Confounding Variables
March 14, 2023 8:58 AM Subscribe
This paper constructs a novel dataset to explore whether Matt Levine’s vacation drives market volatility as suggested by readers and Levine himself. It finds that contrary to expectations, Levine’s vacations have an (almost) statistically significant effect in the opposite direction: decreasing market volatility. This paper provides a series of potential explanations for this observed market reaction. Abstract for: A Matt Levine Effect? [PDF]
Academic trolling is my favorite trolling.
posted by biogeo at 9:27 AM on March 14, 2023 [3 favorites]
posted by biogeo at 9:27 AM on March 14, 2023 [3 favorites]
UGH. What a terrible example of "If I build a causally bad model, it will give bad causal results."
None of the models considered are based off of a full hypothesis of what is going on in the system - just half-arsed controls thrown in wily-nily.
And the misuse of AIC is making me SCREAM. AIC is about prediction, not causality.
I can feel the self-satisfied smugness coming off of this paper, and it's all completely unearned and obnoxious. Bleh.
(and yes, causal thinking and modeling is part of the research I do)
posted by redbeard at 9:39 AM on March 14, 2023 [7 favorites]
None of the models considered are based off of a full hypothesis of what is going on in the system - just half-arsed controls thrown in wily-nily.
And the misuse of AIC is making me SCREAM. AIC is about prediction, not causality.
I can feel the self-satisfied smugness coming off of this paper, and it's all completely unearned and obnoxious. Bleh.
(and yes, causal thinking and modeling is part of the research I do)
posted by redbeard at 9:39 AM on March 14, 2023 [7 favorites]
→
I think you'd disappoint a whole bunch of people in data "science" if you got all picky like that.
posted by scruss at 10:55 AM on March 14, 2023 [4 favorites]
half-arsed controls thrown in wily-nily
I think you'd disappoint a whole bunch of people in data "science" if you got all picky like that.
posted by scruss at 10:55 AM on March 14, 2023 [4 favorites]
Matt Levine commented on this in his Monday newsletter (paywalled), of course, and made the most obvious and direct observation:
posted by angrynerd at 11:05 AM on March 14, 2023 [5 favorites]
The main explanation is that I take vacations at normal vacation times, when everyone else takes vacations and markets are often quiet.We got an example of another factor driving the "Matt Levine effect" last week. Levine was supposed to be on vacation on Friday, March 10, but due to the collapse of Silicon Valley Bank, he came back from vacation and wrote a special column on the topic. So, not only does he normally take vacations at normal times, we have evidence that when something noteworthy happens while he's on vacation, he sometimes stops being on vacation to write about what's happening.
posted by angrynerd at 11:05 AM on March 14, 2023 [5 favorites]
Reads the article… guessing this SSRN manuscript will never graduate to be a peer-reviewed paper.
posted by jilloftrades at 11:18 AM on March 14, 2023 [1 favorite]
posted by jilloftrades at 11:18 AM on March 14, 2023 [1 favorite]
People know this is a joke, right?
posted by praemunire at 11:28 AM on March 14, 2023 [5 favorites]
posted by praemunire at 11:28 AM on March 14, 2023 [5 favorites]
It’s not a joke if it’s true though? One could imagine an influential columnist effecting market behavior by taking days off. In this case though, it appears to be a joke. Unless it’s true.
posted by chrchr at 11:31 AM on March 14, 2023 [1 favorite]
posted by chrchr at 11:31 AM on March 14, 2023 [1 favorite]
This should certainly be bundled with the Inverse Cramer ETF that just launched.
Tired: meme stonks
Wired: meme exchange-traded fonds
posted by Reyturner at 11:58 AM on March 14, 2023 [3 favorites]
Tired: meme stonks
Wired: meme exchange-traded fonds
posted by Reyturner at 11:58 AM on March 14, 2023 [3 favorites]
Metafilter: It's not a joke if it's true.
As they say on the Lions lead by Donkeys podcast (possibly not the origin of the quote: "If your stupid idea works, its not a stupid idea."
Stock markets are casinos, economists/financial analystd are the court astrologers/numerologists. Brokers are dealers, and the believers are the sacrifical lambs.
Its superstition and white noise all the way down. Hell of a way to ruin people's lives and the planet. Hope it was fun.
posted by anecdotal_grand_theory at 12:06 PM on March 14, 2023 [1 favorite]
As they say on the Lions lead by Donkeys podcast (possibly not the origin of the quote: "If your stupid idea works, its not a stupid idea."
Stock markets are casinos, economists/financial analystd are the court astrologers/numerologists. Brokers are dealers, and the believers are the sacrifical lambs.
Its superstition and white noise all the way down. Hell of a way to ruin people's lives and the planet. Hope it was fun.
posted by anecdotal_grand_theory at 12:06 PM on March 14, 2023 [1 favorite]
I didn't realize how popular his writing is until now I guess. This explains why I found it tiresome and unsubscribed however. No one gets popular in finance by being critical of finance.
posted by Nec_variat_lux_fracta_colorem at 12:40 PM on March 14, 2023 [2 favorites]
posted by Nec_variat_lux_fracta_colorem at 12:40 PM on March 14, 2023 [2 favorites]
Redbeard, is your comment a joke or a parody? I honestly can’t tell.
posted by MisantropicPainforest at 12:46 PM on March 14, 2023 [1 favorite]
posted by MisantropicPainforest at 12:46 PM on March 14, 2023 [1 favorite]
Levine is great. Since the newsletter is already out and the full text might be of interest, here's his whole thing on the vacation effect:
-
Money Stuff was supposed to be off on Friday, but then Silicon Valley Bank collapsed and there was an unscheduled Money Stuff. Back when Elon Musk was trying to buy or not buy Twitter, that sort of thing happened a lot, to the point that the Verge published a story titled “A brief history of Elon Musk sabotaging Matt Levine’s time off.” This is enough of a thing that people sometimes joke on Twitter about buying index options whenever I write “Money Stuff will be off tomorrow,” on the theory that the market gods always get up to something crazy whenever I try to take a day off.
Naturally someone did a backtest and turns out it’s a bad idea (not investing advice!). Here is “A Matt Levine Effect?” by Paul Connell, William Fallon and Nicholas Foretek, on SSRN and everything:
This paper constructs a novel dataset to explore whether Matt Levine’s vacation drives market volatility as suggested by readers and Levine himself. It finds that contrary to expectations, Levine’s vacations have an (almost) statistically significant effect in the opposite direction: decreasing market volatility. This paper provides a series of potential explanations for this observed market reaction.
The main explanation is that I take vacations at normal vacation times, when everyone else takes vacations and markets are often quiet. The paper is full of good sentences, including “The measured effect of Matt Levine’s vacations on financial market volatility is negative and of a considerable magnitude,” “This paper contributes to two bodies of research. The first is a collection of studies that examine spurious statistical correlates of important macroeconomic events,” “To our knowledge, we are the first to combine Money Stuff publication dates, VIX closing data, federal bank holidays and sunspot records into a single database” and “Collection of the data for this study was made possible by the authors’ status as graduate students who, if our advisors are asking, definitely spent the whole week before spring break working on our dissertations.”
posted by BlackLeotardFront at 12:57 PM on March 14, 2023 [3 favorites]
-
Money Stuff was supposed to be off on Friday, but then Silicon Valley Bank collapsed and there was an unscheduled Money Stuff. Back when Elon Musk was trying to buy or not buy Twitter, that sort of thing happened a lot, to the point that the Verge published a story titled “A brief history of Elon Musk sabotaging Matt Levine’s time off.” This is enough of a thing that people sometimes joke on Twitter about buying index options whenever I write “Money Stuff will be off tomorrow,” on the theory that the market gods always get up to something crazy whenever I try to take a day off.
Naturally someone did a backtest and turns out it’s a bad idea (not investing advice!). Here is “A Matt Levine Effect?” by Paul Connell, William Fallon and Nicholas Foretek, on SSRN and everything:
This paper constructs a novel dataset to explore whether Matt Levine’s vacation drives market volatility as suggested by readers and Levine himself. It finds that contrary to expectations, Levine’s vacations have an (almost) statistically significant effect in the opposite direction: decreasing market volatility. This paper provides a series of potential explanations for this observed market reaction.
The main explanation is that I take vacations at normal vacation times, when everyone else takes vacations and markets are often quiet. The paper is full of good sentences, including “The measured effect of Matt Levine’s vacations on financial market volatility is negative and of a considerable magnitude,” “This paper contributes to two bodies of research. The first is a collection of studies that examine spurious statistical correlates of important macroeconomic events,” “To our knowledge, we are the first to combine Money Stuff publication dates, VIX closing data, federal bank holidays and sunspot records into a single database” and “Collection of the data for this study was made possible by the authors’ status as graduate students who, if our advisors are asking, definitely spent the whole week before spring break working on our dissertations.”
posted by BlackLeotardFront at 12:57 PM on March 14, 2023 [3 favorites]
Who's Matt Levine?
posted by Billiken at 1:22 PM on March 14, 2023 [3 favorites]
posted by Billiken at 1:22 PM on March 14, 2023 [3 favorites]
> Billiken: "Who's Matt Levine?"
Via Wikipedia:
Via Wikipedia:
Matt Levine [...] is a columnist for Bloomberg News covering finance and business. [...] His newsletter, Money Stuff, is one of the most popular on Wall Street with over 150,000 subscribers as of October 2020.posted by mhum at 1:30 PM on March 14, 2023 [2 favorites]
Okay so as a bit of a joke explainer, this paper uses standard statistical regression and model comparison methods, which as far as I understand it are pretty typical for an econometrics study, to compare a few models for fitting a measure of market volatility. Their "simplest" model, which fits Matt Levine's days off, finds a statistically significant relationship between Levine taking a day off and decreased market volatility. This is also the worst-performing model by far, as it fails to include the known fact that one day's volatility is strongly related to the previous day's volatility. The other models account for that, and in none of them does Matt Levine's vacation significantly explain market volatility. They also ran a model that included sunspot activity as a predictor of market volatility, and this turned out to be the best-performing model overall.
The joke is that all of these are poorly-considered models. Sunspot activity is a prima facie bad predictor, and the fact that the model including it is the best-performing model of the bunch strongly suggests that all of them are just picking up on spurious correlations.
I laughed, anyway.
posted by biogeo at 2:17 PM on March 14, 2023 [3 favorites]
The joke is that all of these are poorly-considered models. Sunspot activity is a prima facie bad predictor, and the fact that the model including it is the best-performing model of the bunch strongly suggests that all of them are just picking up on spurious correlations.
I laughed, anyway.
posted by biogeo at 2:17 PM on March 14, 2023 [3 favorites]
But the sun does make plants grow. They really are onto something. Maybe the market gets a little kick when the sun spots them some cash?
I thought it was good and silly. Also this is why I don't trust data scientists.
posted by crossswords at 2:34 PM on March 14, 2023 [1 favorite]
I thought it was good and silly. Also this is why I don't trust data scientists.
posted by crossswords at 2:34 PM on March 14, 2023 [1 favorite]
Why does this joke make you not trust data scientists
The authors aren’t even data scientists!
posted by MisantropicPainforest at 2:44 PM on March 14, 2023 [2 favorites]
The authors aren’t even data scientists!
posted by MisantropicPainforest at 2:44 PM on March 14, 2023 [2 favorites]
The Columbia Econ department is also a hotbed of causal inference research.
posted by MisantropicPainforest at 3:08 PM on March 14, 2023 [1 favorite]
posted by MisantropicPainforest at 3:08 PM on March 14, 2023 [1 favorite]
The dislike to data science comes from being in a field that has grown enamored with data science and seeing people abuse AIC. Definitely just my own bias, prompted by the author's jokey not-conclusion.
posted by crossswords at 4:03 PM on March 14, 2023 [1 favorite]
posted by crossswords at 4:03 PM on March 14, 2023 [1 favorite]
But like, they’re jot data scientists and they’re not abusing AIC. I’m flabbergasted.
posted by MisantropicPainforest at 4:11 PM on March 14, 2023 [1 favorite]
posted by MisantropicPainforest at 4:11 PM on March 14, 2023 [1 favorite]
Heh, I thought this would be something genuinely interesting, despite not having ever heard of Matt Levine. As soon as I saw 'Sunpots' included in the keywords, I smelled a rat. But I thought this was funny and clever, particularly in light of the well-known fact that 50% of statistics are made up anyway.
posted by dg at 11:41 PM on March 14, 2023 [1 favorite]
posted by dg at 11:41 PM on March 14, 2023 [1 favorite]
Sunspot activity is a prima facie bad predictor
I, for one, welcome the new Sun God and their interest in our economic markets.
Edited to add: actually would be pretty interesting to see the return comparisons between these investment-entertainers vs Sun Spots
posted by LizBoBiz at 8:21 AM on March 15, 2023 [1 favorite]
I, for one, welcome the new Sun God and their interest in our economic markets.
Edited to add: actually would be pretty interesting to see the return comparisons between these investment-entertainers vs Sun Spots
posted by LizBoBiz at 8:21 AM on March 15, 2023 [1 favorite]
« Older Drill Baby, Drill! | Change your mindset Newer »
This thread has been archived and is closed to new comments
posted by JoeZydeco at 9:07 AM on March 14, 2023 [4 favorites]