Japan's construction and real estate firms borrowed heavily from banks in the late 1980s, but many of them became insolvent when the real estate bubble burst in 1991. Instead of recognizing or restructuring problem loans, Japanese banks continued their loans to these "zombie" firms in order to hide the extent of non-performing loans and avoid immediate loss being added to their balance sheets. The literature shows two channels via which zombie lending affects allocative efficiency: (1) loans to zombie firms could be allocated to more productive firms and (2) the presence of inefficient, but subsidized zombie firms deters entry of productive new firms. In this paper, I propose an additional mechanism via which zombie lending undermines allocative efficiencyzombie lending allows unviable construction projects to continue and hoard inputs that may have better alternative uses. Since construction projects mostly hire low-skilled workers without college degree, subsidized loans to construction industry effectively reduce the availability of these low-skilled workers to other industries, which, in turn, disproportionately slows down the growth of low-skilled industries. To empirically probe this mechanism, I put together an industry-level data set on output, employment, and capital utilization in each of Japan's 47 prefectures between 1992 and 1996. The results show that low-skilled industries experienced disproportionately lower growth rate in real value-added, employment, and capital stock in prefectures where zombie lending increased more. In addition, using the detailed data on the number of establishments by age, industry, and prefecture in the same period, I show that zombie lending also disproportionately slowed down firm entry in lowskilled industries. The results suggest that zombie lending disproportionately harmed the growth of low-skilled industries that competed for non-college educated workers with zombie industries in a local labor market in Japan.