The shareholdings of French multinational enterprises (MNE) differ across destination countries and across individual MNE's life cycles. The average shareholdings are larger in countries that are highly-developed, geographically, and culturally closer to France; productive MNEs match with productive local firms; MNEs follow a gradual internationalization process where foreign experience increases the average shareholdings of following affiliates and the probability of wholly-owning. I build and calibrate a dynamic model featuring heterogeneous firms in both mobile know-how (MKH), transfer of local know-how (LKH) from partners, and accumulation of LKH for future entries. The composition of LKH and MKH determines MNE's entry decision. The distribution of LKH and MKH generate a positive assortative matching pattern as in the data. Compared with a static model, the dynamic model predicts a much larger contribution of LKH and production loss to ownership restriction, which cannot be recovered using static data. In a counterfactual where the LKH transfer cost of Romania is set equal to Germany, French MNE production will increase by 80%, where 30% comes from dynamic effects.

Agglomeration Externalities and Industry Concentration Dynamics, with Dan Lu Link

Manufacturing industries are highly concentrated geographically, and firms in larger regions are larger and grow faster. We focus our study on the effect played by the agglomeration: producers benefit from the existence of nearby producers. Empirically, this paper uses Chinese Census of Industries data to provide evidence of region-size premium on entrant number, sales and sales growth rate, and a changing industrial concentration. Theoretically, this paper builds a dynamic model where firms make location choices, and the agglomeration effect determines how fast firms update their technology. In the calibration, we infer natural advantage changes from observed entrant shares and identify the model-based agglomeration effect using the correlation between the average sales growths and entrant share growths. We simulate the model to match China's entrant share changes from 1998 to 2007. Quantitatively, we show that agglomeration effects can explain one-fourth of the firm-level productivity growth, one-eighth of the sales growth, and two-thirds of the average productivity difference in the largest regions (5th quantile) versus the smallest regions (1st quantile).

What Do Alibaba Data Tell Us about Quality Growth in China, with Mark Bils, Min Fang, & Zinbin Huang

Conglomerate Market Power, with Min Fang, & Xiaomei Sui

We study how firms exert market power through their ownership network and its macroeconomics implications. We first document such empirical evidence using European firm-level data. We then build a dynamic model with heterogeneous firms, ownership networks, and endogenously variable markups. In the model, a controlling parent firm expands through both intensive margin in own production and extensive margin in acquiring/establishing subsidiaries. It then influences the production and pricing decisions of its subsidiaries through its corresponding voting right. As a result, a controlling parent firm exerts market power through its ownership network. We calibrate our model to match European firm-level data and quantify the mechanism. We find that market power is substantially underestimated without accounting for the ownership network. Finally, we show how the conglomerate market power affects the aggregate economy with respect to aggregate shocks and policy interventions.