… is from page 117 of Joel Mokyr’s deep and lovely 2002 book The Gifts of Athena [original emphasis]:
The economic history of knowledge suggests that an emphasis on aggregate output figures and their analysis in terms of productivity growth may be of limited use in understanding rapid growth over long periods. The full economic impact of some of the most significant inventions in the past two centuries would be almost entirely missed in that way. One reason for that has been restated by [Brad] DeLong (2000). Income and productivity measurement cannot deal very well with the appearance of entirely new products…. Traditional measures underestimate the rate of progress and do so at a rate that grows over time.
The (superb) DeLong paper cited here by Mokyr is this one.