Compositional change, aggregation, and dynamic factor demand. Estimates on a panel of manufacturing firms
利用美国制造业企业面板数据，估计基于成本最小化和柯布-道格拉斯技术的动态要素需求模型，发现调整滞后模型表现更优，且大企业要素需求对价格更敏感、对产出更不敏感，证实了使用加总数据确实存在问题。
Abstract Most empirical research on investment and dynamic factor demand has used aggregated data. The large number of authors who have cited this as a source of problems strongly suggests possible benefits from analyzing individual firm data. This paper presents an analysis of a panel dataset of US manufacturing firms. Several models, based on cost minimization and a three‐factor Cobb–Douglas technology, are developed. The differences concern whether the technology varies across two‐digit SIC industries, the presence of fixed adjustment lags, and the determinants of adjustment costs. Identification relies on the rational expectations hypothesis, and estimation on non‐linear 3SLS. The estimates indicate that versions with the adjustment lag perform better than others. Conditional elasticities reveal that factor demand responds rapidly to anticipated changes in output and factor prices, a finding consistent with other recent work. It appears that the factor demand of large firms is more price sensitive and less sensitive to output than small firms, consistent with recent work on credit market imperfections. Comparison of the results based on the pooled and the industry varying technologies indicate that the use of aggregate data is indeed a source of problems.