#micro cap

We are often asked how much of a plan’s assets should be allocated to microcap equities. As long-term investors that view the opportunity set through the lens of factors, our answer is usually some version of "probably more than you currently do." Microcap is a very challenging asset class to evaluate. There is little empirical research specific to the intricacies of the space, and common benchmarks cast a shadow on the alpha that is readily apparent in active manager returns and factor spreads. As I have written about previously, true microcap offers substantial opportunity for differentiated alpha generation. This post attempts to provide an alternative framework for approaching and sizing strategic allocations to microcaps.

Optimization Meltdown

Asset allocation typically involves some form of optimization process that requires return, risk, and correlation assumptions. The table below shows common proxies for U.S. equity asset classes. As you move from top to bottom, returns decrease and volatility rises, causing decreasing risk-adjusted return (Sharpe Ratio). While lower return and higher volatility are not enough in and of themselves to eliminate an asset class from inclusion, correlations can. As one can infer from the benchmark statistics below, typical mean-variance optimization will suggest no allocation to the microcap asset class.

A simple test to determine the efficacy of adding an asset class to a portfolio is to look at correlation-adjusted Sharpe ratios.[1] To do this, simply multiply the Sharpe Ratio of an existing portfolio by its correlation with the new asset. The Sharpe of the new asset would then need to be greater than this adjusted Sharpe Ratio to be included. In the table below, both the Russell 2000 and Microcap would fail this test when compared with a 100% Russell 1000 portfolio.

While it would be easy to write off small and microcap as an asset class, a key issue I detailed in a previous post is the poor construction of the commonly used Russell Microcap benchmark.[2] Significant overlap with the Russell 2000® Index, about 88%[3], results in correlation between the indices of 0.96, resulting in little differentiation. This causes microcap as an asset class, defined by Russell, to fail simple tests for strategic inclusion in portfolios. This begs the question of whether cap-weighted benchmarks should always be the de facto measuring stick for asset classes. In my experience researching small and microcap portfolios, it is abundantly clear to that there is ample opportunity to generate return that is distinct from cap-weighted benchmarks.

A Returns-based Approach to Allocation

Many practitioners and academics agree that the “market” is an aggregation of stocks weighted by market capitalization. While this is certainly accurate, it represents a capacity-based view of the opportunity set.

Constructing a portfolio, or index, in proportion to market cap weights is unique in that it provides the lowest cost, highest capacity exposure to the market, regardless of investor size. It requires minimal trading beyond dividend reinvestment and investor driven flows because portfolio weights adjust in proportion to changes in market weights as stocks rise and fall. This minimizes ongoing implementation costs.

For active managers, the proposition is that an alternative portfolio exists which will survive ongoing implementation costs required to maintain exposure to its strategy. As the size of the investable portfolio grows, aligning that portfolio more closely with market cap weights becomes a necessity, not an option, because research has shown that implementation costs rise at approximately the square root of assets.[4]. This implicitly concentrates the bulk of investor equity exposure into more competitive portions of the market—large, liquid names. While this is great for maximizing strategy capacity, alpha becomes more scarce as market cap increases. This disadvantages larger investors relative to their smaller counterparts.

The opportunity for alpha can be defined along two dimensions: consistency and magnitude. Consistency relates to how often alpha opportunities exist; base rates, or batting averages, capture this concept. Strategies that win more often than they lose are predisposed to generate persistent outperformance over time. However, if the average loss is greater than the average win, the power of consistency is diminished. Investors seeking persistent and outsized gains relative to some benchmark attempt to identify situations where consistency is in their favor, and the magnitude of wins is greater than that of losses. When properly aligned, consistency and magnitude have a compounding effect over time. A capacity-based view of the opportunity set runs exactly opposite of this concept, favoring allocations where consistency and magnitude are lowest.

The chart below represents a stylized capacity-based view of the equity opportunity set. Notice how the probability of outperformance is inversely related to the largest allocations.

For all but the largest investors, I would argue a returns-based approach is likely more applicable for assessing the opportunity set. The returns-based view begins by disaggregating the effect of market cap to equal weight the opportunity set. This has the effect of pulling allocations away from the largest, most liquid names.

Viewed as a level playing field, investors would begin allocating to asset classes for which the magnitude and consistency of alpha is aligned and highest. Adjustments related to investor risk tolerances, constraints, and costs of implementation could then be made with a greater understanding of the tradeoffs associated with those decisions.

Factor Spreads as a Proxy for Alpha

To evaluate the opportunity for alpha, two breakpoints are established within the U.S. market. The first breakpoint demarcates the difference between Large Stocks, which have a market cap greater than the average across all investable stocks on the U.S. market, and Small Stocks, which have a market cap less than average. The second breakpoint sets the minimum market cap of $200 million for Small Stocks. Stocks below $200 million and greater than $50 million are designated as Microcap Stocks.[5]

Within each particular asset class, factor spreads serve as a decent proxy for the availability of alpha. Factor spread is defined as the return of a portfolio of stocks falling into the highest-ranked decile of a factor minus the return of a portfolio comprised of the lowest-ranked decile of a factor. In the chart below, which shows the results of investing based on a multi-factor value theme within the micro and large stock universes, respectively, this would be the difference between the excess return of decile 1 and decile 10—28.2% for micro and 12.4% for large stocks. [6] Clearly, the microcap portion of the U.S. stock market has significantly wider spreads, more than twice that of Large Stocks, which suggests the opportunity to generate alpha is higher.

There are two key inferences from the chart above. First, stocks ranking in the cheapest decile within the microcap universe deliver on average nearly 3x the excess return of their large counterparts. They also outperform almost 25% more often in rolling three-year periods. Second, the most expensive stocks underperform their large counterparts by 2x, which suggests the benefits of avoiding the most expensive stocks are much greater. This helps explain why passive allocations in small and microcap fair worse in mega and large cap. The most expensive stocks also underperform with greater consistency, 95.4% of the time versus expensive large stocks which underperform only 79.4% of the time. This suggests a relatively greater opportunity for alpha generation and a wide margin for error in stock selection that provides flexibility to accommodate real world constraints.

Practically, this is reflective of how the results for active management have played out over the last decade. The chart below shows the performance of active managers within the micro, small, and large cap competitive universes.[7] The median micro cap manager outperformed the Russell Microcap benchmark by 2.55%—net of fees—while the median large cap manager underperformed the Russell 1000 by -0.46%. Even the 75th percentile manager in the micro cap universe outperformed the Russell Microcap benchmark by 1.21%.

Notice how the performance of the median manager decreases relative to the universe, and how the spread between the top and bottom quartile manager shifts lower. In other words, the magnitude and likelihood of alpha generation is inversely correlated to market cap, benchmark construction, and ultimately, the competitive nature of the space.

Generating Return “Expectations”

One challenge we often face as quantitative investors is the idea of developing return expectations. Frankly, forecasts give a false sense of precision, and should always be taken with a grain of salt. Though I believe that over very long periods, themes like value, momentum, yield, and quality will offer investors superior risk-adjusted return, there is no way to forecast the next year, three years, or even five years with any certainty.

A good bit of the academic literature suggests that the factor themes mentioned above represent risk “premiums.” The concept of premiums is a bit curious because it is suggestive of something that is always available, whereas, factors historically come in and out of favor. A mean-reverting perspective seems a more accurate depiction of the ebb and flow that is inherent in all factors. Since factor timing is a yet unsolved mystery, a reasonable approach seems to be consistent diversified exposure to multiple factor themes.

To illustrate, I created a hypothetical factor-based microcap portfolio. The portfolio is constructed by starting with a Microcap Stocks universe and eliminating stocks falling into the worst decile by our stock selection themes of financial strength, earnings quality, and earnings growth.[8] The portfolio then focusses in on stocks with the strongest combined score by our momentum and value themes. The portfolio is refreshed monthly based on a rolling annual rebalance.

After generating a return stream for this portfolio for the 35 year period from 1982-2016, the portfolio’s return is regressed on the excess return of the highest-ranking decile of the various factor themes to generate exposures, column 1 and 2 in the table below, respectively. I ran the same process for the benchmark in column 4. The contribution to return from factor exposures are in columns 3 and 5 for the portfolio and benchmark, respectively. Column 6 represents the Active Exposure—the difference in exposure for the portfolio and benchmark. Finally, column 7 decomposes the Factor Impact on the portfolio’s excess return. Using the Value line item as an example, the highest-ranked decile of stocks ranked by the value theme outperformed the universe return of 9.0% by an annualized 11.3% excess return. The portfolio had a 0.34 overweight exposure to Value, which contributed annualized excess of 3.9% to return over the full period.

Based on the results of this three-and-a-half-decade study, I make the intellectual leap that factor excess returns, volatility, and correlation in the future will somewhat resemble those of the past. Over reasonable time frames, this has been a decent assumption.[10]

“Expected” return, volatility, and risk-adjusted return (Sharpe) can be found at the bottom of the table. Since absolute returns are incredibly difficult, if at all possible to predict, the more instructive info is likely the excess return, tracking error, and information ratio. The table is suggestive that this factor based portfolio, which demonstrates strong active exposures to value and momentum, should generate excess return of 6.0% over the long-term. For perspective, this level of excess return would be representative of the 5th percentile manager within the microcap manager peer universe over the last 10 years. A key assumption that cannot be emphasized enough is consistent factor exposure throughout the period. For example, if this hypothetical strategy started buying growth stocks without regard to valuation in the 1990’s, this data becomes irrelevant. Discipline to any strategy is key to avoid behavioral pitfalls on a go-forward basis.

I performed similar exercises to generate factor-based portfolios within the large and small stock universes. To provide some diversification of factor exposure across the universe, the large portfolio uses Shareholder Yield as its final selection factor, and the small portfolio uses the Value composite theme. Results for those portfolios are included below.

In both cases, factor exposures are about as expected, the small portfolio had strong active value exposure with benchmark-like quality and momentum exposure. The large portfolio had strong active Shareholder Yield exposure, and mostly benchmark or better elsewhere. In both cases, the “expected” excess is on par with top active managers over the previous 10 years as illustrated in the chart above.

Determining Allocations

Finally, having generated expectations for excess return and volatility in micro, small, and large portfolios, we can apply the results in the common mean-variance optimization (MVO) framework to determine overall equity portfolio weights.

The inputs required for MVO are expected returns and covariances. I use the expected returns generated in the analysis above as inputs. The return streams for the portfolios are then used to generate a covariance matrix. Implicit is the assumption that covariances are stationary over time. We know this not to be the case in the short-term, so care should be taken to interpret the results only in the context of long-term strategic, not tactical, decisions. The correlation matrices below demonstrate the differentiated return profile that can be generated with carefully constructed factor portfolios as distinct from relying on market cap weighted benchmarks in asset allocation. Notice the decrease in correlation between the Micro and Small portfolios and the Russell 2000 and Russell Micro benchmarks.

Armed with expected returns and covariances, I apply very few constraints to the overall portfolio optimization. The portfolio must be fully invested at all times, and shorting is not allowed. Other than that, no constraints are needed to “force” the optimization into reasonable results. The objective is maximizing risk-adjusted return via the Sharpe Ratio.

The table below displays the results of the MVO process. Weighting to the micro and small portfolios are much greater than most allocators are probably used to at a combined 57% of the equity portfolio. The lower portion of the table includes summary statistics for our hypothetical optimization of the micro, small, and large factor portfolios as compared to the cap-weighted benchmarks at the optimization weights. Contrary to what might be commonly expected, the increased allocations do not result in dramatic increases in volatility. Volatility actually decreases by 1%. With 4.5% annualized excess return and 1% lower volatility, the Sharpe ratio increases dramatically.

As a comparator, I ran a parallel comparison which uses expected returns and volatility for the Russell benchmarks as proxies for cap-weighted portfolios. The results are much more in-line with typical investor allocations, though still probably higher on small cap than expected.

Thus far, we have not considered the risk tolerance of the allocator. While one set of investors may be perfectly comfortable with significant micro and small cap exposure, certain investors probably need to adjust portfolios to their risk preferences.

It turns out that there is a relatively simple way to scale portfolio returns based on volatility. This entails incorporating a penalty factor for portfolios that adjusts based on an investor’s risk aversion. Risk averse investors would incorporate a greater penalty in determining their appropriate policy portfolio. Less risk averse investors would incorporate lower penalties. Risk aversion could be modeled to incorporate any number of different characteristics—investment horizon, sensitivity to absolute and/or relative drawdowns, liquidity needs, etc. For this post, I demonstrate an example based on volatility.[11]

In the table below, a Risk Aversion score of 0 represents the utility of each portfolio for a return-seeking investor, effectively the results we determined in the MVO analysis above. We then apply successively increasing risk penalties based on the volatility of each portfolio. Because the micro and small portfolios are more volatile than large cap, returns decrease to the point at which the utility of the micro and small portfolio are close to indifferent with large by the time a risk aversion score of 5 is reached. Think of the returns below as a proxy for how the investor feels about the level of return given the volatility required to achieve that return. A highly risk averse investor with a score of 10 significantly prefers the large cap portfolio rather than small or microcap.

These utility-adjusted returns can then be used as expected return inputs into additional MVO analysis which adjusts allocations of the total equity portfolio for individual risk aversion. The table below displays these results at each level of risk aversion. As risk aversion increases, the optimal weight dials down exposure to the micro and small strategies in favor of the lower volatility large strategy. In my experience, most investors fall in the 3 to 6 range.

Conclusion

It would behoove of investors to recognize traditional indexes for what they are, factor-based strategies predicated on one factor, market cap. Though market cap has everything to do with low cost implementation, high capacity, and cheap beta exposure, it has little to do with optimal investor allocations for all but the largest plans.

Breaking from the capacity-based, cap-weighted perspective allows investors and allocators to focus on asset classes in which "edges" are apparent and hidden within traditional benchmarks. Allocators should view portfolios through the lens of consistent factor exposures over multiple market cycles. Doing so allows for reasonable "expected" excess returns that are otherwise overshadowed by cap-weighted indexes when used as proxies for asset class returns. Further, poor benchmark construction can, in and of itself, actually eliminate entire asset classes from consideration.

Using long-run factor excess, correlation, and risk aversion inputs in traditional MVO analysis yields surprising results that suggest volatility reduction and return enhancement through inclusion of micro and small cap stocks in equity asset allocation.

---

[1] See appendix for more detail and references.

[2] “Microcaps — Factor Spreads, Structural Biases, and the Institutional Imperative”. August 2017.

[3] As of 12/31/2016

[4] “Asset Growth and Its Impact on Expected Alpha.” Ronald N. Kahn.

[5] Market cap breakpoints are adjusted for inflation historically.

[6] The value theme is defined as an equal-weighted score of ranking based on price to sales, price to earnings, ebitda-to-ev, free cash flow-to-ev, and shareholder yield. Shareholder yield is the combination of share buybacks and dividend yield.

[7] PSN

[8] See Appendix for Factor Theme descriptions.

[9] See Appendix for Factor Theme descriptions.

[10] See Factor Correlations in Appendix

[11] For a detailed explanation of our approximation of risk aversion, see the appendix and Bodie, Kane, and Marcus (2004, p. 168).

 Appendix

Decision to Add the Asset Class

For an investor deciding to gain exposure to an asset class, the decision itself can be addressed through common frameworks that seek to balance risk-return tradeoffs. Blume (1984) and Elton, Gruber, and Rentzler (1987) suggest that the decision to add an asset class to an existing portfolio can be determined by comparing the Sharpe ratio of the new asset class with the correlation adjusted Sharpe ratio of the existing portfolio. The correlation adjustment is important as it incorporates the benefits of risk reduction when evaluating the new asset.

Factor Theme Descriptions

Universe - The market factor is an equal-weighted selection universe for the portfolio.

Value - The excess return of the highest-ranking decile of a Value Composite relative to the selection universe. The Value Composite consists of underlying constituents such as price relative to sales, earnings and cash flows.

Momentum - The excess return of the highest-ranking decile of a Momentum Composite relative to the selection universe. Momentum consists of four underlying constituents—3-month, 6-month, and 9-month momentum, and twelve month volatility.

Yield - The excess return of the highest-ranking decile of Shareholder Yield relative to the selection universe.

Earnings Quality - The excess return of the highest-ranking decile of an Earnings Quality Composite relative to the selection universe. The composite consists of several underlying constituents, which measure the conservatism of accounting choices through accruals.

Financial Strength - The excess return of the highest-ranking decile of a Financial Strength Composite relative to the selection universe. The composite consists of multiple underlying constituents, which assess balance sheet leverage and strength.

Earnings Growth - The excess return of the highest-ranking decile of an Earnings Growth Composite relative to the selection universe. The composite consists of multiple underlying constituents, which measure the consistency of earnings and profitability.

Implementation – A proxy for the cost of implementation is measured using two factors that historically correlate with the cost of trading, such as dollar volume and market cap.

Factor Correlations

The table below includes summary stats for the rolling 36-month correlation of the highest-ranked decile of six factor themes encompassing value, momentum, yield, and quality relative to the microcap universe. Correlations are on average above 0.9, with deviations within reasonable bounds.

Risk Aversion

Risk aversion can be proxied through utility theory. Practically, the return of a portfolio can be adjusted through a penalty factor for increased volatility. Risk averse investors would incorporate a greater penalty in determining their appropriate policy portfolio. Less risk averse investors would incorporate lower penalties based on increases in risk. Risk aversion could be modeled to incorporate a number of different characteristics—investment horizon, sensitivity to absolute and/or relative drawdowns, liquidity needs, etc.

So far in this series, we’ve covered faulty benchmark construction, the wide array of fundamental drivers, and the critical importance of quality in cutting through the noise among micro cap stocks. Now, we turn to the largest factor spreads I’ve come across in any segment of the market, real world considerations for implementation, and why the dichotomy of scale versus alpha could result in a persistent opportunity for outperformance.

Factor investing is more effective in micro than any other cap range

Though factor investing has rooted itself squarely in large cap equities, it significantly more effective in eclectic corners of the market—small and micro cap. Thus far, we touched on quality themes like financial strength, earnings quality, and earnings growth to screen stocks out. Let’s turn our focus to a broader suite of multi-factor themes by bringing value and momentum into the arena. While value and momentum are also effective in negative screening, they are most effective in identifying which stocks to select.

In an analysis of the performance of each multifactor theme from 1982-2016, shown below, there are enormous differentials in the return spread between high and low-ranking stocks. Spreads serve as a proxy for robustness of a factor. In the academic literature, these are hypothetical long-short portfolios that suggest the size of a systematic return premium.

The table below displays the spread between the return of high and low deciles in Large, Small, and Micro stocks on each theme. Using this lens, it is readily apparent that factors are more robust in micro than large, and even small stocks. Within the micro cap space, the smallest spread (Earnings Quality, 14.5%) is wider than the largest in Large Stocks (Value, 12.5%). This again highlights the importance of quality in microcap. As measured by the spread, quality is 3-4x more important in Micro as Large.

The spreads for value and momentum are greater than twice the spread in Large Stocks. The Value spread in Micro suggests one could go long a portfolio of the cheapest stocks and short the most expensive to earn an eye-popping 28.2% annualized return. Practically, this would be virtually impossible due to the operational challenges and costs of managing the short side of a micro cap portfolio. This real world complexity necessitates a focus on not owning the lowest ranked names—as opposed to shorting, and owning the highest ranked names.

The table below continues our previous analysis of adjusting the microcap universe for quality by eliminating poorly ranked names. To that high quality group of stocks, the two rightmost columns display the results from only owning names falling in the highest ranked decile by Momentum and Value, respectively.

As was the case with quality adjustments, a focus on momentum improves return by 3.6% annualized with 10% lower volatility than the quality adjusted group. The addition of value is even more compelling. A focus on value improves return by 6.1% annualized with 17% lower volatility than the quality adjusted group.

Structural features underpin the persistence of factors in micro cap

While I wish I was the only one aware of the massive spreads available in the micro cap space, the reality is that this information is well known. On a recent Masters in Business podcast, Ed Thorp remarked that "Any edge in the market is limited, small, temporary, and quickly captured by the smartest, or best informed, investors."

It is curious that investors have not arbitraged this clear edge. In practice, real world implementation costs quickly erode theoretical alpha if not managed precisely. Thorp also commented "Every stock market system with an edge is necessarily limited in the amount of money it can use and still produce extra returns." Along these lines, there are three inherent structural constraints to scale that hamper professional money managers, thus, protecting the persistence of alpha for dedicated investors at appropriate scale.

Supply of Transactable Stock

Liquidity can be thought of along a spectrum that ranges from the most liquid U.S. Treasury securities (T-bills) to illiquid private businesses (Private Equity). Moving to the illiquid end of the spectrum, the cost of implementation increases, which magnifies the importance of expertise when transacting in scale. The primary considerations as it relates to implementation on the liquidity spectrum are free float and dollar volume of transactions.

Float is the number of shares that are freely available to trade. While a mega cap firm like Apple has a free float of 96% of its shares outstanding, micro cap stocks tend to have the lowest free float as a percentage of the total shares of any market cap range. As of year-end 2016, their average free float was just 72% of shares outstanding. Because of their stage in the business life cycle, micro caps commonly feature large ownership by founders and insiders, and relative to large stocks, could be considered closely held. This feature is important because it reduces the available supply of stock to transact in by 28%. Given Apple’s 96% free float, $737 of its $767 billion in market cap is freely tradeable. Within the micro cap universe of $50-$200 million, just $72 billion of the $100 billion is freely tradeable. This curtails the size of any individual actor in the space, including large institutional investors and product providers.

Volume of Transactable Stock

Dollar volume gives a sense of transaction velocity, and an investor’s ability to enter and exit the market at will. The chart below details total dollar volumes, adjusted for inflation, for large, small, and micro stocks over the last 20 years. The difference in dollar volume is astounding. Dollar volume in large and small stocks is 245x and 43x greater, respectively, than micro cap. The relatively low $420 million volume for micro cap suggests that an active manager employing strategies similar to the ones discussed in this paper would find it rather difficult to oversee assets of significant size, while still being able to transact.[1]

Transaction Costs

While free float and volume constrain the ability to oversee a large amount of assets in the space, implementation costs erode theoretical factor spreads. At scale, these costs can be material. Real world costs have always been, and will likely continue to be, a barrier to entry at scale in less efficient spaces.

The three costs of implementation investors must grapple with are commissions, market impact, and bid-ask spreads. Fortunately, commissions have a relatively low impact on cost given the highly competitive nature of the brokerage business. Most institutional transactions occur at pennies per share (generally not relevant unless transacting in penny stocks). Commissions are the only true explicit cost. The more relevant and hidden share of costs are market impact and bid-ask spreads. Market impact is effectively how much you move the market when transacting at a certain size.

The chart below organizes the U.S. market into liquidity groups sorted from most to least dollar volume to assess the market impact and bid-ask spread of a hypothetical $10 million trade to get exposure to each liquidity group. Overlaid on the chart is the measure of dollar volume across the market. The horizontal axis is the average market cap for each liquidity bucket. From this, one can infer that dollar volume and market cap are highly positively correlated, while cost and dollar volume are clearly inversely correlated. Said another way, the smaller the stock, the lower the dollar volume, the more expensive to trade.

A $10 million trade could be implemented in the most liquid group of U.S. stocks for approximately 5 bps. Sophisticated trading techniques would likely neutralize this impact altogether as smart traders act as liquidity providers when establishing positions. This is made considerably easier with an average $480 million average volume in the most liquid stocks with which to work the trade. Capacity in this part of the market is virtually unlimited. On the other end of the spectrum, stocks in the least liquid bucket bear an all-in cost estimate of 220bps, a 44x increase in cost on 99.97% lower dollar volume with which to trade. Again, sophisticated trading techniques could minimize, but in this case not eliminate, relevant costs. Capacity in this corner of the market is low, but alpha potential remains massive net of transaction costs.

Supply, volume, and cost act as significant barriers to scale in micro cap. Not only do they require a specialized set of skills to implement portfolios in an efficient manner, but they require restraint on the part of money managers as it relates to asset gathering.

Scale Destroys Alpha, Alpha is Expensive to Realize

Objectively, the capacity of a given strategy is a function of supply, volume, and cost of implementation. Subjectively, and most importantly, capacity is determined by the investment manager’s desire for assets under management. Increasing strategy capacity can often lead to conflicts of interest between the business necessity for fee generation and the client necessity for alpha generation. There is a dichotomy in the fact that less liquid micro cap stocks require greater skill in implementation, while also requiring restraint in scale. From a product management perspective, the space is anathema to large money management organizations because asset-based fees on low capacity strategies struggle to support the costs of dedicated teams and infrastructure.

It is not uncommon for micro and small cap managers to creep up the market cap spectrum in order to realize greater capacity. Moving up-market results in smaller factor spreads, and therefore, reduced opportunity for alpha generation. Another alternative is bearing greater market impact costs through larger trade sizes. Both are unappealing options. Pushing the limits of scale could easily detract hundreds of basis points of return. Static, however, are manager fees, which capture a greater proportion of alpha even as the effectiveness of factors are diluted.

Look beyond highly competitive markets for factor exposure

We’ve been conditioned for decades to believe that obvious anomalies will be arbitraged away. Business schools teach the fundamental principles of the efficient market hypothesis even though it clearly does not reflect reality. Most investors readily agree that alpha is scarce. It is hard to find, highly sought after, and requires skill to extract.

Based on this premise and the recent horrendous performance of active managers, many investors establish their beachhead in the most competitive portions of the equity market, large cap, where alpha is scarcest.

I’ll call this the institutional perspective. Though we often mock the Hollywood scene, we are just as guilty of star-gazing. Institutions follow their peers like hawks, and research has shown that herding does occur amongst sophisticated investor asset allocations. For a multi-billion dollar plan, sheer size prevents them from accessing the micro cap. A $5 billion plan would probably need to make a $100 million allocation to micro cap to make a difference to overall plan returns. That’s a large allocation to a constrained space. Going larger, a $30 billion plan? Forget about it! So, instead they pay massive fees for coveted, concentrated access to illiquid private equity markets where their edge cannot be arbitraged away—as easily. But, should smaller investors follow suit?

While large allocators face structural constraints, all else equal, this behavior doesn’t make sense for smaller investors. Just as business schools teach the intricacies of the efficient market hypothesis, students cross the courtyard for their next round of classes in…marketing strategy, corporate finance, competitive strategy, game theory, entrepreneurship…all geared toward identifying and exploiting strategic advantage in business. Investors should start building allocations where competition is low and alpha is less scarce—micro cap. Why not approach allocations from the non-institutional perspective?

At a time when the proliferation of factor investing is being driven by asset gatherers in highly-competitive spaces, my guess is that discerning investors find the research on factors in micro cap quite enticing. In this series on micro cap, we began by reviewing the Russell definition of micro cap, finding that the majority (88%) of what Russell considers micro cap to actually be small cap. The inferior construction methodology of the index—simple market cap weighting—omits critical considerations for quality and the cost of implementation in micro cap. The lackluster results of index returns fail to offer a compelling narrative for micro cap allocations. We then explored the composition of the micro cap universe to shed light on why it is a less competitive and lower quality space. A revolving door of new ventures and fallen angels flank a core group of steady state firms, which cause significant variability in the measurement of underlying stock fundamentals—often leading investors to write off the space as a junk yard littered with poor quality stocks. We then homed in on pure micro cap stocks that offer the potential for risk-adjusted return on par with large stocks through a framework for quality assessment. We noted the significantly greater spreads for the multi-factor stock selection themes of value, momentum, earnings quality, financial strength, and earnings growth in micro cap as compared to large and small stocks. We closed with an argument for the persistence of alpha generation in micro cap based on the structural barriers of supply, volume, and implementation costs to scale.

By breaking away from the institutional paradigm that is heavily aligned with the most competitive portions of the market, avoiding low quality, controlling implementation costs, and focusing in on stocks with strong momentum and value characteristics, I believe investors can realize substantial alpha in this capacity-constrained

In the previous post in this microcap series, I established that the fundamental drivers of micro cap businesses are widely varied, at least in part, due to their state of being—new venture, steady state, and fallen angel. This poses a challenge for stock pickers. It creates a lot of noise in the data, as we saw in the previous post, and requires expertise in many different types of situations—venture, growth, distressed, etc. One could argue that stock picking in microcap requires a broader analytic skill set than for more stable large stocks. This post attempts to cut through some of the noise inherent in micro caps to level the playing field.

Let’s take a step back to build intuition for stock selection regardless of where a company falls on the business life cycle. In our research, we have found several quality metrics to be indicative of good businesses. Generally, businesses should be profitable, growing at a reasonable pace, and appropriately capitalized. Individually, these metrics are effective, but when used together thematically, they provide a powerful framework for eliminating poor quality stocks. The table below compares several characteristics for Large and Mirocap stocks.

In each case, a simple average of characteristics for Micro cap stocks betrays the universe’s lower quality nature relative to Large Stocks. One would assume from looking at the Microcap Stocks column that these businesses are rapidly growing their asset base (Change in Net Operating Assets), not particularly profitable (ROIC), taking on tremendous debt (1-Year Debt Change), and generating negative free cash flow (Free Cash Flow-to-Enterprise Value). All of these would seemingly signal starvation for cash, a particularly bad thing for small businesses.

Change in Net Operating Assets (NOA) measures the growth in assets required to run the business. If a small consumer products company, for example, hit the jackpot with a new contract at a huge retailer and then had to ramp up production to fulfill the order, this metric would increase. Sales growth requires large investment for raw materials, inventory, delivery of finished goods, and equipment for ongoing production. The challenge with growth is that it requires huge cash outlays. This cash is all outlaid before revenue occurs. Dramatic growth in operating assets can be indicative of stress, as it leaves the business in a tenuous cash position. This state of affairs appears to be the norm for micro cap stocks with an average change in NOA of 44.3%—close to twice the rate for Large Stocks.

Few small firms have enough internal capital to fund such large investments. They then turn to capital providers to fund growth—equity offerings or taking on debt. Keep in mind that many micro cap stocks have no analyst coverage, so the ability to tap equity capital markets is limited and expensive. Debt becomes the default source of capital. The average 1-Year Change in Debt for the universe is 32.6%, and debt-to-equity is on par with Large Stocks. The ROIC of just 13.2% indicates that capital, of which debt is a part, is not being as efficiently invested as with Large Stocks. A free cash flow yield of -4.5% suggests economic value is being destroyed, rather than created.

Each of these characteristics are components of multi-factor themes that can be used to assess the quality of a firm: Earnings Quality (NOA), Financial Strength (D/E, Change in Debt), and Earnings Growth (ROIC). To level the playing field for comparison to Large Stocks, we can rank stocks in the micro cap universe based on these themes and eliminate the lowest ranking decile. Firms falling into these poorly-ranked groups tend to be poorly capitalized and have low profitability and weak earnings quality.

By adjusting the micro cap universe, the overall metrics dramatically improve, and in some cases, are actually better than Large Stocks. Quality Adjusted Microcap Stocks reveal much more moderate growth rates in NOA. An average 13.7% is indicative of businesses that are more likely to handle organic business growth without needing to seek substantial funding from debt or equity issuance. The improvement in the 1-Year Debt Change metric after adjusting for quality supports this logic. A large 32.6% increase in debt decreases to just 12.1%—lower than the average for Large Stocks.

Clearly, the universe quality metrics have improved, but how does this translate into investor returns? It turns out that elimination of poor quality boosts the return of our universe by 5.3% annualized with a 0.7% reduction in annual volatility (table below).

Incorporating quality criteria to eliminate stocks from consideration has a dramatic impact on micro cap stocks. Performing a quality assessment highlights the importance of a less appreciated aspect of factor investing. While many researchers focus on the outperformance associated with factors, the avoidance of groups of stocks can be just as positive a contributor to investor returns. After controlling for quality, the risk-adjusted returns available are in-line with large stocks. In the first post of this series, I mentioned that the historical return and risk of the Russell Microcap® Index did not merit an allocation according to mean-variance optimization. This simple quality screen alters the space’s characteristics to such an extent that it becomes a viable source of differentiated return for investors.

For the first installment of this series, I dove into the Russell Microcap® Index to understand it’s construction and behavior. I showed that Russell’s definition of micro cap is flawed in that it is predominantly representative of small cap stocks, and includes highly illiquid names that drag on performance. From an allocator’s point of view, the index return is lackluster when compared to large cap stocks. For this post, I think more fundamentally about what drives the opportunity set of “true” micro cap stocks.

For the remainder of this series, I diverge from the Russell index definitions to get a better sense of the composition of micro cap and the alpha opportunity available. I define micro cap stocks as those trading on U.S. exchanges with an inflation-adjusted market capitalization between $50 million and $200 million.[1] This micro cap universe is also equal-weighted, as opposed to cap-weighted. This provides a “pure” view of the micro cap market that has minimal overlap with small cap stocks. This group of about 1,300 stocks represents a disproportionately small 0.4% of total U.S. market capitalization. With average daily volume of just $700 thousand and a cumulative market cap of about $100 billion, the group is a mixture of exciting growth opportunities and the land of misfit toys. Once I screen out companies with unreasonable liquidity and non-U.S. firms, the list dwindles to about 500 investable stocks.

For comparison purposes, I periodically refer to a Large Stocks universe. Large stocks consist of U.S. firms with a market capitalization greater than the average capitalization for the total market, currently those stocks above an inflation-adjusted $7 billion market cap. This group is instructive as it represents the bulk of investor’s U.S. equity allocation. It is analogous to the S&P 500 Index on an equal-weighted basis.

Unique aspects of the micro cap universe

An investor cannot fully appreciate the micro cap space without understanding how stocks have come to fall on the micro cap spectrum. Whereas, most large stocks have succeeded in attempts to grow their businesses, as recognized by their multi-billion dollar valuations, micro cap stocks are on a completely different playing field. These businesses range from biotech startups to failing businesses that have depreciated to their current middling market cap. From an empirical perspective, the result is a lot of noise in the data.

To demonstrate, let’s look at one of the most fundamental metrics for a firm, sales growth. Though its efficacy as an investment factor is marginal, sales are the lifeblood of any firm and have a cascading effect on all other elements of the financial statements. The chart below compares the distribution of 3-year sales growth across large and micro cap stocks.

Notice the significantly fatter tails for micro cap relative to large stocks. If growth in sales is the most basic assessment of the state of a firm, this suggests much greater dispersion in the underlying metrics of micro caps. The popular rhetoric is often that small and micro cap stocks are junkier than their large cap counterparts. While this is true on average, a wide dispersion in fundamental metrics obscures many phenomenal businesses in meaningless averages.

A deeper dive reveals a disparate group of constantly evolving (and devolving) businesses

Investors have widely accepted that there exist many different types of private equity—angel investing, venture, early stage, late stage, mezzanine, LBO’s, distressed. Interestingly, in the private space, these labels represent the need of the firm receiving the investment. Just as there are many sub-classes of venture capital and private equity, such is the case with micro cap stocks, but for whatever reason, we do not view these businesses with the same categorical lens as we do private investments.

The micro cap universe can be divided into three broad categories: New Ventures that have become revenue generating within the last three years, distressed Fallen Angels that have descended into the micro cap universe from small cap—and sometimes large cap—and those in a Steady State that have been micro caps for at least three years.

From 1982-2016, new ventures represented 25% of the micro cap universe, while 16% were fallen angels, and 59% were steady state. Effectively, 41% of the universe is in some sort of transition—from startup to established firm, or from established firm to potential liquidation. When you think about micro cap, think of a revolving door where firms are constantly entering and leaving for different reasons.

This simplistic perspective on the universe is relevant because it sheds light on the strong inherent biases that skew the underlying fundamental characteristics. Below is the same distribution of sales growth for micro cap broken down by these three categories.[2] These disparate groups possess fundamentally different metrics that when averaged together obscure a lot of noise in Micro cap stocks.

New Ventures, with their small sales bases are highly skewed towards positive sales growth. Unsurprisingly, new ventures tend to be comprised of Information Technology and Health Care stocks—most notably biotech, software and pharmaceuticals. Currently, these industries represent a rather large 20% of the micro cap universe. The average annualized return of this group from 1982-2016 is 4.7%, woefully short of the micro cap universe average of 8.9%. Adding insult to injury, annualized volatility for this group is 27.8%. This likely has to do with the nature of outcomes in the space. Biotech firms generally succeed or fail in what amounts to binary outcomes—leaving investors with staggering gains or maximum losses.

Steady State firms are more centered in the distribution, but still positively skewed. At 59% of the overall universe, a good proportion of steady state firms are Commercial Banks and Thrifts. These two industries represent 20% of the universe currently. Banks are the least volatile micro cap industry and one of the top performers. The remainder of firms in this category tend to be widely dispersed across industries. Steady state firms are the best performing of the three categories with an annualized return of 10.1% and volatility of 22.8%.

Fallen Angels skew significantly in the negative growth direction. This group is a smattering of firms across industries. Currently, the oil & gas industry has the highest representation in this category. It tends to offer representation of groups of stocks that suffered in the previous cyclical business downturn. This group delivered an annualized return of 8.0% from 1982-2016 with volatility of 27.4%.

Our task as factor investors is to develop empirical criteria that enable us to cut through the noise to separate the good from the bad within the space. Given the perspective above, we know that there are reasonable fundamental explanations for the “junk-ish” nature of micro cap stocks. Quite simply, a lot of micro cap stocks possess poor business characteristics, whether that be weak cash flow generation, too much leverage, or dwindling and unprofitable revenues. By categorically identifying and removing firms with poor characteristics, we can improve the investor’s base rate for success.

For the third installment on micro cap investing, I’ll suggest some screening criteria that will aid in parsing out some of the “junkiness” of microcaps. Surprisingly, eliminating low quality, puts risk-adjusted returns for micro caps on par with large stocks.

 [1] This would be analogous to the approximately 2,600th to 4,000th stock using Russell’s ordinal market cap ranking methodology.