Should governments subsidize firms' own innovation or adoption of foreign technology? How does the answer change over different stages of development? To answer these questions, we digitize the universe of technology transfer contracts between domestic and foreign firms in South Korea during its growth miracle period. This data has novel information on the price of technologies. We find that, when the productivity gap between domestic and foreign firms is larger, 1) productivity increases more after adoption, 2) the adoption fee is lower, and 3) domestic firms more often choose technology adoption over innovation. Motivated by these findings, we build a two-country growth model with endogenous adoption and innovation decisions. Foreign firms can sell technologies for an endogenous fee, internalizing the future loss of profit due to stronger competition with domestic firms. By construction, adoption can raise domestic firms at most to the technology level of foreign firms. Therefore, as domestic firms close the productivity gap, the expected productivity gain from adoption decreases, making an adoption subsidy less effective than an innovation subsidy. We evaluate Korea's technology policies since 1973, which started with an adoption subsidy and shifted to an innovation subsidy as the productivity of Korean firms converged with that of foreign competitors. Our result suggests that this state-dependent policy increased consumption-equivalent welfare by 5%, which raises welfare more than time-invariant policies that subsidize only innovation or adoption throughout. Our analysis also shows that the optimal year to switch from an adoption to an innovation subsidy would have been 1985, when Korea's GDP reached 55% of Japan's.

This paper develops a measure of Capital-Embodied Innovation (CEI). The measure counts the number of patents applied to capital goods by matching patent documents with Wikipedia articles on capital goods. Using occupation-level variations on the sets of capital goods from O*NET, we document that CEI is biased toward abstract and non-routine occupations. Furthermore, we highlight the heterogeneous effects of CEI across the capital-occupation relationship. When the capital good performs a similar function as the occupational task (task-substituting capital), the CEI reduces the relative demand for labor. In case the capital good performs a different function than the occupation tasks (task-complementing capital), the CEI raises relative demand for labor. Abstract occupations have disproportionately more CEI on task-complementing capital than non-abstract occupations. A model-based counterfactual implies that the employment growth between the 1980s and the 2010s would be 37% less biased towards abstract-task occupations without CEI. The degree of job polarization would have also been lower without CEI.

We study how the adoption of foreign technology and local spillovers from such adoption contributed to late industrialization in a developing country during the postwar period. Using novel historical firm-level data for South Korea, we provide three empirical findings: direct productivity gains to adopters, local productivity spillovers of the adoption, and complementarity in firms' adoption decisions. Based on these findings, we develop a dynamic spatial model with firms' technology adoption decisions and local spillovers. The spillovers induce dynamic complementarity in firms' technology adoption decisions. Because of this complementarity, the model potentially features multiple steady states. Temporary adoption subsidies can have permanent effects by moving an economy to a new transition path that converges to a higher-productivity steady state. We calibrate our model to the microdata and econometric estimates. We evaluate the effects of the South Korean government policy that temporarily provided adoption subsidies to heavy manufacturing firms in the 1970s. Had no adoption subsidies been provided, South Korea would have converged to a less industrialized steady state in which the heavy manufacturing sector’s share of GDP would have been 15 percentage points lower and aggregate welfare would have been 10% lower compared to the steady state with successful industrialization. Thus, temporary subsidies for technology adoption had permanent effects.