The Impact of Crowding on Alternative Risk Premiums

17.May 2019

Related to all factor strategies …

Author: Baltas

Title: The Impact of Crowding in Alternative Risk Premia Investing

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3360350

Abstract:

Crowding is a major concern for investors in the alternative risk premia space. By focusing on the distinct mechanics of various systematic strategies, we contribute to the discussion with a framework that provides insights on the implications of crowding on subsequent strategy performance. Understanding such implications is key for strategy design, portfolio construction, and performance assessment. Our analysis shows that divergence premia, like momentum, are more likely to underperform following crowded periods. Conversely, convergence premia, like value, show signs of outperformance as they transition into phases of larger investor flows.

Notable quotations from the academic research paper:

"Crowding risk is listed as one of the most important impediments for investing in alternative risk premia. We contribute to this industry debate by exploring the mechanics of the various ARP in the event of investor flows, and study the implications of crowdedness on subsequent performance.

The cornerstone of our methodology is the classification of the ARP strategies into divergence and convergence premia. Divergence premia, like momentum, lack a fundamental anchor and inherently embed a self-reinforcing mechanism (e.g. in momentum, buying outperforming assets, and selling underperforming ones). This lack of a fundamental anchor creates the coordination problem that Stein (2009) describes, which can ultimately have a destabilising effect.

Divergence factor

Conversely, convergence premia, like value, embed a natural anchor (e.g. the valuation spread between undervalued and overvalued assets) that acts as an self-correction mechanism (as undervalued assets are no longer undervalued if overbought). Extending Stein’s (2009) views, such dynamics suggest that investor flows are actually likely to have a stabilising effect for convergence premia.

Convergence premia

In order to test these hypotheses we use the pairwise correlation of factor-adjusted returns of assets in the same peer group (outperforming assets, undervalued assets and so on so forth) as a metric for crowding.

We provide empirical evidence in line with these hypotheses. Divergence premia within equity, commodity and currency markets are more likely to underperform following crowded periods.

All divergence premias

Whereas convergence premia show signs of outperformance as they transition into phases of higher investor flows.

All convergence premias"


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About


Follow us on:

Facebook: https://www.facebook.com/quantpedia/

Twitter: https://twitter.com/quantpedia


 

Continue reading

News Implied VIX Since The Year 1890

9.May 2019

We present an interesting academic paper with a methodology that allows estimating VIX (volatility risk) since the year 1890 …

Authors: Manela, Moreira

Title: News Implied Volatility and Disaster Concerns

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2382197

Abstract:

We construct a text-based measure of uncertainty starting in 1890 using front-page articles of the Wall Street Journal. News implied volatility (NVIX) peaks during stock market crashes, times of policy-related uncertainty, world wars and financial crises. In US post-war data, periods when NVIX is high are followed by periods of above average stock returns, even after controlling for contemporaneous and forward-looking measures of stock market volatility. News coverage related to wars and government policy explains most of the time variation in risk premia our measure identifies. Over the longer 1890-2009 sample that includes the Great Depression and two world wars, high NVIX predicts high future returns in normal times, and rises just before transitions into economic disasters. The evidence is consistent with recent theories emphasizing time variation in rare disaster risk as a source of aggregate asset prices fluctuations.

Notable quotations from the academic research paper:

"This paper aims to quantify this “spirit of the times”, which after the dust settles is forgotten, and only hard data remains to describe the period. Specifically, our goal is to measure people’s perception of uncertainty about the future, and to use this measurement to investigate what types of uncertainty drive aggregate stock market risk premia.

We start from the idea that time-variation in the topics covered by the business press is a good proxy for the evolution of investors’ concerns regarding these topics.

We estimate a news-based measure of uncertainty based on the co-movement between the front-page coverage of the Wall Street Journal and options-implied volatility (VIX). We call this measure News Implied Volatility, or NVIX for short. NVIX has two useful features that allow us to further our understanding of the relationship between uncertainty and expected returns:

(i) it has a long time-series, extending back to the last decade of the nineteen century, covering periods of large economic turmoil, wars, government policy changes, and crises of various sorts;

(ii) its variation is interpretable and provides insight into the origins of risk variation.

The first feature enables us to study how compensation for risks reflected in newspaper coverage has fluctuated over time, and the second feature allows us to identify which kinds of risk were important to investors.

We rely on machine learning techniques to uncover information from this rich and unique text dataset. Specifically, we estimate the relationship between option prices and the frequency of words using Support Vector Regression. The key advantage of this method over Ordinary Least Squares is its ability to deal with a large feature space. We find that NVIX predicts VIX well out-of-sample, with a root mean squared error of 7.48 percentage points (R2 = 0.19). When we replicate our methodology with realized volatility instead of VIX, we find that it works well even as we go decades back in time, suggesting newspaper word-choice is fairly stable over this period.

News Based VIX Index

We study whether fluctuations in NVIX encode information about equity risk premia. We begin by focusing on the post-war period commonly studied in the literature for which high-quality stock market data is available. We find strong evidence that times of greater investor uncertainty are followed by times of above average stock market returns. A one standard deviation increase in NVIX predicts annualized excess returns higher by 3.3 percentage points over the next year and 2.9 percentage points annually over the next two years.

Interpretability, a key feature of the text-based approach, enables us to investigate what type of news drive the ability of NVIX to predict returns. We decompose the text into five categories plausibly related (to a varying degree) to disaster concerns: war, financial intermediation, government policy, stock markets, and natural disasters. We find that a large part of the variation in risk premia is related to wars (53%) and government policy (27%). A substantial part of the time-series variation in risk premia NVIX identifies is driven by concerns tightly related to the type of events discussed in the rare disasters literature."


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see a performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About


Follow us on:

Facebook: https://www.facebook.com/quantpedia/

Twitter: https://twitter.com/quantpedia


 

Continue reading

An Analysis of PIMCO’s Bill Gross’ Alpha

4.May 2019

Bill Gross is probably the most known fixed income fund manager. A new academic paper sheds more light on his track record and sources of his stellar performance …

Authors: Dewey, Brown

Title: Bill Gross' Alpha: The King Versus the Oracle

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3345604

Abstract:

We set out to investigate whether ''Bond King" Bill Gross demonstrated alpha (excess average return after adjusting for market exposures) over his career, in the spirit of earlier papers asking the same question of ''Oracle of Omaha," Warren Buffett. The journey turned out to be more interesting than the destination. We do find, contrary to previous research, that Gross demonstrated alpha at conventional levels of statistical significance. But we also find that result depends less on the historical record than on whether we take the perspective of academics interested in market efficiency, investors picking a fund or someone (say a potential employer) asking whether a manager has skill or is throwing darts to pick positions. These are often thought to be overlapping or even identical questions. That's not completely unreasonable in equity markets, but in fixed income these are distinct. We also find quantitative differences, mainly that fixed-income securities have much higher correlations with each other than equities, make alpha 4.5 times as hard to measure for Gross than Buffett. We don't think our results will have much practical effect on attitudes toward Gross as an investor, but we hope they will advance understanding of what alpha means and appropriate ways to estimate it.

Notable quotations from the academic research paper:

"Superstar bond portfolio manager Bill Gross announced his retirement last week. From 1987 to 2014, his PIMCO Total Return fund generated 1.33% per year of alpha versus the Barclays US Credit index, with a t-statistic of 3.76. For many years his fund was the largest bond fund in the world, and was generally considered to be the most successful. This track record inspired us to take a closer quantitative look along the lines of Frazzini, Kabiller and Pedersen's Buff ett's Alpha (FKP). Gross, like Bu ffett, often publicly discussed what he perceives as the drivers of his returns. At the Morningstar Conference in 2014 and in a 2005 paper titled "Consistent Alpha Generation Through Structure" Gross highlighted three factors behind his returns: more credit risk than his benchmark, more 5-year and less 30-year exposure, and long mortgages and other securities with negative convexity.

We present five main fi ndings:

1. We con firm that those three factors, plus one for the general level of interest rates, explain 89% of the variance in Gross' monthly return over the 27-year period. We further estimate that Gross outperformed a passive factor portfolio by 0.84% per year, which is signi ficant at the 5% level. Gross' compounded annual return over the period was 7.52%, versus 6.44% for the Barclay's Aggregate US Index. So we find that most of his 1.08% annual outperformance of the index was alpha.

Bill Gross' Alpha

2. The FKP paper mentioned above considered one of the best-known track records in the equity asset class, Warren Buff ett's. We compliment this work by examining one of the best-known track records in the fixed-income asset class. Fixed-income investing o ffers a di erent set of challenges and opportunities than equity. We o ffer a novel discussion on the concept of manager alpha including important qualitative and quantitative di fferences in the concept of alpha with Gross versus Bu ffett.

3. The main qualitative di fference is that Gross exploited well known sources of risk and potentially excess return in the fixed-income market, exposures that investors rationally demand additional yield to accept. Bu ffett's performance, for the most part, correlates with factors uncovered long after he began investing and were still not accepted as fully as factors like credit risk or mortgage prepayment risk. Moreover Buff ett's factors probably result from behavioral biases and institutional constraints rather than rational investor preferences.

4. The main statistical di fference is the much higher r2 value in Gross' regression versus Buff ett's (about 0.9 versus 0.3) makes the alpha signi ficance estimate 4.5 times as sensitive to the observed returns on the factor portfolios. Since it is nearly impossible to estimate expected returns – there is considerable debate about the level of the equity premium even with 150 years or more of data – this makes it important to select factors that conform as closely as possible to what Gross actually did, rather than factors that merely have a high return correlation to Gross' results. The closer the factors conform to Gross' practice, the better the chance that any deviations in factor performance from expectation over the period are reflected equally in both Gross' actual results and the factor portfolio results.

5. Gross earned essentially all of his alpha in favorable markets for his factors and had a signi ficantly negative timing ability in the sense that his factor exposures were greater in months the factor had negative returns than in months the factor had positive return. This latter feature could be unfortunate timing decisions or negative convexity in the factor exposures. We discuss whether this can shed light on the source of Gross' alpha, speci fically whether it relates to preferential access to new issues and leverage."


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About


Follow us on:

Facebook: https://www.facebook.com/quantpedia/

Twitter: https://twitter.com/quantpedia


 

Continue reading

Case Study: Quantpedia’s Composite Seasonal / Calendar Strategy

26.April 2019

Introduction

Despite the fact that the economic theory states that financial markets are efficient and investors are rational, a large amount of research is about anomalies, where the result is different from the theoretical expectation. At Quantpedia, we deal with anomalies in the financial markets and we have identified more than 400 attractive trading systems together with hundreds of related academic papers.

This article should be a case study of some strategies that are listed in our Screener, with an aim to present a possible usage of strategies in our database. Moreover, we have extended the backtesting period and we show that the strategies are still working and have not diminished. This blog also should serve as a case study how to use the Quantpedia’s database itself; therefore the choice of strategies was not obviously random and strategies were filtered by given criteria, however, every strategy is listed in the “free“ section, and therefore no subscription is needed.

Since our strategies are connected with the equities, the first filter in our screener is “Markets“, where we have picked equities. Secondly, various characteristics of strategies could be simply found by searching with keywords, in this case, we would choose the "seasonality". Lastly, we would pick “Only Free“ from the Free/Premium filter.

Screener setup

Although the term anomalies in the financial markets might sound too complicated, we have picked simple, yet working seasonal anomalies. Each strategy works well alone and at first, we examine them one by one, but we also show that those strategies could and probably should be considered as building blocks of one bigger strategy. Therefore, this blog is divided into two parts: where the first part is about the examination of the seasonal strategies found by the filtration – Turn of the Month in Equity Indexes, Federal Open Market Committee Meeting Effect in Stocks, Option-Expiration Week Effect and The Payday Effect. As the names are suggesting, the anomalies are seasonal. Another possibility is that we could consider them as calendar anomalies, therefore the analysis could be done simply and in our opinion, the strategies are an ideal way how to dig into the world of anomalies in the financial markets.

As we have previously mentioned, the second part is about the composite strategy made of building blocks. This approach might be simple, but it is also efficient and working. Although some complicated mathematical models may sound more fancy, combining simple strategies can lead to extraordinary results and our goal is to prove that if strategies picked by us are combined together, they are working just in that way. Additionally, for each strategy or the composite one, the investor only needs to invest into S&P500 index, which can be easily made by the ETFs (for example SPY). Last but not least, our aim is to present the strategies from the practitioner's point of a view, for the readers interested in the more theoretical depths we would advise to read the original papers connected with the particular strategies.

Turn of the Month

The turn of the month is a well-known effect on stock indexes, with a simple idea that equity prices usually increase during the last four days and the first three days of each month. Such a pattern was identified by various distinct researches for various time periods both in the Dow Jones Industrial Average and also in the S&P500 index. Apart from the expanded backtesting period and proving that such simple strategy with an easy execution still works in the present, we think that this strategy can be even more simplified by buying the SPY ETF on close at the end of the month, and selling it on close of the first day in the following month.

Despite the simplicity of this anomaly, it is a puzzle for the academic world. The Turn of the Month strategy is a big challenge for the academic world that tries to explain the potential reasons for the functionality. The reason for the functionality is not a risk-based, since higher risk does not appear to explain the turn-of-the-month effect. However, according to the other branch of the literature, the effect may be simply explained by the regularity in payment dates in the United States, because investors receive a preponderance of compensation from employment, dividends and interest at month-ends. Consequently, as investors seek to invest these funds, equity prices are pushed up.

Turn of the Month

As the graph of the strategy for the years 1993-2019 shows, the Turn of the Month is profitable and this pattern is still alive even in the present. One dollar invested in 1993 would be more than doubled to 2,11 dollars in the year 2019, with a yearly performance of 3,01%. Such strategy has a maximal drawdown of 11,97%, which results in the return to drawdown ratio of 0,25.

Federal Open Market Committee Meeting Effect in Stocks

According to the past research, the S&P 500 index average daily returns during Federal Open Market Committee (FOMC) meetings since 1980 (or the year since FED started to be less secretive and more open about its future plans and actions) are outstanding – more than 5 times greater than returns during other average days on market. Since dates of FED meetings are publicly known and available, such effect could be easily utilized in the seasonal strategy that would long the S&P 500 index during these FED meetings. As we have previously mentioned, the simple execution of this strategy could be made by buying SPY ETF on a close day before the meeting and selling it on close after the meeting.


A simple explanation could be connected with the one market wisdom that says: „Don’t fight the FED“. Since the FED's main aims are to address banking panics, maintain the stability of the financial system, contain systemic risk in financial markets and strengthen economic growth, it is highly unlikely that FOMC meetings conclusions would be highly negative for stocks. Moreover, in reality, the actions of FED are the opposite and in average those actions are positive for the stocks, which is the main cause for a positive drift.

FED effect

The graph shows that also this anomaly is alive and working in the present. A dollar invested in 1993 would have resulted in the 1,77 dollars in 2019, with a yearly performance of 2,30%. Although the performance might not seem to be impressive, it is important to bear in the mind that the strategy is invested only a few days during the year and the strategy is able to perform in this way with a relatively small maximal drawdown of 7,17%, what results in the return to drawdown ratio of 0,32.

Option-Expiration Week Effect

As the name Option-Expiration week effect suggests, this effect is another calendar anomaly. This one is connected with the Option-expiration week – a week before options expiration (Friday before each 3rd Saturday in each month). The research suggests that stocks with large market capitalization, that have actively traded options, tend to have substantially higher average weekly returns during these weeks. This leads to a construction of a simple market timing strategy, where an investor buys the SPY ETF on close each Friday before 2nd Saturday in each month and sells it on close again in the next week’s Thursday.

According to the research, intra-month weekly patterns in call-related activity contribute to patterns in weekly average equity returns and this should be the main driver of the performance. Simplified, hedge rebalancing of option market makers trading the largest stocks that have the most actively traded options, should be the main reason for the abnormal returns of these largest stocks. Additionally, during option-expiration weeks, a sizable reduction occurs in option open interest, because the near-term options approach expiration and after that, they simply expire. This reduction in call open interest should be associated with a reduction in the net long call position of market makers, which implies a decrease in the short-stock positions being held by market makers to delta hedge their long call holdings.

Option expiration effect

Out of our seasonal strategies, the Option-Expiration Week Effect has the highest performance, the dollar invested in 1993 would result into 3,36 dollars in the year 2019. Therefore, the strategy has an annual performance of 4,93%. The maximal drawdown of the strategy is 20,39%, which results in a return to drawdown ratio of 0,24. Unfortunately, the high performance is connected also with a greater risk represented by the maximal drawdown, but this could and we would show later that should be enhanced.

The Payday Effect

The Payday effect is similar to the Turn-of-the Month (ToM) anomaly. For the ToM, the research has linked the abnormal return with pay-days. After pay-days, investors seek to invest these funds which causes pushed up equity prices. However, many companies pay their employees twice a month, on the 15th day and at the end of the month, therefore it is natural that on the condition that the pay-day effect holds true for the turn-of-the-month days, there should be a recognizable pattern in the middle of the month as well. Research confirms the aforementioned hypothesis and abnormal returns truly exist in the middle of the month. Therefore, the simply strategy that utilizes this effect consists of buying the SPY ETF on close on the 15th day each month and selling it on close next day.

The reason for the functionality is probably deeply connected with paychecks. Many companies pay their employees twice a month, on the 15th day and at the end of the month, therefore building on that the pay-day effect holds true for the turn-of-the-month days, there should be a clear pattern in the middle of the month as well as at the end of the month. If employees get paid, many of them either automatically invest a portion of their paycheck in the market through retirement contributions or are encouraged to do so by having a surplus of funds with the new paycheck. This causes a temporarily pushed prices of stocks up, moreover it is rational to utilize that temporary situation by investing into S&P500, since these stocks should move up at most. Last but not least, according to the research, the 16th of the month is the 3rd best day in the month overall, therefore it makes a great addition to our pool of strategies and different calendar days.

Pay-day effect

Consistent with the research that has been made, the Pay-day Effect strategy is obviously profitable and this still holds true even in the present. The dollar invested in 1993 would result in approximately 1,68 dollars in 2019. The strategy has an annual performance of 2,08% and is able to perform in this way with a maximal drawdown of 8,62%. This results in return to drawdown ratio of 0,24.

Composite strategy

As we have previously mentioned, we would like to form one bigger strategy out of these smaller ones that represent building blocks. Although there are many options how to form such strategy, a simple approach of investing the whole portfolio into SPY during “anomaly“ days sounds simple, but it is also logical. If some days are overlaying, we do not leverage our portfolio. Considering our composite strategy, one would probably think also about the risk of the strategy and not only the returns. Most of the strategies do not have drastically high maximal drawdowns, however, there is an option how to lower this maximal drawdown. The solution of this task is to add the trend factor into strategy, where the one of the simplest way how to do it is to trade only if the price of SPY is higher than it’s 200 day average. Such addition would reduce the maximal drawdown from 24,02% to 10,14% with only a small reduction in the profits.

Summary strategy

Clearly, the composite strategy without the trend having an annual performance of 9,06% is the most profitable one. But one question arises, is it worth it? What about the ratio of return to risk? The strategy with the added trend has an annual performance of 7,47%, but with the reduced maximal drawdown, the return to drawdown ratio is 0,74. The return to drawdown ratio of the simple summary strategy without the trend is only 0,38. Therefore, as a final strategy, we would propose the strategy with the added trend factor.

Authors:
Radovan Vojtko, CEO, Quantpedia.com
Matus Padysak, Analyst, Quantpedia.com


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About


Follow us on:

Facebook: https://www.facebook.com/quantpedia/

Twitter: https://twitter.com/quantpedia


 

Continue reading

Momentum In International Government Bonds Can Be Explained By Currency Momentum

18.April 2019

A new academic paper related to:

#8 – Currency Momentum Factor

Authors: Zaremba, Kambouris

Title: The Sources of Momentum in International Government Bond Returns

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3332942

Abstract:

This study aims to offer a new explanation for the momentum effect in international government bonds. Using cross-sectional and time-series tests, we examine a sample of bonds from 22 countries for the years 1980 through 2018. We document significant momentum profits that are not attributable to bond-specific risk factors, such as volatility or credit risk. The global bond momentum is driven by the returns on underlying foreign exchange rates. Controlling for currency movements fully explains the abnormal returns on momentum strategies in international government bonds. The results are robust to many considerations including alternative sorting periods, portfolio construction methods, as well as subperiod and subsample analysis.

Notable quotations from the academic research paper:

"The various types of momentum effects have also been documented in government bonds, implying that the fixed-income winners outperform fixed-income losers. Although the finance literature extensively discusses the sources of momentum in an equity universe, the specific explanations for momentum in government bonds are rather scarce.

This paper aims to contribute in two ways. First, we provide new evidence on the momentum effect in international government bond markets. Using cross-sectional and time-series tests, we investigate a sample of government bonds from 22 countries for the years 1980 through 2018.

Second, and more importantly, we offer and test two new explanations of momentum. Our first hypothesis builds on Conrad and Kaul (1998): we conjecture that the momentum in bonds may simply capture the cross-sectional variation in long-run returns. In other words, the top performing assets continue to deliver higher returns because they exhibit excessive risk exposure. In particular, we assume that the winner (loser) bonds may display high (low) exposure to duration and credit risks, which drive the excessive long-run returns. The second hypothesis is that the momentum in bonds might be driven by the returns on underlying currencies.

Fund flows

The primary findings of this study can be summarized as follows. We document a strong and robust momentum effect in government bonds. An equal-weighted portfolio of past winners tends to outperform past losers by 0.24–0.35% per month. The effect is not fully attributable to the risk factors in government bonds, which explain 38–55% of the abnormal profits. Nevertheless, the phenomenon is entirely explained by the momentum in underlying foreign exchange rates, which is consistent with our second hypothesis. Once we control for the currency returns in cross-section or time-series tests, the momentum alphas disappear. The results are robust to many considerations, including alternative sorting periods and portfolio implementation methods, as well as subperiod and subsample analyses."


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About


Follow us on:

Facebook: https://www.facebook.com/quantpedia/

Twitter: https://twitter.com/quantpedia


 

Continue reading

Can We Explain Abudance of Equity Factors Just by Data Mining? Surely Not.

11.April 2019

Academic research has documented several hundreds of factors that explain expected stock returns. Now, question is: Are all this factors product of data mining? Recent paper by Andrew Chen runs a numerical simulation that shows that it is implausible, that abudance of equity factors can be explained solely by p-hacking …

Author: Chen

Title: The Limits of P-Hacking: A Thought Experiment

Link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3358905

Abstract:

Suppose that asset pricing factors are just p-hacked noise. How much p-hacking is required to produce the 300 factors documented by academics? I show that, if 10,000 academics generate 1 factor every minute, it takes 15 million years of p-hacking. This absurd conclusion comes from applying the p-hacking theory to published data. To fit the fat right tail of published t-stats, the p-hacking theory requires that the probability of publishing t-stats < 6.0 is infinitesimal. Thus it takes a ridiculous amount of p-hacking to publish a single t-stat. These results show that p-hacking alone cannot explain the factor zoo.

Notable quotations from the academic research paper:

"Academics have documented more than 300 factors that explain expected stock returns. This enormous set of factors begs for an economic explanation, yet there is little consensus on their origin. A p-hacking (a.k.a. data snooping, data-mining) offers a neat and plausible solution. This cynical explanation begins by noting that the cross-sectional literature uses statistical tests that are only valid under the assumptions of classical single hypothesis testing. These assumptions are clearly violated in practice, as each published factor is drawn from multiple unpublished tests. In this well-known explanation, the factor zoo consists of factors that performed well by pure chance.

In this short paper, I follow the p-hacking explanation to its logical conclusion. To rigorously pursue the p-hacking theory, I write down a statistical model in which factors have no explanatory power, but published t-stats are large because the probability of publishing a t-stat ti follows an increasing function p(ti). I estimate p(ti ) by fitting the model to the distribution of published t-stats inHarvey, Liu, and Zhu (2016) and Chen and Zimmermann (2018). The p-hacking story is powerful: The model fits either dataset very well.

p-hacking model

Though p-hacking fits the data, following its logic further leads to absurd conclusions. In particular, the pure p-hacking model predicts that the ratio of unpublished factors to published factors is ridiculously large, at about 100 trillion to 1. To put this number in perspective, suppose that 10,000 economists mine the data for 8 hours per day, 365 days per year. And suppose that each economist finds 1 predictor every minute. Even with this intense p-hacking, it would take 15 million years to find the 316 factors in theHarvey, Liu, and Zhu (2016) dataset.

This thought experiment demonstrates that assigning the entire factor zoo to p-hacking is wrong. Though the p-hacking story appears logical, following its logic rigorously leads to implausible conclusions, disproving the theory by contradiction. Thus, my thought experiment supports the idea that publication bias in the cross-section of stock returns is relatively minor."


Are you looking for more strategies to read about? Check http://quantpedia.com/Screener

Do you want to see performance of trading systems we described? Check http://quantpedia.com/Chart/Performance

Do you want to know more about us? Check http://quantpedia.com/Home/About


Follow us on:

Facebook: https://www.facebook.com/quantpedia/

Twitter: https://twitter.com/quantpedia


 

Continue reading

Subscribe for Newsletter

Be first to know, when we publish new content


logo
The Encyclopedia of Quantitative Trading Strategies

Log in