, the Morningstar methodology was updated to use a “10-factor model.”  Value is now measured as a blend of price-to-earnings, price-to-book, price-to-sales, price-to-cash flow, and dividend yield scores.  Growth, on the other hand, is now measured based upon actual growth metrics: earnings growth, sales growth, cash flow growth, and book value growth.

Here, we’ll speculate that the dot-com era caused a redefinition of “growth” in popular nomenclature.  Indeed, growth on its own is not necessarily a bad thing: a cheap, growing company might very well be a fantastic investment!

But the value versus growth split permeated product construction.  As we outlined in our commentary Growth is Not “Not Value,”[2] index providers such as CRSP, Russell, S&P Dow Jones, and MSCI used the definitions to bifurcate the investment universe.  Value indices, then, were filled with stocks that exhibited positive value characteristics and negative growth ones; while Growth indices held the exact opposite.

The Confusing Narrative of Value and Growth

Splitting the investment universe into value and growth might make sense from a product sales perspective – after all, why would we want to sell products tracking two indices that hold overlapping securities? – it puts us in a bit of an odd place from an investment standpoint.  Or, at least, from a narrative perspective.

First consider that investment products exist for both value and growth indices that have been constructed with the bifurcation methodology.  If we believe in the value premium – i.e. “cheap” will outperform “expensive” – we really have two choices:

  1. We also believe high growth will outperform low growth, and therefore both value and growth indices are somewhat diluted by introducing the negative of the other.
  2. We do not believe growth screening will have any impact, in which case the growth index is just expensive stocks, which we expect to underperform.

Not a wonderful narrative in either case for those selling these products.

(We should note that there appears to be at least some evidence that the growth factor has positive returns, so we will, with a great deal of reservation, lean towards #1[3])

Second, in so far as we are investing in value because we believe there is a mispricing opportunity (i.e. the value premium is not just compensation for bearing excess risk) we are seeking to take advantage of, the double screen of value and growth means that value is filled with justifiably cheap stocks because we are constraining the universe to be low-growth companies.[4]  The opposite end of the spectrum, then, must be “growth at an (un)reasonable price” companies.  Neither sounds particularly appealing to us.

In our opinion, splitting the world into growth and value leads to very confused index definitions.

Structural Sector Bets in Value and Growth 1.0

A side-effect of the first-generation construction methodology is that it created structural sector over- and underweights in the value and growth indices.  By using “not growth” to define value and “not cheap” to define growth, sectors like Financials end up overweight in value and Technology overweight in growth.  This can lead to performance discrepancies that have little to do with the styles themselves, but are simply due to structural (and, arguably, unintended) sector bets.

We can explore the impact of this decision by decomposing the performance difference between the Russell 1000 Growth Index and the Russell 1000 Value Index into sector and residual components.  To do this, we:

  1. Find the historical quarterly sector weights for each of the indices.
  2. Calculate the long-term average sector weight of each index. Calculate a version of each index where we assume a static allocation to the average weight over time, implemented using the SPDR Sector ETFs.  The difference in performance between these two index versions captures the difference to due long-term, structural sector differences.
  3. Calculate a version of each index where quarterly sector weights are implemented using SPDR Sector ETFs. The difference between these two versions, minus the amount due to structural differences, captures how much performance differential is due to sector tilts away from the average weights.
  4. Finally, taking the performance difference between IWF and IWD and subtracting out the performance due to the long-term sector average differences and the sector changes over time leaves the residual: the performance difference attributed to actual security selection.

The results are plotted below.

Source: CSI Analytics, Bloomberg.  Calculations by Newfound Research.  Past performance is not a guarantee of future results.  Chart decomposes the realized return between the iShares Russell 1000 Growth ETF (“IWF”) and the iShares Russell 1000 Value ETF (“IWD”) from 6/30/2009 to 11/29/2017.  The starting point reflects the furthest date back for which quarterly sector weights were available for both IWD and IWF.  Details of analysis explained above.

Source: CSI Analytics, Bloomberg.  Calculations by Newfound Research.  Past performance is not a guarantee of future results.  Chart decomposes the realized return between the iShares Russell 1000 Growth ETF (“IWF”) and the iShares Russell 1000 Value ETF (“IWD”) from 6/30/2009 to 11/29/2017.  The starting point reflects the furthest date back for which quarterly sector weights were available for both IWD and IWF.  Details of analysis explained above.

We can see that the long-term, structural sector differences (the orange line) have largely driven performance differences over the last 8 years.  This evidence suggests that when you’re buying Value or Growth 1.0, you’re predominately making a sector bet, not a style bet.

Value 2.0

As factor investing has gained popularity, only value has remained in the pantheon of broadly accepted investment styles.  Growth has, however, largely fallen by the wayside (with the exception of showing up as a component in some quality factor definitions).  Accordingly, the definitions of value, in this world, have changed.

MSCI defines their Enhanced Value scoring methodology based upon P/E, P/B, and EV/CFO.  The Fidelity Value Factor Index uses P/FCF, EV/EBITDA, P/B, and P/E.  FTSE Russell defines their Value factor is P/CF, P/E, and P/S.  J.P. Morgan’s U.S. Value Factor Index uses P/E, P/B, P/FCF, and Dividend Yield.

Notably absent from any of these definitions is anything growth related: value is defined entirely on its own.  If it overlaps with a growth 1.0 index, so be it.

Unfortunately, this re-definition does not necessarily address the structural sector issues.  Stocks in certain sectors may be perpetually plagued with “expensive” labels simply because they are high growth stocks.  Technology, again, comes to mind.  In theory, a structural sector bet should be uncompensated.  After all, why would we expect one sector to perpetually under- or outperform?

As it turns out, evidence suggests that the value premium may actually be largely driven by cross-industry differences, not cross-market differences.[5]  In other words, instead of ranking an entire universe of stocks on the same valuation metric, stocks should only be ranked relative to their industry peers, allowing us to focus on picking the cheapest stocks within each sector.

This does not necessarily invalidate prior value research: rather, it might just be that prior value research was a noisy estimate of this more refined metric.  In other words, cross-market price-to-book may have historically worked (as evidenced by the success of the HML factor), but only because it was correlated enough with cross-industry price-to-book.

The cross-sector ranking allows for a sector-neutral implementation.  This means that the value index will always match the sector weights of its parent index (e.g. the S&P 500), neutralizing any unintended sector bets.  Research suggests that while these sector-neutral implementations tend to have higher information ratios, they also have higher turnover and lower annualized returns when compared against their unconstrained peers.[6]

The obvious downside to a sector-neutral approach is, however, in that the market dictates the sector weights, which can be dangerous in environments where the market is overly bullish or bearish on an entire sector.  A sector-neutral implementation of value during the dot-com era, for example, would have forced investors to place north of 30% of their portfolio into technology.

While structural over- and under-weights may not be compensated bets, our own research suggests that value-based sector rotation can be successful[7], indicating that tactical sector bets can potentially add value.  The Shiller Barclays CAPE US Index family is based on this very idea.  Importantly, however, sectors are not chosen based upon having the lowest absolute Shiller CAPE ratio.  Such a method would simply lead to structural over- and under-weights as before.  Rather, the Shiller CAPE ratio for each sector is first normalized against its long-term average.  We use a similar method in our own research, but leveraging current versus historical dividend yield as our valuation proxy.

Briere and Szafarz (2017) find that factor allocation is better for capturing risk premia while sector investing is better for managing risk through diversification, particularly in crisis periods.[8]  This lends credence to the notion that an unconstrained implementation may be sub-optimal in quiet periods, but help manage risk in crisis environments.

That said, Hughen and Strauss (2017) also find that investing in the cheapest stocks of the cheapest sectors outperforms simply investing in the cheapest sectors (firm neutral) or investing in the cheapest stocks (sector neutral) alone.[9]

Conclusion

While Value 2.0 may be a step forward from 1.0, there still seems to be significant disagreement in the details.  And the return devil is always in the portfolio details.

How, then, should we proceed?

First, we should recognize that all of this research agrees on one point: buying cheap stuff works.

We should also recognize that while value-based sector rotation seems to work, structural sector bets do not make much sense.  We likely want to avoid value products that are always underweight technology or always overweight financials simply because of how our valuation metrics are defined.

After that point, the question is not necessarily whether there is a single, objectively optimal implementation.  Rather, the question we have to ask ourselves is: “how do we want to use this strategy?”

Do we plan on employing leverage or taking a long/short approach?  Do we have a risk budget and want to maximize the return generated per unit of tracking error?  In such a case, focusing on an implementation with a higher information ratio may make sense.  Evidence suggests this would imply a more sector-neutral implementation.

Are we simply looking to maximize our compound annualized growth rate, with no consideration for tracking error?  Then we probably want to look at super-concentrated, unconstrained strategies.  Indeed, taking a portfolio view instead of a strategy view and combining an unconstrained value strategy with an unconstrained momentum strategy may allow you to benefit from higher compound growth rates while reducing tracking error through diversification.

Progressing to these more sophisticated value product – or any other factor product – implementations puts more burden on investors to evaluate these strategies and their proper use within a portfolio. However, having strategies that go beyond the 1.0 versions allows us to more finely target objectives in ways that are backed by both theoretical and empirical evidence.

 


 

[1] See methodology document here http://corporate.morningstar.com/US/documents/MethodologyDocuments/MethodologyPapers/MorningstarStyleBox_Methodology.pdf

[2] https://blog.thinknewfound.com/2016/02/growth-not-not-value/

[3] https://www.msci.com/www/blog-posts/has-the-growth-factor-earned-a/0635572694

[4] We should mention that a potential rebuttal to this idea is that by screening on low growth companies, we could potentially take advantage of investors’ anchoring bias, benefiting from surprise positive growth situations since investors are pricing in an overly pessimistic future.

[5] See http://rnm.simon.rochester.edu/research/ROCatXR.pdf

[6] See Finding Value (MSCI, 2015), Does a Sector-Neutral Value Strategy Help Reduce Risk? (Ned Davis Research Group, 2017), Low-Risk Investing without Industry Bets (Asness, Frazinni, and Pedersen 2014), and Smart Beta Efficiency Versus Investability: Introducing the Cost-Adjusted Factor Efficiency Ratio (S&P Dow Jones Indices, 2016).

[7] Does Sector Rotation Work? Newfound Research, 2015.

[8] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2965346

[9] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2699948

Corey is co-founder and Chief Investment Officer of Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Corey is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.

Prior to offering asset management services, Newfound licensed research from the quantitative investment models developed by Corey. At peak, this research helped steer the tactical allocation decisions for upwards of $10bn.

Corey is a frequent speaker on industry panels and contributes to ETF.com, ETF Trends, and Forbes.com’s Great Speculations blog. He was named a 2014 ETF All Star by ETF.com.

Corey holds a Master of Science in Computational Finance from Carnegie Mellon University and a Bachelor of Science in Computer Science, cum laude, from Cornell University.

You can connect with Corey on LinkedIn or Twitter.

Or schedule a time to connect.