The study focused on a cross-section ofcountries observed in the mid-1990s, so the conclusions fromthat e¤ort are beginning to be a bit dated. In addition,signi cant revisions of the data underlying the 2005 paperhave been published. Last but not least, in the interveningyearsbecome aware of ways in which the originalmethodology can be usefully improved and extended. Hence thepresent update and upgrade of the original paper. This paperfocuses on data (mostly) from 2005 and improves on theoriginal methodology in several dimensions. Developmentaccounting compares di¤erences in income per worker betweendeveloping and developed countries to counter-factualdi¤erences attributable to observable components of physicaland human capital. Such calculations can serve a usefulpreliminary diagnostic role before engaging in deeper andmore detailed explorations of the fundamental determinantsof di¤erences in income per worker. The research and policyagenda would then have to focus on technology, allocative e¢ciency, competition, and other determinants of the efficientuse of capital. However because of limitations inthecoverage of the test results, author also present resultswhere human capital is only measured from years of schoolingand health. It turns out that, at least in my preferredcalibration, the addition or omission of cognitive skills(as measured by test scores) does not greatly affect thequantitative results.