Industrial Policy and Competition Aghion, Dewatripont, Du, Harrison, Legros

Transcription

1 Industrial Policy and Competition Aghion, Dewatripont, Du, Harrison, Legros Discussion by Jean Imbs Paris School of Economics

2 The Paper is Rich Theory with growth effects of industrial policy Empirical predictions tested on Chinese firm-level data and already Famous.

3 Theory Firms choose between high vs. low growth sectors. Economy can be diversified (one firm in each sector) or focused (specialized? both firms in one sector). Specialization means both firms operate in high growth sector, and must innovate to compete. Growth is high. Diversification means firms operate in different sectors, which offers shelter from competition but at cost of innovation (and growth). Specialization is growth enhancing but does not necessarily obtain in laissez faire.

4 Theory Under diversification, a firm will only shift to high growth if (i) The innovation gains are large (ii) The prospects of collusion are high (i.e. competition is low) If these conditions are not met, diversification will endure, and growth be low. Industrial policy can improve growth if it induces specialization in the high growth sector, i.e. enforces competition, i.e. entices innovation. The growth gains will be larger, the more competitive the sector, since then collusion is unlikely, and innovation must ensue.

5 Empirics lntfp ijt = α 1 Subsidy ijt + α 2 Comp jt + α 3 Subsidy ijt x Comp jt + ε ijt i indexes (Chinese) firms, j sector, t time. Subsidies observed directly. Battery of fixed effects (i, t) Battery of controls: at firm level (ownership ) at sector level (tariffs on inputs/output) TFP measured using Olley-Pakes Estimated in levels or in first differences Alternative LHS: a measure of proportion of output that comes from new products.

6 Empirics lntfp ijt = α 1 Subsidy ijt + α 2 Comp jt + α 3 Subsidy ijt x Comp jt + ε ijt 1) Presumably subsidies have negative correlation with TFP. 2) Key Prediction: Interaction positive subsidies have larger positive effect on TFP when imposed in competitive sector. 3) Interaction is especially positive when subsidies are given equally to all active firms i.e. when they seek to maintain competition. Check marks on all accounts!

7 Comment on Theory What structural transformation are we talking about? Here, a country is not specialized in low growth sectors. It is either diversified, or specialized in high growth sectors. Paper is about a later stage of development, where economies are diversified already. Or at least are not stuck in low growth activities raw materials, low R&D, unskilled manufacturing.

8 Comment on Theory What is missing? Market access. Demand and trade are frictionless No big push issue, market size is irrelevant. Not really about infant industries. Still, about industrial policy in middle income countries.

9 Comment on Empirics lntfp ijt = α 1 Subsidy ijt + α 2 Comp jt + α 3 Subsidy ijt x Comp jt + ε ijt Worry is the endogeneity of subsidies. Suppose subsidies are targeted to low TFP firms. Then α 1 <0. Ceteris paribus, suppose subsidies are lower if competition is low (i.e. subsidies target low TFP firms facing keen competition) Then α 3 >0, since subsidies take (relatively) low value in low competition sectors and low TFP firms. Is this empirically implausible? I don t know.

10 Comment on Empirics lntfp ijt = α 1 Subsidy ijt + α 2 Comp jt + α 3 Subsidy ijt x Comp jt + ε ijt But then why would α 3 be more positive when subsidies are equally allocated across firms? Because there will be many low TFP firms in low competition sectors. And since these are low TFP firms, they will benefit from subsidies.

11 Reinventing hot water lntfp ijt = α 1 Subsidy ijt + α 2 Comp jt + α 3 Ω + ε ijt Obviously authors recognize the importance of endogeneity! 1) Compute Ω 1 =corr(subsidy r jt, Comp r j0) over time, with r denoting city. Replace interaction term with this. City level variables exogenous to firm-level lntfp ijr t. But not clear solves the problem. What if low TFP firms are clustered geographically, across cities. High TFP in Shanghai low TFP in Kashgar. Is this empirically plausible? I don t know.

12 Reinventing hot water lntfp ijt = α 1 Tariff jt + α 2 Comp jt + α 3 Ω 2 + ε ijt 2) Compute Ω 2 =corr(tariff jt, Comp r j0) over time, with r denoting city. Replace interaction term with this, and subsidies with tariffs. Now tariff is sectoral and national. Sector level tariffs exogenous to firm-level lntfp ijr t. Much more convincing. But again, subsidies (tariff) policy may well be driven by sector-specific concerns. Are tariffs in China motivated by some sector specific concerns with low TFP and/or high competition? I don t know. Would be interesting to know more about aggregation level at which tariffs are observed.

13 Conclusion Important paper Useful, tractable, intuitive theory. Probably applicable to structural transformation in middleincome countries. Impressive data, impressive empirical work, impressive results. Would be useful to show us more of the data to set endogeneity concerns to rest.