中文摘要
长期以来，人们认为新机器会提高对技能劳动力的相对需求，然而企业层面的证据并未在投资企业中发现这种转变。我们通过将视角延伸至企业边界之外来解开这一谜题。通过关联比利时关于资本投资、买方-供应商交易和工人技能的数据，我们表明新机器的技能偏向完全产生于外部供应商。在机器投资激增之后，从供应商处的采购使企业对高STEM工人的总需求提高了45%，是其他技能群体涨幅的三倍，而内部雇佣对此需求的提升不到5%，且在各技能群体间均匀分布。这种外部需求首先来自机器集成商，随后来自软件等互补性服务，并且在规模较小的企业、更集中的投资以及快速进步的资本中更强，这与内部雇佣高STEM工人所面临的固定成本和调整成本相一致。在基于我们事件研究估计的一般均衡模型中，2003年至2021年人均机器量的上升使对高STEM工人的相对需求提高了8.8个对数点，并使其工资溢价提高了3.1个对数点。若仅以工资单衡量，二者都会下降。
Abstract
New machinery has long been argued to raise the relative demand for skilled labor, yet firm-level evidence finds no such shift at investing firms. We resolve this puzzle by looking beyond firm boundaries. Linking Belgian data on capital investments, buyer–supplier transactions, and worker skills, we show that the skill bias of new machinery arises entirely at external suppliers. Following machinery investment spikes, purchases from suppliers raise firms’ total demand for high-STEM workers by 45%, three times the rise for other skill groups, while in-house hiring raises it by less than 5%, evenly across skill groups. This external demand comes first from machine integrators and later from complementary services such as software, and is stronger for smaller firms, lumpier investments, and rapidly advancing capital, consistent with fixed and adjustment costs of hiring high-STEM workers in-house. In a general-equilibrium model estimated on our event studies, the rise in machinery per worker from 2003 to 2021 raises relative demand for high-STEM workers by 8.8 log points and their wage premium by 3.1. Measured on payrolls alone, both would fall.