There are currently two ways of estimating relative levels of per capita income back to the nineteenth century. The most commonly adopted approach, popularized in particular by Angus Maddison and labeled “long-span” estimates by Marianne Ward and John Devereux, is to use time-series projection from a recent benchmark.Maddison, Monitoring; and Ward and Devereux, “Measuring British Decline.” The alternative procedure, labeled “direct estimates” by Ward and Devereux, is to use direct benchmark estimates for earlier years. Ward and Devereux provide some new “direct estimates” which they claim over-turn the conventional chronology of U.K./U.S. relative per capita income levels for the period 1870–1990, derived using “long-span” projections. One important purpose of this note is to dispute the new estimates, and hence to reassert the conventional chronology. Along the way, however, I wish also to make a more general methodological point about the use of time series projections and direct benchmark estimates of relative per capita income levels. In my view, it is unsatisfactory for researchers to note a large discrepancy and simply claim the superiority of one type of evidence over the other. Rather, a satisfactory account of the evolution of relative per capita incomes over a long period should be able to encompass both sorts of evidence.I am grateful to Nick Crafts, Douglas Puffert, Mark Harrison, Douglas Irwin, Angus Maddison, and Mary O'Mahony for comments on an earlier version and to Knick Harley for editorial advice. Any remaining errors are my responsibility.