Hello, welcome back to Papers with Backtest podcast. Today we dive into another algo trading research paper. Today we're digging into shared analyst coverage unifying momentum spillover effects. It's by Ali and Hirschleifer from 2019. Right. And this one seems to touch on quite a few areas looking at those GEL codes, G12 asset pricing, G14 market efficiency. Exactly. And, you know, G24 financial institutions.
It really revolves around momentum spillovers, how one stock's momentum might predict another's. They call it CF momentum. Yeah, CF for connected firm momentum. The whole idea is linking firms, but specifically through analyst co-coverage. Okay, let's unpack that. So the core idea, if I'm getting this right, is that all these different ways momentum seems to spill over, like between industries or firms in the same region, supply chains. Just here in tech, yeah, all of those.
The paper suggests maybe they aren't separate things after all. but they all trace back to shared analyst coverage. That's the hypothesis. The analyst covering the same set of firms is sort of the underlying connection driving these spillovers we see. Interesting. So our mission for this deep dive then is to really get into the weeds of this CF momentum strategy. Yeah. We need to understand the actual trading rules. Right. How they built it step by step and crucially what the back tests showed.
And how it stacks up against the other momentum effects we already know about. It sounds like it could be a, well, simpler way to think about. cross-asset momentum, maybe. Potentially, yeah. A more unified view instead of, you know, a whole zoo of different effects. Okay, so first things first, defining connected firms. The key is this analyst co-coverage, right? How exactly do they define that? Yep, that's central.
Two stocks are connected if there's at least one analyst who covered both of them. And covered means? It means the analyst issued at least one earnings forecast, like for the current or next fiscal year, sometime in the past 12 months. They used IBES data from late 83 to late 2015. What's the significance of using this specific definition? Why not just use industry groups? Well, the authors argue it's more granular, more firm specific.
It's based on what analysts actually do, which firms they choose to follow together. It might capture subtler fundamental links than just being in the same broad sector. Okay, that makes sense. So once you know which germs are connected to which, How do you calculate this CF portfolio return for a stock? Right, the CFRET. For any given stock, say Stocki, its CFRET is basically a weighted average return of all the stocks it's connected to, all the stock Js. Weighted average?
How are the weights determined? It's based on the number of analysts covering both Stocki and Stockj. So if a connected Stockj shares more analysts with Stocki, it gets a higher weight in that average return calculation. Ah, okay. The idea being that more shared analysts signifies a stronger connection. Precisely. And just for context, they found the average stock was connected to about 86 other firms.
86. Wow. Yeah. And they also noted, which isn't too surprising, that analysts tend to cover larger stocks. So these connected firms often had higher market caps. OK, so we have this CFRAT number for every stock reflecting the recent performance of its analyst linked peers. How do you trade on that? What's the strategy? It's a pretty standard momentum approach in structure.
At the end of every month, you take all your stocks and you sort them into quintiles, five groups based purely on their CFRET over the past one month. So rank them by how well their connected firms did last month. Exactly. Then you form a long short portfolio. You go along the top quintile, the stocks whose connected firms did best, and you short the bottom quintile, the ones whose connected firms did worst. And rebalance monthly. Rebalance monthly, yes. All right.
The crucial part, the back test. What happened when they ran this in the U.S. market? Did it make money? Oh, yeah. The results were pretty compelling. They looked at the alphas, the excess returns after accounting for the usual suspects, market risk, size, value, momentum, the FAMA French factor. Four-factor model. Right. And even adding a fifth factor for short-term reversal, the strategy still showed significant positive alphas. How significant are we talking? What were the numbers like?
Well, for the valuated portfolio, the five-factor alpha was 1.19%. per month. That's substantial. The T-stat was way up there, 6.71. 1.19% per month. That's huge. It is. And the equal weighted version was even stronger, 2.1% per month alpha, T-stat of almost 12. Wow. And did the profits show up on both the long and short sides? Yes. They mentioned it was profitable on both legs of the trade, which is always a good sign. And there was a clear monotonic relationship.
The higher the CFRET quintile, the better the future stock performance. Did the effect last? Or was it just a one month blip? That's another key, finding it persisted. The positive returns continued over the next 11 months after forming the portfolio. So it wasn't just immediate mean reversion or something? No, exactly. Over a full 12 months, the cumulative return for the long short strategy was 3.21% value weighted and 6.68% equal weighted.
This persistence really points towards market under reaction to this information. The market is slow to price in the implications of how connected firms are doing. That seems to be the story, yes. And they noted that if they used deciles, 10 groups instead of five, the results were even more pronounced. Okay, so CF momentum looks strong on its own. But the paper's big claim is that it unifies other effects. How did they test that? How does it compare?
Right, this is where it gets really interesting. They did spanning tests. They looked at seven other known cross-asset momentum anomalies. Like industry momentum. Yep, industry, geographic momentum, customer momentum. supplier industry momentum, even links between single and multi-segment firms, and technology momentum. Quite a list. OK, so they built factors for all of these. They did, using similar methods ranking based on linked firms' past returns, forming long-short portfolios.
And then, did CF momentum get explained away by these others, or did it do the explaining? It was definitely the latter. First off, the CF momentum factor itself was highly profitable, even against the five-factor model, 1.68% alpha per month. But the crucial part When they added the CF momentum factor to the regressions, trying to explain the returns of those other seven momentum factors, their alphas disappeared. Pretty much.
They became small, statistically insignificant, sometimes even negative. It strongly suggested that CF momentum was capturing the same underlying phenomenon. It subsumed them. While the reverse wasn't true, the other factors couldn't explain CF momentum. Correct. CF momentum remained highly profitable even when controlling for those other factors. It seems to be capturing something more fundamental that drives these other effects. It's like finding the common denominator.
That's a good way to put it. They also did cross-sectional regressions, right? Looking at predicting individual stock returns. Yes, Fama-McBeth regressions. And the story was consistent. Past one month CF return was a strong predictor of next month's stock return. And adding it to the model weakened the other momentum variables. Significantly.
Especially for larger stocks, things like industry momentum, geographic momentum, customer momentum, lost their predictive power once you accounted for the stock's CF return. Okay, so strong results in the U.S. Did this travel? What about international markets? Yes, they tested it internationally too. Found strong CF momentum pretty much across the board in developed markets. Really? Same pattern?
Yeah, significant alphas in 10 out of the 11 developed countries they looked at, even after controlling for local industry momentum. And And interestingly, just like in the U.S., adding CF momentum often made the industry momentum factor insignificant in those countries. So CF momentum seems to dominate industry momentum internationally as well. It appears so. They even showed a profitable global ex-U.S. CF momentum strategy.
Were the results sensitive to how they define things like industries or the time period? They checked that, used different industry definitions, different geographic definitions like state level and the results held up. They also split their sample period in half and CF momentum was strong in both halves. Seems quite robust. Did they compare it to any other similar ideas like maybe peer momentum? Yes, they did compare it to a measure from Israelson's 2016 paper on peer momentum.
The CF momentum strategy still generated significant alpha even when benchmarked against that Israelson measure. In fact, in their analysis, the Israelson measure itself wasn't showing significant returns. So again, see if momentum seemed to subsume it. Okay, let's talk practicalities. Trading costs. Monthly rebalancing sounds like it could generate high turnover. It does, especially for the one-month strategy.
They acknowledged the high turnover, but they calculated the break-even transaction costs. Meaning how high costs could be before wiping out the profit. Exactly. And the levels suggested that, well, for large institutions, large arbitragers with low trading costs, it could still be profitable net of fees. What about the longer-term effect? You mentioned And it persisted for 12 months. Right.
And they pointed out that if you implemented a strategy based on, say, 12-month CF momentum instead of one month, the turnover would be much lower. And potentially similar net returns after costs. Potentially, yes. Lower gross alpha maybe, but much lower transaction cost drag. So wrapping this up, the big picture seems to be that this shared analyst coverage idea isn't just another factor, but potentially a unifying explanation. for a lot of momentum spillover effects we see.
That really is the core message. Shared analyst coverage identifies this potent CF momentum effect that seems to drive, or at least absorb, many previously documented cross-asset momentum anomalies. And the back tests show it's generated significant persistent returns. And for you, the listener, trying to get a handle on all these different market effects, this research suggests a potentially more parsimonious view.
Focusing on analyst connections might give you a simpler way to think about these complex relationships. Right. Instead of tracking a dozen different types of momentum spillover, maybe focusing on this analyst linkage gives you a more fundamental signal. The real aha moment here is just how powerful that seemingly simple metric, which analysts cover which stocks turns out to be, that it could explain this whole zoo of momentum effects is, well, it's quite surprising. It definitely is.
And it leaves you with something to think about. How could you use this idea of shared analyst coverage in your own analysis? Could looking at these connections offer a more holistic view of momentum? And does that apparent underreaction the paper finds signal real opportunities? Definitely food for thought. 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.paperswithbacktest.com. Happy trading.