We survey the literature on the accuracy of econometric software. We also assess the advantages of open source software from the point of view of reliability and discuss its potential in applied economics, which has now become fully dependent on computers. As a case study, we apply various accuracy tests on GNU Regression, Econometrics and Time-series Library (gretl) and demonstrate that the open source nature of the program made it possible to see the cause, facilitated a rapid fix, and enabled verifying the correction of a number of flaws that we uncovered. We also run the same tests on four widely-used proprietary econometric packages and observe the known accuracy errors that remained uncorrected for more than 5 years.