Essays on the Economics and Econometrics of Production Networks in Development

• Department of Economics

PhD candidate Daniel Prosi will defend his thesis "Essays on the Economics and Econometrics of Production Networks in Development"

This thesis studies interactions between firms in developing countries and how they affect firm-level and aggregate outcomes in the economy. A special emphasis is placed on the role of foreign-owned firms. This is motivated by the widespread use of industrial policies to attract foreign investment.

The thesis consists of three independent chapters. The first chapter empirically studies direct interactions between foreign-owned and domestic firms in Rwanda. It shows that sales interactions with foreign-owned firms induce positive spillovers to domestic buyers of intermediate inputs – so-called forward linkages. The literature instead emphasises backward linkages – domestic firms supplying inputs to foreign-owned firms. The chapter argues that in input-constrained economies with market-seeking investment forward linkages are the dominant channel of productivity spillovers.

The second chapter studies indirect interactions between firms. The analysis of upstream and downstream spillovers is complemented with horizontal effects operating through the competitive structure of the economy. The chapter derives a model of production with intermediate inputs and an endogenous measure of market power in a production network. It then analyses how changes to the productivity or size of a firm propagate over the production network. The model is applicable to a large class of production technologies. The chapter then studies whether the positions of foreign-owned firms in the economy of Rwanda generate monopoly rents that offset the positive spillovers via efficiency gains downstream and demand-creation upstream.

The third chapter explores how treatments of units interacting over an observed network can be identified from an econometric perspective. It proposes a synthetic control estimator that incorporates high-dimensional information from the observed network in the estimation of the synthetic control group. The proposed estimator accounts for interference of treatment assignments over the network and yields estimates for the direct treatment effect in settings when standard synthetic control methods fail.

The event will take place in hybrid modality