Shocks ” undertaken at Research Institute of Economy, Trade and Industry (RIETI). The authors are grateful to

Extending the endogenous growth model proposed by Young (1998), we construct a two-sector growth model that explains the observed pattern of structural change. Unlike existing studies, we assume neither non-homothetic preferences nor exogenous differential in productivity growth among different sectors. Our key assumption is that any two goods produced by firms in a sector are closer substitutes than those produced by firms in other sectors, which gives rise to an endogenous differential in productivity growth among different sectors. When the two composite goods are poor substitutes, the share of employment gradually shifts from the sector with a low markup to that with a high markup. To test the validity of the prediction of our model, we also estimate industry-level total factor productivity (TFP) growth and markup using Japanese firm-level panel data. The empirical results show a negative correlation between estimated markup and long-term TFP growth, and a positive correlation between the growth of industrial labor input and markup, which supports our theoretical results. Finally, in contrast to the previous studies of structural change that consider competitive economies, we study the socially optimal allocation and characterize the optimal ta