20 April 2013: Krugman on Reinhart/Rogoff
Paul Krugman shows how a bug in software was used in 2010 by governments in support of a switch from Keynesian policies to “austerity”. Two respected Harvard economists claimed to find a “tipping point” when debt exceeds 90% of GDP but other researchers couldn’t duplicate their result; when their spreadsheet was examined they found three flaws described by Krugman in this post, and the most important flaw was a coding error in the spreadsheets.
The user of a modern spreadsheet (of which Excel is practically the only example, the need for compatibility having given Excel monopoly power owing to what Krugman himself calls “network externalities”) is given a vast amount of flexibility in implementing the logic of the spreadsheet as well as its appearance.
But as a result, Excel usage is a form of programming and most people, including professional programmers, find programming hard. It requires an infrastructure of reusable tools and standardized, verifiable work methods.
However, Excel was designed by and for business majors who had nothing but contempt for working people (they took a managerial viewpoint and its contempt for labour) and who were as such overconfident to the point of arrogance about their spreadsheets.
Typically, the spreadsheets they used to get their results were not, unlike other forms of computer software, reviewed for correctness. This was the source of the Reinhart-Rogoff bug, in all probability.
Around the time of the invention of Excel, management-oriented trade journals such as Computerworld and Datamation would constantly call upon programmers to think, not like mathematicians or programmers, but like managers. The problem was, for me anyway, what and how managers think at all.
I’d go out to lunch with managers and be astounded by the low culture and intelligence of managers whose pronouncements never made any sense, but with two children you don’t say anything you have another drink.
Computerworld lauded new software whose “parameters” could be controlled by the all-mighty User (an always singular and male but never quite defined deus ex machine: less a person than a rhetorical device for resolving arguments) and in fact I had great success in using my knowledge of compiler design to design “little languages” that did empower the user. On one project for complex, interlinked real estate appraisal I found myself reinventing Excel and its algorithm for updating all cells by visiting each once, reverse-engineered from Lotus and Excel.
But I found that at a certain point, giving the user a language to design his own solution is cheating; using the parameters, the spreadsheet or the “little language”, the programmer needs to implement a solution that works “today” for the end user before demonstrating how to modify the parameters or statements in the “little language”. And someone, whether the economist/scientist/manager/etc. end user or the programmers, has to take responsibility for correctness.
There seems to be a correlation between right-wing “austerity” politics and an only superficial knowledge of software and a failure to take responsibility for its correctness. Elites jumped enthusiastically on the Reinhart/Rogoff results, and ignored the Reinart/Rogoff bug, because they didn’t want to do the hard work, whether of ensuring that the correct analytical tools were used in forecasting or of just redistribution.
It’s unlikely, but there’s no reason why bugs, not added by end users such as Reinhart and Rogoff, but pre-existing in the software used by them, should not further distort their results, because as I’ve said, successful software packages tend, as Excel has tended, to be winner-take-all, and push out all competition: there’s probably a lot of people today that have never even heard of Lotus 1-2-3. Once a software package such as Excel gains a 100% monopoly, the need to invest in control of bugs decreases as users start to trust the package overmuch, and as here, even rely on the bugs to be feature. The need to rely on such a shaky platform creates insistence on there being no alternative, and a reluctance to be open to alternatives; Bradley Manning and Julian Assange offered the biggest scoop in history to the media in the form of the Wikileaked Apache murder of Reuter’s journalists, but they weren’t interested.
The “vox clamant in deserto” is created, such as my father on the subject of brain death. My father showed that more criteria were needed in practice, but for isomorphic reasons to those of the elites in wanting to reverse Keynesian job creation to austerity (saving money on hospital beds), it was decided to force doctors in the 1980s to “call” death, as is shown on medical TV such as ER, using a simple flatlined EKG. This made and makes men like Julian Assange, Bradley Manning, myself and my late father each a vox clam anti in deserto, whether on Wikileaks, software reliability or brain death.
When you know that your own tools and sources cannot be trusted you insist all the more on their veracity because you are responsible for choosing them. Elites, who wanted to continue to be the comfortable class, with no tax increases, preferred the results from the buggy Excel spreadsheet.