Before we start the series on the individual factors that underly our timing model, we are going to address a legitimate concern that was discussed in a recent paper by Amit Goyal, Ivo Welch, and Athanasse Zafirov, “A Comprehensive 2022 Look at the Empirical Performance of Equity Premium Prediction” (available at ssrn.com).

They re-examined a set of variables that have been claimed to predict the equity risk premium and found them to perform poorly. Very poorly. About half the features had insignificant power even in sample, and most of the others were poor out of sample. This seems to be a damning indictment of our research and investment philosophy.

Readers might be aware of a similar critique of factor investing, where various groups of stocks (for example small cap stocks) are postulated to outperform on a relative basis. Early research examined a small group of factors such as value, size, beta, and momentum. But the number of factors soon exploded. “A Census of the Factor Zoo” by Campbell Harvey and Yan Liu (also available at ssrn.com) documented over 400 factors that had appeared in reputable journals and showed that many were insignificant.

We agree strongly that most published predictors are poorly performing both in terms of statistical tests and practical investment. But we disagree with the authors’ conclusion, even without doubting their methodology. The situation isn’t as bleak as it may first appear.

• The fact that most of the predictors weren’t helpful shouldn’t obscure the more important fact that some were. Predicting the market is hard. Most things won’t work. Even most things that survive enough tests to be published, probably won’t really work. Data mining is a hell of a drug.

• None of the things that work are independent of each other. If they are all successfully predicting the same instrument, they must be in agreement a lot of the time. But the relationships are not simple, are non-linear, and vary over time. So, building a portfolio from the signals is non-trivial and important. Portfolio construction can’t turn poor signals into good signals, but bad portfolio construction can ruin things even if the components are good.

• Even things that work will have long periods when they don’t. The long-term positive return to common stocks is the most well documented empirical fact in finance, and this fact isn’t wrong just because the S&P 500 was unchanged from 2000 to 2013. If the strongest risk premium of all can fail to perform over more than a decade, isn’t it unrealistic to think that other predictors won’t also have bad periods?

Also in the paper, the authors caution on believing a result more because it is based on a “strong theoretical basis”. Again, we generally agree with an important caveat. Caution is absolutely required here, as almost any result can be justified by appealing to some behavioral theory. The most caricatured example of the bad approach would be to data mine an effect and then justify it by sifting through the list of behavioral “anomalies”. There are now hundreds of these, so almost anything can be explained (it seems like most things could be explained through just over-confidence or under-confidence).

However, because something can be done poorly doesn’t mean it shouldn’t be done well. We strongly believe that a prior theoretical basis is a necessary part of finding good predictors. Note the qualifier, “prior.” Post-hoc rationalization is a different story.

This is a very good paper. Reading something that challenges your assumptions, processes, and conclusions is more useful than reading something that agrees with you. And, just as our models themselves are adaptive, so is our overall research process.

Disclaimer

HTAA, LLC serves as the investment advisor to the fund. The Fund is distributed by Northern Lights Distributors, LLC., which is not affiliated with HTAA, LLC or any of its affiliates. HTAA is not affiliated with Ultimus Fund Solutions, LLC or any of its affiliates.

Carefully consider the Fund’s investment objectives, risk factors, charges and expenses before investing. This and additional information can be found in the Fund’s prospectus, which may be obtained by visiting www.hulltacticalfunds.com. Read the prospectus carefully before investing.

Investing involves risk, including the possible loss of principal. Investments in smaller companies typically exhibit higher volatility. The Fund will invest in (and short) exchange-traded funds (ETFs). The Fund will be subject to the risks associated with such vehicles. The Fund may also invest in leveraged and inverse ETFs. Inverse and leveraged ETFs are designed to achieve their objectives for a single day only. For periods longer than a single day, leveraged or inverse ETFs will lose money when the performance of the underlying index is flat over time, and it is possible that a leveraged or inverse ETF will lose money over time even if the level of the underlying index rises or, in the case of an inverse ETF, falls In addition, shareholders indirectly bear fees and expenses charged by the underlying ETFs, as well as the Fund’s direct fees and expenses. The Fund may invest in derivatives, including futures contracts, which are often more volatile than other investments and may magnify the Fund’s gains or losses.

The Fund is an actively managed ETF and, thus, does not seek to replicate the performance of a specified passive index of securities.

The Fund may take short positions. The loss on a short sale is theoretically unlimited. Short sales involve leverage because the Fund borrows securities and then sells them, effectively leveraging its assets. The use of leverage may magnify gains or losses for the Fund.

There is no guarantee that any investment strategy will produce positive results. There is no guarantee that distributions will be made.