The traditional asset classes that fund managers use to build portfolios have better data-driven substitutes. When I combined multiple return "factors" into ETFs and then optimize their portfolio weightings, I found a recipe for effectively building wealth with less risk. In particular, I can show that certain ETFs structurally outperform their lower-quality and more popular peers. Whether you invest primarily in individual stocks or ETFs, you should be able to use the principles of factor investing to improve your returns and reduce your risk.
For example, academic research has shown that value stocks and the stocks of small-cap companies have been shown to outperform the S&P 500 over time. While much of the "smart-beta" movement consists of funds with slick marketing and high fees, understanding which factors drive returns and how to exploit them allows you to choose a portfolio with superior performance (and lower fees).
What is factor investing?
Andrew Ang, former finance department chair at Columbia Business School, defined factors as "broad and persistent drivers of return that research has proven to be historically enduring."
Ang now works for BlackRock, which owns the iShares line of ETFs. He also has a checklist for what factors are and why they're allowed to persist. I found it to be pretty enlightening.
Here's Ang's checklist:
1. Rewarded risk - Some factors have earned higher long-term returns as a reward for taking on more risk.
2. Structural impediment - Market rules or other constraints can place restrictions on certain investors, such as pension funds. Those off-limit investments can become opportunities for others.
3. Investors’ biases - Over time, opportunities are generated by investors not behaving perfectly rationally. Taking a contrarian view can be beneficial.
Certain drivers of equity returns are structural rather than cyclical. For example, when institutional investors have policy constraints, they tend to bid up assets that allow them to achieve their goals and fit within their limitations. For example, after the Fed cut rates to zero, institutional investors piled into high yield bonds, giving those assets a poor risk/reward. Pension funds similarly pile into high-beta stocks without regard to quality, causing them to often be overvalued relative to safer stocks.
For example, high-quality companies (as defined by factors like profitability, positive earnings growth, lower leverage, and less volatility) outperform low-quality companies with remarkable consistency. The top 10 percent of quality scored companies have been shown to outperform the bottom 10 percent by 6 percent per year, with less risk to boot.
It's a bizarre thing to watch play out in the stocks and funds I track because the outperformance of quality is more like a few dollars consistently per day than other proven long-term factors like the value premium, which see multi-year periods of underperformance.
Value stocks are obviously another segment of the market that has been shown to outperform over time, as are small-cap companies. But it's only when you start combining multiple factors at once that you find the secret sauce. I've spent the past few months working on finding ETFs that exploit various drivers of outperformance, and now I feel like I have it close enough to perfect to share with my readers.
Factors are the driving force behind smart beta funds, which seek to outperform the market for any given level of risk. Smart beta funds are a little hard to pin down, however. For example, if a fund has "smart beta" in its title, it's probably not actually a consistent outperformer. Chances are that the harder a fund is marketed, the less intelligent the construction actually is. For example, equal-weight ETFs have been marketed as a holy grail for investors, but have failed to deliver their promised outperformance. True outperformers tend to gather assets with minimal marketing. Superior performance tends to sell itself.
For the portfolio, I used a Markowitz (Efficient Frontier) algorithm to find dividend-adjusted returns and risk for various asset classes and optimize a portfolio off of them. In the process, I found that I have a superior, data-driven alternative to each traditional asset allocation category.
It's intuitive, it's simple, and most of all, it works. Every point along the line is a rational portfolio based on return and risk profiles. Note that time periods were not chosen to cherry pick, but because data isn't available for all time periods for all ETFs.
Feel free to play around with my assumptions in the model (link here). I also ran asset allocation calculations starting in 2008, which I describe in the notes section (there were very few changes for the pre-crisis portfolio). Also, note that I didn't select the tangency portfolio for my illustration because it's unrealistic to be able to borrow at the risk-free rate.
Read some debate about the tangency portfolio here.
What I found was that some asset classes were dominant (in the game theory sense) over others. Unsurprisingly, proven factors such as small-caps and value stocks found themselves in the efficient portfolio, dominating lesser strategies in the risk/reward calculations. Dividend funds seem to have a superior exposure to the value premium than the value sector itself due to higher exposure to the quality factor (which may be the hidden third variable driving value's returns). Additionally, quality focused ETFs outperformed their underlying sectors with less risk over every time period I looked at.
Another surprising result was that long-term bonds tend to find their way into the optimal portfolio due to their reasonable returns and negative correlation with equities.
The portfolio that beats the S&P 500 with less risk ended up with this allocation:
25 percent VYM, 30 percent VGT, 20 percent TLT, 25 percent IJR.
Return: 13.6 percent, Standard deviation = 9.6 percent. Sharpe ratio of 1.41.
(If you want to take the same risk as the S&P 500, you can beat the S&P by even more by going 30 percent VGT, 35 percent IJR, 30 percent VYM, and 5 percent TLT.)
Now, let's explain why some ETFs tended to get better risk-adjusted returns than others.
Vanguard High Dividend Yield ETF (VYM)
Where the alpha comes from - VYM succeeds due to its exposure to the value and quality factors. I caught more than a little flak about declaring the death of value investing last week. Value investing isn't actually dead, but value funds do a lousy job of actually capturing the value premium. By tilting towards companies with ample cash flow and liquidity to pay dividends, VYM sustainably outperforms its pure value cousin, the Vanguard Value ETF (VTV). They're both filled with value stocks, VYM just tilts towards the higher quality ones, as evidenced by its lower standard deviation and higher return.
Although VYM is branded as a dividend fund, it actually fits the intersection of the value and quality factors like a glove. Investors seeking income should be pleased to know that it yields roughly 3.3 percent, vs. VTV at 2.6 percent. It also outperforms VTV by 0.5 to 0.75 percent per year, with less risk. This alpha gets back to superior index construction. I noted in my original article that value ETFs tend to buy everything cheap irrespective of value. This hurts their ability to outperform.
VYM's construction is elegant. Vanguard takes every dividend payer in the Russell 1000 and throws out every stock with below the median dividend yield. Then they weight it by market cap (this avoids over-allocating to the highest yielding value traps, which have been shown to underperform).
Ironically, VYM is slightly cheaper at 14.7 times prospective earnings than VTV is a 14.9X.
Rather than putting your faith in one market anomaly, VYM allows you to exploit two. Dividend payers have been shown to outperform over time due to their exposure to the quality factor, and the index tilts heavily towards the known value factor. I also looked at Deutsche Bank's new quality/value ETF, Xtrackers Russell 1000 US QARP ETF (QARP), for this slot, but there isn't a long enough track record or sufficient liquidity yet to incorporate it into the model. VYM only costs 0.08 percent per year in fees, compared to far more for ETFs marketed as "smart beta."
VYM found its way into both portfolios due to its superior risk-adjusted return. It earned a 25 percent weight in the main portfolio.
Vanguard Information Technology ETF (VGT)
Where the alpha comes from - Momentum, quality, and international revenue exposure.
This may be a little more controversial than the other holdings, but investing in quality technology stocks is a data-driven alternative to investing in international stocks.
It's a popular asset allocation move to put money in international stocks and/or emerging markets, but what most investors don't know is that nearly 60 percent of tech sector revenue comes from abroad. This number is additionally likely to rise in the future. As tech companies get more and more of their revenue abroad, they can replace foreign stocks for exposure to global GDP growth. While it can be hard to determine which companies will succeed in the future, the sector-wide EPS numbers will continue to grow faster than any other sector.
My research has found that political stability is a better predictor of returns than GDP growth, so tech may be the most efficient way to play the emerging market thesis. Valuations are reasonable at roughly 19 times earnings – not bad at all considering technology is the main driver of productivity growth at this point.
Technology has been the best performing segment of the market over the last 8-10 years (note that VGT excludes FANG stocks by design, and it's still the best performer amongst Vanguard sector funds). Unlike the late 1990s, this time around the returns have been driven by earnings, and valuations are attractive. Tech benefits from exposure to the factors of beta, quality, and momentum.
Technology dominated the optimal portfolio so much that I put a constraint at a max 30 percent allocation. It's roughly 25 percent of the S&P 500 at large, so a 30 percent pure allocation to tech doesn't scare me. VGT clocks in with a 0.10 percent yearly expense ratio.
Tech maxed out its allocation at 30 percent.
iShares Core S&P Small-Cap ETF (IJR)
Where the alpha comes from - The small-cap premium, the quality factor, and from avoiding the high-frequency trading circus around the annual Russell 2000 rebalancing.
I've written about it before, but IJR is my secret sauce for boosting portfolio returns. Most asset allocation models are based on the returns of the Russell 2000 index, which is rigged to underperform and was from the day it started. Read all about it here. IJR tracks the S&P 600, which has a simple quality screen for inclusion (all the companies have to do is make a profit!) beats the Russell 2000 (IWM) every month, quarter and year in my dataset with less risk. Every day I check, it's up at least a couple of basis points more than the Russell. The difference amounts to roughly 2 percent per year in extra return and 1 percent less standard deviation per year. It's a free lunch and it's not going away.
This matters to you because if you base your investment decisions on the Russell 2000, then, by definition, you're not realizing the full extent of the small cap premium. When you combine small caps with a basic test to weed out failing companies, the resulting product finds itself in nearly every optimal portfolio calculated since its inception.
Notably, the Russell 2000 doesn't win an allocation at any point, even if you remove the IJR from the equation. There is no statistical uncertainty around the IJR vs. Russell 2000 choice. IJR has an expense ratio of 0.07 percent per year.
IJR wins a 25 percent allocation of the main portfolio in the Markowitz algorithm.
Bonds: iShares 20+ Year Treasury Bond ETF (TLT)
Where the alpha comes from - Strong negative correlation with equities and higher return than the overall bond market.
This inclusion was a surprise to me at first. However, long-term treasury bonds dominate other bonds in the algorithm due to their negative correlation to stocks in times of stress. They also have a higher return than shorter-term bonds due to their higher yields and capital gains from rolling down the yield curve. This creates a positive expected return over time, which compensates for most of the risk of rising rates hurting your portfolio. Even with the current (relatively flat) yield curve, roll-down effect contributes roughly 0.5 percent per year in windfall capital gains. Read more about the roll-down effect here.
TLT is interesting because it tends to dominate the Vanguard Total Bond Market ETF (NASDAQ:BND) when portfolio bond holdings are under 35-45 percent. As bond holdings go over that threshold, you tend to see a greater and greater tilt towards BND, but all three ETFs tend to be in the portfolio until you get to very conservative (80+ percent bond) allocations.
The reason why TLT tends to win over BND is its higher expected return and strong negative correlation with equities. If an investor is only going to have a small allocation to bonds in their portfolio and is invested for the long term, data shows that long-term government bonds beat every other type of bond when accounting for risk, return, and correlation with equities. BND's correlation with equities is anywhere from about -0.3 to -0.5 over the time periods studied.
Long-term bonds also serve as a counter-cyclical investment. When you need liquidity most, you can depend on them to hold their value. If you like to rebalance your portfolio, you gain optionality from this. An excellent case study for this is from 2007 to 2009. During this time, long-term government bonds surged as money flowed out of risky assets.
How would you like to have had a holding in your portfolio have been up 35 percent in 2007-2008? The world isn't such a scary place when you have a hedge. And, unlike other strategies such as buying puts on the SPY or QQQ, bonds actually pay you to wait.
The BND doesn't work like this as it has exposure to corporate credit risk and has shorter-term bonds. Long-term government bonds work so well as a hedge because of their nearly unassailable credit quality and long duration. The "volatility" actually benefits you because falling interest rates are correlated with economic recessions. The same change in interest rates that might move the BND 6-7 percent upward will cause TLT to surge ~20 percent.
Here's a very academic study on how this works for those who are interested. I will note that, as a counterpoint, long-term treasury bonds are likely to suffer if inflation and rates rise dramatically. However, research from Vanguard shows that a potential shock may not be as disastrous over the long run as you think. To this point, if you have a mortgage, the effects of inflation on your investments and the effects on inflation to shrink the burden of your mortgage will cancel each other out.
Another point I would like to note is that owning short-term bonds while carrying a mortgage is not smart. By buying long-term bonds rather than short term, we match the maturity of our debt to the maturity of our assets and avoid borrowing at the long end of the curve to invest at the short end (very suboptimal).
If you find yourself tempted to invest a lot in short-term bonds, resist the temptation because you can make a higher risk-free return from paying down principal against your mortgage. Long bonds behave a little differently, which is to our benefit.
While we are still less creditworthy than the U.S. Treasury and pay a negative carry by owning long bonds while holding a mortgage, owning TLT gives us optionality that holding intermediate-term bonds doesn't (we make a big profit in our bonds and can separately refinance our mortgage if rates drop, but the US Treasury can't refinance their debt unless they buy it back). If the economy tanks and the Fed initiates another round of quantitative easing, the value of our bonds will surge 18 percent for every one percentage point interest rates fall, according to TLT's published duration. Conversely, if inflation surges, you more than likely stand to gain more from the reduction in the present value liability of your mortgage debt than you'd lose in long-term treasuries. When you count roll-down, optionality, and the strong negative correlation between long bonds and stocks, the roughly 1 percentage point negative carry is worth paying.
Owning long bonds gives you protection via a flood of liquidity in a bear market when deflation often threatens to rear its head. Stock prices will fall during a recession, so we can potentially bargain hunt with the liquidity the bonds give us. You may want to scale into TLT as long-term interest rates rise towards 4 percent, but when the next recession hits, you'll be happy to see your portfolio holding steady. After the December rate hike will be an excellent time to buy some TLT to hedge. You obviously have to take valuation into account (I wouldn't have bought TLT when QE was in full bloom), but being able to invest countercyclically is huge. TLT carries an expense ratio of 0.15 percent per year but is the best choice due to having the largest size and most liquidity (bond funds with lower published fees may have higher actual costs due to the costs of buying and selling).
TLT won 20 percent of the optimized portfolio weight.
Points to consider for your portfolio:
1. Another surprise was that REITs (VNQ) were continually dominated by other asset classes. REITs have a lot of risk, which is only partially offset by their returns. REITs may be a ripe place to pick individual holdings.
2. Note that nothing in this article should dissuade you from doing bottom-up homework on stocks and investing in companies you believe in. I'm writing about which sectors to invest in, and there are lots of ways to get to the finish line in investing. However, if want to buy individual stocks, make sure they're good ones, as 2 in 3 stocks underperform the index over the long run.
3. For reference, the tangency portfolio had these holdings:
10 percent VYM, 18 percent VTV, 9 percent VGT, 11 percent TLT, 4 percent IJR, 48 percent (IEF)*.
Return: 8.2 percent, standard deviation, 4.6 percent with a Sharpe ratio of 1.74.
4. VTV made a comeback in the tangency portfolio due to its low correlation with other assets in the portfolio, especially IEF.
5. Extending the model to start the data in 2008 only changed the portfolio to favor TLT and IEF more and select VYM less. Every other factor was remarkably consistent.
*The tangency and 2008 dataset portfolio included portfolios allocated much more heavily to intermediate-term government bonds. If you're going to have a large bond allocation, the risk/reward is better from intermediate bonds for the majority of your allocation, because you would far have less equity risk to offset. Substitute municipal bonds (MUB) for intermediate-term bonds if you choose to be at this point on the curve and make over $200,000 per year as the model doesn't account for taxes. Many investors don't have to pay tax (foreigners and pension funds combine for over 50 percent of US stock holdings and don't have to pay tax on interest), so market prices on intermediate-term bonds don't necessarily reflect taxes.
Thanks for reading! If you've made it this far in the article, you might as well follow me!
Logan C. Kane
Disclosure: I am/we are long IJR, VGT, VYM, BND, MUB, VNQ.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.