April 17, 2010
Hal Varian passes along this amusing result if you query Google Insights for Search for searches on "mixed drinks" and "hangovers."
Posted by James Hamilton at April 17, 2010 07:33 AMdigg this | reddit
Good one. Now try "mixed drinks" vs "stds". Scary.
Posted by: Peter at April 17, 2010 08:08 AM
The mid-Jan to mid-Feb data are interesting in that hangover searches spiked both same day as mixed drink searches spiked, and the day after. My half-serious theory/hypothesis would be that during this period people planning to make/order a lot of mixed drinks (or choose between mixed drinks and other drinks) that evening still had memories of that particularly bad New Year's Day hangover fresh in their minds, so they also researched hangovers prior to their evening drinking to have a better sense of risk/reward, but of course once they were out and had consumed a couple of drinks, rationality went out the window and they ended up with a hangover the next day, searching for something to mitigate the pain from both the hangover and from kicking themselves.
Posted by: Brooks at April 17, 2010 10:55 AM
Makes sense, especially with the blue spikes mostly either leading or coinciding with the red ones. I also like how the blue spikes most often occur on Saturdays!
Posted by: Ironman at April 17, 2010 12:49 PM
Check out the inverse relationship between searches on "irs" and searches on "guns" http://www.google.com/insights/search/#q=irs%2Cguns&geo=US&date=today%203-m&cmpt=q
I was wondering if there would be a positive correlation, but I don't have a theory/hypothesis as to why there would be a negative correlation. Any thoughts?
Posted by: Brooks at April 17, 2010 12:57 PM
Economists finally discovered Saturdays!
Posted by: Ernesto at April 17, 2010 02:24 PM
Cleary, economists have a lot of spare time.
Interesting tool, though, tricky to use correctly.
Posted by: Steven Kopits at April 17, 2010 02:43 PM
It looks like precisely a one day lag. That sounds about right.
Posted by: Brian Quinn at April 17, 2010 04:06 PM
Hypotheses for Brooks:
(1) Info on how to deal with the IRS and guns are substitutes.
(2) After trying to read incomprehensible tax code, people become angry and increasingly interested in guns (hence the lag)
(3) After reading up on IRS, people realize they have tax rebates coming.
(4) (most likley) The gun searches spike on weekends; this is recreational searching. The IRS searches spike on Mondays and slowly decline -- this is work-related.
Posted by: Charles N. Steele at April 18, 2010 08:44 AM
Also for Brooks: "irs" and "fishing" have a similar pattern to "irs guns," hence my conjecture that its simply work vs. recreation.
Posted by: Charles N. Steele at April 18, 2010 08:47 AM
Better yet, try "mixed drinks" and then "morning after pill."
Posted by: Anonymous at April 18, 2010 06:22 PM
To Brooks: "irs" & "cooking" is almost identical to "irs-guns."
To Anonymous: "mixed drinks-morning after" ... beautiful. You're turning this into an art form.
Posted by: Charles N. Steele at April 18, 2010 07:16 PM
Charles -- good point, thanks.
Posted by: Brooks at April 18, 2010 07:34 PM
My failure to think of weekdays vs. weekends reminds me of something George Carlin said. (Paraphrasing) A year is a natural unit of time. A month is a natural unit of time. A week is NOT a natural unit of time...which is why the pigeons in the Financial District can't figure out why no one is there to toss them bread on weekends.
Posted by: Brooks at April 18, 2010 09:57 PM
Try “work” vs “play”...
Most of these oscillations reflect the weekly nature of work/play cycle.
Posted by: bio at April 19, 2010 12:16 AM
"day after pill" also peaks on weekends, or right after
Posted by: Dino Ovo at April 19, 2010 07:03 AM
Posted by: Ivars at April 19, 2010 09:44 AM
Interesting. Try "fuel prices" and you will notice two distinct peaks in 2005 and 2008. If you compare "fuel prices" to "fuel economy", you will note a lone spike in "fuel economy" in 2009.
"Terrorism" exhibits an interesting downward trend.
Posted by: GNP at April 20, 2010 05:55 AM
Omitted variable bias!!!
Posted by: carlo at April 20, 2010 09:52 AM
It actually looks like if you plot ANYTHING with ANYTHING else, you more or less get this pattern, i.e. there is some sort of seasonality within the day. We'd probably need to strip out the cycle.
Posted by: Spooky at April 29, 2010 07:23 AM