In theory, there’s a strong case for building portfolios based on risk factors. In practice, the jury’s out.
The transition from academic research to real-world portfolios, as usual, is a rocky affair. The results you achieve will depend on the factors you hold (and don’t hold), the funds you select to represent the factors, the weights you assign the factors, and the rebalancing schedule. Given that a huge range of portfolio designs are is possible from adjusting those variables, it’s not surprising that results will vary, perhaps dramatically.
As a simple test, let’s build a smart-beta portfolio using ETFs and compare the results with a conventional S&P 500 ETF. As a preview, the results don’t look encouraging: the multi-risk-factor portfolio more or less tracks the S&P. That may be due to the naïve design of the factor portfolio, which we’ll define in a minute. But as a first step in exploring how an ETF-based factor portfolio performs let’s begin with a strategy that’s intuitive in the sense that it throws together a broad mix of the obvious risk exposures via the following eight funds:
* iShares Edge MSCI Min Vol USA (USMV) – low-volatility
* Vanguard High Dividend Yield ETF (VYM) – high-dividend yields
* Guggenheim S&P 500 Equal Weight ETF (RSP) – small-cap bias within large-cap space
* iShares Edge MSCI USA Quality Factor (QUAL) – so-called quality stocks
* iShares Edge MSCI USA Momentum Factor (MTUM) – price momentum
* iShares S&P Small-Cap 600 Value (IJS) – small-cap value stocks
* iShares S&P Mid-Cap 400 Value (IJJ) – mid-cap value stocks
* iShares S&P 500 Value (IVE) – large-cap value stocks
The motivation for holding a broad set of factors rather than just one or two? Risk management. Some analysts recommend diversifying across risk factors, although not everyone agrees. The standard approach is to use, say, a small-cap value fund to supplement exposure to standard large-cap equity allocation to boost expected return. The question here is whether holding a broad allocation of different risk factors is superior to owning the stock market via a conventional index?
For the test, we’ll equal weight this mix and rebalance back to equal weights at the end of each calendar year. The glitch is that the historical data is limited. The underlying concept for targeting specific risk factors has been around for decades, but several of the smart-beta products in the list above are only a few years old. As a result, the portfolio test below begins in July 2013.
Yes, that’s a ridiculously short time span and so the results below should be taken with a grain of salt. But let’s ignore that caveat for now and see what the cat drags in.
The main question: how does the factor strategy compare with a conventional index of US equities? Let’s use the SPDR S&P 500 ETF (SPY) as the benchmark. As the chart below shows, the factor strategy and the S&P 500 track one another closely (correlation is 0.99).
Based on a start date of July 13, 2013 through yesterday (March 22, 2017), the factor strategy generated an annualized 11.4% total return, which is slightly below SPY’s 11.6%. Note, however, that the results are before subtracting the trading costs for rebalancing the factor strategy. The S&P portfolio, by contrast, is buy and hold and so there are zero rebalancing expenses.
Adjusting for risk doesn’t change the comparison. Profiling via Sharpe ratio, for instance, reveals virtually identical results: 0.92 for the factor strategy and 0.91 for SPY.
What happens if we increase the frequency of rebalancing to a monthly schedule? Not much. The factor strategy’s risk/return profile is roughly the same regardless of whether the portfolio is rebalanced monthly or at the end of each calendar year.
Does this mean that factor investing is worthless? No, not at all. But what the test above suggests is that earning a sizable premium over a conventional cap-weighted strategy may be tougher than it appears from the research.
To be fair, the history above is too short to draw conclusions. Also, it’s debatable if owning eight different factor funds is optimal. Would we do better by selecting a subset or dynamically managing the weights?
Keep in mind, too, that the S&P 500 has enjoyed an unusually strong run in recent years, which raises the possibility that a broadly diversified factor strategy may deliver more encouraging results in relative terms in the years ahead if the S&P’s performance reverts to its lower long-term track record.
In future posts I’ll consider alternative tests for analyzing ETF-based factor strategies. Meantime, the results above suggest that swapping conventional cap-weighted funds for factor funds may not be a short cut for materially boosting performance and/or lowering risk.
Pingback: Building a Portfolio Based on Risk Factors - TradingGods.net
You could probably go much further back by looking at the underlying indexes and making the comparison vs. SPTR rather than using funds. Just a thought.
ETF Man,
Yes, indeed, I agree completely. And just to be clear, serious analysis of this (or any other strategy) deserves a deeper look along the lines you suggest. This post, by comparison, was far less ambitious–really just an excuse to explore a hunch based on some obvious ETF proxies.
–JP
I quit reading on encountering this sentence which would not be acceptable in grammar school: “Given that a huge range of portfolio designs are possible from adjusting those variables, it’s not surprising that results will vary, perhaps dramatically.” This says “a range . . . are possible”. TILT! There is no rule of English grammar that says a verb should be singular or plural according to whether the most recent preceding noun is singular or plural. Not a big point, but it suggests the author is not careful about little things, not of master of basic education, and, therefore, not a good source of advice. I have been retired for years, living well, following the advice of Bogle, with no attempt to beat the market.
Another shortcoming to this analysis is that the S&P 500 value etf is not the same as the value factor etf. We have looked back at portfolios created with factor indices and going back to about 2000, certain combinations do outperform. Most notably size, value, momentum, and quality. But you do have to use the actual factor strategies not just large cap value.
Besides the fact that none of the factors had a significant premium in the US since March 2009 (they performed better in Emerging Markets) and therefore no fund based on these factors performed well, your test shows two problems: the listed funds are long-only, therefor only capturing a part of the premium (usually only half of it), and secondly, the funds expense ratio plus their rebalancing costs eat up much of the remaining premium.
Additionally it might be possible that the performance of US funds since 2013 is mainly explained by external factors, with only little alpha to gain from specific stock selection – a chart displaying the average correlation of each stock to the S&P500 would provide input, and in case of high correlations explain why these factors didn’t deliver a premium on the long side in the past years.
As we don’t know how these funds will perform in the next bear market (or: would have performed in 2008/2009) and shortly after, it’s hard to tell if they are worth their money. They might reduce drawdown, or outperform in an early bull market.
Regarding the selection of funds: based on the recent publication by Andrew Berkin and Larry Swedroe (“Your Complete Guide to Factor-based Investing”) only four factors are considered to be useful: size, value, momentum and profitability/quality. They also conclude that multi-style funds have the advantage of a lower turnover, as e.g. a value stock gaining momentum would be sold from the value fund and purchased by the momentum fund, whereas might be kept by a multi-style fund. They suggest to buy funds from AQR, Bridgeway and DFA (but we do not know who financed the publication), but they all don’t perform too well since 2013 either.
In case you’ll continue your tests, I suggest adding smart beta funds covering developed and emerging markets (e.g. Mebane Faber constructed a few value/momentum funds based on country indices instead of U.S. stocks), and comparing that to a broad world index. While it might not be viable for the average investor to hold a few dozen funds, this test will remove the current focus on a single, very specific (and aging) bull market.
Pingback: What I am reading today (26 Mar 2017) – FinSummary
Ted,
You are correct — 10 lashes for this writer and no pudding for dessert. Grammatical errors do slip into the text from time to time, which is a hazard of working solo. You’ll find, however, that the oversight has been corrected and it’s now safe to read the rest of the post. But remain vigilant–grammatical trouble could arise without warning. It’s a jungle out there!
–JP
Alec,
Good points. Everyone should keep your comments in mind in this corner of investing.
–JP
Bill,
That’s an interesting point — factor strategies work best across the capitalization spectrum. Maybe, although there’s quite a lot of folks who focus on small-cap value, for instance.
–JP
Widely corroborated anaylsis by serious academics have shown that factor investing over the long term has added value. My own analysis, using index data going back 20 years also reached the same conclusion. Looking at ETF’s in a throw away portfolio over the past 3 years is hardly solid analysis. Certain factors are stronger than others. And weightings should provide the tilts that access the strongest factors. Your conclusion, over a short time frame, refutes many serious academics studies. Perhaps your conclusion should have read ” this portfolio, over this time frame, added no value”. That is different than implying factor investing does not work……which seemed to be your implied thesis in this article.
John Russell,
I agree that there are many academic studies showing that factor analysis has been productive over the long haul. The 3-factor Fama-French model comes to mind, which is arguably the most influential, successful, and widely used framework for factor investing. I’m not trying to refute such research, certainly not in this blog post. Rather, this is merely an attempt to consider how an equal-weighted mix of a broad set of US equity factor ETFs has fared against the S&P 500. Curiousity is the motive. As I pointed out in the article, there are several caveats, starting with the reminder that the time frame is too short to draw any serious conclusions. But as the factor-ETF product lineup expands, it’s useful to take a closer look at how the real-world results compare with the academic studies. Should investors form an opinion about factor investing generally from this piece? Absolutely not.
–JP
While there might be little reason to question factor premiums, it still is unclear if an ETF only covering the long side of the factor(s) is able to generate a premium for the buyer, after all costs & expenses. The theoretical average long term (1927-2016) factor premium is 3-9%, resulting in 2-5% on the long side, before all costs.
As we’re talking about *average* premiums, history has shown that all factors have long periods (up to a decade or even longer) of underperformance. Since 1965, only two periods had a significant factor premium: 1973-1987 and 2000-2008.
The resulting question is: will the average investor keep a factor ETF in his portfolio after e.g. ten years of underperformance, being told by “experts” that the factor stopped working? I doubt it. And in that case: when, if ever, is the appropriate time to buy a smart beta ETF? Is there a way to anticipate a rising premium for a factor? Is it suggested to wait for the next 20% correction before buying, are there external indicators to watch that increase chances that a specific factor will generate a premium? (e.g. historically value stocks increased more than growth stocks when the money supply increased)