Ruble, Oil, Chinese Stocks, & Evolving with Sentiment Regimes

 It Is Not So Cheap 

One thing: you have to walk, and create the way by your walking; you will not find a ready-made path. It is not so cheap, to reach to the ultimate realization of truth. You will have to create the path by walking yourself; the path is not ready-made, lying there and waiting for you. It is just like the sky: the birds fly, but they don't leave any footprints. You cannot follow them; there are no footprints left behind.
~ Osho

It’s been over a decade since MarketPsych set out to decipher how sentiment impacts global markets.  We dove into text-analytics in 2004 after we saw research indicating that - even when the overall payoff odds are known by investors, they will often choose investments with lower (or even negative) payoffs due to the words used to describe such investments.  Such frenzies of excitement and despair about markets - leading to wealth-destroying investor behavior - are expressed in social and news media, and they seemed to us (and many others) to be correlated with market price patterns.  We realized that price and fundamental data wasn't enough.  We wanted to understand markets on the level of group behavior - consensus and the information flow that drives it.  

After good preliminary results from trading social media sentiment on paper from 2006-2008, we ran the MarketPsy Long-Short Fund LP - the first social media-based hedge fund - from 2008-2010.  There were some media reports on our fund with an interview here and an overview here.  But while our trading was profitable - our fund beat the S&P 500 by 24% over its lifetime and it beat its benchmark by even more - our ability to keep our tech operations capitalized through the financial crisis was not.   Turning away from fund management and into data distribution and research, we were fortunate to enter into a partnership with Thomson Reuters, pairing our unique style of analytics with their world-class project management, data handling expertise, and information flow.

As we look back over a decade of our development, we see exciting changes ahead.  We're planning to release important new products in 2015, and every week we're finding novel insights into the role of sentiment and information in driving asset prices.  

In today's newsletter we explore how sentiment impacts markets with examples from Russia, Crude Oil, and the Chinese stock market.  We then explore how sentiment is itself dynamic.  The best investing strategies differ in bull versus bear markets, and this excellent Thomson Reuters white paper reveals a dynamic technique to take advantage of this effect.  We then explore how emotionality among investors, versus the analytical discussions of fundamentals, is actually one of the best predictors of stock prices.  Goal setting for investors is far different than for people in other fields - we wrap-up with a guide to investment goal-setting for 2015.

How Sentiment Impacts Markets:  Overreaction in Russia and Underreaction in Oil

Contrariwise, if it was so, it might be; and if it were so, it would be; but as it isn't, it ain't. That's logic.
~ Lewis Carroll

Sentiment patterns in markets are not logical, which is perhaps the key to their continued existence.  They are typically contrarian. Fortunately, academic research is rapidly creating an logical framework for understanding the impact of sentiment on markets.

There are two well-known sentiment-based patterns in financial prices:  overreaction and under-reaction.  An event is called overreaction if prices mean-revert (bounce) in a predictable way.  It’s called under-reaction if a trend (momentum) is likely to be born. Of course, the terms are easy to apply after-the-fact.  These patterns are more difficult to predict than the names imply.  

Both over and underreaction patterns result from information cascading through markets.  When traders all panic, as we saw with the Russian Ruble in December, prices drop steeply and climax in a frenzy of selling.  Once such a frenzy slows a bit, prices bounce.  A V-bottom is characteristic of this pattern.  We saw this pattern in the Turkish Lira last year:



We something similar with the Russian Ruble now.  As with the Turkish Lira last year, our Buzz index is the most dramatic representation of the selling climax in the Ruble - note the spike in the black line.  Up to 40,000 relevant news and social media mentions in one 24 hour period:



Underreaction results when markets take their sweet time digesting surprising new information.  Unexpected news creates an initial shock, followed by efforts to adjust mental models and expectations.  The market gradually awakens to the importance of the new information.  For under-reaction, we usually see trends.  We had previously found an excellent predictor of crude oil trends that is based on underreaction:


 
Using this reliable predictor of oil updated to this weekend, it looks as if support may be arriving for cude prices (see the far far right of the crude oil price chart below).  A bounce may be in store, but I'm still a bit dubious, and the indicator is not yet green.


As seen in this chart, the Energy sector is looking to have some struggles ahead.  It often takes time for such adjustments as the recent decline in energy prices to work its way through the sector (i.e. underreaction), occasionally bankrupting those who took on too much debt.

Oil prices respond slowly to consensus as they trend.  We see something similar in the Japanese Yen, which is prone to 
underreacction and is the most consensus-driven asset we have identified.

Overrection and underreaction can occur in sequence, witness China's stock market.  China's phenomenal stock rally continued through December, and based on our indicators, it is likely to persist on an upwards trajectory a bit longer.  The indicator below captures momentum, and similar to the Crude Oil indicator above, it has continued to predict the Shanghai Composite with a decent degree of accuracy since we first featured it under our Country tab.  Note in this 2011-2015 chart the waves of enthusiasm that have infected investors about China over the past several years (the green surges mark overreaction when they roll over).  All enthusiasm previously met with disappointment.  The current one has not peaked yet and is still trending higher:


While over and underreaction are difficult to predict systematically over an entire business cycle, comprehensive new research is identifying how sentiment regimes contain the keys to properly trading on such patterns in markets. Beyond the obvious impacts of sentiment above in Russia, Crude Oil, and China, we also see more subtle findings, including those in a  ground-breaking paper by Thomson Reuters’ Senior Quantitative Research Analyst, Elijah DePalma PhD.

Strategy-Shifting Across Sentiment Regimes

Logic: The art of thinking and reasoning in strict accordance with the limitations and incapacities of the human misunderstanding.
~ Ambrose Bierce

During 2008-2009 when we traded stocks for our hedge fund - the MarketPsy Long-Short Fund LP- weekly overreaction (mean-reversion) strategies performed phenomenally well for us in 2008 and 2009.  Since the financial crisis, beyond short periods of default fears ("Not Dubai!"  "Oh no, Greece!"), momentum strategies have dominated returns for sentiment-based strategies.

Research by Elijah DePalma documents the effects of sentiment regimes on the predictable returns from market anomalies (patterns).  DePalma significantly expands the work of academics such as Livnat and Petrovits (2009), who find that post earnings announcement drift is significantly greater when market sentiment is opposite the direction of the earnings surprise. As DePalma puts it, “During periods of high (low) sentiment investors generally expect good (bad) news, and if a firm reports earnings contrary to these expectations then investors’ under-reaction to the earnings surprise may be magnified.”  When the baseline mood of investors is positive, disappointments are more negatively impactful to prices than good surprises are positive.

One of Dr. DePalma's most stunning findings is that high beta (more volatile) stocks do in fact outperform low beta stocks, but only in negative sentiment environments.  The opposite relationship is true in positive environments.  This stunning image shows the significance of low versus high-beta stock performance in each regime.  On the left, when monthly sentiment is negative, high beta stocks (the furthest right blue bar) signficantly outperfom over the following month.  The opposite is true when a positive sentiment climate is present.



When we used to trade our hedge fund, we would short stocks that both 1) experienced a sharp weekly price rally and 2) showed a peak in Joy - what traders call a blow-off top.  However, this pattern became quire unprofitable for us in 2010 - many stocks rallying on high joy were continuing to rally, leading to pain for our fund.

It is not the strongest or the most intelligent who will survive but those who can best manage change.
~ Charles Darwin​

In order to avoid being out of step with the market’s best strategies, Dr. DePalma proposes a dynamic methodology in which strategies are selected for trading capital depending on the predominant sentiment environment.  We must change our strategy depending on the overall market sentiment.  His model allocates investment capital to strategies depending on 1) the current sentiment environment and 2) how they are expected to perform in the current sentiment climate.  He demonstrates the results of such a dynamic strategy in his paper.

It turns out that overall market sentiment is the best predictor of which price and fundamental-based strategies will outperform.

Studying Emotion Vs Fact

When dealing with people, remember you are not dealing with creatures of logic, but creatures of emotion.
~ Dale Carnegie

Our text-analytics software uses a sophisticated process to identify and quantify how investors and the news media are describing individual companies.  We convert those references into time series, and the end result is the Thomson Reuters MarketPsych Indices (TRMI).  The TRMI are quantitative indexes of business news and social media content from 1998 to the present.

While we do produce a pure Sentiment index, one of our more interesting indexes is our new emotionVsFact index.  This index is a ratio between emotional (fear, anger, joy, etc...) versus purely factual (fundamental, accounting, earnings, etc...) commentary on each company.

In order to study our indexes, our Head of Research CJ Liu creates simple rotational models.  In the following examples, his software identifies the top 200 US companies with the highest Buzz in the news over the past 1 year.  It then ranks those 200 companies by the average value of one of the sentiment indexes (e.g., emotionVsFact, fundamentalStrength, earningsForecast, etc…) for the past year.  In the examples below, it simulates a portfolio that buys the top 20% of the most emotional companies, and it shorts the bottom 20% of the most factual companies.  It then calculates the returns of this 40 long versus 40 short portfolio daily, and it re-ranks and re-enters positions with 1/12 of the portfolio monthly from 1999 to summer 2014.  This model is performing emotional arbitrage.

Using the technique described above, we see that buying the most emotional and shorting the most factual quintile of stocks as described in the news produces the low volatility equity curve below:

 

Interestingly, this effect is slightly stronger (although more volatile) with data coming only from social media.


 
This effect is even more powerful on a monthly versus an annual basis.  The below equity curve was generated using a 1 month holding period with monthly position rotation.  Transaction costs were not factored in, which would diminish the outperformance of monthly versus yearly returns in a live strategy.



None of the individual emotion indexes we produce (Fear, Anger, Joy, Gloom) accounted for the stability and magnitude of the performance we see above.  It turns out that much of the outperformance of the emotionVsFact index was due to specific factual topics.  

A mind all logic is like a knife all blade. It makes the hand bleed that uses it.
~ Rabindranath Tagore

We consider fundamental conversations to be factual.  The following equity curves show that the fundamental indexes show good performance when selecting the top 200 by Buzz, then ranking those stocks by the following indexes and arbitraging the top versus bottom 40.   In this case, users of the data should short the stocks with the highest earnings forecasts and the best fundamentals as described in the news, while buying those with negative earnings outlooks and weak balance sheets.



In prior newsletters we showed similar equity curves for Anger and Fear which are positively correlated with future returns.  What we see above is that regimes can be automatically arbitraged by understanding the ratio of emotion versus fact in the media.

These results are reflective of what Dr. DePalma found.  When negative emotions dominate (as during a bear market), we want to buy the most negative stocks.  Investors and media will contain high levels of negative emotion expressed about bad stocks, and we should buy these.  When positive emotions dominate - and in bull markets emotional commentary focuses on the happiest, trending stocks - we shoud buy those.  In both cases, the most emotional stocks are generally negative during bear markets (predisposing to mean-reversion bounces) and positive during bull markets (predisposing to positive upwards trends).  The emotionVsFact index appears to self-correct for sentiment regimes, dynamically capturing the best trading approach through both bull and bear markets.  

Goals

Mastering others is strength. Mastering yourself is true power.
~ Lao Tzu

In prior years our January newsletters were dominated by goal-setting advice.  Like trading appropriate to the regime, good goals are emotionally rooted.  We won’t repeat what we’ve written in prior years, but you can see some of our still-valid advice, specifically under “Keeping Resolutions:  An Unconscious Perspective.”  A more detailed description of goal-setting for fund managers is in last year's January newsletter.  And here is a Kiplinger’s article on the topic.

Goal-setting, just like investing, should consider the sentiment regime in which it is occurring.  If you’re happy and positive when you set goals, you may set them too high, settling yourself up for disappointment.  The best New Year’s resolutions are those set in consideration of how you will act on your worst day.  If the resolution will survive you at your worst, then it’s tenable.

 Closing

Man's mind, once stretched by a new idea, never regains its original dimensions.
― Oliver Wendell Holmes Sr.

We hope our monthly newsletters help you gain insight into yourself and markets, leading to more profitable investments and a richer life.   In 2015 we anticipate the launch of several new data products and predictive models.  

Please contact us if you'd like to see into the mind of the market using our Thomson Reuters MarketPsych Indices to monitor market psychology and macroeconomic trends for 30 currencies, 50 commodities, 130 countries, 50 equity sectors and indexes, and 8,000 global equities extracted in real-time from millions of social and news media articles daily.

We love to chat with our readers about their experience with psychology in the markets and with behavioral investing!  Please send us feedback on what you'd like to hear more about in this area.

We look forward to continuing to share our insights with you.​

Here’s to a Healthy and Prosperous 2015!
Richard L. Peterson, M.D. and the MarketPsych Team

 References

- Depalma, E.   2014.  Sentiment & Investor Behavior.  Thomson Reuters Machine Readable News.  White Paper.
- Livnat, J., & Petrovits, C.  2009.  Investor sentiment, post-earnings announcement drift, and accruals. AAA.