Alpha Architect

Relative Sentiment and Machine Learning for Tactical Asset Allocation: Out-of-Sample Results

We examine Sentix sentiment indices for use in tactical asset allocation. In particular, we construct monthly relative sentiment factors for the U.S., Europe, Japan, and Asia ex-Japan by taking the difference in 6-month economic expectations between each region's institutional and individual investors. These factors (along with one-month forward equity returns) then serve as inputs to a wide array of machine learning algorithms. Employing combinatorial cross-validation and adjusting for data snooping, we find relative sentiment factors have robust and significant predictive power in all four regions; that they surpass both standalone sentiment and time-series momentum in terms of informational content; and that they demonstrate the ability to identify the subsequent best- and worst-performing global equity markets from along a cross-section. The results are consistent with previous findings on relative sentiment, discovered using unrelated datasets.

Can We Measure Inflation with Twitter

Drawing on Italian tweets, we employ textual data and machine learning techniques to build new real-time measures of consumers’ inflation expectations. First, we select keywords to identify tweets related to prices and expectations thereof. Second, we build a set of daily measures of inflation expectations around the selected tweets, combining the Latent Dirichlet Allocation (LDA) with a dictionary-based approach, using manually labeled bi-grams and tri-grams. Finally, we show that Twitter-based indicators are highly correlated with both monthly survey-based and daily market-based inflation expectations. Our new indicators anticipate consumers’ expectations, proving to be a good real-time proxy, and provide additional information beyond market-based expectations, professional forecasts, and realized inflation. The results suggest that Twitter can be a new timely source for eliciting beliefs.

Momentum Everywhere, Including in Factors

Managed portfolios that exploit positive first-order autocorrelation in monthly excess returns of equity factor portfolios produce large alphas and gains in Sharpe ratios. We document this finding for factor portfolios formed on the broad market, size, value, momentum, investment, prof- itability, and volatility. The value-added induced by factor management via short-term momentum is a robust empirical phenomenon that survives transaction costs and carries over to multi-factor portfolios. The novel strategy established in this work compares favorably to well-known timing strategies that employ e.g. factor volatility or factor valuation. For the majority of factors, our strategies appear successful especially in recessions and times of crisis.

Global Factor Performance: July 2022

Standardized Performance Factor Performance Factor Exposures Factor Premiums Factor Attribution Factor Data Downloads

Does Intangible-Adjusted Book-to-Market Work?

The book-to-market ratio has been widely used to explain the cross-sectional variation in stock returns, but the explanatory power is weaker in recent decades than in the 1970s. I argue that the deterioration is related to the growth of intangible assets unrecorded on balance sheets. An intangible-adjusted ratio, capitalizing prior expenditures to develop intangible assets internally and excluding goodwill, outperforms the original ratio significantly. The average annual return on the intangible-adjusted highminus-low (iHML) portfolio is 5.9% from July 1976 to December 2017 and 6.2% from July 1997 to December 2017, vs. 3.9% and 3.6% for an equivalent HML portfolio

Wes Discusses Merging Education, Investing, and the Marines w/ Lagniappe Podcast

In this episode, Wes talks with Doug and Greg about why Alpha Architect is abnormal, the method to their madness, how Wes challenged Eugene Fama (and nearly won), how Alpha Architect is responding to today’s volatile markets, and what the future looks like for value investors.

Combining Factors in Multifactor Portfolios

Reschenhofer’s findings demonstrate the important role that portfolio construction rules (such as creating efficient buy and hold ranges or imposing screens that exclude stocks with negative momentum) play in determining not only the risk and expected return of a portfolio but how efficiently the strategy can be implemented (considering the impact of turnover and trading costs)—wide (narrow) thresholds reduce (increase) portfolio turnover and transactions costs, thereby increasing after-cost returns and Sharpe ratios. His findings also provide support for multiple characteristics-based scorings to form long-only factor portfolios, encouraging the combination of slow-moving characteristics (such as value, investment and/or profitability) conditional on fast moving characteristics (such as momentum), to reduce portfolio turnover and transactions cost. Fund families such as AQR, Avantis, Bridgeway and Dimensional use such an approach, integrating multiple characteristics into their portfolios conditional on momentum signals.

Can Machine Learning Identify Future Outperforming Active Equity Funds?

We show, using machine learning, that fund characteristics can consistently differentiate high from low-performing mutual funds, as well as identify funds with net-of-fees abnormal returns. Fund momentum and fund flow are the most important predictors of future risk-adjusted fund performance, while characteristics of the stocks that funds hold are not predictive. Returns of predictive long-short portfolios are higher following a period of high sentiment or a good state of the macro-economy. Our estimation with neural networks enables us to uncover novel and substantial interaction effects between sentiment and both fund flow and fund momentum.

Using Institutional Investor’s Trading Data in Factors

The authors investigate how the interaction between entries and exits of informed institutional investors and market anomaly signals affects strategy performance. The long legs of anomalies earn more positive alphas following entries, whereas the short legs earn more negative alphas following exits. The enhanced anomaly-based strategies of buying stocks in the long legs of anomalies with entries and shorting stocks in the short legs with exits outperform the original anomalies, with an increase of 19–54 bps per month in the Fama–French five-factor alpha. The entries and exits of institutional investors capture informed trading and earnings surprises, thereby enhancing the anomalies.

Does Emerging Markets Investing Make Sense?

The analysis above suggests that portfolios that include or exclude emerging allocations are roughly the same. For some readers, this may be a surprise, but for many readers, this may not be "news." That said, even if the data don't strictly justify an Emerging allocation, the first principle of "stay diversified" might be enough to make an allocation.

Of course, the assumptions always matter.

Relative Sentiment and Machine Learning for Tactical Asset Allocation

The application of machine learning models to Sentix relative sentiment data appears to extract more predictive information than our original, simplistic approach was capable of.

Factors Investing in Cryptocurrency

We find that three factors—cryptocurrency market, size, and momentum—capture the cross-sectional expected cryptocurrency returns. We consider a comprehensive list of price- and market-related return predictors in the stock market and construct their cryptocurrency counterparts. Ten cryptocurrency characteristics form successful long-short strategies that generate sizable and statistically significant excess returns, and we show that all of these strategies are accounted for by the cryptocurrency three-factor model. Lastly, we examine potential underlying mechanisms of the cryptocurrency size and momentum effects.

The Unintended Consequences of Single Factor Strategies

Since the 1992 publication of “The Cross-Section of Expected Stock Returns” by Eugene Fama and Kenneth French factor-based strategies and products have become an integral part of the global asset management landscape. While “top-down” allocation to factor premiums (such as size, value, momentum, quality, and low volatility) has become mainstream, questions remain about how to efficiently gain exposure to these premiums. Today, many generic factor products, often labeled as “smart beta”, completely disregard the impact of other factors when constructing portfolios with high exposures to any single factor. However, recent research, such as 2019 study “The Characteristics of Factor Investing” by David Blitz and Milan Vidojevic, has shown that single-factor portfolios, which invest in stocks with high scores on one particular factor, can be suboptimal because they ignore the possibility that these stocks may be unattractive from the perspective of other factors that have demonstrated that they also have higher expected returns.

Global Factor Performance: June 2022

The following factor performance modules have been updated on our Index website.[ref]free access for financial professionals[/ref]

Short-term Momentum

We document a striking pattern in U.S. and international stock returns: double sorting on the previous month’s return and share turnover reveals significant short-term reversal among low-turnover stocks, whereas high-turnover stocks exhibit short-term momentum. Short-term momentum is as profitable and as persistent as conventional price momentum. It survives transaction costs and is strongest among the largest, most liquid, and most extensively covered stocks. Our results are difficult to reconcile with models imposing strict rationality but are suggestive of an explanation based on some traders underappreciating the information conveyed by prices.

S&P 500 Valuations with and without Technology

A few quick charts for our readers. As we all know, technology-related sectors and names have been crushed. But is blood in the streets? Not really.

Strategies to Mitigate Tail Risk

Investors care about more than just returns. They also care about risk. Thus, prudent investors include consideration of strategies that can provide at least some protection against adverse events that lead to left tail risk (portfolios crashing). The cost of that protection (the impact on expected returns) must play an important role in deciding whether to include them. For example, buying at-the-money puts, a strategy that eliminates downside risk, should have returns no better than the risk-free rate of return, making that a highly expensive strategy.

Value Investing: Headwinds, Tailwinds, and Variables

Investing is no different. A question we regularly get in the current environment is "How does inflation affect value stocks?" Well...it depends. I could show you some data on how value stocks did in the 70's (period of high inflation) versus how they did in the 90's (low inflation). But if WW3 broke out tomorrow, wouldn't that variable quickly top all other variables? Probably. So let's table that variable.

Trend Following: Timing Fast and Slow Trends

A large body of evidence demonstrates that momentum, including time-series momentum (trend following), has improved portfolio efficiency. Research has found that there are a few ways to improve on simple trend-following strategies. Techniques that have been found to improve Sharpe ratios and reduce tail risk include volatility scaling and combining fast and slow signals as well as combining long-term reversals. These have been incorporated by many fund managers into investment strategies. Cheng, Kostyuchyk, Lee, Liu and Ma provided evidence that machine learning could be used to further improve results. With that said, a word of caution on the use of machine learning is warranted. The powerful tools and the easy access to data now available to researchers create the risk that machine learning studies will find correlations that have no causation and thus the findings could be nothing more than a result of torturing the data. To minimize that risk, it is important that findings not only have rational risk- or behavioral-based explanations for believing the patterns identified will persist in the future, but they also should be robust to many tests. In this case, investors could feel more confident in the results if their findings were robust to international equities and other asset classes (such as bonds, commodities and currencies).

Form 3 and Form 4 Alpha: Focus on What Insiders Don’t Trade

Some individuals, e.g., those holding multiple directorships, are insiders at multiple firms. When they execute an insider trade at one firm, they may reveal information about the value of all—both the traded insider position and not-traded insider position(s)—the securities held in their “insider portfolio.” We find that insider “not-sold” stocks outperform “not-bought” stocks. Implementable trading strategies that buy not-sold stocks following the disclosure of a sale earn alphas up to 4.8% per year after trading costs. The results suggest that even insider sales that are motivated by liquidity and diversification needs can provide value-relevant information about insider holdings.

Are You Ready to Become a Better Investor?

Get Updates Now

Recent Posts

Important Disclosures