Entering text into the input field will update the search result below

Seeking Alpha Crowd Wisdom Predicts Future Stock Returns

Mar. 19, 2014 5:17 PM ET197 Comments
Eli Hoffmann profile picture
Eli Hoffmann


  • A recent academic study, highlighted in today's WSJ, validates that crowd-sourced research is predictive of future stock prices.
  • SA articles and comments were predictive of future stock prices and earnings surprises over all time-frames studied, from one month to three years.
  • Traditional media, sell-side research, and other social platforms don't appear to have similar predictive value.

The notion that the collective wisdom of the crowd is superior to that of any individual is well-documented. Here's one of the classic wisdom-of-the-crowd findings:

At a 1906 country fair, eight hundred people participated in a contest to estimate the weight of a slaughtered and dressed ox. Statistician Francis Galton observed that the median guess, 1207 pounds, was accurate within 1% of the true weight of 1198 pounds.

This concept, as I discussed in a recent article, shaped Seeking Alpha's vision of being an open, curated platform for diverse stock-market opinion, in which all sides have a say, and in which the audience plays a key role in the discussion.

But investors care more about results than they do about noble visions, which prompts the question: Is crowd-sourcing predictive of future returns within the sophisticated arena of stock investing, or are above-average returns the domain of a few gurus?

A recent academic study highlighted in today's Wall Street Journal puts that question to rest: Seeking Alpha outperforms broader markets, the sell-side, financial media, and other social platforms over every time frame studied.

Here's what the Journal has to say (with light edits for brevity/clarity):

A new academic study lends credence to the idea that the “wisdom of crowds” phenomenon applies not just to encyclopedia entries and restaurant reviews, but also to stock market predictions.

Researchers from City University of Hong Kong, Purdue University and Georgia Institute of Technology found that the tone of stock market opinion blogs published on investor forum seekingalpha.com predicted stock returns, as well as earnings surprises, above and beyond what was evident from Wall Street analyst reports and financial news articles.

Seekingalpha.com is a forum for investors who write opinion pieces about stocks for the site. An editorial board vets the quality of the blogs and posts up to 250 articles every day.

Researchers analyzed about 100,000 Seeking Alpha articles and commentary published between 2005 and 2012 for the paper “Wisdom of Crowds: The Value of Stock Opinions Transmitted Through Social Media,” which is forthcoming in the Review of Financial Studies.

The potential to discover market-predicting information in social media has been tantalizing to investors big and small. Several startups, including companies like Dataminr, Gnip and DataSift, have talked about providing feeds of Twitter and other social-media outlets to hedge funds and other money managers to inform their investment strategies. But the connection between what’s happening on social media and what’s happening on the stock market has had little proof to date. One study showed that the general mood of the public, as exhibited in aggregate tweets, can predict the movements in the Dow Jones Industrial Average. That study didn’t look at individual stocks, but the new study did.

“Seeking Alpha is the only platform to date that we have shown can predict individual stock returns,” said Yu Jeffrey Hu, associate professor at Georgia Tech’s Scheller College of Business and one of the authors of the study. Other authors of the study are Hailiang Chen, Prabuddha De, and Byoung-Hyoun Hwang.

Dr. Hu said that research was in no way sponsored or facilitated by Seeking Alpha. The company did provide a proprietary data stream for one portion of the research, he said.

The researchers looked for the fraction of negative words in the articles and commentary on a particular stock on a particular date, and then tracked the performance of the stock after the blog publication. It turned out that the more negative the blogs and blog comments were on a stock, the more likely the stock was to perform worse than similar stocks in the next several months. Similarly, the negative tone of Seeking Alpha articles predicted earnings surprises, which are the earnings results reported by companies relative to the average of analysts’ earnings forecasts.

“We cannot say this particular stock will go up or down in absolute. But it will perform better than a portfolio of roughly similar stocks,” Dr. Hu said.

The researchers created a virtual investment strategy where they went long on stocks most liked and went short on stocks most disliked by the Seeking Alpha community, and showed a multi-year return of some 40%, Dr. Hu said. Notably, the virtual profit kept growing through the 2008 market crash.

Does this mean sell-side financial analysts are useless? “I wouldn’t say [Seeking Alpha blog writers] are absolutely better than the highly paid Wall Street analysts,” Dr. Hu said. But “Seeking Alpha sentiment has additional insight above and beyond financial Wall Street analysts,” Dr. Hu said.

The researchers also noted that Seeking Alpha predicted stock returns above what was evident from news articles. The report used news stories published on Dow Jones News Service, which is a part of Dow Jones & Co., publisher of The Wall Street Journal. Dow Jones declined to comment.

Overall, the findings fit with prior analysis in other fields on the way crowds can outsmart, or at least be just as smart, as professionals. Studies have shown, for example, that Wikipedia accuracy is similar to that of Encyclopedia Britannica.

Crowd platforms have other advantages over professionals, besides accuracy - Seeking Alpha, for example, covers a greater stock universe than stock analysts, and Wikipedia has more subject entries than Encyclopedia Britannica. Plus, both get updated and revised as things change more quickly than their professional counterparts, Dr. Hu said.

“It’s clear that we’re sitting on valuable data,” Seeking Alpha CEO David Jackson said. “We need to mine and share more data about what the community is thinking and doing.”

The article doesn't fully describe something I found remarkable: The researchers looked at pageviews and reads-to-end (how many readers finished reading an article) for articles that were destined to be predictive of future returns, and for articles that were destined to be counter-predictive of future returns. They also measured the acrimoniousness of those articles' comments. They found, astonishingly, that during the first 48 hours after publication, articles that were destined to be predictive of future returns had above average page-views, above average reads-to-end, and unacrimonious comment streams, while articles that were destined to be counter-predictive had below average page-views, below average reads-to-end, and acrimonious comment streams. In other words, crowd behavior was a leading indicator of which predictions would be right, and which would be wrong.

Other points from the study that weren't obvious from the Journal article:

  • Researchers discounted stock performance during the first 48 hours after publication in order to remove from performance data immediate reactions to articles being published.
  • The researchers looked at the ability of Seeking Alpha articles to predict not only future stock returns, but also future earnings surprises. This was in order to discount for the possibility that the phenomenon was simply a case of a large platform influencing future stock prices rather than predicting them. Future prices could in theory be influenced by crowd behavior, but investors have no ability to influence future earnings surprises.
  • SA articles and comments predicted stock returns over every time-frame examined: three months, six months, one year and three years. This was not true of previous studies of the predictive value of Twitter and Internet message boards, which demonstrated no predictive value. Previous studies of sell-side research demonstrated some predictive value over short time frames that disappeared over longer time frames.
  • Further demonstrating the value of crowd-sourced research, they found that in cases of broad-based disagreement between authors and the community, community sentiment was more accurate in predicting future stock prices and earnings surprises.

Hats off to David Jackson, our CEO, for having the vision to build such a powerful platform. And to everyone inside Seeking Alpha who works tirelessly to make SA a great website. But most of all, to our contributors and readers: this study is a validation of the collective wisdom you provide that makes us all smarter investors.

This article was written by

Eli Hoffmann profile picture
Seeking Alpha mourns the loss of Eli Hoffmann, beloved manager, friend and guiding light.

Recommended For You

Comments (197)

T-III profile picture
26 Apr. 2017
Wow that last part of the article is incredible. Personally, I think these findings from Seeking Alpha are more than just a neat statistical outcome. Like most full-time traders and investors, I get my info from a variety of sources: news, research papers, blogs, SA-type "social media", etc. In my old life I was a science-oriented Aeronautical Engineer with 20 years experience - and my job required the regular use of fairly advanced mathematics. My Bachelor's was in Physics, and at least in a casual way I try to keep up with findings in various quantum fields (unbelievable stuff). So I read this last paragraph again:

"They found, astonishingly, that during the first 48 hours after publication, articles that were destined to be predictive of future returns had above average page-views, above average reads-to-end, and unacrimonious comment streams, while articles that were destined to be counter-predictive had below average page-views, below average reads-to-end, and acrimonious comment streams. In other words, crowd behavior was a leading indicator of which predictions would be right, and which would be wrong."

What? So readers were acting as a whole... AND it wasn't that we each went on to making the right decision (no surprise there), but we WERE sensing when articles were CORRECT. I've not read all the comments yet (will in the morning). Sorry if any of this is redundant. But I would bet someone is working right NOW on how to harness our powers! Lol.

It is becoming well known even in the general public, that some very unusual interaction effects are now known to occur in quantum physics (which defy standard physics) AND there is an increasingly undeniable relationship with human consciousness. Heavy stuff, I know, but it started over a century ago with Heisenberg's Uncertainty Principle and Schrodinger's cat, the stories about Einstein's uneasiness in accepting "spooky action at a distance", etc. Back then, those theories remained theories because, short of "thought experiments", there was just no technology to really test them.

Fast forward to the modern era and "spooky action" is called Quantum Entanglement, and has been PROVEN. Again this is only at the sub-atomic quantum level, but it appears that two "entangled" particles will continue to interact (i.e. communicate) even when separated by ANY measure of distance (and maybe time). Even more bewildering is something called the "Observer Effect". This one does not have a consensus interpretation, but it is also proven. The essence is that when trying to disprove entanglement the experiment outcome is literally and directly effected BY THE FACT THAT IT WAS OBSERVED. We can predict it until we "watch it" - at which point it collapses into the "other" model of possible outcomes. So maybe we won't be so easy to "harness"? Lol. Not saying that, I have no idea.

But I'm not referencing some kind of "confirmation bias". Picture some high-tech experiment, just a "scientist" observing the read-out from sensors to computer screens. From machines that measure some matching binary-type aspect of two entangled particles - now impossibly separated. Flip the "switch" on one, and instantly the other one flips too. I mean INSTANTLY. Oh, unless a scientist is WATCHING. You can't make this stuff up. You could separate those particles by the galaxy (literally) - and they stay "linked".

So what force connects or "entangles" them? Short answer: Quantum mechanics. Certainly no force known in classic physics. And we know that one way to UN-entangle them is to just introduce an "observer"... and they will revert to a "statistically-determi... relationship (still not classical physics). Talk about a mind bender.

Meanwhile, in the field of Social Sciences: they thought their models of human decision-making were working great (same ones since the 1970's), used for determining what we "should" do and what we "actually" do. However, they had started to realize that when an element of uncertainty is added (volatility?) - suddenly the ability to predict us starts to really deviate from the model. Can you guess where that went?

They tried application of more complex quantum methods - as a model to predict human decision making (within a specifically defined decision-making scenario), and WITH extra "uncertainty" - and the results were much more accurate. It's a new field, but it's not fiction and it's NOT all theoretical. There's already been research on HFT applications. They've even got "quantum computers" - already WORKING (I think IBM is in the lead). They use a type of crystal instead of chips (if you didn't know, Google). One of the advantages IS this ability to "couple" two different (microscopic) "bytes" - knowing you can separate throughout a network at the speed of light. But retain control over one, and each "pair" is now an INSTANT switch. Quantum computers also have unreal capability for memory storage.

So what in the WORLD does it have to do with Seeking Alpha?? Lol.

There are significant parallels to consciousness. And some of this research has already investigated decision-making related to market investing. If you're interested, but don't care to read science journals, there's some great documentaries. So, is it just a coincidence that human thinking - as a "complex system" - lends itself to the specific mathematics of quantum physics? As mentioned in one of the above articles, many of the experiments they did to prove Entanglement theory required the ability to generate TRULY random numbers. I mean that was CRITICAL, and the lengths they went to are amazing (like triggering off the flashes from a distant star, i.e. using "input" that originated far away in both space AND time. (It's almost funny how badly physicists did NOT want this to be real! Lol). But interestingly that random number part of the set-up is called the "Free Will" machine .

Apparently, our universe of not-quite-particle / not-quite-wave, vibrating and interchangeable forms of energy and matter - has a type of "free will" (quantum level). Who knows, maybe we do too. :-) If you're still awake and want more, start down the rabbit hole here:





snowcoat profile picture
So can we start a weekly poll with target prices on SPY, gold, oil and bonds?
Glenn Abrett profile picture
On another note. There was a rather notorious Tesla bear named John Patterson who posted well written and apparently well-research negative articles on Tesla up until around six months ago. I am almost certain that his articles almost always made the stock go up the next day, probably because every point was always refuted at great length and in great detail by an army of tesla bulls. I wonder if there has been any research that shows that some authors are 'contrarian' -- in that the bull/bear bias of there articles result in short term swings the other way.
Glenn Abrett profile picture

In more modern times there has been extensive research (sorry, don't have links but you can google it) on the bottle of marbles conundrum which is similar to the weight of the ox crowdknowledge experiment you discuss. A large glass water bottle is filled with marbles. People are asked to estimate how many. The more people asked, the closer the average of the answer is to the truth. It doesn't matter if the group of people is comprised of world-class mathematicians or homeless people -- it is always the same result.

The most interesting and rather frightening thing about this is that NO ONE KNOWS WHY THIS IS TRUE. It is almost Jungian -- the collective unconscious or something weird like that. There is absolutely no rational explanation. Makes one think that ones sure knowledge of the universe is suspect.
AlphaCoils profile picture
Very fascinating. I wish that commentators had more power. (for example the ability to include their own images and graphics when responding.) Also wish the article itself had a like and dislike voting button and the comments had a dislike button. Often comments are far more valuable than the article and it would be nice if there were better tools to make those comments more visible and influential. Also I think it's unfair how an SA headline for a precisely timed or engineered article can easily make prominent position next to a stock symbol in Google finance (or other outlets) whereas the reading and commentary behind it is hard to get to (most stock traders don't even bother to get an SA account so they never get past the first article page or to the comments.) It's great for SA itself of course from an advertising point of view but at times this borders on market manipulation.
Retired Securities Attorney profile picture

Good suggestions except for the dislike button. Too many commentary streams degenerate into opinionated personal contests as it is.
AlphaCoils profile picture
@Retired Securities Attorney: Dislike! (just kidding...)
jack789 profile picture

I just can't resist a political comment here. It's also why democracies work better than top down rule from elites. And also why the market would have worked better than President Obama and his advisors' insistence that they were smarter than free markets for ways to revive the economy. But as we know, Pres Obama admires China and its ability to so easily enact massive infrastructure projects (most of which, after a few years, are now seen to be failing, so much so that China is returning to the free market to get the job done).
moves profile picture
25 Mar. 2014
When will we see an IPO for Seeking Alpha ? :-)
advisor4 profile picture
97,000 artilcles and comments were analyzed 1/3 of the total for 4 years

and we can conclude that "SA predicts stock prices"

SA readers as a group are very distrustful of academic articles even when decades of similar studies over extended time periods produces results that confirm the first research.

But this one is one upon which we can make strong conclusions.

Crowdsourcers please explain you rationale
Sumflow profile picture
> authors and the community, community sentiment was more accurate in predicting future stock prices and earnings surprises.<
Well that comes as no surprise because the authors in many cases sound like they are turning in reports from high school projects a week after they dropped out of school. Some do no research at all, and don't even know enough to have the right to have an opinion. The community can take them down anytime.
The community has money on the line, whereas Seeking Alpha authors often are thinking about putting money on the line. A completely different thing. Hats off the Seeking Alpha.

They give people the opportunity to make fools of themselves and the community to correct them.
Todd Renfro profile picture

There is a small button at the bottom of the article (above the comments) that says track new comments. If you check that, it will let you know when there are new comments.

Conversely, when you are tired of following a thread,you can uncheck that.

inkyx profile picture
I'm a little late in posting as it's way after the two day window of comments, but could SA flash me notices on articles where I have just "liked" a comment. It flashes me when I comment, but there are many articles I read, "like", but don't comment that are hard to reference afterwards.

There seems to be no other way to suggest to editors of SA than through these posts on articles they write.

Thanks for SA and all the good work that comes out of your site. I'm really surprised that WSJ published such a positive article on SA. It was after cancelling my subscription to the Journal and focusing on SA that I really started making serious money. I like reading the Journal, but it is mostly (99.9%) irrelevant to making money.
Sumflow profile picture

inkyx :> I'm a little late in posting as it's way after the two day window of comments<
What two day window of comments, is this some rule that you made up. If you have something to contribute say it. If you think you have evidence that contradicts the theory of time and space lets hear it.
You concept of time may be in error. When I search in Google I do not see this two day window that constricts your world. If something is in error, have your say and correct it.
There is mystery and magic everywhere.... But not here.
Petrarch profile picture
The comments, the article, the study
Classic SA. Classic

So experts suck? What a surprise.
The collective wisdom of the many is predictive? Wow. Please tell Jack Bogle.
The Efficient Market Hypothesis works? Don't tell the SA crowd. They don't like it.

The period of the study is very very short.
Statistically speaking we are still talking about a random event. Basically luck

Still - I believe it should hold over time in the general sense. The specifics about stocks are tenuous. Will take a lot of time and many more observations to prove.

Based on this though you should rename the site Seeking Beta - since that is what it really demonstrates. Congratulations.

Classic SB.

I think the study generally indicates that the content of SA tends to be very useful to evaluate stocks and the economy, and that the contributors and commenters on SA have better than average knowledge and insight. Indeed, I love SA for many informative articles and opinions and the comment stream, which also often contains valuable additional information and opinions. I frequently recommend SA to friends and acquaintances as being clearly the best investment website. However, regarding actual stock selection or market timing, unless SA creates a continuously updated index for each stock and index referenced, which shows the status of an objective measurement of SA net opinion using the methodology of the subject study, I doubt if I could correctly infer the crowd opinion for a stock or index or have the time or correct methodology to use to measure the vast SA content to determine by myself the crowd net opinion. Apparently, the methodology of the study involved frequency of "positive" and "negative" words, "acrimoniousness," page views, page views to end, etc. It seems to me some of those things involve subjective judgment and are difficult to consistently quantify from one article to the next or one comment stream to the next, so it might be very troublesome to create an accurate index that can be conveniently measured.
Sumflow profile picture
>contributors and commenters on SA have better than average knowledge and insight.<
Does anyone dispute this?
advisor4 profile picture
The funny thing is this book Extraordinary Delusions and The Madness of Crowds first published in 1841 http://bit.ly/VNXcUH
is considered an investment classic

every great investor and every rules based trading system has one thing in common buying when nobody wants to own a stock and selling when everyone wants it

I know I know...it's different this time
Hmm. Wouldn't one have expected this?
We read allot about how sell side analysts on the street water down their sentiments since their firm may want to do ibank business with the target at some point. Also, they might be denied access to management if the comments were too negative. I would think that means less negative commentary. S A analysts would not suffer from this bias.
advisor4 profile picture
IMO anyone that actually reads the study would conclude that there is absolutely no profitable investment strategy that can be implemented based on the study. But in the interest of crowdsourcing I am open to being corrected.

Also I am not sure what 4 years of data can tell you.Do others think that is enough of a sample upon which to draw conclusions ?

crowdsourcers pitch in and help me out.
whidbey profile picture
The research is almost too good to be true or should I say accurate. The Wisdom of Crowds has always been suspect as only another expression of the law of large numbers. Crashes in markets must be some kind of an exception. Panics are the nature of crowds (they stop thinking critically and just react?) What affects behavior the most: "considered opinion" v. panic? Differences between those who are invested and those who are not invested has been shown to influence the prevailing general opinion of an investment. Why? lethargy? transactional costs?

Processing SA opinions is time consuming for investors. It is not best to stay invested in stock indexes showing strength (since the market rises about 70% of the time) and set stops at the RSI prevailing reading, or at some variance measure?

One can strain his arm while patting one's back.

I would tend to be cautious of this research.
Sumflow profile picture

whidbey :> The Wisdom of Crowds<
Do you think it helps investors to have access to other shareholders?
Davis said exactly what I think: "I find SA is a fantastic resource for gathering varying perspectives. This open forum nurtures organic idea generation and openly welcomes debate." It is a breath of fresh air of what a free market is. I use it to confirm and validate my own trend/sentiment tracking model.
Paul Leibowitz profile picture

Your reference to "What Strategy from the 1930' brokers don't want you to know" is appreciated, as it was a fun read.

For those who need the link, it's here: http://tinyurl.com/lw8...

Question. This is/was true for "specialists" (or, "market makers") but, today, how much of an impact do specialists still have?

For example, trades on NASDAQ are automated. It's the New York Stock Exchange that uses designated specialists.

The Nasdaq is younger than the New York Stock Exchange but actually much bigger in terms of daily dollar volume, number of issues listed and total market capitalization, or price multiplied by number of shares.

The Nasdaq does not operate from a single physical location or a trading floor where specialists meet, as does the New York Stock Exchange. The Nasdaq is a sort of virtual stock market, a vast electronic agglomeration of stocks, exchange traded funds, warrants, preferred shares and other investments, each with an identifying ticker symbol and each quoted and traded through a network that automatically matches buyers with sellers. Nasdaq actually is made up of three exchanges. http://bit.ly/1g9mszV

When I think about it, I tell myself it's likely still true for companies listed on the NYSE but only to the extent that other boards process fewer trades.

The NYSE Euronext contracted as market structure changed, and became more electronic.

More on that, here: http://tinyurl.com/crg...
inkyx profile picture
Thanks on the link, Paul. It's not unlike the long/short comments on some stocks here on SA.
Disagree with this article? Submit your own. To report a factual error in this article, . Your feedback matters to us!
To ensure this doesn’t happen in the future, please enable Javascript and cookies in your browser.
Is this happening to you frequently? Please report it on our feedback forum.
If you have an ad-blocker enabled you may be blocked from proceeding. Please disable your ad-blocker and refresh.