Smart Beta And The Portfolio Construction Puzzle

The portfolio puzzle The Rubik’s cube has become a popular metaphor for the marketing teams of ETF providers. With good reason. For each client there’s a portfolio construction puzzle to be solved with building blocks, representing geographies, sectors, asset classes, factors and styles. There has been rapid expansion from providers of ETFs tracking main-market indices, with the largest institutional providers capturing the lion’s share of flows, owing to their ability to deliver on four key ETF governance criteria — consistency, liquidity, transparency and, of course, price. This means that ETFs for main market cap-weighted indices are increasingly commoditized. After all, there doesn’t seem to be anything overly smart about replicating market beta, other than the smartness of saving on fees relative to ‘closet-tracker’ active funds. Traditional cap-weighted index investing is a preference: either out of philosophy or necessity. Innovation Means Smarter? Hence R&D of institutional investors, index providers and ETF manufacturers alike has focused more on “smart beta.” This has triggered a slew of innovation – both superficial and substantive. At a superficial end, age-old alternative weighting strategies (e.g. value indices that screen stocks for low book values, or dividend-weighted indices) have been re-branded as being “smart.” In these cases, for “smart” read “non-market-cap weighted.” In fairness, this rebranding is part of broadening of alternative weighting strategies that are factor-based. More substantively, research programs such as EDHEC-Risk Institute’s Scientific Beta have been instrumental in promoting fresh thinking in the field of both factor-based and risk-based smart beta strategies. Factor-Based Approach As a result, providers are focusing on making building blocks smarter. Instead of relying on the ‘traditional’ factor of market capitalization for index inclusion, smart beta indices (and related ETFs) look at alternative factors: book value, dividend yield, volatility, for example. In that respect, the FTSE Russell 1000 Value Index launched in 1987 is probably the oldest factor index on the block. More recent factor indices are stylistic: Both iShares (Oct-14) and Vanguard (Dec-15) have launched global equity factor ETFs focusing on liquidity, minimum volatility, momentum and value. The sophistication of factor-based index construction will continue to increase with the increase in data availability and computing power. Risk-Based Approach Portfolio strategists meanwhile can apply quantitative rules-based approaches to portfolio construction, creating static or dynamic asset allocation strategies from a growing universe of both cap-weighted and alternatively weighted index tracking funds. These strategies — such as maximum Sharpe, minimum variance, equal risk contribution and maximum deconcentration — offer an alternative to the standard but troubled single period mean variance optimization (MVO) approach. MVO’s limitations The single-period MVO approach remains the traditional bedrock of very long-run investing in normal market conditions where the sequence of returns does not matter. However it runs into difficulty in the short-run when markets are non-normal and sequence of returns matters a lot. So unless you are a large endowment with an infinite time horizon, or perhaps can afford to invest for yourself and your family without ever needing to withdraw any capital, relying entirely on the MVO approach for asset allocation gives false comfort. For cases where there are constraints that challenge the MVO model – due to multiple or limited time horizons, expected capital withdrawals, risk budgets, and unstable risk/return/correlation profiles of asset classes (collectively known as real life) — portfolio construction requires a smarter, more adaptive approach that observes, isolates and captures the reward from shifting risk premia over time. Risk-based portfolio strategies attempt to achieve this and are designed to offer a liquid alternative approach to investing that is uncorrelated with traditional single-period MVO strategies. What’s the Problem to Solve? Whether assessing factor-based ETFs or risk-based ETF strategies, at best these new developments all seem very smart. At worst it’s just a bit different. However, as ETFs get smarter and the strategies that combine them become more sophisticated, there’s a risk that the key question in all of this gets lost in an incomprehensible barrage of Greek. The key question for portfolio managers nonetheless remains the same. What client outcome am I targeting? What client need am I trying to solve? For portfolio strategy, whether using a discretionary manager that relies on judgment, or a systematic rules-based approach that relies on quantitative inputs, the key client considerations remain return objective, time horizon, capacity for loss and diversification across asset classes and/or risk premia. Broadening the Toolkit A portfolio strategy has little meaning without an objective that focuses on client outcomes. Factor-based ETFs and risk-based ETF portfolio strategies offer an alternative or additional set of tools to help deliver on those outcomes, in a way that is systematic, liquid and efficient. Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article. Additional disclosure: This article has been prepared for research purposes only.

Smartening Watson: IBM Supercomputer Bolsters Cybersecurity Job Gap

Skynet isn’t becoming self-aware, but IBM ‘s ( IBM ) supercomputer, Watson, aims to become more aware … of cybersecurity attacks. To combat a growing number of Black Hats and a shortfall in their counterpart White Hats — the company says forecasts see as many as 1.5 million unfilled cybersecurity positions by 2020 — IBM was set early Tuesday to announce a year-long project with eight universities to smarten Watson. Watson is already filtering Big Data to bolster cancer research, create learning tools and improve business operations. The next frontier? Teaching Watson to scour the 80% of unstructured online data to suss out cyberthreats. IBM’s security operations center (SOC) already receives 20 billion pieces of raw data per day detailing potential cyber mischief, says Caleb Barlow, IBM Security vice president. On average, companies spend $1.3 billion annually, or 21,000 hours, chasing false positives. “Some can be mundane, like a user was locked out after 10 password tries,” Barlow told IBD. “Or, we could get data about an ATP (advanced threat protection) attacker. … It’s not a matter of looking for the needle in the haystack. It’s a matter of a looking for the needle in a stack of needles.” Enter Watson On Unstructured-Data Front Watson is capable of digesting structured data, Barlow says. He likens it to a paramedic responding to a car accident. Watson can take the vitals, but it cannot look for the crack in the windshield where the victim hit his head (unstructured data). It’s the difference between analysis and insights, Barlow says. Humans can do both, but the sheer volume of data is overwhelming. “Security data is in unstructured data — blogs, wikis, articles, white papers, presentation notes,” he said. “How do we take that experiential data, data we can only get from a human and apply that to this challenge?” First, IBM will team up with students from California State Polytechnic University, Pennsylvania State University, Massachusetts Institute of Technology, New York University, University of Maryland, University of New Brunswick, University of Ottawa and University of Waterloo. There will be 200 IBM staff members and students working on the project. “The partnership between IBM and Penn State is an ideal opportunity for our students to experience the kinds of bleeding edge knowledge management that will drive technology in the next century,” Penn State professor Patrick McDaniel said via email. “At the same time, it is a wonderful chance for Penn State to showcase its exceptional student engineers.” Under instruction from IBM experts, the students will process 15,000 documents per month including threat intelligence reports, cybercrime strategies and threat databases. Watson will slowly begin to learn that unstructured data. It’s almost childlike, Barlow says. “You have to sit down with Watson and explain the language,” he said. “Then, we go through, ‘Here, you were right,’ or ‘Here, you were wrong.’” The difference is that Watson won’t forget, he says. Still, a human analyst remains necessary to respond to developing attacks — whether that’s blocking the hacker, watching malware inside the network or plugging holes. “Watson is not replacing the analyst,” he said. “But if I can get Watson to ask all those questions and prioritize that, I can be asking millions of questions (to suss out legitimate cyberattacks) I would not be able to ask otherwise.” Demand Outstrips New Talent That’s more valuable than finding the needle, Barlow says. More than 10,000 security research papers and 60,000 security blogs are published each year and each month, respectively. The National Vulnerability Database has received reports of 75,000-plus software vulnerabilities. Coupled with that, varying reports place the current paucity in cybersecurity skilled employees at 200,000 to 1 million. It was a huge topic at the RSA Conference in February in San Francisco. But it’s not that students aren’t interested, Barlow says. “Universities have shown me their growth statistic,” he said. “It’s a hockey stick. Their challenge is, they are running out of facility space.” The problem is immediacy. Twenty years ago, cybersecurity wasn’t at the forefront of IT concerns. They, today there aren’t enough skilled professionals. The chief information security officer (CISO) is newest entrant to the C-Suite. “These are not skills people have historically had,” Barlow says. “It’s IT-centric computer science skills. It requires a collision of those traditional computer science skills with forensics and investigative skills.” He added: “No matter how aggressively universities turn out new talent, they won’t be able to meet the demand.”

The Small-Cap "Alpha" Myth

There is a common misconception about “alpha” in the small-cap market within the United States. Many professionals believe that once we step out of the mega-cap world of companies like Google, Wal-Mart, Coca-Cola and Apple where there is an army of analysts digging into the vast amounts of data and pricing stocks accordingly, that there is opportunity in its smaller counterparts given the perceived market inefficiency. The story goes that there are fewer analysts covering these particular companies and, therefore, there is an opportunity to produce superior risk-adjusted returns. Whenever we want to research a particular topic in investing, it is always best to start looking into peer-reviewed academic research. In fact, we published an article all the way back in 2001 that covered this particular topic. Our analysis was based on a research paper entitled “The Small Cap Myth” produced by Richard M. Ennis and Michael D. Sebastian of Ennis Knupp Associates, one of the largest pension consulting firms in the country. Based on a sample of 128 small-cap managers, they concluded that once we adjusted for (1) management fees, (2) improper benchmarking, and (3) survivorship bias within the sample, the average “alpha” fell to virtually zero. Aon Hewitt, another large consulting firm, recently published its own research on the small-cap “alpha” myth in January of this year entitled “The Small-Cap Alpha Myth Revisited.” Based on the eVestment Database of small-cap equity managers, the researchers found that the median performance of these managers was worse for the 10-year period ending June 30, 2015 than the original analysis in 2001. The median performance across all styles in the small-cap market was less than 1% (originally around 4%). Once the researchers adjusted for survivorship bias, back-fill bias, liquidity and transaction costs, which the researchers estimated to be almost 200 basis points, the median results were actually negative. Click to enlarge Similarly, we can compare the average performance of all 479 actively managed small cap funds (as classified by Morningstar) against commercial benchmarks like the Russell 2000 Index and S&P Small Cap 600 Index. If we then add small-cap index funds from Dimensional, Vanguard and iShares, we have a nice comparison chart over the 15-year period ending 12/31/2015. As you can see below, the average actively managed small-cap fund underperformed the Russell 2000 Index by 0.24% per year and the DFA U.S. Small Cap Fund by 2.0% per year, net of fees. These results not only highlight the ” arithmetic of active management ” that Nobel Laureate Bill Sharpe reminds investors of, but also the potential benefits of utilizing a strategy, such as the one offered by DFA, that can better capture the small size premium by designing their own DFA small-cap index that has a smaller weighted average market capitalization than other indexes. Click to enlarge How can different index funds produce significantly different performances if they are all targeting the same asset class? In short, differences in performance come from differences in indexes. For example, the Russell 2000 Index focuses on the bottom 2000 companies in terms of market capitalization in the Russell 3000 Index. DFA, on the other hand, defines its Small-Cap Index as a market-capitalization-weighted index of securities of the smallest US companies whose market capitalization falls in the lowest 8% of the total market capitalization of the eligible market ( see details here ). The eligible market is composed of securities of US companies traded on the NYSE, NYSE MKT (formerly AMEX), and Nasdaq Global Market. Exclusions include non-US companies, REITs, UITs and Investment Companies and companies with the lowest profitability and highest relative price within the small cap universe. Profitability is measured as operating income before depreciation and amortization minus interest expense scaled by book. You can find an even more detailed explanation of the historical composition of their indexes in the footnotes below. It is an important reminder that DFA is not new to the indexing industry. In fact, it is one of the pioneers of understanding and implementing index-based strategies. There is no “right” answer, but DFA’s approach seems to better capture the small-cap premium. It is a delicate balance between maintaining strong diversification, pursuing the small cap premium, and keeping trading costs as low as possible. The chart below displays the historical annualized return and standard deviation for a few DFA and Russell Indexes over the last 37 years. You can see that DFA generates a higher return than Russell by better capturing risk premiums in the stock market. Click to enlarge In its own words, Aon Hewitt summed up belief in the small-cap “alpha” with the following: “The widely held assumption that inefficiencies within the U.S. small- cap equity market should lead to greater opportunity for active management than the large-cap equity market appears to be just as mythical in 2015 as it was in 2001. The growth in actively managed assets within the small-cap space over the past 14 years may be significantly contributing to the lack of inefficiency that many market participants erroneously assume.” We couldn’t agree more. Click to enlarge IFA Painting: The Size Premium Disclosure: I am/we are long DFSTX. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.