- Speaker #0
Hello, welcome back to Papers with Backtest podcast. Today, we dive into another algo trading research paper.
- Speaker #1
We are. This time, we're taking a close look at a paper called Are Return Seasonalities Due to Risk or Mispricing? Evidence from Seasonal Reversals.
- Speaker #0
OK, seasonality. So we're talking about that thing where certain stocks just seem to perform well or poorly in the same month, year after year.
- Speaker #1
Exactly. And this paper. Well, it really digs into the why behind that. Yeah. Is it some underlying risk that changes seasonally?
- Speaker #0
Or is it something else? Like maybe the market just gets the price wrong temporarily, some kind of mispricing.
- Speaker #1
That's the core question. Are these predictable monthly wiggles driven by, you know, fundamental economic risk? Or are they maybe signals that the market's having a bit of a hiccup?
- Speaker #0
Right. And this is where it gets really interesting for us, especially if you're thinking about actual trading strategies. Because I guess if it is mispricing, you'd expect some kind of balancing effect, wouldn't you?
- Speaker #1
Well, that's what the paper argues. If a stock's price gets, say, pushed up too high in January just because it's January, then maybe its returns in February, March, the rest of the year should be a bit lower to compensate. It can't just stay over-valued forever.
- Speaker #0
Ah, OK. So this idea of seasonal reversals.
- Speaker #1
That's the term they use. The thinking is if these monthly patterns are just temporary mispricings. Then a month where you'd expect high returns should probably be balanced out by months where you'd expect lower returns. And the other way around, too, over the whole year, it kind of evens out.
- Speaker #0
And that could be a really valuable insight for traders, couldn't it? If it's not just random noise, but something predictable that corrects itself.
- Speaker #1
Definitely.
- Speaker #0
So for this deep dive, we're really going to zoom in on the trading rules the paper looked at and maybe most importantly, what the back test showed.
- Speaker #1
Sounds good. Let's start with that season reversals concept then. Okay. So what they found basically is that stocks that tend to do well in, say, April. Yeah. Often tend to do less well in the other 11 months. And the opposite is true too. Stocks weak in one month might be stronger in others, like an ebb and flow.
- Speaker #0
Okay, a predictable ebb and flow. So how did they actually measure this? How did they identify these monthly habits and these reversal effects?
- Speaker #1
Well, they used a lot of historical data, obviously. Right. For each stock. They calculated its average return for January over many, many years. Right. And then they also calculated its average return for all the other months combined, February through December. And crucially, they made sure to skip the most recent year's data when doing these calculations.
- Speaker #0
Why is that?
- Speaker #1
To avoid look-ahead bias. You don't want your strategy simulation to accidentally use information that wouldn't have actually been available at the time you were making the trade decision.
- Speaker #0
Gotcha. Makes sense. Only use past data. Okay, so they've identified these tendencies. Now, the million dollar question, did trading on them actually, you know, work?
- Speaker #1
Let's look at the back tests. First, they tested a strategy based purely on the same month average return, the simple seasonality. They call it ANN.
- Speaker #0
How did that do?
- Speaker #1
Surprisingly well, actually. The average return was 0.61% per month.
- Speaker #0
Okay, 0.61.
- Speaker #1
Which, you know, might not sound earth shattering on its own, but the T value was 8.37.
- Speaker #0
Wow, 8.37. That's statistically very significant, isn't it?
- Speaker #1
Extremely. It strongly suggests this isn't just luck or random chance. There seems to be a real persistent pattern there.
- Speaker #0
OK, so trading on the month's typical performance looks promising. What about the other side of that coin, the reversal idea? Trading based on how a stock does in the other months. That's the NANN factor.
- Speaker #1
Exactly. NANN stands for non-annual. So here, the strategy is betting on that reversal. You buy stocks that have historically done poorly in the other 11 months.
- Speaker #0
Hoping they'll revert in their good month.
- Speaker #1
Kind of, or maybe just identifying stocks whose bad months are particularly bad, suggesting the good month effect is more pronounced, and you sell short the ones that do well in the other months.
- Speaker #0
Right, the reversal play. And how did that perform?
- Speaker #1
That one came in with an average monthly return of 0.45%.
- Speaker #0
Still positive.
- Speaker #1
Still positive, yes. And the T-value was 4.89.
- Speaker #0
Okay, also statistically significant, though not quite as high as the first one. Still very solid, yeah. It supports the idea that these seasonal highs in one month seem connected to relative lows in the other months.
- Speaker #1
So both strategies kind of work on their own. Did they try putting them together? Sure. Like looking at the difference between the same month return and the other month return. That was the AMN factor, annual minus non-annual?
- Speaker #0
They did. And this combined approach, it gave the best results of the lot.
- Speaker #1
Really? How much better?
- Speaker #0
The average monthly return jumped up to 0.67%.
- Speaker #1
Okay. Higher than either individually. Yep.
- Speaker #0
And the t-value shot up to 9.93.
- Speaker #1
Whoa, nearly 10. That's huge.
- Speaker #0
It really is. Extremely statistically significant. It strongly suggests that considering both the seasonal tendency and the reversal pattern together, gives you a much more powerful signal.
- Speaker #1
That feels intuitive actually. If you know a stock tends to go up in May and it tends to underperform the rest of the year, that combination paints a clearer picture maybe.
- Speaker #0
Seems that way. And importantly, the paper notes that neither the basic seasonality factor, ANN, nor the reversal factor, NNN, fully explains the other.
- Speaker #1
Meaning they contain some independent information. They're not just perfectly mirrored images. Each one captures a slightly different nuance of the expected return pattern.
- Speaker #0
Interesting. Okay, so we have these potentially profitable seasonal strategies. How do they fit in with the rest of the factors, you know, market risk, size, value, momentum, the usual suspects?
- Speaker #1
That's a great question for portfolio building. They check the correlations. And the combined AMN factor showed pretty low correlations with those traditional factors.
- Speaker #0
Low correlation. That's good news for diversification, right?
- Speaker #1
Potentially, yes. If this seasonal strategy zigs, when your value or momentum strategy zags, It could help smooth out your overall ride.
- Speaker #0
Right. Uncorrelated returns are often highly sought after. But did they generate alpha? Did these strategies produce returns after accounting for exposure to those common risk factors?
- Speaker #1
They did look at that. They ran regressions against the Carhartt four-factor model that's market, size, value, and momentum. And the result? Significant alpha across the board. The simple seasonality factor, ANN, had an alpha of 0.64% per month. T-STAT, 8.79. The reversal factor. and A&M had an alpha of 0.35% per month, T-stat 6.17.
- Speaker #0
Still significant.
- Speaker #1
Very much so. And the combined A&M factor delivered an alpha of 0.66% per month, with a T-stat of 9.70.
- Speaker #0
Almost identical alpha to the combined return and that massive T-stat again.
- Speaker #1
Exactly. It means that even after you account for the returns you'd expect, just from being exposed to market movements, small caps, value stocks, or momentum stocks, These seasonal strategies still generated significant excess returns.
- Speaker #0
That really strengthens the case for mispricing, doesn't it? If it was just risk, the standard factors should have explained more of it away.
- Speaker #1
That's certainly how the evidence seems to lean. It suggests these patterns aren't just capturing known risk premiums.
- Speaker #0
Now, you mentioned reversals. How does the seasonal reversal compare to, like, standard long-term reversal strategies, you know, where docs that have been beaten down for years tend to bounce back?
- Speaker #1
Good point. They did compare it. They looked at a typical long-term reversal factor. LTEV. The long term reversal factor had a lower average return about 0.29 percent per month. T-stat around 2.95.
- Speaker #0
OK. Lower returns, less statistical significance compared to the seasonal one.
- Speaker #1
Right. And more importantly, when they regressed LTEV on just the Fama French three factor model market size value, its alpha wasn't statistically significant.
- Speaker #0
Ah. So long term reversal seems largely explained by standard factor.
- Speaker #1
Whereas these seasonal reversals are not. They're distinct phenomena. The seasonal ups and downs aren't just a mini version of long term mean reversion.
- Speaker #0
So it looks like these seasonal factors offer something genuinely different. Did the paper touch on what this might mean for overall portfolio performance like risk adjusted returns?
- Speaker #1
It did. They did some analysis looking at maximum sharp ratios.
- Speaker #0
The measure of risk adjusted returns.
- Speaker #1
Exactly. The results suggested that adding these seasonal factors, particularly the combined AMN factor to a portfolio of traditional factors, could potentially lead to a noticeable improvement in the overall Sharpe ratio.
- Speaker #0
So better bang for your buck risk wise.
- Speaker #1
That's the implication. Yeah. Improved risk adjusted performance.
- Speaker #0
OK, let's try to boil this down then. The research presents pretty compelling evidence for these predictable seasonal reversals. Stocks strong in one month, often offset by weakness in others and vice versa. And it seems more likely linked to temporary mispricing than just shifting risk.
- Speaker #1
That's the main thrust, yes.
- Speaker #0
And the trading strategies built on this, especially that combined AMN factor, showed really strong backtest results, significant returns and crucially significant alpha even after accounting for standard factors.
- Speaker #1
Precisely. The numbers, particularly the T-stats and the alpha results are quite persuasive.
- Speaker #0
So for you listening in, this definitely gives some food for thought. How might these recurring seasonal glitches, these mispricings, fit with other market anomalies you track or even your own strategies?
- Speaker #1
Right. Could weaving in a seasonal perspective actually give you an additional edge? It's definitely something worth mulling over.
- Speaker #0
It seems like paying attention to the calendar might be more important than some people think.
- Speaker #1
This research certainly suggests it could be a fruitful area to explore, looking beyond just the usual factors for potential opportunities.
- Speaker #0
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 HTPS.PapersWithBacktests.com. Happy trading.
