The following is a guest article written by Estimize Founder and CEO Leigh Drogen.
Here’s why we did it: real problems cannot be solved with government regulation, only market solutions. The fact of the matter is that our government and its associated regulatory regimes have been so thoroughly captured by the industries they regulate (read financial sector) that it is nearly impossible to effectively police them.
For example, after the tech bubble burst in 2000, the government faked an attempt at limiting the ability for sell side analysts to manipulate stocks by publishing Buy/Sell/Hold recommendations or price targets to the public while telling their largest institutional clients to get the hell out (yeah they knew it was a bubble). In support of investment banking fees from the same companies they were researching and rating, they lied their butts off.
But when the government wrote new rules around the conduct of analysts — Chinese walls between research and banking, Reg FD, etc. — they knew that if the rules were actually ever enforced, it would put the equity trading side of the banks and brokerages out of business.
And herein lies the issue. You can’t solve true problems with government regulation, only market solutions, because market solutions don’t give a hoot about incumbent businesses.
Five years ago I started a company called Estimize to solve a piece of this problem. Instead of relying on the biased sell side analysts for earnings and revenue estimates on a quarterly basis, we built a platform to collect them from the crowd (hedge fund analysts, corporate finance professionals, independent traders, students, etc.).
We now have over 20,000 contributors to our data set, and Estimize Consensus is more accurate than consensus estimates 74% of the time. Half a dozen published academic papers (including one by the Deutsche Bank quant team) prove that the Estimize Consensus is the truest expectation for a company’s upcoming earnings report. We’ve signed deals with CNBC and other major financial platforms to have our estimates used as the consensus, and sell our data to institutional quantitative and discretionary buy side clients all over the world.
We solved the regulatory problem by producing a better data set, by crowdsourcing it, not by attempting to change the behavior of a group of people that had no economic incentive to do so. We’re reducing the influence of sell side analysts on the market as our Estimize Consensus is replacing it in trading models and investors’ decision making.
Along the way we’ve been asked over and over again to do the same for Buy/Sell/Hold recommendations, to remove the influence of the sell side there as well. And we’ve said no because Buy/Sell/Hold is an awful system that was never meant to produce any real information. It was designed as marketing so that sell side analysts could push the stocks that investment bankers were raising capital for. Or worse yet, the stocks their institutional clients were invested in!
What does “Buy” even mean? Buy how much? Buy when? What if the stock drops, should I sell it? When should I take profits? What does sell mean? Should I short the stock? Sell when? Keep selling? What should I do when you move a sell to a hold? The whole system is rotten from the start.
There is zero efficacy to this ratings system, and that has been confirmed in the data over the past 40 years. There is no persistence of accuracy among analysts across time, because the system itself is random and useless. Not only this, 80% of stocks are rated buy! Even if you assume that buy doesn’t really mean buy, it means “overweight” relative to the weighting in its index, you still suffer from the Lake Wobegon effect. 80% of things can’t be rated overweight, it’s statistically impossible! (Shakes head in disgust).
We can do better. So we did. The best way to judge which stocks an analyst expects to outperform or underperform is to ask the question correctly. We developed a new crowdsourcing system, as a competition that looks very similar in nature to Fanduel (then we patented it).
Each week you can find various competitions on Forcerank that include 10 stocks from sectors such as Social Media, E-Commerce, Apparel, or Biotech. There are also competitions for a list of 10 Commodities, US Equity Sector ETFs, Global Index ETFs, and Forex ETFs. The only thing we ask participants to do is to rank the 10 stocks in each contest in order of how they believe the stocks will perform over the coming week (by % change). If you think Nike is going to outperform Under Armour and the other 8 Apparel stocks this week, you place it #1 in your Forcerank.
Each contest has an entry fee ($5, $25, $100, etc.) and a prize pool (entry fee times number of competitors). We score each competitor at the end of the week based on how accurately they forecastedwhat actually happened in the market, and pay real cash to the top half of the leaderboard (they double their entry fee).
Here’s the thesis. Investors in general are great at stock selection, but terrible at market timing, risk management, and position sizing. The latter three are not intuitive things, are extremely difficult to learn, and go against so many embedded heuristics based on our evolutionary biology (fear, greed, etc.). But investors really are good at equity selection.
If I asked you how Apple was going to trade this week or month, you’d probably shrug your shoulders and say I have no clue. But if I asked you whether you think Apple will outperform Google, you’ll probably have a decent opinion. Why? Because I’m not asking you to opine on the direction Apple’s stock will move over that short period of time, I’m only asking the relative question of whether it’ll simply do better than Google, a much easier question.
By asking this question of 10 stocks with homogeneous characteristics from thousands of people and giving them a monetary (and ego) incentive to be accurate, Forcerank collects the best possible sample of what stocks investors believe will outperform the rest.
There are many possibilities you can do with the data. You could build a consensus rankings list of all stocks in the S&P 500 each week and look at how stocks move up or down that list. You could build an ETF that was not market cap weighted, but sentiment weighted. If the data was accurate (my bet is that it will be, just like Estimize) you could go long the top five stocks and short the bottom five stocks each week (beta neutral) and generate a ton of alpha.
Or you could take the top 20 percentile of stocks based on their consensus ranking and call those your “Overweight”, the bottom 20 percentile and call them underweight. And to close the circle from the beginning of this piece, you would remove the influence of sell side analyst ratings, solving the regulatory problem with a market solution.
It will take time to prove out our hypothesis regarding the predictive nature and representativeness of the data (6-12 months). It will take time to grow the number of contests and list of stocks being ranked. Some people who don’t understand securities or online gaming laws will say what we’re doing is illegal (it’s not in any way). But I am supremely confident that at scale Forcerank will replace Buy/Sell/Hold.