I estimate the local government spending multiplier to be 14 at the prefecture city level in China during 2001-2019. To overcome identification challenges, I construct a novel instrument for local government spending using the fraction of unoccupied raw land in the downtown area in 2000. After the land market was formalized in 2000, a higher fraction of unoccupied raw land implies higher net profits from land sales for local governments. The fraction is uncorrelated with either the level or the growth of economic activities and infrastructure before 2000, or net land supply after 2000. The multiplier is larger for cities with higher GDP per capita and GDP growth rate before 2000, and does not depend on the benefit from WTO entry or the initial level of infrastructure. There are large positive spillovers within but not across cities. The large spending multiplier can be explained by the facts that local government spending has been productive. The increase of local government spending increases labor demand and wages, firm entry and local firms’ productivity and output, especially for industries with more use of transportation services. It also generates positive spillovers through technology and supply chains.