Econometrica: Apr, 1959, Volume 27, Issue 2

The Validity of Cross-Sectionally Estimated Behavior Equations in Time Series Applications

https://doi.org/0012-9682(195904)27:2<197:TVOCEB>2.0.CO;2-J
p. 197-214

Edwin Kuh

Regression estimates from cross-section and time series sample data are often different. Some reasons why these discrepancies arise are presented, along with quantative results for three investment functions, explaining investment by gross internal funds and the firm's capital stock, and using individual firm observations. A substantial analytical advantage can be gained from having both time series and cross-sections for an identical group of firms so that the error variance structure for estimates based on both sorts of data can be analysed efficiently.