Although their respective proportions of errors were roughly equal according to the Coverity Scan 2012 Report, the 2013 edition of the report shows that quality is markedly higher now in open source code than in its proprietary equivalent.
After the previous year's Coverity Scan 2012 Report attributed a defect density per 1,000 lines of code of 0.69 to open source software, compared with 0.68 in proprietary software, the Coverity Scan 2013 Report mentioned defect figures of 0.59 per 1,000 lines of code in open source software but rising to 0.72 for proprietary software. In other words, the proportion of errors in the latter has increased, while decreasing amongst open source solutions.
As a reminder, Coverity develops (proprietary) software that performs statistical analyses of C, C++ and Java source code.