Several recent works have explored Wassily Leontief’s distinction between standard econometrics, which he called “indirect statistical inference,” and a “direct induction” he called “direct observation.” These works usually understand Leontief’s direct induction through the lens of input-output analysis. I argue that this is too narrow a perspective. Instead, I show how this distinction stemmed from Leontief’s (1929, 1932a) econometric work, when he developed a statistical technique for determining supply and demand curves. From lesser-known published texts by Leontief from this period, as well as unpublished material from the archives, it appears that Leontief’s distinction was in part borrowed from Jacob Marschak (1931) when they were both in Germany. Like Marschak, Leontief distinguished between two epistemic strategies: indirect, using data from the marketplace, i.e., price-quantity data; or direct, using specific data separately on buyers (e.g., household surveys) and on sellers (e.g., plant surveys). This result fundamentally revises our understanding of Leontief’s view of econometrics.