Dodgy Spreadsheets
I've never particularly liked spreadsheets. The dislike goes back a long way. I think the underlying reason is because there is no way to see what the assumptions underlying the figures are. As we in the computer business are wont to say, "Garbage in, garbage out" (aka GIGO). Put crap data into a spread sheet, and you will get crap out, no matter how clever the processing you do with the spread sheet.
Another problem is that spreadsheets never make explicit the underlying assumptions that they are working on. The output may look convincing, the input may be meaningful and accurate, but the assumptions can be way out, and there is no way, from looking at the sheet, that you can tell what the assumptions are.
Then, of course, spreadsheets and their manipulators have no recognized way of testing the instructions they've programmed into the spreadsheet to make sure there are no mistakes. It's taken the programming community a long time to understand the importance of, and the best way to implement, tests, and even now we still have serious problems. The aficionados of the spreadsheet world haven't even reached the starting blocks yet.
Which brings me to the question of the economy... Much of the theoretical justification for austerity is based on the work of a couple of famous economists who used a spreadsheet to predict, in a paper in 2010, that there was a critical threshold - a tipping point - for the level of government debt. Go over that level (90%) and Kaboom! the bottom drops out of economic growth. As western government debt approached this level in the last couple of years, the paper has been used to provide the theoretical justification for austerity policies.
Whether or not you agree with austerity policies or not is another matter, but the paper has been widely quoted as justification. There was one niggling little problem, though. No one else had been able to get the same figures, showing a 90% tipping point out of their own spreadsheet programs. Eventually, the paper's authors allowed researcher to look at their original spreadsheet, and, lo and behold, the examination revealed all three types of error referred to earlier!
In the first place, the authors had omitted some of the data. Secondly, they used unusual and somewhat questionable statistical procedures, and finally, they made an error in the spreadsheet calculations. And that doesn't even cover the issue of correlation as opposed to cause and effect, which I've covered previously...
So, next time someone tells you that their spreadsheet proves something 'conclusively' just ask them to prove it, before the economy gets really messed up...
http://www.bbc.co.uk/news/magazine-22213219
Alan Lenton
28 April 2013