Factor Investing: Evidence Based Insights

Factor Investing: Evidence Based Insights

June 22, 2017 Factor Investing, Research Insights
Print Friendly
(Last Updated On: June 24, 2017)

I will be talking on the Factor Investing panel at the upcoming Evidence-Based Investing Conference in Dana Point, CA next Sunday –Tuesday. I am excited for the opportunity to chat, and figured I would highlight a few thoughts we have on the topic going into the event.

First, what is “evidence based investing?” What does that even mean?

If we read a index factsheet, review a long-term live track record, or read an academic finance paper, are we now an evidence-based investor? No.

To me, evidence based investing means reviewing all the relevant research on a topic, viewing this body of research as a whole, understanding the economic/human underpinnings of the finding, and then making an informed decision that is as objective as possible. Achieving the title of being an “evidence based investor” sounds easy, but the reality is that this is incredibly difficult. This difficulty is only magnified by the sheer amount of information and (so-called) research being distributed into the marketplace. To solve this information overload problem investors often get complacent and rely on a single firm’s research (or research group) to inform their opinion on all matters. This approach will likely turn you into a faith based investor, not an evidence based investor (see here for a piece on the subject). And when it comes to factor investing, we often see well-meaning investors, who want to be evidence-based investors, quickly devolving into faith based investors who are no longer critically reviewing and assessing the available research.

With factor investing there are a few topics worth discussing:

  1. Factor Premiums — Are they here to stay?
  2. Factor Portfolio Construction – Does it matter?
  3. Factor Investing Funds – What are you actually buying?

Factor Investing Premiums – Are they here to stay?

Are factor premiums here to stay or have they been arbitraged away?

This is a common question we are asked. As advocates of factor-investing(1), how does one respond to the countless articles citing the evidence that factor/active funds are losing to the passive index?

Our response is not satisfying: active (factor) investing is simple, but not easy.

As we have highlighted in the past (I recommend you read the entire article), active factor portfolios are notoriously challenging for investors. Consider value investing, the most well known factor approach. Value investing has had numerous periods of underperformance, especially during the Internet Bubble. Using Ken French’s data, we examined just how painful it was to be a value investor in the late ’90s.

We examine the returns from 1/1/1994-12/31/1999 for a Value portfolio (High B/M quintile, VW returns), Growth portfolio (Low B/M quintile, VW returns), Risk-Free return (90-day T-Bills), and SP500 total return. Results are shown in Figure 1 below. All returns are total returns and include the reinvestment of distributions (e.g., dividends). Results are gross of fees.(2)

Figure 1: Summary Statistics (1994-1999)

Value and Growth 1994-1999
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Talk about a bad stretch! Looking at the annual returns (shown in Figure 2), value investing lost almost every year to the market!

Figure 2: Annual Returns

Annual Performance 1994-1999
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

What would their hypothetical returns look like in the long run? As you can see below in Figure 3, value quickly recovers and outperforms the entire time period thereafter. Here are the returns to the same portfolios from 1/1/2000 – 12/31/2014, the 15 years following the 5-year underperformance:

Figure 3: Summary Statistics (2000-2014)

Value and Growth 2000-2014
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Sticking with the value strategy, although painful, was richly rewarded with a 4.67%+ edge–per year–over the market benchmark from 2000-2014.

Over the entire cycle, patient, disciplined investors were also rewarded. Here are the results (Figure 4) over the entire time period, measured from 1/1/1994 to 12/31/2014:

Figure 4: Summary Statistics (1994-2014)

Value and Growth 1994-2014
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Similarly, here is a more recent example using momentum. A common argument is that there are too many supercomputers / Ph.D.s / CFAs in the market so momentum investing is dead. For example, here is a chart on the performance of momentum over the past several decades:(3)

Is momentum dead? Blue line is a portfolio long winners, the dotted line is a portfolio that is long losers. They are converging, rapidly. The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

The chart certainly follows along with the story — but correlation is not causation. Here is a chart extended back to 1801, which shows that decade long periods of poor relative performance are not uncommon.

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

If we subscribe to the theory that better technology, smarter people, and more competition are associated with arbitraging away the momentum factor premium, we also need to then explain why momentum went on a 10-year drought following the Civil War.

So why do we highlight how painful factor investing can be?

Any active factor investor should expect the following (in the short-run):

  1. They will lose to the market with a 100% probability over some time period.
  2. They will experience a multi-year relative performance drought.

If factor investing were easy, and allowed everyone to beat the market every year, the premiums would certainly disappear — no pain to invest in factors, but all gain…why not!

Market experience suggests that markets are very competitive so situations of no pain, but all gain are transitory. To the extent that one believes a factor has a premium above and beyond the market return, one needs to identify the pain associated with investing in a factor — and be ready to embrace it. This pain could be higher fundamental risk and/or some relative performance pain, but there has to be a catch for it to be sustainable.

Once one understands the potential pain of their factor investments, they can identify if they are well suited to take on this pain to reap the potential gains. Some people are, but most are not. For example, if an investor is going to examine their Value portfolio in 2 years and compare it to the market, we would highly recommend they switch to passive Vanguard funds.(4) Otherwise, the risk of this investor selling a factor at the wrong time is too high. These investors will end up contributing to factor premiums, not exploiting them.

But back to the question: will factor premiums exist in the future? We’ve already highlighted that factor investing is noisy, and takes place over long time cycles. In addition, academics have good arguments that factor premiums are driven by a combination of (1) poor investor behavior(5) and (2) a risk premium, i.e. the stocks are inherently riskier(6). So to sit here today and say that factor premiums will be zero in the future, you must have expert knowledge on (1) investor behavior and (2) investor risk preferences.(7)(8)

Now here is another practical application. Consider an investment in a Vanguard US Index fund. Few would question this purchase. However, this investment is implicitly betting on the US equity risk premium, or the spread in returns between equities and bond/cash investments. The US equity risk premium is often considered an “anomaly,” just like value or momentum investing. The equity risk premium is obviously volatile, but perhaps there is a reason to believe it will decrease because everyone already knows that stocks beat bonds over the long-term.

Common reasons for the equity risk premium center around the same arguments – investor behavior and investor risk preferences. Luckily, there are many pundits that can predict future factor premiums based on (inherent) assumptions on (1) investor behavior and (2) investor risk preferences.

So what if we allow machine learning algorithms determine that we should have 100% of our capital in the US equity premium — this will decrease the premium because everyone (using their machine learning algorithms) will do the same and follow this “rational” advice.

So I will leave you with this question, which I do not know the answer to.

For anyone who wants to argue that factor premiums, such as Value and Momentum, will go to zero, I ask you the following – will the equity risk premium exist in the future?

Factor Investing – Does portfolio construction matter?

We can argue about factor investing in the future, and what the premiums might look like over the long haul. These are important discussions. To inform our opinion on these debates we can infuse it with some data from the past. Investors are lucky nowadays because they can lean on the backs of thousands of Ph.D. students grinding on the stock databases. Here is a link to one (of many) papers highlighting all the anomalies in the literature (here is a replication paper).

A few factors tend to pop out from the literature and from our own research – Value, Momentum, Low Volatility, Quality, and Size. These factors all exhibited a “premium” over the value-weight market portfolio in the past. In general, these factors appear to be the most robust, but none of them are perfect (arguably why they exist).

One missing component from the factor debate is related to portfolio construction. Academic papers often analyze zero-investment long/short portfolios which are impossible to implement in the real world. We investigated.  I decided to test what where the returns to long-only Value and Momentum portfolios when one varies the number of firms in the portfolio, as well as the holding period (how often the portfolio is refreshed). The basic idea is the following – every month take the top 1,000 firms based on market capitalization. Then select the top 50, 100, 150, etc. firms based on either a Value measure (EBIT/TEV)(9) or a Momentum measure (12_2 momentum).(10) In addition to varying the number of firms in the portfolio, I also wanted to vary the holding period. In my tests, I allow the holding period to vary from one up to twelve months.(11)

I have updated the results, which now run from 1970 – 2016. All returns are to value-weight portfolios (12) and are gross of any transaction costs or fees.(13)

Value Portfolios (1970 – 2016)

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Momentum portfolios (1970 – 2016).

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Now to be clear, the results shown above are before any management fees or transaction costs.(14) We allow the reader to judge an appropriate transaction cost to apply, as many argue over this number (see here and here). However, if one assumes 0.25% per rebalance, the results are pretty clear for Value and Momentum – in the past, portfolios with more concentration (less stocks) and more frequent rebalancing (such as quarterly), outperformed a larger-stock portfolio that is only rebalanced once a year.

As long-term investors looking to capture the largest equity risk premiums should we all switch to concentrated Value and Momentum portfolios that are rebalanced more than one a year?

No.

But why not?

Because the career risk and psychological pain may be too great. One way to measure this risk is via tracking error, which is essentially the standard deviation of the difference between a particular strategy and the benchmark (in this case the SP500).(15)

The tracking errors numbers for the various factor portfolios are below:

Value Portfolio Tracking Error (1970 – 2016)

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Momentum Portfolio Tracking Error (1970 – 2016)

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Most investors like to track the broad market indices. However, one thing is clear from the charts above — a strong positive relationship between factor returns and tracking error. In short, if you want more expected returns, you need to embrace the risks associated with tracking error.

This finding has implications for a variety of investors. Consider an independent financial advisor building a portfolio on behalf of a client. Part of the advisor’s job is to understand the client’s goals as well as the client’s mindset when it comes to investing. If the advisor believes the client is going to ask every quarter (or ever year) how they did relative to the index, and if they underperformed they will quit the strategy. In this case, it is probably not a great idea to invest in concentrated factor portfolios. Perhaps the advisor sprinkles in a small dosage of concentrated Value or Momentum  (5% or less) alongside a bunch of cheap beta? Maybe they go for a more diluted version of Value and Momentum with less tracking error and put 50% in the “smart beta” form of the factor and 50% in cheap beta?

An evidence based investor should make portfolio decisions based both on the knowledge of the (1) portfolio construction and (2) the mindset of the investor.

Factor Investing – What am I buying?

Now let’s assume we’ve read all the academic research on factor portfolios and factor premiums. We know factor investing is –and has to be — painful at times. We know factor investing portfolio construction has effects on determining the expected excess returns. Now let’s turn to the market to purchase a smart-beta factor portfolio that incorporates all the knowledge we have on our plate.

Unfortunately, this can be harder than anticipated.

Investors are faced with a serious challenge these days. Even if you have a PhD in finance, it can be hard to identify what you are buying based on the marketing materials and analytical tools available in the market. That said, one way an investor can identify what a portfolio delivers to an investor is to examine the portfolio factor loadings over time (a DFA favorite). However, factor models can be finicky, and changing how one defines the factors (e.g., the definition of the “Value,” or HML factor) can have massive impacts on the conclusions one makes. Another approach is to understand how a portfolio is formed (e.g., construction analysis above), and how this will drive portfolio characteristics (e.g., B/M of .5 versus 2), which will arguably be more predictive of expected excess returns than the factor loading. (see here for some research on the topic). This approach is useful, and we highlight the results above, but this approach is also unrealistic for the typical investor unless they are extremely sophisticated and can program. Another tool (aside from tracking error)  is to examine the “active share” of the portfolio to determine how different the factor-investing fund is from the market-cap weighted index.(16) Active share calculations are hard to come by and this measure is not perfect, as being highly active does not necessarily imply outperformance. One still needs to understand the characteristics of the portfolio. For example, one can purchase a high active share overpriced growth stock portfolio, but it is unclear whether or not a long-term factor investor wants a high active share bet on what has historically been a long-term loser factor.

We’ve even attempted to create our own tools to help investors make better decisions. We call it visual active share (were developing this further over the summer). The mission of visual active share is to create a 1) simple, 2) intuitive, and 3) powerful way for investors to understand what they are buying. Currently, the tool allows investors to select (U.S. based equity) ETFs and view the underlying holdings, where the holdings are ranked on a laundry list of  common factor characteristics, such as market-cap, Book-to-Market, Momentum, etc.

We also include “fake” academic anomaly portfolios which more closely reflect how anomaly portfolios are created in academic literature which documents these factor premiums.(17) For example, the B/M effect may be characterized by sorting the universe of stocks on the 10% cheapest set of firms on B/M, rebalanced annually, and comparing this to other portfolios (the 10% most expensive firms) to determine if the factor portfolios is “anomalous.”

To see how the academic portfolios work, start typing “academic” in the “add ETFs” box of our tool. Let’s stick with the classic value measure in academic literature — price-to-book. We type in “PB_BTM” or “academic bottom decile P/B.” Out pops a portfolio. We percentile rank the underlying firms along the x-axis on the Price-to-Book measure (inverse of book-to-market) and along the y-axis on market capitalization.

Image as of 6/19/17

As one can see, the academic portfolio (not an actual ETF) buys the cheapest 20% of stocks based on price to book (we eliminate micro-caps).

Now let’s select the SPY ETF, which invests in the SP500 stocks.

Image as of 6/19/17

As one can see, SPY is a large-cap fund with little deference to value or growth characteristics (as defined by P/B).

Let’s say we want a little value exposure. We pick a Value ETF, the Vanguard S&P 500 Value ETF (VOOV).

We plot it out on visual active share:

Image as of 6/19/17

From the images above, one can see that the Vanguard Value ETF is nothing like the academic factor portfolio, which arguably captures the factor premium as it is documented in academic research. So what are we buying? As the picture highlights, we’re buying a portfolio very similar to the S&P 500 exposure.(18)

Now to be clear, I am not picking on Vanguard here – I believe they are a great institution that has done more for investors than any other financial institution in the last 40 years. And to be fair, their Value portfolio uses numerous value measures, so they shouldn’t exactly line up with the academic portfolio. Second, they are Vanguard – they are taking in billions of dollars a day, so they need to invest in larger stocks that have enough liquidity. By construction, they almost have to be close to the S&P 500.

However, I think most would agree that their portfolio is not capturing the risk/rewards of the academic value portfolio.

So to the extent that an investor wants to capture the academically documented factors (such as Value and Momentum) in unique and differentiated ways, our tool will help to assess ETFs currently available in the marketplace.

Conclusion

I covered a few topics pertinent to factor investing.

First, will factor investing premiums be here in the future? I believe they will exist due to investor behavior, but I could be wrong. In the future if the headlines state that “investors plow record amounts into vastly under-performing Value funds” – I may be worried. However, people generally chase performance, and the same goes for factors. Second, portfolio construction mattered in the past – it is something an evidence-based investor should understand. Third, assessing real-world portfolios can be difficult, so we are trying to help build tools to help the investing public.

I look forward to the factor investing panel at the evidence based investing conference next week. If you are going to be there and want to meet up, please let me know!


Note: This site provides NO information on our value investing ETFs or our momentum investing ETFs. Please refer to this site.


Join thousands of other readers and subscribe to our blog.


Please remember that past performance is not an indicator of future results. Please read our full disclosures. The views and opinions expressed herein are those of the author and do not necessarily reflect the views of Alpha Architect, its affiliates or its employees. This material has been provided to you solely for information and educational purposes and does not constitute an offer or solicitation of an offer or any advice or recommendation to purchase any securities or other financial instruments and may not be construed as such. The factual information set forth herein has been obtained or derived from sources believed by the author and Alpha Architect to be reliable but it is not necessarily all-inclusive and is not guaranteed as to its accuracy and is not to be regarded as a representation or warranty, express or implied, as to the information’s accuracy or completeness, nor should the attached information serve as the basis of any investment decision. No part of this material may be reproduced in any form, or referred to in any other publication, without express written permission from Alpha Architect.


Definitions of common statistics used in our analysis are available here (towards the bottom)

References   [ + ]

1. which is a form of active investing, just done in a systematic fashion
2.

Statistics Definitions

  • CAGR: Compound annual growth rate
  • Standard Deviation: Sample standard deviation
  • Downside Deviation: Sample standard deviation, but only monthly observations below 41.67bps (5%/12) are included in the calculation
  • Sharpe Ratio (annualized): Average monthly return minus treasury bills divided by standard deviation
  • Sortino Ratio (annualized): Average monthly return minus treasury bills divided by downside deviation
  • Worst Drawdown: Worst peak to trough performance (measured based on monthly returns)
3. Two Centuries of Price Return Momentum by Chris Geczy and Mikhail Samonov.
4. Of course, the factor investing portfolio may win, but it is not guaranteed
5. Example being LSV 1994
6. Fama and French
7. I wish I had this knowledge!
8. There are some factor premiums which can be argued will go away due to lower limits to arbitrage. I am focusing on Value and Momentum which do not fall into that category
9. See our paper examining all the Value factors here, and the updated data here.
10. The 12_2 momentum measures the total return to a stock over the past year ignoring the last month
11. Note for any holding period over one month, overlapping portfolios are used.
12. Equal-weight portfolios yield similar results
13. Note the Universe returns are slightly different due to the Value portfolios requiring fundamental information that Momentum portfolios don’t require.
14. Transaction costs were much larger in the past then they were today. We discuss competing research on momentum transaction costs here, here, and here.
15. Here and here are some primers on the concept.
16. Here is a tool by Professor Marijn Cremers. We also have a tool on our website.
17. You can read more about this here.
18. Something that sits to the bottom right of the charts in the prior section on portfolio construction



About the Author

Jack Vogel, Ph.D.

Jack Vogel, Ph.D., conducts research in empirical asset pricing and behavioral finance, and is a co-author of DIY FINANCIAL ADVISOR: A Simple Solution to Build and Protect Your Wealth. His dissertation investigates how behavioral biases affect the value anomaly. His academic background includes experience as an instructor and research assistant at Drexel University in both the Finance and Mathematics departments, as well as a Finance instructor at Villanova University. Dr. Vogel is currently a Managing Member of Alpha Architect, LLC, an SEC-Registered Investment Advisor, where he heads the research department and serves as the Chief Financial Officer. He has a PhD in Finance and a MS in Mathematics from Drexel University, and graduated summa cum laude with a BS in Mathematics and Education from The University of Scranton.


  • Govind

    Too bad there won’t be someone at the Conference to present the opposing point of view regarding capacity limitations and frictional costs.

  • Jack Vogel, PhD

    We highlight some of the competing research above (see footnote 14).

  • Hijo de la Luna

    according to some researchers past data on long-term returns of the stock market is suspicious

    https://www.researchgate.net/publication/247907549_The_Myth_of_1926_How_Much_Do_We_Know_About_Long-Term_Returns_on_US_Stocks
    wouldn’t this affect the reliability of any study of factors which uses this data as evidence?

  • possibly, but this is why out of sample tests are important.

  • Hijo de la Luna

    Ok. By the way great point on the equity risk premium above. In the end one could push skepticism to doubt nearly everything.

  • Mitesh Patel

    What do the value and momentum VW CAGR surfaces look like if you take the holding period to the extremes? That is, all the way down to one day (ignoring frictional costs, of course!) and out to years.

  • Jack Vogel, PhD

    Great question. I don’t have the exact numbers, but daily rebalance gives best results for both Value and Momentum (before trading costs!). Value works out to a few years, while momentum does not (compared to the universe returns).

  • Mitesh Patel

    Thank you. I don’t think we’ll be seeing a Daily Inverse 2X Momentum ETF just yet.

    Ritholtz quotes a chart from Northern Trust’s Bob Browne that shows cycle lengths of some factors:

    http://ritholtz.com/2017/07/factors-still-learning/

    Do you agree with these lengths?

  • Jack Vogel, PhD

    Unfortunately the details on the cycle length calculation are not in the report. One thing is clear, from that report and from others, is that factor strategies go in waves (but they are almost impossible to time!).