Computers and Productivity: Are Aggregation Effects Important?

This paper examines the empirical implications of aggregation bias when measuring the productive impact of computers. To isolate two specific aggregation problems relating to "aggregation in variables" and "aggregation in relations," we compare various production function estimates across a range of specifications, econometric estimators, and data levels. The results show that both sources of bias are important, especially as one moves from the sector to the economy level, and when the elasticity of all types of non-computer capital are incorrectly restricted to be equal. In terms of computers, however, the estimated elasticity is surprisingly stable between industry and sector regressions and does not appear to be biased by the incorporation of a restrictive measure of non-computer capital. The data consistently show that computers have a large impact on output.