Blog: The Scientific Investor

Welcome to the Scientific Investor

This blog should fill two purposes: It aims to provides access to financial research recently published in the top academic journals and should stimulate readers to dig deeper into a scientific topic (also in the form of a potential Bachelor, Master, and PHD thesis for students). 

Moreover, the objectiveis to inspire practitioners to implement these academic insights in the real business world. It goes without saying that empirical results shown in the papers are based on historical data and there is no guarantee that a proposed investment strategy will be successful in the future.

THe SCientific Investor - PArt 03/2024: Does Alternative Data Improve Financial Forecasting? The Horizon Effect

The digitalization of information has increased the volume and diversity of alternative data available to investors and analysts. So far, empirical studies show that these new data sources have value for short-term investment decisions; however, it is not clear whether they also bear value for long-term predictions.

In light of this research gap, Dessaint, Foucault, and Fresard (Journal of Finance, 2024) study the accuracy of analysts’ earning forecasts over different horizons (from 1 month to 60 months) in the period 1983 to 2017. To measure the informativeness of an analyst’s forecasts at a given time for a horizon h, they compute the R2 of a regression of realized earnings at this horizon (across the firms she covers) on her forecasts of these earnings. Hence, a higher R2 at horizon h means that the analyst’s forecasts at this horizon are more informative.

What are the main findings of the paper?

(i)  Over the whole sample period, the informativeness of analyst’s earnings forecasts is decreasing in the forecast horizon. As an example, the informativeness R2 is over 60% for one-year forecasts, whether it is below 40% for four-year forecasts. (Probably, this is expected by the reader.)

(ii)  The accuracy of SHORT-term forecasts (i.e., forecasts of earnings over more than two years) has INCREASED by 10% over the period 2000 to 2017 compared to 1983 to 1999.

(iii)  Surprisingly, however, the accuracy of LONG-term forecasts (earning of earnings for less than two years) has DECREASED by 10% in the second period compared to the first period.

How can one explain these opposite results in (ii) and (iii)?

The authors provide evidence that the origin is the more frequent use of alternative data from analysts in the second period. Most alternative data sources (such as web traffic, credit card, or social media sentiment data) are short-term oriented and have become easily available so that the analyst optimally substitutes effort away from collecting long-term information.
In line with his notion, the paper finds that the change in informativeness on the short- and the long side is (i) more pronounced in industries covered by analysts applying more alternative data, and (ii) analysts who are more exposed to alternative data generated on StockTwits.

Do you want to learn more? Inspect the paper on SSRN and the published version forthcoming in the Journal of Finance

THe SCientific Investor - PArt 02/2024: From "Man vs. MAchine" to "Man + Machine"

 The story that IBM’s Deep Blue beat the then reigning grand master in 1997, Garry Kasparov, is well known. What is far less known is that humans, despite having lost interest in “men vs. machines” chess contests, engage in “men + machines” matches, i.e., chess games in which a player equipped with AI assistance (a “centaur” player) competes against AI. Up to today, the centaur has kept an upper hand against machines; and even more encouragingly, there have been more and better human chess players with the advent of affordable Ai-based chess. 

If AI can help humans to become better chess players, it is likely that AI can help to improve performance in skilled jobs, such as medical doctors and investment advisors. In their Journal of Financial Economics (2024) research paper, Sean Cao, Wei Jiang, Jumbo Wang, and Baozhong Yang empirically compare the accuracy of 12-month stock price forecasts between three groups: (i) Real-life professional analysts (“Man”), (ii) A self-trained AI analyst with access to past firm prices and fundamentals, macroeconomic + industry information, and SEC reports (“Machine”), and (iii) The AI analyst in (ii) with additional information about past analyst reports (“Man + Machine”). Their sample period is from 2001 to 2018. 

Their empirical analysis reveals the following results: 

(a)  The self-trained AI analyst (“Machine”) is able to beat human analysts (“Man”): The AI analyst outperforms in 54.4% of the stock price predictions made by all IBES analysts. 

(b)  The machine’s advantage arises from its superior ability to process information and immunity from predictable behavioral biases. 

(c)   The AI analyst with analyst information (“Man + Machine”) outperforms 54.8% of the forecasts of the AI-only model. 

(d)  The improvement based on analyst information comes from information from illiquid firms, firms with more intangible assets, and firms with higher distress risk. 

(e)  The “Man + Machine” model avoids about 90% of extreme errors made by human analysts and 40% of those by the pure AI model. 

Do you want to learn more? Read the feely available on SSRN here. The published version is forthcoming in the Journal of Financial Economics. 

THe SCientific Investor - PArt 01/2024: PEAD.txt

The post-earnings-announcement drift (PEAD) is one of the most studied asset pricing anomalies on financial markets. It corresponds to the tendency of a stock that – after the company having announced abnormal positive (negative) earnings – to drift in the direction of the surprise over the next weeks. Hence, a stock with a positive earnings surprise is not only outperforming on the announcement day, but also over the upcoming weeks. 

Why does the PEAD exist on financial markets? The majority of papers attribute the effect to underreaction to news of market participants. The magnitude of PEAD is also related to the proportion of institutional investors holding the stock, arbitrage risk, and revenue surprises. In recent years, the effect of the PEAD anomaly has weakened up to a point where it is not statistically significant from zero any more.

In their Journal of Financial and Quantitative Analysis (2023) research paper, Vitaly Meursault, Pierre Jinghong Liang, Bryan R. Routledge, and Madeline Marco Scanlon study a new form of PEAD, which they term PEAD.txt. They first propose a new numerical earnings surprise measure based on the text of earning calls without explicitly incorporating the earnings number. This measure, labelled SUE.txt in their manuscript, is computed from a prediction model based on a regularized logistic text regression that extracts “good news” and “bad news” from earnings call text.

The main result is that the text-based post-earnings-announcement drift (PEAD.txt) is larger in magnitude than the classical PEAD based on pure earnings surprises. On average, the difference is growing each quarter following the release of the earnings call text with magnitudes of 2.87%-1.54%, 4.61%-2.7%, 6.51%-3.87%, and 8.01%-4.63%. 

Using the model’s predictions, the authors construct a profitable trading strategy that goes long in firms whose calls contained the “best news” and shorts the firms with the “worst news”. The generated alpha is statistically significant when adjusted for standard risk factor models.

Are you keen to get more information about this research paper? It is available on SSRN and the Journal of Financial and Quantitative Analysis.

THe SCientific Investor - PArt 07/2023: Do Social NEtworks Facilitate Informed Option trading?

An important trading style of hedge funds is concerned with M&A transactions where they aim to capture the “merger arbitrage spread”, that is, one buys the target and sells the acquirer AFTER the transaction deal is announced. But would it not be more profitable to buy the target firm even BEFORE the deal is announced?

In their Journal of Financial & Quantitative Analysis (2022) article, Harvey Cheong, Joon Ho Kim, Florian Münkel, and Harold D. Spilker III investigate how private information transmits through social networks and can lead to profitable investments for hedge funds. The authors manually collect university background data from firm directors and hedge fund managers and check whether they have attended the same university and are in the same alumni reunion group (for example, alumni celebrating their 5th, 10th, 15th, etc. reunions are often invited to the same reunion event every 5 years.). 

Their empirical analysis shows that connections in the social network strongly affect hedge funds’ trading decisions:

(a)    Hedge fund managers connected to directors of firms engaged in merger deals increase call option holdings on target firms by more than 100% before deal announcements. This holds for directors working at the acquirer or the target firm.

(b)    Hedge fund managers connected to directors of firms engaged in merger deals decrease put option holdings on target firms.

(c)     These effects are not explained by confounding factors such as geographic proximity, fund managers’ differing ability, and latent factors with the quality of shared alma maters.

(d)    Certain types of directors are more likely to provide information: Independent directors, directors with small ownership stakes in the firms, and directors with short tenure. These findings are consistent with the notion that directors who do not expect a significant increase in compensation from a successful merger or are constrained in their own insider trading may be prone to leak information to outside investors.

(e)    There is no consistent effect of alumni reunion connections on informed trading in the targets’ stock holdings as well as the acquirors’ option holdings.

Do you want to learn more? Read the feely available on SSRN. The published version is published in the Journal of Financial & Quantitative Analysis.

THe SCientific Investor - PArt 06/2023: What MAtters To individual Investors?

The academic finance literature offers no shortage of theories about individual investors’ motivations and beliefs which translate into aggregate market outcomes. However, testing these theories with observational data has been difficult so far.

In their Journal of Finance (2020) article, James J. Choi and Adriana Z. Robertson survey a representative sample of more than 1000 U.S. individuals to examine how well leading academic investment theories correspond to their financial beliefs and decisions. They put emphasis which factors (i) determine the portfolio share in equity, (ii) explain the equity risk premium, (iii) direct investors to purchase actively managed mutual funds, and (iv) influence beliefs on risk/return considerations in the cross-section of stock returns. The survey was completed in December 2016.

The survey shows the following main results:

(a) Portfolio share in equity:
- The amount of an individual’s wealth, years left until retirement, and a person’s background risk are the most important determinants for the portfolio share invested in equity.
- Individuals acknowledge that equity historically delivered higher long-term returns, but also elevated risk. When being confronted that their retirement will be delayed by 10 years, respondents are ten times more likely to increase allocations in equity than to decrease it.

(b) Equity risk premium:
- The main factor of equity risk for investors is related to rare disasters of the economy, where GDP shrinks by a large amount (comparable to a decrease which happened in the Great Depressions).
- The cornerstone of neoclassical asset pricing models, i.e., the covariance between an asset’s return and the marginal utility of money matters for 35% of all respondents. Hence, this concept is also practically important (but not as important as rare disasters).

(c) Actively managed mutual funds:
- Individuals invest into actively managed funds because they think that these funds earn higher returns than passive funds and because their advisor tells you to do so.
- 46% of respondents agree that high past returns of an actively managed fund are a signal of good stock picking kills (which is not necessarily true!).

(d) Cross-section of stock returns:
- There is no clear tendency whether growth or value stocks are expected to show higher returns in the future. However, most respondents perceive growth stocks to be riskier than value stocks.
- Respondents expect high momentum stocks to outperform low momentum stocks in the future; this comes with a slightly perceived higher risk of momentum stocks.

There are much more interesting results embedded in the paper. It is feely available on SSRN. The published version is found in the Journal of Finance

THe SCientific Investor - PArt 05/2023: The Rate of Return On Everything

What is the aggregate real rate of return in the economy? Is it higher than the growth rate of the economy and, if so, by how much? And which particular assets have the highest long-run returns?

Questions like this are important for investors and policymakers around the globe and require the use of a precise and long-term dataset of financial returns. In their Quarterly Journal of Economics (2019) article, Oscar Jordà, Katharina Knoll, Dmitry Kuvshinov, Moritz Schularick, and Alan M. Taylor introduce such a database on the annual total rates of return for the major asset classes -- equities, housing, bonds, and treasury bills -- across 16 developed countries (among others, the USA, UK, Germany, and Switzerland) starting from 1870. They take care to construct clean long-term return series for housing which was – up to now – frequently neglected.

When analyzing this data, they draw the following main results:

(a) Risky assets: Over the past 150 years, residential real estate and equities have shown average real returns of 7% per annum. In terms of risk-adjusted returns, housing outperformed the stock market because the asset class showed lower volatility.  This holds for the pre-1950 and post-1950 time period.

(b)  Where is the return from real estate generated? The main part of the real housing return comes from the yield component (rents) and not from capital gains (price changes in houses).

(c) The highest real housing returns over the past 150 years occurred in Finland, Sweden, and Denmark. The lowest real housing returns were earned in Italy, Spain, and the UK.

(d) The highest real equity returns over the past 150 years have occurred in Finland, the US, and Sweden. The lowest real equity returns were earned in France, Portugal, and Norway.

(e)  Equity and housing returns are moderately positively correlated (average correlation of 0.3). Post-1950 correlations are lower than pre-1950 correlations.

(f)  Housing showed to be a more robust hedge against inflation than equity.

(g) Safe assets: Bonds and treasury bills have earned an average real return of 1%-3% per annum over the past 150 years. Volatility of these asset classes was surprisingly high and frequently exceeded the volatility of real estate.

(h) Returns minus market growth rate: The financial return in the economy was much higher than the economic growth rate. The gap amounts to average 3-5% per annum which worsens wealth inequality.

Learn more about the historical returns of nations in the free available SSRN working paper or in the published version in the Quarterly Journal of Economics. The dataset is available here.

THe SCientific Investor - PArt 04/2023: MEASURING INVESTOR Sentiment by COmbining MAchine Learning and News

A picture is worth a thousand words. Does this also hold for measuring investor sentiment?

Numerous studies in the empirical finance literature investigate the impact of investor sentiment on future stock returns. Up to now, most papers rely on a text-based approach to capture sentiment which misses an important part of humans’ recognition: VISUAL information transmitted by pictures or photos.

In their Journal of Financial Economics (2022) article, Khaled Obaid and Kuntara Pukthuanthong are the first to develop a daily market-level sentiment index based on photos from the Wall Street Journal. To do so, they apply a convolutional neural network (CNN) for classifying photos in the newspaper from 2008 to 2020 and examine how the aggregate sentiment of daily photos are related to future market returns. Their main variable, photo pessimism (PhotoPes) is computed as the proportion of news photos predicted to be negative on a given day.

Their results are as follows:

(a) PhotoPes negatively predicts the next day’s market return. In terms of magnitude, the average impact of a one standard deviation increase in PhotoPes leads to a market decline of 4.2 basis points.

(b) In contrast to the initial effect, PhotoPes is positively related to market returns over the remaining trading week (hence, there is a return reversal). A trading strategy that takes advantage of both effects earns a statistically significant 5.45% alpha per annum.

(c) PhotoPes and pessimism embedded in text are positively correlated and interact. Specifically, pessimism in photos and text seem to substitute for one another; photos capture attention from text during periods when photos are salient.

(d) The negative effect of PhotoPes is particularly pronounced (i) during periods with high levels of market fear and (ii) for stocks that are difficult to arbitrage.

(e) High levels of PhotoPes are not only related to market returns, but also predict increases in the next day’s abnormal trading volume.

Learn more about this exciting topic in the freely available working paper or the published version in the Journal of Financial Economics

THe SCientific Investor - PArt 03/2023: Inflation and Individual Investors' Behavior

Inflation is among the most important economic risks faced by individual investors. After the outbreak of the COVID 19 pandemic, inflation resurfaced in many countries with staggering rates up to 10% p.a. in the UK, the USA, and Germany. During such turbulent times it is important to understand how individual investors respond and how they change their portfolio accordingly. 

From a normative point, academic literature posits that – during periods of high inflation rates – investors should increase their share in real assets, such as stocks and real estate. This is because real assets are assumed to be increasing in line with the level of goods as well as services and hence hedge the investor against general price increases. This mechanism is called the hedging hypothesis

However, it is also possible that investors are prone to money illusion. This hypothesis states that, when valuing real assets based on the Discounted Cash-Flow or the Dividend Discount models, agents adjust nominal rates (using the high inflation rate!!!) but ignore that firms’ cash-flows also should grow with inflation. Consequently, prices of stocks are assumed to decrease. 

So, how do investors behave during periods of high inflation? In their forthcoming paper in the Review of Financial Studies (2023), Fabio Braggion, Felix von Meyerinck, and Nic Schaub dig deep into the archives and use actual security transaction data from a large German bank as well as local inflation rates from the German hyperinflation between 1920 to 1924 to answer this question. 

Their main finding is that individual investors behave according to the money illusion hypothesis. They BUY LESS and SELL MORE stocks when local inflation is strongly increasing. In numbers, one shows that a 1% increase in local inflation is associated with a significant decline in the buy-sell imbalance for stocks of 3.5% .

Furthermore, the paper shows that: 

(a) Money illusion is related to investor sophistication. The negative relationship between local inflation and buy-sell imbalances is attenuated for sophisticated individual investors, while it even turns positive for institutional investors. 

(b) The buy-sell imbalance is particularly negative for stocks with low dividend yield and “high-tech” stocks. 

(c) The decision to buy less and sell more stocks in inflation periods leads to severe economic losses for individual investors. Stocks sold by investors tend to be undervalued rather than overvalued. 

Do you want to learn more about this exciting topic? The working paper is available for download at SSRN

THe SCientific Investor - PArt 02/2023: Stocks for the long run? Evidence from a broad sample of developed markets

“A diversified stock portfolio is a safe investment vehicle for the long-run, it will always provide you with a positive real return over holding periods of more than 10 years.” Maybe you have heard this statement from your investment advisor. But, is this claim really true?

According to a new research paper published in the Journal of Financial Economics (2022), it is not! To the contrary, the authors, Aizhan Anarkulova, Scott Cederburg, and Michael S. O’Doherty observe that long-horizon returns of equity markets show a much higher downside risk than previously thought. To draw conclusions about medium- and long-term stock return distributions, they sample historical equity index returns from 39 developed countries over the period from 1841 to 2019 provided by the database Global Financial Data (GFD). This leads to a final stock return sample spanning almost 32,000 months or 2,700 years.

When simulating return distributions from the historical worldwide sample, the authors document the following main results:

(a) An investment of USD 100 in the stock market is expected to provide a final (inflation-adjusted) wealth level of USD 738 after a 30-year horizon (good news!). This amount is in line to previous studies that mainly rely on past US equity returns: Here, the final wealth level has been expected to be USD 891.

(b) The risk of stock market underperformance relative to inflation is MUCH larger than previously suspected. Based on the worldwide historical sample, this likelihood is estimated to be 12 out of 100 (bad news!). Compared to the US sample, this likelihood of losing money in real terms by investing in the stock market is amplified by a factor of 10!

(c) In line with previous results on the topic, the paper finds that the risk of equities underperforming against inflation gets smaller, the longer the investment horizon. It ranges from 43.2% for a one-month horizon to 36.8% (28.3%, 21.5%, 15.5%, and 12.1%) for a one-year (five-years, ten-years, twenty-years, and thirty-years) horizon.

(d) The likelihood of equities underperforming against inflation is dependent on the sample stating date. It ranges from only 3.83% (for the sample that starts in 1983) to 18.0% (for the sample that starts in 2000).

(e) Investors using the historical worldwide sample for calibration should optimally invest less in stocks (and more in cash) compared to investors who sample solely from US returns.

In summary, it seems that stock investments are riskier (on the medium- and long-term) than previously thought. Are you interested to learn more? The research paper is available on SSRN and the Journal of Financial Economics.

THe SCientific Investor - PArt 01/2023: Thinking about Prices versus Thinking about Returns in Financial Markets

Prices and returns are alternative ways to present information in financial markets. Based on rational decision theory, expectations about prices and returns of the same security should contain the same information. But is this really the case? In other words, is it the same to ask an analyst to forecast prices or returns? Or do investors understand a chart of a fund’s performance in the same way when performance is displayed based on return bar charts or price line charts?

In their Journal of Finance (2019) research paper, Markus Glaser, Zwetelina Iliewa, and Martin Weber examine the question how individuals receive the terms security PRICES and RETURNS in three experimental forecasting tasks.

The main results are the following:

(a) When being asked to predict security returns as opposed to prices, forecasts for returns are significantly HIGHER. The magnitude of the effect is large and ranges between 1.1 to 2.4 percentage points per month.

(b) When being shown return bar charts as opposed to price line charts for the same security results in significantly LOWER expectations. Again, the magnitude of the effect is substantial and ranges between 1.0 to 1.7 percentage points per month.

The results are stable for different degrees of financial experience across candidates (i.e., they are robust for students and finance professionals), different amounts of information, and alternative incentive schemes. A psychological foundation of these findings is provided by “intuitive number sense”, i.e., the way intuition affects the processing of numbers. When subjects are shown price charts, they extrapolate only from the most recent past; in contrast, when they are shown return bars, they take the entire available past information into account.

The paper’s results imply that the format design in providing financial markets’ information can be used as a powerful tool to alter investors’ expectations. As an example, digital asset managers that mainly communicate by means of graphical information, could manipulate clients’ willingness to buy/sell a security by providing a certain format of performance history.

Are you keen to get more information about this research paper? It is available on SSRN and the Journal of Finance

THe SCientific Investor - PArt 08/2022: Institutional Investors and Infrastructure Investing

Over the past decade, there has been a surge in the allocation of institutional investor assets to infrastructure investments. As a result, institutional investors are becoming increasingly important alongside governments in the provision of the capital that finances infrastructure projects with favorable regulations. For example, changes in the Swiss pension fund regulation adopted in 2020 separates infrastructure from other alternative assets and allows an allocation of up to 10% of total asset only to infrastructure (and imposes a joint maximum cap of 15% to all other alternative assets).

What do institutional investors desire when they invest into infrastructure? According to the Preqin Investor Outlook surveys in 2018, 2019, and 2020, the three most common reasons are a) reliable income streams, b) low correlation to other asset classes, and c) inflation hedging. But – so far – there is no large-scale investigation whether infrastructure can fulfill these requests.

In their Review of Financial Studies (2021) article, Aleksandar Andonov, Roman Kräussl, and Joshua Rauh investigate the performance of the asset class “infrastructure” using historical data from Preqin (and Burgiss) in the period from 1990 to 2020.  As investors obtain exposure to infrastructure mainly through closed-end private infrastructure funds (CEPIFs), the authors focus on these investment vehicles.

Their main results are as follows. CEPIFs:

a) exit from profitable stakes quickly (similar as private equity funds) instead of earning long-term reliable income streams of their investments,

b) deliver procyclical cash-flows which limit diversification benefits,

c) do not show a hedge against inflation (note: the sample period is from 1990 to 2020, so inflation levels were moderate), and

d) underperform the equity market, as well as private equity and real estate funds based on a risk-adjusted basis.

Despite both the weak performance and the failure to match the supposed characteristics of infrastructure investments, CEPIFs have increased their assets under management from USD 59 billion in 2008 to USD 486 in 2019. How is this possible?

The authors find evidence that the increase of assets is particularly driven by public investors (public pension funds, sovereign wealth funds, etc.) that seek exposure to suitable ESG or impact investment projects. Indeed, public investors that have signed the United National Principles for Responsible Investment (UN PRI) make 0.759 more infrastructure investments per year as compared to investors that have not signed the UN PRI.

Do you want to get more information on the relation between institutional investors and infrastructure investing? Reed the freely available working paper or the published paper in the Review of Financial Studies.

THe SCientific Investor - PArt 07/2022: visual finance: The Pervasive Effect of Red on Investor Behavior

The financial industry relies heavily on the visual communication of information. For example, Bloomberg displays historical stock price data in blue, while Yahoo! Finance exhibits the same information in red (for losses) and green (for gains). Moreover, historical graph charts are commonly drawn in blue color (e.g., in the Wall Street Journal).

Standard finance theory suggests that the color in which data is visually represented should not matter for investment decisions. But is this really the case? Visual perception is a complex task which involves biological and neural mechanisms: Hence, it is likely to subconsciously affect the way how individuals perceive financial information.

In their Management Science (2021) article, William J. Bazley, Henrik Cronqvist, and Milica Mormann investigate the effect of the color, in which financial data is displayed, on (i) risk preferences, (ii) expectations of future returns, and (iii) trading decisions of investors. To do so, they run a series of experiments and trading games with a total of 1,451 participants.

Their main results are as follows:

a) In a standard lottery task, individuals reduce risky behavior when potential financial losses are shown in RED COLOR. Specifically, individuals take approximately 20% LESS RISK in financial gambles when losses are shown in red (compared to other colors).

b) When individuals view negative historical stock price data in red, their expectations about future returns are significantly reduced (compared to data displayed in black color). Investors expect future declines of negative red price information, while they expect negative black price information to reverse.

c) Overall, displaying past negative stock data in red color leads to a decreasing willingness to purchase equity by almost 15%, all else equal.
It is interesting to note that the impact of red color on financial decisions is not observed for experiment participants from China. This can be attributed to the notion that “red” symbolizes “prosperity” in Chinese culture and is generally not used to represent losses. Moreover, the impact of color on financial decisions is both observed for participants with low and high financial literacy and sophistication.
Do you want to learn more on the impact of colors on financial decisions? Have a look at the freely available working paper on SSRN or the published paper in the journal Management Science

THe SCientific Investor - PArt 06/2022: (Re-)Imag(in)ing Price Trends

“Nevertheless, technical analysis has survived through the years, perhaps because its visual mode of analysis is more conducive to human cognition, and because pattern recognition is one of the few repetitive activities for which computers do not have an absolute advantage.” (Lo, Mamaysky, and Wang, 2000)
Not any more! In their forthcoming Journal of Finance (2022) article, Jingwen Jiang, Bryan Kelly, and Dacheng Xin investigate the relationship between a stock’s past price patterns and future returns. In contrast to other studies on past price information (such as momentum or short-term reversal), the paper is the first to use IMAGES from stock-level price charts as the predictor variable.
To process information on stock level price charts and relate them to future returns, the authors apply so-called CONVOLUTIONAL NEURAL NETWORKS (CNNs). The input to a CNN is the image of a stock’s past price information represented as a matrix of black and white pixel values. Being trained on historical future stock returns, the input is then transformed and smoothed on different layers to produce a binomial forecast on the future development of the stock price (i.e., it predicts whether the stock will rise (outcome 1) or fall  (outcome 2) in the next period).
The empirical analysis revolves around a panel prediction model for US stock returns from 1993 to 2019 and yields the following results:
a) Image-based CNN predictors are powerful and robust predictors of future returns. A weekly portfolio that is long the decile with the highest predicted CNN-returns and is short the decile with the lowest predicted CNN-returns yields an annualized gross Sharpe ratio of 7.2 (1.7) on an equal-weighted (value-weighted) basis. The strategy remains profitable when taking account longer holding frequencies and transaction costs.

b) The CNN predictions are not well described by predictions from traditional predictors, such as momentum, dollar volume, short-term reversal, or illiquidity. These traditional predictors only explain about 10% of the cross-sectional variation in CNN forecasts. Hence, CNN captures new and unexplored information!

c)  The return-predictive patterns by the CNN extrapolate to contexts outside of the main data set of daily US stock prices. CNN model estimates from US data also lead to significant investment profits in 26 foreign markets around the world.

d) Successful CNN-patterns at high (i.e., daily) frequency are similar to successful CNN-patterns at low (i.e., monthly or quarterly) frequency. A CNN model to predict 5-day returns from images of 5-day prior price data sampled every 12 days outperforms a CNN trained directly on quarterly data.
The study shares a new perspective on the investing style of technical trading. It is the first to apply VISUAL data sources for investment recommendations. As the majority of asset institutional asset managers use some form of technical analysis in the decision process, the practical implications of the study seem to be of great value to the profession.

The research paper can be read for free on SSRN. The paper is also forthcoming at the Journal of Finance. Have a look and learn more about this exciting form of technical trading!

THe SCientific Investor - PArt 05/2022: do individual investors trade on investment-related internet postings?

Many people share investment ideas online. More recently, individuals can also observe detailed trading strategies of “experts”, FOLLOW these strategies, and interact with the experts. This type of fintech development is called SOCIAL TRADING and aims to provide (especially, inexperienced) individuals the possibility to copy trading strategies from professional investors. The idea sounds promising, but…

In their Management Science (2021) article, Manuel Ammann and Nic Schaub investigate the benefits/problems of social trading from an investor’s perspective. Specifically, they concentrate how experts communicate with followers by asking the following questions: (i) Do individual investors rely on internet posting of experts when making investment decisions? (ii) Do postings help individual investors identify investment strategies that deliver superior performance in the future? To answer the questions, the authors obtain data from a leading European social trading platform in the time period from 2013 to 2014.

The paper yields the following main results:

(a) The average trading performance of “experts” on social trading platforms is weak. Social trading platform experts investing their money in equity significantly underperform corresponding indices, such as the MSCI Europe and the MSCI World.

(b) Communicating with followers is beneficial for the expert: The posting of a comment leads to an increase of 6% compared to the average daily net investments from followers. Comments with a positive tone lead to more net inflows. Moreover, most inflows subsequent to postings come from unsophisticated investors.

(c) However, following experts that post a lot is not beneficial for the investors: Neither the number nor the tone of postings is related to subsequent investment performance.

The study raises considerable doubt of whether it is useful to invest money via social trading platforms. To the contrary, the picture emerges that experts lure unsophisticated investors to invest their money into their trading strategies by posting comments on specific trades. However, the performance of these trades is subpar, and investors would be better off by flocking their money into passively-managed indices.

The published version can be found in the journal Management Science. Have a look and learn more about social trading platforms!

THe SCientific Investor - PArt 04/2022: A NEW TEST OF RISK FACTOR RELEVANCE

Why do some assets have higher expected returns than others? The usual textbook explanation in asset pricing is that investors like assets that insure them against bad states of the world, frequently approximated by different risk factors (i.e., by consumption growth, market volatility, or market illiquidity). Consequently, an asset’s correlation to these risk factors should determine its expected price and return. As an example, an asset with a high sensitivity to consumption growth has higher expected returns compared to an asset with a low sensitivity, because the latter hedges investors in times of low consumption growth (i.e., during bad states of the world).

In their forthcoming Journal of Finance (2022) article, Alex Chinco, Samuel Hartzmark, and Abigail Sussman are tackling this fundamental perspective in asset pricing. By doing so, they develop a survey to examine whether different types of investors (professional investors, retail investors, and students) care about an asset’s correlation to a very prominent risk factor, that is, aggregate consumption growth. The survey asks the participant how he/she would allocate an endowment between a portfolio of stocks and a riskless bond based on data describing stock returns and consumption growth. The consumption growth data is the same for each participant, but different stock data is simulated with randomly chosen mean, volatility, and consumption growth parameters.

The survey yields the following two main results:

(a) Participants DO respond to changes in the mean and volatility of returns as predicted by theory. They invest more in stocks when average stock returns are higher and invest less when stock returns are more volatile.

(b) Participants DO NOT respond to changes in consumption-growth correlations. To the contrary, most participants INCREASE their demand for stocks when stock returns are MORE correlated to consumption growth. This is in stark contrast to asset pricing theory.

A key contribution of this paper is to highlight survey evidence for checking fundamental assumptions in asset pricing theory. In principle, the framework in the paper can be also applied to test the relevance of alternative risk factors in the literature (such as the SMB-size and HML-book-to-market factors of Fama and French, as well as the momentum factor).

Have a look at this interesting research paper. It can be read for free on SSRN. Its published version can be found in the Journal of Finance.

THe SCientific Investor - PArt 03/2022: Engineering Lemons

Financial intermediaries engineer new products that investors demand, and as a result, this will improve social welfare. Oh, wait a second… is this really the case?

In a study recently published in the Journal of Financial Economics (2021), Petra Vokata analyzes a large sample of more than 28,000 structured products – so called yield enhancement products (YEPs) – over the period from 2006 to 2015 on the US market. YEPs are typically characterized by two features: First, they offer an attractive coupon rate (e.g., 8% p.a.) to the investor during the maturity of the products (which is usually up to one year). Second, to receive the yield, the investor must be willing to bear downside risk in case of a declining price of the underlying. As an example, the investor participates on downside movements of the asset when it falls below 80% of the initial price when purchasing the YEP. 

Do YEPs deliver attractive returns to investors over the inspected sample period? No, they don’t. Issuers of YEPs charge disproportionally high implicit fees to these products amounting to more than 6% p.a. These fees are large enough, so that the ex-ante and ex-post returns of YEPs are, on average NEGATIVE! Hence, any risk-averse investor would be better off to invest in the risk-free asset than investing in these products. 

Additional findings of the study are the following:

(a) Embedded fees are the highest for short-term YEPs (with a maturity of 2 – 4 months, average annualized fee of 11%) and more modest for longer-term YEPs (with a maturity of > 8 months, average annualized fee of 3.97%).

(b) YEPs are also not suited for hedging purposes. More than 40% of the structured products are state-wise dominated by simple combinations of listed options available to investors on exchanges (when taking account of transaction costs).

(c) In line with increasing price pressure, embedded fees for YEPs have been decreasing over the more recent years 2010 – 2015. Nevertheless, fees in these years were large enough for their expected returns to be negative.

Why are investors buying these products? Evidence from previous research on the topic finds that banks are smartly targeting these products to retail and unsophisticated investors who do not understand the construction of structured products. In this manner:

(a) The high coupons of the products are saliently advertised in the name of the product, whereas to quantify the possible losses, investors need to use option-pricing techniques.

(b) The possible loss of the initial of the principal at maturity is not emphasized as a capital loss but reframed as the delivery of the underlying.

(c) The underlying stocks of the YETs are not chosen at random, but overrepresent highly volatile stocks with high downside losses (which give rise to a seemingly high enhanced coupon).

Read this working paper for free on SSRN and the published article in the Journal of Financial Economics.

THe SCientific Investor - PArt 02/2022: It's not so bad: Director Bankruptcy experience and Risk-Taking

Existing evidence suggests that individual traits, especially risk attributes, are shaped by past salient experiences. Without doubt, the experience of a bankruptcy is such an event which has the potential to affect directors’ attitudes toward risk, and hence, the advice they provide on a range of corporate decisions.

In their Journal of Financial Economics (2021)  research paper, Radhakrishnan Gopalan, Todd A. Gormley, and Ankit Kalda retrieve data on directors that were engaged in multiple US board seats from 1994-2013 and study the research question whether bankruptcy experience at one firm affects the director’s advice for risk-taking at another firm.  The main finding is surprising: Firms take MORE risk (i.e., increasing net leverage, equity issuances) when one of their directors has EXPERIENCED a corporate BANKRUPTCY at another firm when they concurrently serve as a director. This also leads to higher cash flow and return volatility of the respective firm in the future.

The increase in risk-taking, while surprising at the first blush, is reasonable if the bankruptcy experience lowers a director’s assessment of distress costs (“it was not so bad…”). In line with this notion, the authors document that higher risk-taking is concentrated among firms where a director experienced a bankruptcy with lower distress costs. Hence, directors update their views regarding distress costs downward following less severe bankruptcy experiences. Moreover, additional results show that the careers of directors are not negatively affected by less expensive bankruptcies (“again, it is not so bad…”).  The findings of the paper are emphasized among firms where the primary role of the interlocked director is to provide advice and when the director is expected to have a greater influence over the board. There is no evidence that there is a shift in monitoring of the director when having experienced a bankruptcy.

Read the full working paper on SSRN and the published article in the Journal of Financial Economics. 

THe SCientific Investor - PArt 01/2022: Financial Statements and Lazy Prices

Is it possible to predict long-term future returns of a company based on soft information from its financial statements? 

The semi-strong form of market efficiency implies that new information is immediately priced into stocks, so it is expected that the publication of new company information only has a very short-term impact. However, the length and complexity of financial reports have strongly increased over time, so that investors face difficulties to immediately incorporate all relevant information into stock prices. 

In their Journal of Finance (2020) research paper, Lauren Cohen, Christopher Malloy, and Quoc Nguyen show that prices are indeed “lazy”, and that textual information of financial statements only gradually flocks into asset prices. In their empirical analysis, the authors measure “changes” in the language and construction of firms’ 10-K reports using different similarity metrics. They then investigate the impact on future returns and real effects. 

The main findings of the paper are: 

(a)   Firms that change language and construction of the most recent financial report (compared to older reports) yield lower future returns than “non-changing” firms. A trading strategy that sells “changers” and buys “non-changers” generates monthly returns of more than 7% per annum. 

(b)   Most reporting changes are concentrated in the “Management Discussion and Analysis” (MD&A) section of the 10-K report. The most informative section in terms of stock price predictability is found, however, in the “Risk Factor” section: A trading strategy based on this section as described in (a) yields returns of more than 20% per annum. 

(c)   Why do changes in language and construction of a financial report predict negative returns? There is strong empirical evidence (from natural language processing, NLP) that financial statement changes go together with “negative sentiment”. More precisely, 86% of changes consist of passages that contain a majority of negative-classified words by an NLP algorithm.

(d)   The predictability of soft information is not limited to future stock returns. The paper also shows that statement changes forecast future earnings, profitability, and bankruptcies of firms. 

Are you keen to get more information about this exciting research paper? It is available on SSRN and in the Journal of Finance

THe SCientific Investor - PArt 12/2021: Do People feel less at risk? Evidence from disaster experience

Do individuals perceive more or less risk after having experienced catastrophic disasters? Academic research generally shows that financial disasters (such as stock market crashes) and terrorism attacks (such as the 9/11 assault) makes agents more risk averse. However, no study has looked at households’ risk perception to catastrophic events in comparison to a sensible reference point which households could form. 


In their Journal of Financial Economics (2020) research paper, Ming Gao, Yu-Jane Liu, and Yushui Shi change this shortfall in the literature and examine the risk behavior of Japanese households after the occurrence of local earthquakes in the time period from 2007 to 2012. 


To assess the perceived riskiness of a household in a certain year, they use the households’ life insurance-to-portfolio ratio from a National Survey, which is computed as the amount of life insurance relative to total financial wealth. Households with a higher (lower) life insurance-to-portfolio ratio are supposed to be more (less) risk averse to future catastrophic events. The reference point for expected disaster losses is based on historical fatalities of earthquakes from 1930 to 2006 in the households’ geographical region as well as seismological quantities.

The paper’s results can be summarized as follows: 

(a)   In line with previous literature on the topic, the authors find, on average, that the perceived risk of individuals increases after fatal earthquakes and households increase life insurance in their portfolios by 6.74%. 

(b)   REFERENCE POINTS  matter. Individuals disproportionally increase life insurance when the experienced earthquake is more fatal than expected. HOWEVER: Individuals also DECREASE life insurance when the experienced earthquake is less fatal than expected. In this case, a one standard deviation increase in the negative experience shock is associated with a 1.71% decrease in the households’ life insurance-to-portfolio ratio. Hence, individuals perceive less risk when the experienced earthquake is less fatal than expected. 

(c)   Households adjust their perceived risk on prior experiences: As an example, if a household experiences a high-fatality earthquake first and non-fatal earthquakes later, the effect of reducing life insurance from the portfolio is intensified.

Are you interested to learn more about risk perceptions of households after disasters? Study the working paper on SSRN  or the published paper in the Journal of Financial Economics.

THe SCientific Investor - PArt 11/2021: Divergence and Rewriting of ESG Ratings

Research on environmental, social, and corporate governance (ESG) topics has exploded over the past years, as more and more asset managers have committed to integrate ESG considerations into the investment process. In consequence, from 2010 to 2020, assets managed by ESG investment strategies of US institutions grew from USD 3 trillion to more than USD 17 trillion.

I summarize two important caveats that investors should be aware of when dealing with third-party ESG ratings: (a) ESG ratings from different providers disagree substantially and (ii) historical ESG ratings are ex-post adjusted and rewritten in a systematic way. 

In their working paper “Aggregate Confusion: The Divergence of ESG Ratings”, Florian Berg, Julian Koelbel, and Roberto Rigobon investigate the divergence of ESG across six prominent rating agencies: KLD (MSCI Stats), Sustainalytics, Vigeo Eiris (Moody’s), RobeccoSAM (S&P Global), Asset4 (Refinitiv), and MSCI. They observe that ratings disagree substantially: Overall rating correlations between providers are on average 0.54 and range from 0.38 to 0.71. Interestingly, the highest divergence in rankings is found in the governance dimension, where some correlations are shown to be even negative. When digging deeper, the authors find that scope and measurement are the main drivers of the difference in ESG ranking, i.e., third-party providers assign ratings based on different attributes and different measurement criteria. While scope divergence must be expected as long as different views on ESG exist, measurement divergence can be seen as problematic. ESG ratings should ultimately be based on objective observations that can be ascertained. 

In their working paper “Is History Repeating Itself? The (Un)predictable Past of ESG Ratings”, Florian Berg, Kornelia Fabisik, and Zacharias Sautner make an interesting observation and document widespread and repeated changes to the historical ESG ratings of Asset4 (Refinitiv). The rewriting of the ESG ratings significantly alters the classification of firms into different ESG rankings. Particularly troublesome, the authors find the rewritten ESG scores to have a significantly positive relation to future returns ex-post, whereas the original ESG scores show no significant association. This observation has far-reaching consequences for investors: When back-testing trading strategies, researchers would find a positive association between ESG ratings and future returns of stocks, which is not existent from an ex-ante point of view.

The underlying working papers of my summary are available on SSRN (Paper 1, Paper 2).  Have a read on those!

THe SCientific Investor - PArt 10/2021: Risks and Returns of Cryptocurrencies

Cryptocurrencies are virtual currencies that are secured by cryptography, usually run on blockchain technology and not issued by a central authority.  While the primary purpose of cryptocurrencies is to enable payment exchange in a secure form, both private and institutional investors flock money into different crypto coins for speculation and diversification purposes. As a consequence, the market for cryptos has been rapidly developing and reached a market capitalization of over two trillion USD (with Bitcoin, Ethereum, and Binance being the most important coins based on market cap in August 2021). 


Over the past ten years, investments into (most) cryptos can be characterized by high average returns and high volatility, Consequently, it would be of high value to investors (and market observers) to uncover the drivers of cryptocurrency prices and to investigate whether crypto returns are predictable. 

In their Review of Financial Studies (2021) research paper, Yukun Liu and Aleh Tsyvinski thoroughly examine the risks and returns of cryptocurrencies based on trading data from Coinmarketcap.cop in the period from 2011 to 2018. For this purpose, they construct a coin market return defined as the value-weighted return of all underlying major coins and study its drivers. 

Their main results are as follows:

(a)   Factors capturing crypto demand (i.e., the number of wallet users, active addresses, transaction count, payment count) are important for the price behavior of cryptocurrencies. 19% of the variance in the coin market return are explained by these network factors.

(b)   Factors capturing crypto supply (i.e., costs of mining, electricity costs, computing costs) are not related to crypto returns. 

(c)   Cryptocurrency returns are predictable by: 

  • Past returns. High weekly returns in the past go in line with high weekly returns in the future. 
  • Investor attention: Google and Twitter search queries related to the word “Bitcoin” are related to higher future weekly returns. 
  • Negative investor attention: A ratio between google searches for the phrase “Bitcoin hack” and “Bitcoin” negatively predicts future weekly returns. 

(d)   Cryptocurrencies are not predictable by valuation ratios (e.g., user-to-market ratio, payment-to-market ratio), other traditional asset returns (such as stocks, bonds, and currencies), and macroeconomic factors. 

Are you interested to learn more about the risks and returns of cryptocurrencies? Study the working paper on SSRN or the published paper on the Review of Financial Studies.

THe SCientific Investor - PArt 09/2021: Family Descent and Mutual Fund Performance

One of the most important tasks of investment advisors is to select the most capable and hardworking fund managers for their agents. However, selecting managers from an ex-ante point view is difficult as fund performance is not persistent and only weakly correlated to other fund determinants (such as fund age, turnover, and fund fees). 


A new approach to improve manager selection is to extend the set of predicting variables to fund manager characteristics. In their Review of Financial Studies (2018) research paper, Oleg Chuprinin and Denis Sosyura examine a fund manager’s family descent as a predictor of fund performance. Their sample covers more than 600 solo-managed US mutual funds in the time period from 1975 to 2012; personal census household data of 387 managers (i.e., information in which households fund managers grew up) are merged to this sample. 

The main results of the article are as follows: 

(a)  There exists variation in fund managers’ family wealth: While the bottom quintile of managers sorted on parents’ wealth comes from families with incomes below the national average, the top quintile comes from ultra-rich families with incomes above the 99th national percentile. 

(b)  Fund managers from wealthy families underperform managers from poor families by more than 1% per year. This result controls for the quality and type of managers’ education and demographics, as well as a large set of fund characteristics. 

(c)   Managers born poor face higher entry barriers into asset management than managers born rich. As an example, the negative wealth-performance relation is stronger for managers whose college was located further away from their parents’ home. 

(d)   After having obtained a job in asset management, managers endowed with greater family resources face less stringent performance thresholds in their career progression. 

(e)  Managers from poor families seem to exert more effort in their portfolio strategies visible by higher turnover, shorter holding horizons, and contrarian trading behavior. 

(f)   Neither capital flows nor management fees are higher for funds run by managers from wealthy managers. 

The article touches upon a very important topic in business research by providing evidence that unprivileged individuals are better managers because they face higher screening standards throughout their lives.  Have a look at the working paper on SSRN or the published paper in the Review of Financial Studies.

THe SCientific Investor - PArt 08/2021: Open Source CRoss-sectional Asset Pricing

A 2016 poll of 1,500 scientists conducted in the journal Nature reported that 70% of them had failed to reproduce at least one other scientist’s experiment (including 87% of chemists, 77% of biologists, 69% of physicists and engineers, and 67% of medical researchers). 

Recently, also several voices in finance academia and practice have raised doubt on the truthfulness of empirical studies on the cross-section of average stock returns. As an example, Hou, Xue, and Zhang (2020) claim that more than 50% of characteristics – previously shown to be significantly related to average stock returns – fail the hurdle to display statistical significance at the 5% level in univariate portfolio sorts. 

In a new research project, Chen and Zimmermann (2021) tackle the “replication crisis” in finance and advocate the idea of open-source cross-sectional asset pricing. They provide open-source code (in R and Stata) as well as data to replicate the relationship between 319 characteristics and the cross-section of average stock returns in the USA (during the original sample period and there-after). 

The findings of the open-source research project are surprising and contradict the findings of previous meta-studies:

(a)  For the 161 characteristics that were clearly significant in the original papers, 158 (98%) characteristics are shown to display an absolute t-statistic above 1.96 in univariate portfolio sorts (i.e., they are significant at least at the 5% level). 

(b)  For the 44 characteristics that had mixed evidence in the original papers, the reproductions find absolute t-statistics of 2.00 on average. 

(c)  A regression of reproduced t-statistics on original long-short portfolio t-statistics finds a slope of 0.88 and an R2 of 0.82. Hence, the replicated t-statistics match the original t-statistics based on economic significance. 

(d)  Abnormal returns due to the 161 characteristics decay after the initial sample periods of the original studies, but generally remain positive. 

Hence, the paper takes a large step toward restoring trust in the literature of empirical asset pricing concluding that nearly 100% of the literature's predictability results can be reproduced. Being open-source, the research project also shows how these results are achieved.

The data and code of the research project can be found here. The working paper version with all empirical findings can be retrieved on SSRN.

THe SCientific Investor - PArt 07/2021: Doctoral Education and Investment Performance

Pursuing a doctoral degree in Finance/Economic is a tough, but also rewarding route for ambitious students which consists of years of studying, researching, and teaching. Although usually designed for a career in an academic institution, a significant part of graduating PhDs make their way into asset management firms to manage other people’s money. 
Are investment professionals with a PhD successful in doing so? And does their doctoral degree help them to outperform other investment managers (who do not have a graduate education)? 

In their Management Science (2020) research paper, Chaudhuri, Ivkovic, Pollet, and Trzcinka empirically investigate the performance of investment products managed by firms in which PhDs play a key role (e.g., CEOs or CIOs) in comparison to products managed by otherwise similar firms. Their sample covers data from 59 quarterly snapshots from investment management firms in the period from 1993 to 2007; among others, they examine detailed performance statistics from 531 individuals holding a PhD degree. 

The findings of the article can be summarized as follows:
(a) Financial performance of investment products managed by PhDs is significantly higher than performance of investment products by non-PhDs over the sample period. The difference in monthly (annualized) Carhart (1997) alphas amounts to 0.064% (0.768%).

(b) This positive performance spread gets larger if one considers the investment performance of products from firms which are founded by a PhD as the CEO or partner.

(c) Investment performance of PhDs with a Finance/Economics background is superior to performance of PhDs with a STEM (science, technology, engineering, and mathematics) background. Hence, field-specific training matters.

(d) PhDs who were successful in publishing in top scientific journals are also better investment professionals than PhDs whose scientific papers remain unpublished (i.e., unsuccessful PhDs).

(e) The performance difference between Finance/Economics professionals and STEM professionals is particularly large for unsuccessful PhDs; however, the difference is close for the top PhD students across disciplines. Thus, field-specific training is less important for the very top students across fields. 

Do you want to dig deeper into the relationship between doctoral education and investment performance? Have a look at the working paper on SSRN or the published paper in Management Science.

THe SCientific Investor - PArt 06/2021: how do venture capitalists make decisions?

Over the past 30 years, venture capital (VC) has been an important source for innovative companies. Firms that were supported by VC included Amazon, Apple, Facebook, Google, Netflix, and Starbucks – firms that had and have a tremendous impact on the US and global economy.

But how do venture capitalists (VCs) make their decisions? For their Journal of Financial Economics (2020) research paper, Gompers, Gornall, Kaplan, and Strebulaev surveyed 885 institutional venture capitalists (VCs) to learn how they form their choices among pre-investment screening, structuring investments, post-investing monitoring, and advising. The average VC firm in the survey is small, with 14 employees and five senior investment professionals.

The evaluation of the survey leads to the following important insights.

In General:

(a)    The typical VC partner works 55h per week and spends 22h per week networking and sourcing deals as well as 18h per week working with portfolio companies.
(b)    The average VC deal takes 83 days to close; the VC firm spends 118 hours on due diligence over that period and calls ten references.
(c)    15% of a VCs exits are through IPO, 53% are through M&A, and 32% are failures.

Pre-Investment Screening:
(a)   A VC firm screens approximately 200 companies and makes only four investments in a given year.
(b)    Over 31% of deals are generated through the VC’s professional networks and 28% of deals are pro-actively self-generated.
(c)     When selecting investments, VCs place the greatest importance on the management / founding team, followed by the business model and the product. Company valuation is less important for early-stage deals, but becomes more important for later-stage deals.
(d)    Only few VCs use discounted cash-flow or net present value techniques to evaluate their investments. The most commonly metric is the Multiple of Invested Capital (MoIC).
(e)    9% of all VCs and 17% of early-stage investors do not use any quantitative deal evaluation metrics. 44% of all VCs state that they also make “gut” investment decisions.
(f)     The average expected IRR of a VC investment is 31% (!!!) p.a.

Post-Investment Monitoring:
(a)    After the initial investments, VCs provide a large number of services to portfolio companies, with strategic guidance, connecting investors, connecting customers, and operational guidance as the most important ones.
(b)    When considering which of VC activities are most important for value creation, VCs rank deal selection as most important, closely before post-investment value-added and deal flow.
(c)     When asked VCs what factors contributed most to their successes and failures of investments, the management / founding team is by far the most important factor.

Are you interested to dig deeper into the topic? Have a look at the working paper on SSRN (without charge) or the published paper in the Journal of Financial Economics

THe SCientific Investor - PArt 05/2021: Innovation and Product Differentiation of Mutual Funds

Believe it or not, there were more than 9,000 different mutual funds in the USA in 2020. Based on this large number, it is of first-order importance for funds to differentiate themselves from competitors. A potential way to position a fund as a “unique” investment vehicle is to apply a distinct wording in the fund prospectus. 

As an example, the 2012 founded Patriot Fund was described as follows:

“The investment seeks growth of capital. The fund generally invests in common stocks included in the Standard and Poor's 500 Index, excluding those that fail the adviser's "patriotic" investment screen. The adviser's patriotic investment screen eliminates common stocks issued by companies doing business in nations supporting terrorism (“Terror Nations”), as defined by the U.S. State Department. Currently, there are several countries designated under these authorities: Cuba, Iran, Sudan and Syria.”

In their Journal of Finance (2020) research paper, Kostovetsky and Warner examine innovation and product differentiation using a uniqueness measure based on textual analysis of fund prospectuses. The textual uniqueness metric counts how many words a specific fund prospectus shares with other funds in the same Morningstar category. Consequently, funds that show a low word overlap are labelled as “textual unique”. As an example, the Patriot Fund is among the top 5% textual unique funds in the data sample.

What are the main findings of the study?

(a) Textual uniqueness identifies funds of certain sub-investment styles of the Morningstar category, such as, “thematic”, “socially responsible”, or “tax-aware” funds.

(b) Textual unique funds are particularly offered by small family firms: The start rate for families in the lowest size quintile is 168% higher than the start rate for families in the highest quintile.

(c) Textual unique funds are expensive: A one-standard deviation shock to text-based uniqueness is associated with a 12 basis points increase in the annual expense ratio.

(d) For new funds, textual uniqueness is related to higher money inflows: In the first 36 months, first with above-median textual uniqueness attract about 50% more in asset flows then comparable below-median textual uniqueness funds. 

(e) The relationship between textual uniqueness and future performance is slightly negative; a one-standard increase text-based uniqueness is associated with a decrease in future average returns of approximately 10 basis points.

(f) Finally, textual unique funds have a lower flow-performance sensitivity, i.e., money inflows and outflows from investors are less dependent on the fund’s performance.

Are you interested to dig deeper into the topic of product differentiation of mutual funds? Have a look at the working paper on SSRN (without charge) or the published paper in the Journal of Finance.

THe SCientific Investor - PArt 04/2021: attention to global warming and Stock Returns

On December 28, 2017, former US president Donald Trump twittered:

In the East, it could be the COLDEST New Year’s Eve on record. Perhaps, we could use a little bit of that good old Global Warming that our country, but not other countries, was going to pay TRILLIONS OF DOLLARS to protect against. Bundle up!» 

In his quote, Trump confuses a short-term local weather phenomenon with global climate change. Indeed, this is a very common mistake made from humans around the world: One extrapolates personal experience about attention-grabbing weather events to long-lasting global trends. As a consequence, abnormally warm local temperatures often serve as “wake-up calls” that remind investors to the alert of global warming.

In their Review of Financial Studies (2020) research paper, Choi, Gao, and Jiang test how people react to abnormal local temperatures by examining their attention to climate change and stock prices. This is done using a worldwide sample from 2001 to 2017 based on 74 cities with major stock exchanges (among them New York City, HongKong, and Frankfurt).

The findings of the paper can be summarized in the following way:

(a) During abnormally warm months in a particular city, investor attention to climate change significantly increases. To measure investor attention to climate change, the authors use the volume of Google searches to the term “global warming”.

(b) During these abnormally warm months, local stocks of carbon-intensive firms underperform firms with low carbon emissions by approximately -0.4%. This result is consistent to the idea that if investors direct their beliefs towards global warming, they may buy stocks from low-emission firms and sell stocks from high-emission firms. They do so because they either want to ethically avoid holding social-irresponsible stocks or are concerned about lower future cash-flows through tighter environmental regulations for high-emission firms.

Which investor group is driving this effect? In line with the conjecture that individuals are particularly prone to limited attention and own personal experience, retail investors (and not institutional investors, such as pension funds) decrease their holdings of high-emission firms in an abnormally warm month the most.  It is interesting to note that all these results only hold in a more recent time period and not in a “placebo time period” from 1983 to 2000, when global warming was less of an issue.

Have you become interested in the fascinating relationship between attention to global warming and stock returns? Read the full paper on SSRN (without charge) or the published paper in the Review of Financial Studies.

THe SCientific Investor - PArt 03/2021: Empirical Asset Pricing via Machine Learning

Machine learning (ML) has emerged as a useful device in our everyday life, enabling the recognition of images and speech, performing medical diagnosis, and improving product recommendations. Is machine learning also useful for the design of investment strategies and can we use it to predict future returns of financial securities?

What is ML? The definition of this term is frequently dependent on the field. For research purposes in asset pricing, I define ML as a “collection of high-dimensional models for statistical prediction where (i) the risk of in-sample model overfitting is mitigated and (ii) efficient algorithms search among potential model specifications.” 

In their Review of Financial Studies (2020) research paper, Gu, Kelly, and Xiu apply ML methods to predict future returns using a sample of 30,000 individual US stocks over a period of 60 years from 1957 to 2016. To predict future returns for each stock they use 94 characteristics, eight aggregate time-series variables, and 74 industry-sector dummy variables (totaling more than 900 prediction signals). 

The following ML models are applied:
a) Baseline: Simple unconstrained linear regression
b) Penalized linear regression: Punishes the inclusion of new predictor variables and reduces potential overfitting of the model (models: lasso and ridge regression)
c) Dimension reduction techniques: Average the impact of all potential predictor variables to an aggregate predictor (principal component regression and partial least squares)
d) Penalized generalized linear models that allow for nonlinearities in the predictor variables
e) Boosted regression trees and random forests: Nonparametric models that allow for interactions between the predictor variables
f) Neural networks (deep learning): Nonparametric models that use activation functions and different layers to account for nonlinearities and interactions between predictor variables 

The authors train these different models over the first 18 years in the sample, validate them in the consecutive 12 years, and use the remaining 30 years of data for an out-of-sample analysis based on monthly forecasts. 

The results of the study are in favor of ML being helpful in predicting future stock returns. In particular, neural networks are able to substantially increase the accuracy of future return forecasts. Speaking in numbers, a value-weighted long-short portfolio that takes positions based on stock-level neural network forecasts earns an annualized out-of-sample Sharpe ratio of 1.35. This is twice the performance of a leading regression-based strategy from the literature.

What are the important predictor variables that boost accuracy of return forecasts? There is strong evidence that price trends (past momentum and reversal), liquidity (stock size and volume), as well as stock volatility are driving the results. Moreover, interactions between these variables and the aggregate book-to-market ratio seems to be important for the success of the documented neural network trading strategy.

Note that the investigation refrains from the impact of trading costs which are likely to be substantial (as the portfolio formation of the investment strategy requires a high turnover).

Are you interested to dig deeper into the topic? Have a look at the working paper on SSRN or the published paper in the Review of Financial Studies.

THe SCientific Investor - PArt 02/2021: anomalies across the globe: once public, no longer existent?

Factor investing has emerged as a major trading style over the past decade. Investment companies cater increasing demand for factor strategies from institutional and individual investors through actively-managed funds and (quasi-) passive smart-beta ETFs. 

What is factor investing? I define factor investing as “the selection of securities based on certain attributes that have produced risk-adjusted outperformance over a prolonged period of time." Such attributes are also called anomalies and are either due to the compensation for systematic risk or due to mispricing. Prominent anomalies with established factor premia can be constructed, among others, based on a security’s size, value, momentum, and reversal (I will come back to some individual anomalies in later episodes of this blog). As an example, a self-financing portfolio consisting of US stocks with high momentum (i.e., past annual performance) on the long side and US stocks with low momentum on the short side delivers a risk-adjusted return of 9% in the period from 1927 to 2019. The number of observed anomalies has grown tremendously over the past decade with a current count of more than 450 in the US stock market.

In this blog post, I focus on the question whether stock market anomalies are long-lasting or whether their profitability is destroyed after being made public by an academic research paper. I first turn to evidence for the US market. In their Journal of Finance (2016) paper, McLean and Pontiff document that – when comparing the initial in-sample period and the after-publication out-of-sample period – the magnitude of average factor premia decreases from approximately 7% to 3% per year, i.e., a decline of more than 50%. This finding provides evidence that investors learn from academic research which makes factor premium profitability disappear.

How is the situation in other stock markets worldwide? Jacobs and Müller (2020, Journal of Financial Economics) investigate the profitability and predictability of factor premia in 39 international stock markets. Their main findings are:
a) Most of the factor premia that are documented in the US are also observed on stock markets worldwide. Hence, investors around the globe have the tendency to (mis-)price stocks in a similar way.

b) Average factor premia of the initial in-sample period are slightly higher in the USA (7% per annum) than in the rest of the world (5% per annum).

c)  The decay in factor premia in the after-publication period is significantly larger for the US than for most non-US countries. Hence, on average, factor premia tend to be more persistent in non-US stock markets. (This finding is consistent with the idea that learning about factor premia is particularly pronounced by sophisticated institutional investors that trade on US markets).

To summarize, academic research documents that academic research diminishes the magnitude of factor premia in the USA, but not so much for most other international countries. Consequently, a long-term factor investing strategy (based on known stock market anomalies) should be performed on a diversified global sample. 

Note that in this article, we have refrained from two additional important practical considerations: (i) trading costs in the composition of the long-short factor portfolios (which can substantially decrease the net performance of factor premia) and (ii) short-selling restrictions of stocks (which can make the composition of the long-short factor portfolios impossible).

 To dig deeper into the topic, retrieve the complete working papers on SSRN (McLean and Pontiff; Jacobs and Müller) and the published papers in the Journal of Finance and the JFE.

THe SCientific Investor - PArt 01/2021: Cross-Asset Signals and Time Series Momentum

A lot of practical investment ideas are based on technical analysis, i.e., one uses past returns (and other characteristics) of an asset to predict its future return development. Let us extent this idea a bit: What about if we depend the decision to buy/sell an asset also on the past returns of another asset? In other words, can we use cross-asset signals to successfully derive a systematic trading rule? 

In their Journal of Financial Economics (2020) paper, Pikäjärvi, Suominen, and Vaittinen investigate this idea based on a cross-asset strategy of the equity market (E) and the bond market (B). They show that (1) high bond market returns predict high future equity market returns and (2) low equity market returns predict high future bond market returns on a monthly horizon and beyond.

Why do we observe these relationships? The reason for this cross-asset momentum is mainly based on the notion of slow-moving capital and the perspective that equity is more risky than bonds. During general market upturns, investors shift their investment allocations from bonds to equity to receive a higher premium. During general market downturns, they divest from equity and invest into bonds to reduce potential future losses. However, these transactions do not occur immediately but are slow-moving and hence generate medium-term cross-asset momentum.

A simplified equity-bond cross-asset trading strategy at the beginning of month t  can be constructed as follows: Compute the past 12-month equity return (E past) and the past 12-month bond return (B past). If:

a) E past is positive and B past is positive: Buy equity
b) E past is negative and B past is negative: Sell equity
c) E past is negative and B past is positive: Buy bonds
d) E past is positive and B past is negative: Sell bonds
e) Options a) to d) are not representative, invest in the risk-free rate.
Hold the portfolio for one month and then repeat the same procedure in month t+1.

Pikäjärvi, Suominen, and Vaittinen (2020) show that a diversified cross-asset strategy over 20 countries (including the USA, Germany, and Switzerland) from 1980 to 2016 outperforms pure equity or bond momentum strategies. It also has a significant higher Sharpe ratio than a buy-and-hold portfolio consisting of bonds and equities.

To dig deeper into the topic, retrieve the complete working paper (without charge) on SSRN or the published paper in the JFE