Hedge-fund managers that do the most research will post the best returns
cnbc.com
cnbc.com
With those two simple adjustments, I saw that some big banks were in the hole by (combined) tens of billions of dollars.
The market prices for these banks made it clear that major investors either didn't read or didn't understand the data in the public filings.
Fun fact 1: My bank WaMu which was itself was maybe $0-$10B in the red so I immediately withdrew $2K in panic. I found that money hidden in my filing cabinet about 5 years ago.
Fun fact 2: What was even more crazy is that a famous private equity firm (I think TPG) had just dumped in billions of dollars into WaMu but that still didn't fill the hole (and why would such a big firm be less sophisticated than a nobody like me?)
Fun fact 3: FDIC had ~$50B at the time, which maybe wouldn't cover the losses at the pessimistic end of my estimates.
Fun fact 4: In accounting classes, I identified 3 very shady areas of accounting: options, off-balance sheet entities, and pensions. I believe the first two have since been fixed but I think pension accounting is still very shady so... beware.
Disclaimer: I studied finance and accounting in school and read 10-Ks in my spare time in the early 2000s so I have some knowledge of accounting shenanigans. I also had some experience in real estate. But that's about all the expertise it took.
(Several edits made to improve readability.)
Also, an accounting loss isn't always a "real" loss. For example, with the recent tax overhaul, some companies booked unexpected losses because they had significant carry forward losses. Those were reduced in value because the corporate tax rate went down. But essentially nothing changed -- yes the tax base and book value of these losses were adjusted, but no cash was lost.
They can also push revenue or profits back or pull it forward. It really is a jungle out there.
I do know that 1) a payment system with transactions that take more than a few seconds and cost more than a few pennies is not a good payment system for most things. 2) I've never heard anyone talking about using crypto for anything but speculation or paying for ransomware. Even on the internet, I've mostly heard of speculation, ransomware prostitution and buying drugs. 3) Regulators don't like payment systems for ransomware, prostitution and buying drugs. 4) Assets whose prices are based on speculation eventually fall.
I could be missing something. As I said, I don't know much about crypto-currencies. Please nobody bet on my comment alone. Also, timing collapses is difficult. Bubbles can last for more than a decade.
Since the GFC a lot of macro bets in both US and EU have been bets on political will and central bank actions. In late 2010 Bank of America was technically insolvent (based on analysis similar to yours), but the Fed went to work and backstopped the market in a multitude of ways and the valuation soared. In 2012 I would have bet the farm that Greece was headed into bankruptcy, but the Greek people didn't revolt (as much as anticipated) and so the ECB effectively nationalized most Greek debt across the rest of the Eurozone. I'm no expert but I try to apply these lessons to all new modern crisis I hear about. Underfunded pensions will destroy certain states, or social security will be broke and unable to pay out benefits? Sure; we'll see.
What empirical evidence exists to support this belief?
That's not the evidence requested.
What's done is done. It'll prove out either way in the next few years.
Bubbles existed before Keynesian policies.
The market prices for these banks made it clear that major investors either didn't read or didn't understand the data in the public filings.
That's impressive work, but I interpret that a little differently. There was definitely irrational exuberance going on, but plenty of major investors understood what was happening well before 2008. The reason the valuations were still out of whack is because having the correct data and the correct analysis isn't enough, you have to correctly forecast how the market will react. So there was an unfortunate feedback loop: many investors savvy enough to see the problem were not betting against it because there are far easier and more consistent ways of trading profitably.
This is why the most successful funds don't really try to replicate the process you're talking about. Their process is data and hypothesis agnostic. A lot of their work happens to align with the sort of analysis you've described here, but they don't start from the same place.
1) if you are a fund you just cant' sit out 3 years. Your fund will shut down as everyone will yank their money,. if its your own money then people will leave as you won't pay bonuses.
2) If you went short in 2006 then you wouldn't have survived until teh crash of late 2008.
> It's pretty easy to make way above average returns on the stock market.
This is just an absurd statement along the lines of its easy to build a billion dollar company just mimic what all the other billion dollar companies do.
Sorry for being so negative to your comment, but come on. That statement can't be seriously defended in any resonable manner.
Do you want to spend time finding the right investors or do you want to spend time reading 10Ks?
The fact that funds do worse than indexed etfs says more about fund incentives and investor savviness than it does about the effectiveness of this strategy.
The one distinguishing factor is that they have terrific brand value because of Musk, and this could mean that consumers would want their cars despite all the troubles they have. I think this unusual level of brand value is the confounding factor that muddies predictions and is unquantifiable. Brand value makes consumers make non-rational decisions. If it was any other car company with Tesla's numbers you'd be crazy to bet on them succeeding.
People who hear that say, Lexus are doing badly and have severe problems with their cars are going to go buy an Audi. People who want a Tesla so far are just waiting for a Tesla.
If you're buying something where a large part of the value is in the image/brand it's usually not fashionable to complain about it.
When a S10 breaks it's GM's fault. When a Tacoma breaks the owner didn't treat it right.
When a Kia breaks it's unreliable. When a BMW breaks that's just part of owning a BMW.
Tesla has cultivated an image (or rabid bunch of fanboys, depending on your perspective) where you're expected to put up with shoddy build quality and less than prompt service as part of the ownership experience so Tesla gets a pass on those things.
I'm not sure what they're investing in and how much more they have to invest so I don't have an opinion.
It probably wasn't rational but I was really scared.
Buyout => What would it cost you to pay an insurer to take the liability off your hands (discount rate will basically be the replicating portfolio of government bonds).
Accounting => Depending on standard, the discount rate might just be the expected return on your portfolio
On a 30 year duration liability that could be a difference of over 200% between the different measures. Which is realistically right though?
A friend and I used to run a website that tracked activist short sellers and their campaigns, and published all that information as a nice centralized database basically.
Hedge funds were, by far, the most interested in this - which was surprising to us, our original target audiences were auditors and legal firms. My impression from this experience is that the more successful hedge funds are just mass-ingesting data from everywhere they can, and somehow trying to make sense of it.
It's totally an arms race too: Once one of them is ingesting data from a new source, everyone else has to start doing it as well lest they fall behind. It was significantly easier to convince new clients once we had a few clients already (very much like the FOMO people speak about when fundraising).
We actually did have quite a few people scraping our homepage for some reason (which was rarely, if ever, updated) but they were mostly other aggregators as well, not potential clients.
It's humorous to me that most people who end up selling data to hedge funds more or less "fall" into that industry by accident. Your story is exactly how it typically happens: you develop a new data-enabled product for one market, then a bunch of hedge funds find it and realize it's useful. Most companies pivot from their original market once they realize how lucrative selling data can be.
Yeah, it was a very sudden shift. We always knew they were potential customers, we just didn't realize how quickly they would take to the product (as opposed to legal firms which you can imagine are significantly more bureaucratic about such things).
It's ambiguous, no doubt, but actually pretty easy to discern in practice. There are a few big names that dominate the landscape, as well (e.g. Citron).
Neither result would be surprising, that is why this issue requires research.
Otherwise, the weaker postulate of EMH is sort of redundant. That's not to discredit Fama's work: we know far more now then we did back then, and it's legitimately useful as a theory of the relationship between market state and information asymmetry. But its utility has nothing to do with whether or not the market actually is efficient.
I'd guess that's the value you were providing for them.
I was mostly just surprised at how quick the process was. Hedge funds are much more afraid of missing out on something than they are of paying a monthly fee to someone.
That might sound a bit silly (what business isn't afraid of missing out on something?), but that's definitely NOT the case in many other enterprise sales discussions.
For others reading here and in case curious, the hedge fund I was working at wasn't interested in paying for access to tracking of activist short campaign data for the short trade ideas themselves. There is little interest in copying or jumping onto another's trade like that. The real interest the hedge fund wants to get alerts and track these campaigns was to keep an eye on what other hedge funds are doing - it was more about competitor and industry colleague research than securities research.
Characters like Ackman, Shkreli, etc all sell the idea that they know what they're doing, when the reality is it's simply gambling with other people's money with minimal risk. Luck will favor some, and others...not so much. The entire industry is essentially a collective delusion/sham.
- Warren Buffett
As one of my coworkers explained, there are investors of varying caliber wandering around and they often aren’t interested in your company. They’re getting the gestalt from the show floor by picking any brain that will give them the time of day.
Honest question: isn't that a contradiction in terms? Public information stands for "stuff that everyone is presumed to know", right? Like, economists using an assumption of "perfect knowledge of the market regarding public information" (or something like that, I'm obviously not an economist) as an approximation in their knowledge. But the fact that some managers do look this up, and some don't, means that this approximation does not hold. The premise of what is implied with "public knowledge" is undermined by the very thing being measured.
Assuming I understand this correctly, it looks to me like they essentially correct for the error caused by this assumption of perfect knowledge, which is the opposite of unexpected (but still a very neat finding!)
Most "public" information is actually woefully underutilized. The only real requirement for data to be public is that everyone could know it, not that everyone does know it.
Thank you (and Retric) for explaining, but this phrasing does not quite convince me: it does not explain how this diffusion is supposed to take place, and I for one cannot imagine a way of this knowledge to be distributed among the market as a whole without individuals knowing this information and freely sharing it together. Which again seems to be fundamentally undermined by what was measured.
I mean, yes, there are situations in which you can distribute knowledge among people where no individual can see the whole picture, but together they manage (for example, the knowledge required to turn crude oil into plastic requires knowledge among every step of production, diffused among individuals - nobody truly "knows" how to turn crude oil into plastic if you define that as knowing everything involved in the process). I do not see that kind of context apply here though.
In other words, every time a security is traded, its price is updated with a very small amount of new information. It's not important that any single individual has all the information, it's important that the market has all the information. That is the essential function of the market - price discovery. No single investor (or even a team) could hope to accurately encapsulate everything there is to know about a given security. The market "prices" new information into a security, which is precisely what it means for data to be diffused.
Stated another way, novel information about a security represents a potential price inefficiency. But that price inefficiency is essentially lost once the data has become public.
I'll be honest with you: when I read this my first response was "what does this even mean? What is "the market" and how does it "have information"?" I guess the sentences preceding it are supposed to explain the structure that the market has built up to automagically capture this information, but I don't quite follow.
I'm not criticising your comments, I'm very grateful that you are trying to explain how this works and the logic behind it, but so far it's black-box abstractions all the way down.
A current price aka price at last trade of 10$ and nobody willing to sell for less than 10$ and nobody willing to buy for more than 9.99$.
A new seller shows up, if they are willing to wait they might get more than 9.99$ or they can accept 9.99$.
Or a new buyer shows up they can buy for 10$ or wait for a lower price.
Now, without effort people can just look at the history of trades and see the price rise or fall. This is independent of whatever reason that causes more people to suddenly want to buy or sell, just the fact that people are buying or selling in it's self moves the price.
> In other words, every time a security is traded, its price is updated with a very small amount of new information.
The update happens after trading, so before that the price is by definition outdated, meaning it has that bit of "inefficiency" you spoke about. Hence "doing your homework" as a hedge-fund manager would in theory allow you to spot this inefficiency and make better trades.
This is not to say that most hedge funds are right to ignore any given set of meaningful data. It just means that they have to make decisions about which data to use and how much they'll prioritize it. Then they have to see if they have the requisite infrastructure to trade on that information before anyone else.
Filings are subject to change due to errors, updates, etc. It's cheaper, easier, and less error prone to just scrape and overwrite data than search for any updates.
I may have been heavyhanded in saying that the firms are rescraping the entire universe of filings on a daily basis, but I guess my point here is that these numbers can very easily be skewed and are generally a pretty poor indicator of actual data access.
Filings are subject to change but they are restated, not modified in place. That said, anything goes on the site.
The only real reason not to scrape everything on a continuous loop is that there is a rate limiter in place, so you may end up impeding your intake of new data by loading yourself down with old jobs to handle.
That said, many of these firms also ingest data from aggregated sources, so their research would activity be very difficult to accurately track via IP presence.
The reason for not having your own computers is because you can't afford them. A consideration for upstart funds, not a problem for RenTec.
Because you don't want anyone to have access to your code and data.
If you found out that Facebook is running on AWS, wouldn't you find that strange?
Therefore, A) companies that operate at a scale where the only upside disappears, and B) companies where the cloud saving outweights their expected liability cost, both will have their own datacenters. Facebook is clearly in category A, RenTech is clearly in category B. This is my understanding anyway.
1. https://www.sec.gov/edgar/searchedgar/accessing-edgar-data.h...
Interesting...
https://www.npr.org/sections/money/2016/03/04/469247400/epis...
1. Large shops employ a lot of fundamental analysts and larger funds only get large through outperformance.
2. Less on the fundamental side, funds regardless of size that have automated retrieval (and likely processing of public filings) probably have a higher than average disciplined investment process which leads to outperformance.
The more interesting story is that if you can get this data (MITM or some other means), you could front run a fund by figuring out who they’re researching.
Comparatively speaking, the SEC filings are just a blip next the rest of the data gathered by quantitative funds. Any conclusion that could be drawn from what's contained in this report would be hopelessly misleading.
"As discussed in the introduction, public information may be profitable if sophisticated investors like hedge funds are skilled information processors. Alternatively, private information may be more valuable when used in conjunction with public information." (pg. 19)
The paper goes on to show evidence suggesting that the predominant channel is the latter complementary private information mechanism.
It does seem counterintuitive that publicly-available information does indeed generate alpha, but the real valuable insight here is that hedge funds are able to generate alpha from using public data synergistically with private data.
Even more than that: the best funds can generate alpha by simply combining different sources of public data without necessarily using any nonpublic data.
But they measured which managers actually do and do not use it, so that assumption already does not hold.
Assuming I understand what you are saying correctly, only surprise would be that managers don't actually make efficient use of public information.
> By mapping hedge fund IP addresses to those accessing financial filings, the team identified information gathering by hedge funds such as Renaissance Technologies and AQR
What is interesting for me is that there is a possible correlation between people who mine more and more data - even from publicly available sources - and their returns.
Important to note that correlation is not causation
Also interesting to see that the SEC would give out IP address information of who accessed their filings.
BUMP Form 4 is insider trading disclosures per the original paper. That is even more interesting.
Turns out maybe the students aren't actually that good at knowing if they read the material sufficiently well, and you can get information about who will do better at the test just by looking at who looked at the class material the most.
The surprise is that the students are worse then we thought.
Does this actually reflect some experience of yours? It certainly contradicts the existing research on "test prep".
The holy grail of investing would be a quantitative model which automatically changes and consistently outperforms everyone else.
When most funds say they're "quantitative", what they really mean is that they use huge amounts of data to inform fundamentally manual trading strategies (this includes most of the places widely considered to be "top" firms). They develop trading algorithms, and those trading algorithms are often successful. But the algorithms are developed manually and then deployed. Their researchers and engineers actively seek out new sources of data and try to compete on novel sources of untapped information. But the reality of what happens is that they simply drown in the data. They can't clean it or process it nearly fast enough to maintain long term trading strategies, nor can they even begin to find a way to automate the trading strategy extraction. If you're working with hundreds of terabytes of data, you cannot selectively formulate hypotheses and test them. It's far too slow. You will find dramatically fewer novel insights than a fully automated process.
In other words, they're a step above traditional "fundamental" hedge funds, but they focus on the wrong problem (but not for lack of trying!). In contrast, the truly successful quant funds have automated the data processing and feature extraction pipeline end to end. The data is a pure abstraction to them. They don't bother with forming hypotheses and trying to find data to test them, they allow their algorithms to actively discover new correlations from the ground up. So many quantitative funds advertise how much data they work with, and how they have all these exotic sources of data at their disposal...but the data does not matter. The models for the data do not matter. The mathematics of efficiently processing that data are what matters.
As a result of their consistent profitability, most of the jobs you see listed for the really successful funds (if they have a website) are not "real" in the strictest sense of the word. You can apply to them, but they only keep active careers pages to attract the best researchers. Their only incentive to hire is to 1) keep someone who is actually exceptional from joining a competitor or 2) keep an academic researcher from re-discovering their work when they seems like they're getting close to it. This is why they primarily focus on quantitative PhDs in information theory, high energy physics and computational mathematics (especially information geometry).
To be completely frank, Renaissance is an outlier, but not just because of their returns. They're an outlier because of how public they are. Most of the funds with comparable returns not only don't take any outside investor capital, they only have 25 - 50 employees. They virtually never hire because they don't have to. If your work is fundamentally interesting, novel and applicable to what they're doing (even if you can't immediately see why), they will call you.
___________________________
1. Other more secretive (but equally successful) funds have done this, but they are much more under the radar.
The funds I'm talking about (including RenTec) take in as much unstructured data as they can possibly find, almost indiscriminately, and they tune their processing pipeline to the point that it requires neither manual classification nor munging. In most cases, a trading strategy is sufficiently multidimensional that any particular set of data can be completely public. Exclusive data is helpful, but not required. In many cases people become too dependent on exclusive data and lose sight of the methodology.
This is precisely what I mean: many people think that these firms differentiate based on the sources of data they use. They do not. They differentiate on their ability to automatically extract signals hiding in plain sight. Whether or not the data is public makes very little difference, because the signals come from tens of thousands of indicators combined together.
Are there any firms outside of finance using an analogous approach to data processing and signal construction?
[1] https://www.bloomberg.com/news/articles/2017-08-16/renaissan...
Second, you mention knowing the names of various comparable firms to RenTech. Can you name some of those here?
1. I'm friends with multiple people who used to work at Renaissance, and I've directly spoken with folks who are currently there (among other, similar firms).
2. I've read the research published by professors and post-docs before they were hired.
3. I have first hand experience developing forecasts for various market research firms and many hedge funds. I've seen first hand what the difference is between the firms that say they're quantitative and the ones that are quantitative.
Not to discourage you (and you probably already know this) but: since you've already worked in the industry, Renaissance is very unlikely to take you seriously as a candidate. I'm not sure if that's what you meant when you said you're interested, but I figured I'd put it out there.
In any case, contrary to popular belief it's possible to connect the dots on what firms like RenTec do. But just having a high level idea of how they work isn't nearly enough to replicate their success. Their success relies on a large number of interdisciplinary scientists cooperating with the support of an incredibly specialized infrastructure. Getting the cliff notes on how they achieve e.g. dimensionality reduction doesn't come close to cutting it.
Do you have any other keywords? “Learning from few examples” comes to mind...
But reputedly they neither allow external money into Medallion nor do they let employees compound their existing stake in the fund but pay out ginormous 20%-40% annual dividends exactly because the strategy is not scalable.
The paper explicitly addresses the point that large-scale systematic collection of public records may be indicative of the kind of fund that outperforms, rather than an indication that the public records add alpha in and of themselves - it's in the abstract
> The effect is not due to differences in fund type as the results hold within-fund.
I think you need a more convincing argument than just saying "There are a great number of additional factors at play" and "Trying to link public information to an increase in performance is far fetched".
This is how the SEC characterizes HFT firms.
1. Use of extraordinarily high speed and sophisticated programs for generating, routing, and executing orders.
2. Use of co-location services and individual data feeds offered by exchanges and others to minimize network and other latencies.
3. Very short time-frames for establishing and liquidating positions.
4. Submission of numerous orders that are cancelled shortly after submission.
5. Ending the trading day in as close to a flat position as possible (that is, not carrying significant, unhedged positions overnight).
From pages 4-5 https://www.sec.gov/marketstructure/research/hft_lit_review_...
> we are not high-frequency traders
https://www.aqr.com/Insights/Perspectives/High-Frequency-Der...
Also, they seem to hire a fair few ex-HF guys in senior positions (e.g. https://nypost.com/2015/07/31/aqrs-head-of-trading-on-paid-l...)