There was a peer-reviewed academic paper in the Journal of Finance (JOF), which we were reminded about today in a discussion about evidence-based investing. This was an article that one of our followers, and a highly-renowned columnist/editor at the Wall Street Journal paper, had shared a short while ago. The paper contains a provocative conclusion, as its Abstract plainly articulates: "(there is) a strong inverse link between daily stock returns and hospital admissions". Could this be true? As it turns out, their own limited evidence does suggest this is simply spurious. And it's a lesson in how probability and statistics plays a vital role in assisting with everyday analysis. As the editorial board member of one of the foremost peer-reviewed mathematical statistics journals, the errors exposed in the JOF paper (linked to above) would have been caught early, and we'd question the entire merits of the JOF paper's findings. Additionally the social media engagement on the JOF article, along with our rebuttal summarized below, was quite substantial.
First let's look to the right, at their complete exhibit data underlying their paper. The lesson they are making is clear. As markets fall, hospital admissions go up. Everyone recognizes the iconic date of Black Monday (incidentally my birthday): October 19, 1987. So is this all the statistics we need for a peer-reviewed paper?
The first of many issues is that they are showing admissions to a local California hospital only, not admissions for the entire United States. They might as well have charted hospital admissions for a city that was struck by a tornado on that day. The second issue is that they are showing a time series of just 7 data points, from a stock market history of tens of thousands. One doesn't need to be a statistics professor to know that this is injudicious.
If that's not enough, while on October 19 the markets fell at a geometric (or a continuous) rate of 23%, the Friday before (October 16) the markets fell at a horrifying rate of more than 5%! Even during the global market rout last summer, the S&P never dropped so much on a single day. At the apex of August 24, on what is informally named China's Black Monday, the S&P dropped "only" 4% on a continuous rate.
Yet on the JOF article, on this Friday before October 19, hospital admissions were smartly down! In fact the market had fallen on all three days at the start of the chart above (October 14, 15, and 16). Perhaps citizens are perfect market timers, and suspended curing their illness until after the market completely crashed. Of course we know that this is an absurd idea that only an economist would attempt to demonstrate. Additionally three days after the October 1987 crash, on October 22, the markets fell 4% and again hospital admissions were smartly down.
But in all seriousness, this is our third issue with the article, and it also completely invalidates the premise that large drops in the markets lead to higher hospital admissions. For more details on the daily drops during the worst market crashes in history, please read to our popular research article here. Should the markets fall dramatically again later this year it could serve as a useful reference!
The fourth issue we'll raise is that if we want to consider the entire link between stock market returns and hospital admissions (and assuming we had better data), then we must consider what the overall correlation is before and after large market changes. Let's to the left, chart all of the shown changes in the stock market (horizontal axis) and the change in hospital admissions (vertical axis). These should be obvious notions but the JOF authors, University of California economists, wanted them concealed from you!
And what we see further invalidates the JOF's findings. That we have the presence of an outlier (October 19) that has the debilitating function of merely tricking us into the JOF's analysis (that there is a strong inverse link between daily market returns and abnormal hospital admissions). Nothing could be further from the truth. Without this outlier, we get an even stronger link of the opposite conclusion. Strong market performance (not weak market performance) leads to higher abnormal hospital admissions. But of course this doesn't fit the academic economist's predetermined thesis, so we must invent new ideas that fit their preconceived opinions. Neither does the fact that suicides are at a record high, just as the market and the economy is as well! And were this not a published research article concerning two topics of great concern to most people (their money and their health), this wouldn't be a critical enough topic on which to engage.
On an aside: It's been a while since we've updated on our statistics progress and recent citations of this blog. If not mentioned earlier, I'm now teaching as an adjunct professor at both graduate schools of NYU, Georgetown, and possibly soon at Columbia University. So in addition to day jobs, am getting little sleep. Additionally have been invited as a featured speaker at various academic and media conferences, with details forthcoming.
Great share by Barry Ritholtz, and a Wall Street Journal contributor George Papadopoulos, a NYC pension chief Sanford Rich, Powell's Regis Media, and others: https://mobile.twitter.com/feeonlyplan…/…/730049814717730816 https://mobile.twitter.com/RegisM…/status/730578415649234945 https://mobile.twitter.com/sanfor…/status/730191479277277184
With the help of two Bloomberg columnists (Faye Flam, Barry Ritholtz), retracted two flawed science policy papers. Continues to be enjoyed, including by J.Walker Smith (chairman of a leading marketing firm), and Lance Gravatt (chairman of a leading bio-pharmaceutical firm).
And a movie as horrific as Jim Cramer’s stock picks. Money Monster (#moneymonster) debuts as the 2nd worst movie opening for the year. 3 Oscars Academy Awards winners (George Clooney, Julia Roberts, Jodie Foster) and they still couldn’t prettify Bozo the Clown. The ONLY movie that had a worse opening (out of >20) was the $2-million pornographically-titled Hardcore Henry. Stunning. Cramer of course recently told The Wall Street Journal that this movie was a "winner"; sample his other money-losing "winner" ideas: http://www.cnbc.com/…/cramer-defends-buffett-his-lucks-abou…
Fantastic read in Bloomberg/Bloomberg View. "I asked Salil Mehta about a number of problems with drawing predictive conclusions. 'Seeing a short-term pattern but for the wrong reason,' he said, is often the result of (invariably) small samples." http://www.bloomberg.com/view/articles/2016-05-20/election-uncertainty-isn-t-messing-with-markets
As brought to my attention by my friend Connor Roso, an SEC wonk and law professor, a self-professed scam concerning Donald J. Trump. Enter the national boogey-man, masquerading as an elite political "scientist". Still doesn’t appreciate the bigger story, which is that he is luckier than he is a mastermind. 49 of 50 states correct in 2008? Come on. Any child simply picking 2004’s winners per state would have gotten 48 of the 50 correct. Now provides just less than the same odds in general election that I made THREE MONTHS AGO. http://statisticalideas.blogspot.com/…/the-flip-flopping-po… http://fivethirtyeight.com/…/how-i-acted-like-a-pundit-and…/ Friends, professors, and former and current star IMF economists Allan Brunner and Prakash Loungani (respectively), among those who have enjoyed my various related articles.
My "benchmark >> passives > actives" article widely shared, including by a Chief Investment Officer for BlackRock's FutureAdvisor, CFA Institute, ETF Trends, Abnormal Returns, and Quantocracy. https://twitter.com/simonwmoore/status/734778780628094976 https://twitter.com/cfa_ch/status/734152499536498688 https://twitter.com/abnormalretur…/status/734750760194318341 http://quantocracy.com/?blog=statistical-ideas
Happy to see citations in both Financial Advisor Magazine, and in CFA Institute's special holiday-weekend read list! See them both: http://www.fa-mag.com/…/election--uncertainty--isn-t-messin… https://blogs.cfainstitute.org/…/weekend-reads-the-seersuc…/
Our previous article contrasting GDP nowcasts from New York and Atlanta Federal Reserve banks was enjoyed by both Federal Reserve and the U.S. Labor Department economists, CFA Institute, and Roberts' of KSEV radio.
Lastly blog followers and U.S. Treasury Department executives (Harold Barnshaw, Brian Donovan, etc.) amusingly ask if this is what I meant in our lottery article. I retort why would a couple stealing $175k then immediately play lottery? http://statisticalideas.blogspot.com/…/a-losers-lottery.html http://m.torontosun.com/…/cops-duo-used-stolen-cash-to-buy-…
Remember a sample history of all citations can be found here.
The first of many issues is that they are showing admissions to a local California hospital only, not admissions for the entire United States. They might as well have charted hospital admissions for a city that was struck by a tornado on that day. The second issue is that they are showing a time series of just 7 data points, from a stock market history of tens of thousands. One doesn't need to be a statistics professor to know that this is injudicious.
If that's not enough, while on October 19 the markets fell at a geometric (or a continuous) rate of 23%, the Friday before (October 16) the markets fell at a horrifying rate of more than 5%! Even during the global market rout last summer, the S&P never dropped so much on a single day. At the apex of August 24, on what is informally named China's Black Monday, the S&P dropped "only" 4% on a continuous rate.
Yet on the JOF article, on this Friday before October 19, hospital admissions were smartly down! In fact the market had fallen on all three days at the start of the chart above (October 14, 15, and 16). Perhaps citizens are perfect market timers, and suspended curing their illness until after the market completely crashed. Of course we know that this is an absurd idea that only an economist would attempt to demonstrate. Additionally three days after the October 1987 crash, on October 22, the markets fell 4% and again hospital admissions were smartly down.
But in all seriousness, this is our third issue with the article, and it also completely invalidates the premise that large drops in the markets lead to higher hospital admissions. For more details on the daily drops during the worst market crashes in history, please read to our popular research article here. Should the markets fall dramatically again later this year it could serve as a useful reference!
And what we see further invalidates the JOF's findings. That we have the presence of an outlier (October 19) that has the debilitating function of merely tricking us into the JOF's analysis (that there is a strong inverse link between daily market returns and abnormal hospital admissions). Nothing could be further from the truth. Without this outlier, we get an even stronger link of the opposite conclusion. Strong market performance (not weak market performance) leads to higher abnormal hospital admissions. But of course this doesn't fit the academic economist's predetermined thesis, so we must invent new ideas that fit their preconceived opinions. Neither does the fact that suicides are at a record high, just as the market and the economy is as well! And were this not a published research article concerning two topics of great concern to most people (their money and their health), this wouldn't be a critical enough topic on which to engage.
On an aside: It's been a while since we've updated on our statistics progress and recent citations of this blog. If not mentioned earlier, I'm now teaching as an adjunct professor at both graduate schools of NYU, Georgetown, and possibly soon at Columbia University. So in addition to day jobs, am getting little sleep. Additionally have been invited as a featured speaker at various academic and media conferences, with details forthcoming.
Great share by Barry Ritholtz, and a Wall Street Journal contributor George Papadopoulos, a NYC pension chief Sanford Rich, Powell's Regis Media, and others: https://mobile.twitter.com/feeonlyplan…/…/730049814717730816 https://mobile.twitter.com/RegisM…/status/730578415649234945 https://mobile.twitter.com/sanfor…/status/730191479277277184
With the help of two Bloomberg columnists (Faye Flam, Barry Ritholtz), retracted two flawed science policy papers. Continues to be enjoyed, including by J.Walker Smith (chairman of a leading marketing firm), and Lance Gravatt (chairman of a leading bio-pharmaceutical firm).
@ritholtz: @UniofOxford @prioritization @TheAtlantic all retract w ERRATA given our critique https://t.co/3wNZ6jBx55 https://t.co/Pd2PPbjd2N— Statistical Ideas (@salilstatistics) May 12, 2016
And a movie as horrific as Jim Cramer’s stock picks. Money Monster (#moneymonster) debuts as the 2nd worst movie opening for the year. 3 Oscars Academy Awards winners (George Clooney, Julia Roberts, Jodie Foster) and they still couldn’t prettify Bozo the Clown. The ONLY movie that had a worse opening (out of >20) was the $2-million pornographically-titled Hardcore Henry. Stunning. Cramer of course recently told The Wall Street Journal that this movie was a "winner"; sample his other money-losing "winner" ideas: http://www.cnbc.com/…/cramer-defends-buffett-his-lucks-abou… Fantastic read in Bloomberg/Bloomberg View. "I asked Salil Mehta about a number of problems with drawing predictive conclusions. 'Seeing a short-term pattern but for the wrong reason,' he said, is often the result of (invariably) small samples." http://www.bloomberg.com/view/articles/2016-05-20/election-uncertainty-isn-t-messing-with-markets
As brought to my attention by my friend Connor Roso, an SEC wonk and law professor, a self-professed scam concerning Donald J. Trump. Enter the national boogey-man, masquerading as an elite political "scientist". Still doesn’t appreciate the bigger story, which is that he is luckier than he is a mastermind. 49 of 50 states correct in 2008? Come on. Any child simply picking 2004’s winners per state would have gotten 48 of the 50 correct. Now provides just less than the same odds in general election that I made THREE MONTHS AGO. http://statisticalideas.blogspot.com/…/the-flip-flopping-po… http://fivethirtyeight.com/…/how-i-acted-like-a-pundit-and…/ Friends, professors, and former and current star IMF economists Allan Brunner and Prakash Loungani (respectively), among those who have enjoyed my various related articles. My "benchmark >> passives > actives" article widely shared, including by a Chief Investment Officer for BlackRock's FutureAdvisor, CFA Institute, ETF Trends, Abnormal Returns, and Quantocracy. https://twitter.com/simonwmoore/status/734778780628094976 https://twitter.com/cfa_ch/status/734152499536498688 https://twitter.com/abnormalretur…/status/734750760194318341 http://quantocracy.com/?blog=statistical-ideas
Happy to see citations in both Financial Advisor Magazine, and in CFA Institute's special holiday-weekend read list! See them both: http://www.fa-mag.com/…/election--uncertainty--isn-t-messin… https://blogs.cfainstitute.org/…/weekend-reads-the-seersuc…/
Our previous article contrasting GDP nowcasts from New York and Atlanta Federal Reserve banks was enjoyed by both Federal Reserve and the U.S. Labor Department economists, CFA Institute, and Roberts' of KSEV radio.
Lastly blog followers and U.S. Treasury Department executives (Harold Barnshaw, Brian Donovan, etc.) amusingly ask if this is what I meant in our lottery article. I retort why would a couple stealing $175k then immediately play lottery? http://statisticalideas.blogspot.com/…/a-losers-lottery.html http://m.torontosun.com/…/cops-duo-used-stolen-cash-to-buy-…
Remember a sample history of all citations can be found here.
No comments:
Post a Comment