Imagine you run a restaurant kitchen. You hire a new head chef who immediately reorganizes the line cooks — they're faster, sharper, less idle. But the new chef doesn't upgrade the ovens, and actually wastes more ingredients. You got a labor upgrade, not a kitchen upgrade. That's the core finding here: privatization in China's manufacturing sector was a personnel reorganization masquerading as a technology transformation. The committed claim: SOE privatization raises labor-augmenting productivity immediately after transition, but leaves capital-augmenting productivity unchanged and modestly lowers materials-augmenting productivity. Technological change in Chinese manufacturing 1998–2008 was strongly biased toward capital and labor, not materials. This matters because most prior work treats productivity as a single number — total factor productivity (TFP) — which obscures WHERE gains actually land. Wang estimates firm-level factor-specific (capital, labor, materials) augmenting productivity using China's Annual Survey of Industrial Enterprises, covering the entire manufacturing sector from 1998 to 2008. The method decomposes the standard production function into three separate technology channels rather than collapsing everything into a single residual. Private firms sit above SOEs on capital and labor dimensions but below on materials. The privatization effect on labor is front-loaded — it shows up right after transition, then fades across the full panel, though it persists among 1998 incumbent firms. The architecture is structural production function estimation with factor-augmenting technical change — the Oberfield-Raval / Doraszelski-Jaumandreu family. This is not a reduced-form regression; it requires functional form assumptions about elasticity of substitution and how technology enters the production function. The identifying variation comes from the ownership transition itself, comparing pre- and post-privatization productivity within the same plant. The SOE sector's catch-up, interestingly, comes not from privatized firms improving but from the state restructuring the plants it kept and exiting overstaffed operations. Integrity is mixed. The data source — China's industrial survey — is a workhorse in the field, but self-reported firm data in China during this period carries known measurement issues (reported output, intermediate inputs). There's no pre-registration, no external replication, and the full-panel fade of the labor effect is a genuine ambiguity: is the front-loaded gain real and persistent (as the 1998-incumbent subsample suggests), or does it dissipate (as the full panel shows)? The author is honest about this tension, which is a good sign, but the reader should note that the subsample result doing the heavy lifting is also the one with stronger survivorship selection. The milestone question is about non-neutral productivity measurement becoming standard in development economics. Right now, most policy evaluations of privatization, trade liberalization, and industrial policy use single-index TFP. If factor-specific decompositions become routine — requiring better data and more structural assumptions — we'd get much sharper answers about WHO benefits from reform and through what channel. We're maybe 5–10 years from that being standard practice, depending on data availability in major developing economies. The obvious experiment not run: extending this to services and non-manufacturing sectors, and — critically — estimating the WELFARE implications of the biased technical change pattern. If privatization raises labor productivity but lowers materials efficiency, what's the net welfare effect for workers, consumers, and the state? My honest read: the data don't exist at this quality for services, and the welfare calculation requires a general equilibrium model that's a separate paper entirely. This is being saved.
