As if the London Whale nearly bankrupting Chase Bank wasn't lesson enough, we now hear that an Excel error impacted the conclusions of a major economics paper that influenced the recent austerity policies in the US, UK and elsewhere. Matt Frost says enough is enough, and implores to everyone to stop using error-prone point-and-click tools, and instead use a reproducible scripting language — namely, the R language — for serious statistical analysis:
The slides are awesome, I just wish there was video of the live presentation. My favourite lines from the deck:
"I'm not a programmer!"
"You're not an accountant either, but you seem to use a lot of spreadsheets."
Check out Matt's presentation below, and share it with your spreadsheet-using friends to prevent more formula disasters.
Matt Frost: Stop Clicking, Start Typing (via the author)
Earlier in my career I saw engineers use Excel for all sorts of modelling, including analysis of beam steering networks using FFTs and all. This was mainly necessitated by the lack of other suitable tools (such as Matlab) to do such modelling due to the organisations they worked for being too stindgy to invest in proper tools. I think this has changed with the open source revolution. There is really no excuse now for using a screwdriver to crack a walnut!
Posted by: Leith Mudge | April 17, 2013 at 18:20
I'm continually amazed at the insurance industry, to name a big user of Excel, where an Actuary (one of those guys who's passed all the exams), who has to know all sorts of Higher Math and Math Stat in order to sit in that Aeron Chair, then uses something not much better than a Friden calculator to figure the numbers for billions of dollars of capital.
May be The Great Recession, which is viewed as a Black Swan Event, doesn't happen every other week by sheer luck. May be not having Great Recessions is the Black Swan?
Posted by: Robert Young | April 18, 2013 at 06:36
How would we have done modeling without computers? In other words, how did we do it before computers? Excel? R Language? Blaming the software for poor modeling is illogical. Was the Excel model not verified before relying on it so heavily? Excel, or any other computer software is much faster and reliability accurate than calculations by hand; however, they must be verified as accurate.
Posted by: Mark Schreiber | May 09, 2013 at 05:32
What about "the finest command-line interface 1992 has to offer" for favourite lines?
Posted by: Tom | November 05, 2013 at 00:40