Morgan Slade of CloudQuant shows how New Constructs' core earnings data can generate alpha in long-short portfolios. The signal works strongest in small caps and extreme quantiles, and a 25-day execution lag doesn't kill returns. Common risk factors don't explain the outperformance.
Chapters
Morgan, good to see you. Thanks for being with us. What Morgan and his team have brought to the table is a very succinct paper that shows exactly how you use our data to generate alpha. We've loved being a partner with him. We're happy to have him here today to talk about how to get idiosyncratic alpha out of our data.
David, thank you for the intro. We've been working with David for a number of years and we've had a lot of success using his datasets to identify interesting investment opportunities. We're excited to go over some of our research results. A little bit about CloudQuant. We're a data research and technology provider. We work with lots of different datasets for institutional investors. We try and identify how to use them to make money, how to find edge. We're gonna look at some scenarios where we look at Companies that have at least $100 million in market cap and they trade $1 million a day at least. Let's dive into it. We did a number of experiments, which I'll skip over here. What we're really trying to do is start out with something simple and build up to something more complex that might be what you would use if you're trying to come up with ideas to run your own portfolio or manage your own portfolio. So if we start with just the total earnings distortion, we actually do get a little bit of performance here. Up and to the right is what we want. For the all-cap, we've got a little bit here, but we're going to try and improve that by changing a few simple things. And then the small caps seem to have a little bit more edge. So this is looking at, at the top 10% going long and the bottom 10% going short. Taking the difference between those returns is what generates this curve. All right. So it's known as a long-short portfolio. And the small cap outperforms. If we say, well, let's be a little more specific and granular on the small cap names, let's only trade the top 2% and bottom 2% of the score that we came up with. CloudQuant did come up with a signal based on New Constructs underlying data, and these are the backtests from that signal. So we're looking at the top 2%, bottom 2%. We see that we get even higher performance when we look at the outliers, the companies that are really building up or really drawing down their reserves and their earnings. Cushion that they're either building up or drawing down. These are people who are doing extraordinary things to their earnings. What if we look at sectors where we don't really think this type of analysis will work? And so what if we eliminate those? Well, we can improve the results even a little bit more. That's the limited sectors in blue. Let's look at what if we lag our signals by 25 days? So this is a true test of like, is there something fundamental here? We lag the signals that we came up with from new constructs data 25 days, and that is the bottom one on the legend here. And you see how the performance is equivalent to if I traded it almost immediately after earnings came out. That's a really impressive thing, and it tells us that there's a huge amount of capacity for large investors, but there's also a lot of time for individual investors to do their research and think about these signals before they put the trades on. And so you can really add additional information above and beyond just what we came up with, with our signals, to the analysis as a mosaic to come up with a better answer. Across the board, you see this up into the right pattern, uh, that tells us that this is a very robust, very simple approach with a single factor. A lot of people are looking at multiple factors at once. They're combining things together. In this particular case, we didn't even need to do that. We just took new constructs, did our magic, turned it into a signal. We ended up with something that seems pretty rational and to make sense. Let's look at recent performance. This thing's been on fire lately. I mean, look at this. We're seeing really great performance across the board. On the left, we have the very simple approach, and then further to the right, we did more and more detailed filtering. We're getting returns up to 70%. In 2023, that's truly impressive and unique and definitely worth a look. We did do some more backtesting. This is just a very simple backtest using an equal weight performance or portfolio. We generate these standard reports with our backtesting platform. So that's another look of another strategy we came up with. Is this unique? We do a risk analysis. We try and decompose the returns to see if they're explained by any other factor we could. Invested in the market. And this result here is really telling us that there is nothing that we really could buy to replicate these returns. Nothing available that's tradable that we are aware of that you could use to actually get exposure to this kind of return. So if you want these kind of returns, you would have to buy the data. And that's what it boils down to. So here's a couple of takeaways. We've got a great data set here that's helping us look at and focus with our eyes on the prize and the core earnings. We've got coverage for the top 2,500 equities, 10 years of history. It's all point in time, and it's now being processed within 24 hours of the earnings report. We did show it was most effective in small caps, which is good because there's not a lot of information out there on small caps. It's really hard to get good research for small caps cuz nobody covers 'em. It's not as much investment banking going on. And, and then extreme quantiles have the highest edge. We showed that if you get further and further towards the tail of the distribution of the scores that we came up with, that we can get much, much higher edge. It had a really tiny exposure to the traditional risk factors and smart beta ETFs. We really couldn't find much of a weighting on those, which shows that this is very novel and should be part of your decision-making process. Most effective in consumer staples, materials, industrials, and consumer discretionary. And we even showed that you could wait around for 25 days before you even put the trade on and still get the same results as if you traded right after the earnings reports come out. That's a summary of what we found. I do see a question about the signal lag. I can address that real quick.
Mm-hmm.
We just paused making any investment decisions for 25 days. We got the information we needed. We knew we were going to go long or short, but we waited 25 days to actually execute the trade. And we showed that you could wait that long and still get the same great results, which gives you plenty of time to incorporate other information from other sources to even be more selective or more informed.
Morgan, that's always fascinated me. And can you go a little deeper on what does that mean? Does that mean that This footnotes data, nobody's even looking at it, or it's not being impounded in the market for at least 25 days?
Yeah, that's what it means. It means there might be some short-term funds that are looking at this and maybe they're executing their trades and getting in and out quickly, but there seems to be a big gap and opportunity in the market because somewhere in between a day or 2 and 25 days, it's going back to where it started and you're getting another bite at the apple, which is really unusual in the world. I'm not going to get into the equations, but the Sortino ratio here is higher than the Sharpe ratio. That tells me my winners when I trade with the new constructs data are bigger than my losers. And that's very unusual unless you have a circumstance where nobody's paying attention to it and you don't have too many people chasing. It's not crowded. So you have a great opportunity here to get into something that's not crowded from our perspective quantitatively. And you have the ability to come up with your own strategy, come up with your own ideas without feeling like you're the last person to the party.
Thank you, Morgan. That was great. It's good for the integrity of the capital markets for people to package this stuff up and help people take advantage of increasing number of insights to make the market more efficient.
