Land use regulations have been pursued by many governments around the world. When making land development policies for cities that feature high population density and limited land supply, one question looms large: where and how much to build across different localities of a city? To answer this question, it is essential to understand how land development policies shape (or reshape) the spatial distribution of population and economic activities within a city because such internal urban structure has fundamental and persistent impacts on urban economic development and welfare (Lucas and Rossi-Hansberg, 2002; Rossi-Hansberg, 2004). Ahlfeldt, Redding, Sturm, and Wolf (2015) and Heblich, Redding and Sturm (2020) use a spatial GE framework to demonstrate how historic shocks to urban transportation networks reshaped the urban structure of Berlin and London respectively and in turn influenced their economic development and urban growth. However, there has been few attempts to study the impact of land use regulations (especially heterogeneous land development policies within a city) on the spatial distributions of urban population and economic activities and the consequent welfare implications. This paper aims to fill the gap.Our study faces two main challenges: On the one hand, high building density can accommodate more residents and employment, which will through agglomeration economies further enhance local amenities and attract even more population and employment. On the other hand, high residential and workplace employment densities are likely to generate crowdedness effect and lengthen commuting time and cost. While the former issue has been studied in the literature, the latter one has seldom been addressed in a spatial equilibrium framework. In order to conduct a comprehensive welfare evaluation of land use regulation policies, it is crucial to incorporate the above two forces into a unified framework. This paper develops a spatial general equilibrium model that incorporates both the positive externality of agglomeration economies and the negative externality of commuting congestion. We then calibrate the model. Using it as workhorse, we carry out counterfactual analyses to evaluate the effects of various land regulation policies on urban internal structure and welfare. As a sensitivity check, under the present land use regulations, we find that our predicted equilibrium outcomes can fit the current data quite well, much better than the model without the commuting crowdedness effect. This paper focuses on Shanghai specifically, the most populous city in China. Unlike other cities in market economy countries, the city government of Shanghai has been regulating housing supply across various localities in Shanghai, mainly through limited land supply and building density (floor-to-area ratio) constraints. According to Shanghai’s 2017-2035 master plan, by 2035 the whole city is going to build up 270 million square meter residential floor space in addition to the current 1.05 billion square meter total floor space. Where to build up these new floor space? The plan also delineates the spatial boundaries of one core urban area around the city center and 14 sub-centers surrounding the core urban area. Those center or sub-center areas may enjoy the priority of land development or redevelopment. However, our data suggests that the fundamental (exogenous) local living and production amenities in those center /or subcenters are not necessarily higher than other tracts in Shanghai. Moreover, employment densities in those areas are already high which alerts commuting congestion. Given this, we try various alternative ways of allocating the new floor space across different tracts in Shanghai and evaluate their corresponding equilibrium outcomes in our counterfactual analyses.We first construct a comprehensive data set based on various data resources that contains information on the residential and workplace employment, housing development densities and prices, land use regulations, wages, and bilateral travel flows and times at the tract level (jiedao or zhen in Chinese) within the city of Shanghai. Using this data, we estimate and calibrate the parameters of the model, including commuting cost parameter with respect to travel time, worker utility dispersion parameter, the values of tract-level fundamental production and residential amenities, and congestion elasticities with respect to both residential and workplace employment densities. Next, using the calibrated and estimated model parameters, we conduct counterfactual analysis to evaluate alternative ways of allocating the additional floor space to different locations of the city. The analysis allows us to identify the contributions of various mechanisms to the improvement of welfare resulting from building up more floor space, including agglomeration externalities, commuting costs, and housing costs. Our counterfactual analyses generate three sets of findings. First, allocating floor space to locations with more advantageous local fundamentals (e.g., transport connections, schools, hospitals, green space) can generate welfare gains. Relative to the initial welfare level which is our benchmark and is normalized to one, if we re-allocate the initial floor space to each tract according its own amenity levels, the welfare gain is 70%. This is consistent with the argument of Sturm et. al (2022). Informed by the above result, in all of the following counterfactuals, we allocate the new floor space across tracts according to each tract's fundamental amenities in the respective targeted areas to be developed. Second, compact development can generate a higher utility level than spreading-out development. Specifically, we divide the city into several 10 km-wide belts: 20-30km, 30-40km, 40-50km, and 50-60km. We experiment with allocating all the new floor space to each different 10 km-wide belts respectively. We find that they all generate higher welfare gains (around 12% relative to the initial level) than spreading out the new floor space all over the city which generates a welfare gain of 9.58%. This is because compact development can generate more supply of housing within a smaller geographical scale and in turn facilitate the rising of agglomeration economies without increasing the housing price. Thirdly, decentralized development works better that centralized ones for Shanghai. Among all the 10 km-wide belts along the radius from the city center, the one that is nearest to the city fringe (i.e., the 50-60km belt) is the best location for new housing development. Allocating new floor space to this area according to the values of local fundamentals can generate a welfare gain of 15.42%, relative to the initial level. Within this belt there locate several contiguous tracts that have fundamental amenities above the city average. Meanwhile, they are sparsely populated now. So they have great potential to accommodate more population and firms in the future.By contrast, allocating new floor space to either 0-10km or 10-20km belts works poorly. Further, if we focus on developing the government-specified core urban area (i.e., the 35 tracts near the city center), we can get a welfare gain of 12% relative to the initial level, which is also significantly lower than the above more decentralized development pattern. Why? This is because the core is already very crowded. If increasing building density here, the positive agglomeration effect will be dominated by the negative commuting crowdedness effect. If we decentralize the new floor space to the sub-centers (Counterfactual C8), we can get a welfare gain of 14.12%, relative to the initial level. The wage gain from agglomeration effect is the strongest, which is not surprising since those subcenters have high productivity and amenities. However, the higher housing price somehow drags down the overall welfare gain. In sum, our findings suggest that a more decentralized yet compact land development according to individual tract amenities seem to be a sensible way to build up Shanghai in the future
