Saving software from oblivion

In early 2010, Harvard economists Carmen Reinhart and Kenneth Rogoff published an analysis of economic data from many countries and concluded that when debt levels exceed 90 percent of gross national product, a nation's economic growth is threatened. With debt that high, expect growth to become negative, they argued. □ This analysis was done shortly after the 2008 recession, so it had enormous relevance to policymakers, many of whom were promoting high levels of debt spending in the interest of stimulating their nations' economies. At the same time, conservative politicians, such as Olli Rehn, then an EU commissioner, and U.S. congressman Paul Ryan, used Reinhart and Rogoff's findings to argue for fiscal austerity. □ Three years later, Thomas Herndon, a graduate student at the University of Massachusetts, discovered an error in the Excel spreadsheet that Reinhart and Rogoff had used to make their calculations. The significance of the blunder was enormous: When the analysis was done properly, Herndon showed, debt levels in excess of 90 percent were associated with average growth of positive 2.2 percent, not the negative 0.1 percent that Reinhart and Rogoff had found. □ Herndon could easily test the Harvard economists' conclusions because the software that they had used to calculate their results—Microsoft Excel—was readily available. But what about much older findings for which the software originally used is hard to come by? □ You might think that the solution—preserving the relevant software for future researchers to use—should be no big deal. After all, software is nothing more than a bunch of files, and those files are easy enough to store on a hard drive or on tape in digital format. For some software at least, the all-important source code could even be duplicated on paper, avoiding the possibility that whatever digital medium it's written to could become obsolete. □ Saving old programs in this way is done routinely, even for decades-old software. You can find online, for example, a full program listing for the Apollo Guidance Computer—code that took astronauts to the moon during the 1960s. It was transcribed from a paper copy and uploaded to GitHub in 2016.