Hello, welcome back to Papers of Backtest podcast. Today we dive into another algo trading research paper. We are, and today we're zeroing in on a study called Analyst Days, Stock Prices, and Firm Performance. It's by Diwu and Amir Yarin. Right, from October 2018. Yeah. And it digs into these things called analyst days. What exactly are those, based on the paper? Well, they're basically events hosted by a company. They invite equity analysts, institutional investors, you know, the big players.
Okay. And they share information, updates, plans, that sort of thing. But the key is because of regulation fair disclosure or a reg FD. Oh, reg FD. Exactly. Anything significant they disclose has to be released publicly at the same time. So no selective disclosure. Got it. So it's a burst of public information. And the paper says these have become more popular since around 2004. Yeah, they've seen a rise in usage. Firms seem to find them valuable.
Okay. So our mission today then is to unpack the trading rules and the backtest results the researchers found. Basically, how does the market react after these analyst days? Exactly. What happens to the stock price? And could you potentially treat on that information? And they had a good amount of data to work with, I gather. Oh, yeah. Quite comprehensive. They looked at 3,890 analyst day events. Wow. U.S. listed firms between 2004 and 2015.
And they linked all that event data with stock prices, accounting data. Really thorough. And they even collected the text from the announcement so they could see what was being discussed. Okay, great foundation. Yeah. Let's get straight to the core findings then. What happened to stock prices after these analyst days? Were there any signals for traders? The main takeaway, firms holding these events saw, on average, significantly higher abnormal returns afterwards.
Abnormal returns meaning they beat the market, right? Precisely. They performed better than you'd expect just based on general market trends or their typical risk profile. Can we put a number on that? How much better? Well, one of the first back tests they ran was pretty simple. Buy the stock on the analyst day, hold it for 20 trading days. Okay. Straightforward enough. That strategy earned a market adjusted return of 1.6%. So 1.6% above the market over those 20 days on average.
1.6% in about a month. That's, whoa, that's certainly noticeable. Did they look at it from a portfolio perspective too? They did. They simulated building a portfolio of these stocks around their analyst days, a calendar time approach. That portfolio showed a one-month four-factor alpha of 1.8%. Okay, alpha. So that's return adjusted for market risk, size, value, momentum. Yeah. The usual factors. Exactly.
It suggests there was some genuine excess return there, not just compensation for standard risk factors. Right. And 1.8% alpha per month is quite substantial. And was this just a quick blip? Or did it last? That's the interesting part. It wasn't just a flash in the pan. The abnormal returns stayed significantly positive for up to six months. Six months.
Wow. Yeah. And they didn't find evidence that this was because the stock suddenly got riskier or that it was just some kind of temporary bounce back, you know, mean reversion. So no obvious increase in risk and the gains held for a while. Right. Now, eventually, after about 250 trading days, roughly a year, the returns weren't statistically significant anymore. But they didn't see a big downward trend either. It just sort of faded. That implies the market is, well, maybe a bit slow to catch on.
It takes months to fully price in the information. That's exactly what the researchers suggest. They talk about market underreaction. The cumulative abnormal returns, the SARs, kept trending upwards for the first three months. And trading volume, did that change? It did. Volume was higher in the first 20 trading days post-event, which, again, fits that picture. of investors gradually digesting and reacting to the news. Okay, so the action seems to be after the event.
What about trying to get in before? Did the market anticipate these events? Apparently not, or at least not consistently. The abnormal returns in the month leading up to the analyst day weren't significantly different from zero. So just knowing an analyst day was scheduled didn't really help you from a trading perspective beforehand? Doesn't look like it based on their data. The value seems to be in the information revealed on the day itself.
which means the opportunity window opens right around the event, or maybe even slightly after. That seems to be the case. They found you could realize the average 1.6% monthly abnormal return by buying on the analyst day. But even buying up to a week later still seemed profitable. The SARA-RR, the cumulative return from day T plus 5 to day T plus 20 was still around 1%. So you didn't necessarily need lightning fast reactions. There was some time. A little bit of time, yes.
The under reaction created a slightly longer window. And they check this using different ways to measure returns, just to be sure. Robustness checks. Absolutely. They used Carhartt four-factor model returns, buy and hold abnormal returns, BHRs, looked at various windows like T20 to T plus 60. And the results held up. Largely, yes. Similar positive trends across the board.
They did note maybe a tiny hint of mean reversion in one BHR measure between 30 and 60 days, but the standard errors were pretty high, so not a strong signal. Okay, so the short to medium term picture looks... It's pretty solid. What about even longer term using that calendar portfolio approach? Yeah, looking out further, up to 60 months, it confirmed those strong positive abnormal returns on the day and for up to six months after. Right.
They calculated annualized alphas reaching as high as 8% in those first two months post-event. 8% annualized alpha. That's significant. It is. And again, they checked the risk. The post-event betas weren't significantly different from pre-event betas. So it wasn't just taking on more risk. And beyond six months. The alpha tended to drift back towards zero, especially when you held for more than 250 trading days. Oh, and an interesting side note.
These stocks tended to have negative commovement with momentum stocks. Meaning when momentum strategies were hot, these analyst day stocks maybe didn't do as well and vice versa. Kind of, yeah. An interesting portfolio diversification aspect, perhaps. Let's circle back to the trading volume patterns. Yeah. You said it picked up after the event. Right. Not much happening before, maybe a slight pickup a day or two prior. But then, bang on the event day, volume jumps about 7% above normal.
The day after. Even higher, about 9% above normal. And this elevated volume stuck around for about 20 trading days, averaging maybe 3.4% higher overall during that period. That really paints a picture of information slowly diffusing and triggering trades, doesn't it? Supports the underreaction idea. It really does. Increased interest. More trading as the news sinks in. Now, did the type of information shared at these analyst days matter? Did they all have the same effect? Ah, good question.
They actually looked into that using the text analysis. They use a technique, latent Dirichlet allocation, LDA. Okay, fancy term for grouping topics. Basically, yeah. They classified the announcements into four main types. Product announcements, reviews of past results, discussions about strategy, and talks about technology and markets. And did the market react differently? Very much so. The big winners were analyst days focused on new products or discussing technology and markets.
How much better were they? Well, product-related days showed suckars up to 8% in the two months following the event. That's huge. 8%? Yeah. Compared to? Compared to events just reviewing past financial results. Those didn't generate significantly positive returns at all. Interesting. So forward-looking news, like products and markets, had a much bigger impact than just rehashing old numbers. Exactly. Strategy-related events also showed a positive drift, maybe 1.6% CAR in the first 20 days.
And the technology-market-focused ones had cars up to 4% over 60 days. But product news was the clear standout. That makes intuitive sense, I suppose. New products often mean new growth potential. Precisely. It's about future prospects. The paper also touched briefly on actual firm performance and analyst behavior after these events, right? Not just stock price? It did, yeah.
Just briefly, they found firms holding these days tended to show significantly higher revenue growth, better earnings per share, and higher dividend yields for up to two years afterwards. So the events often preceded actual improvements in the business itself? It seems correlated, yes. And analyst behavior changed too. After an analyst day, you typically saw more analysts covering the stock. Increased coverage.
Higher earnings estimates, higher price targets, and importantly, less disagreement among analysts, lower forecast dispersion. So the events seem to provide clarity and maybe some optimism to the analyst community. That seems to be the effect. Yes, more coverage, more agreement, more positive outlooks. Okay, so let's wrap this up for someone listening who might be interested in the trading angle. What's the main takeaway?
The core takeaway is that this study reveals significantly positive abnormal returns following analyst days. There seems to be a market underreaction. Meaning there's potentially an edge there. Potentially, yes. especially for analyst days focused on those forward-looking topics like product announcements or market updates. The opportunity might persist for days, weeks, even a few months after the event. But and this is the crucial caveat. Absolutely.
This is based on historical averages from 2004 to 2015. Markets change. Any specific strategy needs serious individual research, backtesting on current data. You know the drill. Past performance is no guarantee, et cetera, et cetera. But it definitely highlights analyst days as events. worth paying attention to. Definitely. It's a specific type of information event that, historically at least, seems to have had a predictable, albeit delayed, market impact. Fascinating stuff.
A clear example of potential market inefficiency, even if it's temporary. Indeed. Makes you wonder what other scheduled information events might have similar underappreciated effects. Thank you for tuning in to Papers with Backtest podcast. We hope today's episode gave you useful insights. Join us next time as we break down more research. And for more papers and backtests, find us at https.paperswithbagtest.com. Happy trading!