Rethinking industrial policy. Philippe Aghion

Transcription

1 Rethinking industrial policy Philippe Aghion

2 In aftermath of WWII, many developing countries have opted for trade protection and import substitution policies aimed at promoting new infant industries

3 Classical infant industry argument Some new activities involve high costs at the beginning, but learning by doing reduces these costs over time There are knowledge externalities between these activities and the rest of the economy Then there may be positive dynamic externalities from temporarily protecting and/or subsidizing the new activities

4 Industrial policy has acquired a bad name It allows governments to pick winners It thus increases the scope for capture of governments by vested interests It thus bias competition but we know how important competition and free trade are for growth and innovation (my argument with Dani!)

5

6 Recent period: three motives for a rethink New post-crisis realism: laissez-faire complacency by several governments has led to inefficient growth of non-tradable sectors at the expense of tradables Climate change: path dependence in the direction of innovation leads firms that have innovated dirty in the past will keep on innovating dirty in the future, hence role for government to redirect technical change China: a big deployer of sectoral aid, whose success induces other countries to try and emulate its economic policies

7 The theoretical case for industrial policy in industrialized countries Free entry and growth of firms limited by insufficient financial development even in developed countries There are knowledge externalities between sectors even in developed countries Environment and directed technical change Industrial policy can be made competition friendly and may even help enhance competition

8 Governing industrial policy The question is not so much whether or not we should forbid or preclude industrial policy, but rather how industrial policy should be designed and governed. Some new ideas Selection of sectors: skill-biased (Nunn-Trefler (2010)); competitive sectors (this paper); Governance: do not focus aid on one firm in a sector, minimize concentration of aid (this paper). 8

9 Why industrial policy is not inconsistent with competition Industrial policy may push firms to compete more closely: while firms naturally try and differentiate in order to reduce the intensity of competition, here industrial policy means inducing them to focus on some specific problems/technologies. 9

10 Some suggestive evidence

11 Evidence from EU Set of regressions using Market Share as dependent variable: EU15 member States excluding Portugal The data covers the period running from 1992 to 2008, except Sweden, Austria and Finland for which the period covered is and Luxembourg due to data availability. Set of regressions using Patents as dependent variable: EU15 member States The data covers the period running from 1992 to 2008, except Sweden, Austria and Finland for which the period covered is Main sources: DG Comp European State Aid Scoreboard and OECD databases.

12 The variables (1) «Market share» refers to share of exports of country i to total exports within EU15 in manufacturing and service «SA» refers to total state aid of country i to industry and service. Unit : in millions of euros «Decent» refers to a proxy for decentralisation of state aid. We used the rates of decentralisation of subsidies to «economic affairs», i.e. the numerator is composed of subsidies coming from local or regional states, whilst the denominator is the sum of subsidies from central, regional and local government.

13 The variables (2) «Private» refers to private credit by deposit money banks and other financial intermediaries to GDP. «PMR» refers to the average value of product market regulation using 1998, 2003 and 2008 values. «Patents» refers to total applications filed under the Patent Cooperation Treaty, at priority date for the year and at the applicant s country of origin for the country, using fractional counts. «BERD» refers to total R&D expenditures in the Business entreprise sector. The unit of measure is in million of euros.

14 Market share on state aid and decentralisation Log(Market Share) Log(SA) 0.167*** Log²(SA) Decent SA * Decent (0.063) *** (0.004) ** (0.05) 1.22e-05*** (4.48e-06) Observations 202 Standard errors in parenthesis «***» indicates significance at 1%, «**» indicates significance at 5%, «*» indicates significance at 10% Fixed effects regression controlling for time effects, initial GDP, and labour costs. A constant was included.

15 Market share on state aid and private credit Log(Market Share) (1) (2) Log(SA) 0.14*** 0.168*** (0.048) (0.049) Log²(SA) *** *** (0.003) (0.003) Log(Private) 0.135** 0.128* (0.067) (0.067) Log²(Private) 0.044*** 0.049*** (0.013) (0.014) Log(SA) *Log(Private) -0.02** * (0.009) (0.01) Observations (1) Fixed effects regression, without any controls. (2) Fixed effects regression, controlling for time effects. In both (1) and (2), a constant was included

16 Market share on state aid and product market regulation Log(Market Share) Log(SA) 0.046*** (0.009) Log (PMR) (dropped) SA *PMR -1.39e-06* (7.12e-07) Observations 222 fixed effects regression controlling for time effects and initial GDP. A constant was included.

17 Patents on state aid and decentralisation Patents SA *** (0.034) Decent ( ) SA * Decent 0.632*** (0.101) Observations 176 fixed effects regression controlling for time effects, initial GDP, and BERD. A constant was included.

18 Patents on state aid and private credit Patents SA (0.066) Private ( ) SA * Private * (0.084) Observations 199 fixed effects regression, controlling for time effects and BERD. A constant was included.

19 Competition and sectoral policy using Chinese data Current work with Mathias Dewatripont, Luosha Du, Ann Harrison, and Patrick Legros Panel data of Chinese firms, Industrial firms from NBS: annual survey of all firms with more than 5 million RMB sales Regress TFP on: Subsidies received by firm as a share of sales COMP=1 - LERNER INDEX Sector-level controls, firm and time fixed effects

20 Estimation lntfp ijt = α + β Z ijt + β S + β SUBSIDY * COMP jt jt + β SUBSIDY + α + α + ε i 3 t ijt ijt + β COMP 4 jt Z=Vector of firm-level controls, including state and foreign ownership S=Vector of sector-level controls, including input and output tariffs, sectoral foreign shares. All specifications allow for firm fixed effects and time effects. Three Approaches: OLS, OLS with fixed effects, Olley-Pakes approach to measuring TFP in first stage

21 Rethinking industrial policy Findings are that: The higher competition, the more positive (or less negative) the effect of subsidies on average TFP The overall effect of subsidies on TFP is positive if competition is sufficiently high and/or subsidies are not too concentrated among firms in the sector

22 Results Table 1 (1) (2) (3) (4) (5) (6) VARIABLES lntfp (based on Olley-Pakes regression) Stateshare ( ) ( ) ( ) ( ) ( ) ( ) Horizontal 0.322*** 0.335*** 0.323*** 0.335*** 0.178* 0.198* (0.0756) (0.0793) (0.0755) (0.0793) (0.0947) (0.101) Ratio_subsidy *** *** *** *** *** *** (0.0279) (0.0276) (1.769) (1.404) (1.748) (1.392) Competition_lerner (0.533) (0.535) (0.535) Interaction_lerner 8.212*** 6.724*** 8.074*** 6.773*** (1.818) (1.441) (1.796) (1.429) Backward 0.779*** 0.762*** (0.278) (0.273) Forward (0.0991) (0.0990) LnTariff ** ** ** ** (0.0162) (0.0166) (0.0162) (0.0166) (0.0214) (0.0213) LnbwTariff (0.0174) (0.0172) (0.0174) (0.0172) (0.0194) (0.0189) LnfwTariff ( ) ( ) ( ) ( ) ( ) ( ) Constant 1.726*** 1.213** 1.725*** 1.242** 1.699*** 1.274** (0.0315) (0.534) (0.0314) (0.535) (0.0412) (0.533) Observations 1,072,034 1,072,034 1,072,034 1,072,034 1,072,034 1,072,034 R-squared Notes: Robust clustered standard errors are shown in the parenthesises. Firm fixed effect and time effect are included in each specification. To exclude foreign-invested and state-owned firms, we estimate the results based on the sample of domestic non-state-owned firms.

23 Interacting with Herfindahl Table 2 (1) (2) (3) (4) (5) (6) Dependent: lntfp (based on Olley and Pakes regression) The second quartile: more dispersion in subsidies Ratio_subsidy * ** *** *** *** *** (0.0962) (0.0937) (4.884) (4.037) (4.813) (4.031) Competition_lerner (1.286) (1.285) (1.308) Interaction_lerner 16.63*** 12.24*** 16.88*** 12.19*** (5.096) (4.186) (5.023) (4.178) The fourth quartile: least dispersion in subsidies (most concentrated) ratio_subsidy *** *** ** ** ** ** (0.0625) (0.0627) (3.615) (2.854) (3.710) (2.860) competition_lerner (0.981) (0.982) (1.042) interaction_lerner 9.320** 6.069** 9.107** 6.238** (3.628) (2.883) (3.727) (2.888) Horizontal Yes Yes Yes Yes Yes Yes Forward & Backward No No No No Yes Yes Tariffs Yes Yes Yes Yes Yes Yes

24 Product innovation Here, we use the new product ratio as the dependent variable. New product ratio is defined as the share of output value generated by new products to the total output value. Table 6 (1) (2) (3) (4) (5) (6) Dependent: Ratio_newproduct The second quartile Ratio_subsidy * ** * ** (0.0390) (0.0388) (0.821) (0.755) (0.816) (0.755) Competition_lerner (0.0789) (0.0780) (0.0720) Interaction_lerner 1.562* 1.755** 1.568* 1.744** (0.841) (0.780) (0.837) (0.780) The fourth quartile ratio_subsidy (0.0351) (0.0352) (1.475) (1.442) (1.468) (1.432) competition_lerner 0.117* 0.114* 0.122* (0.0662) (0.0657) (0.0622) interaction_lerner (1.503) (1.470) (1.495) (1.460) Horizontal Yes Yes Yes Yes Yes Yes Forward & Backward No No No No Yes Yes Tariffs Yes Yes Yes Yes Yes Yes

25 Climate and directed technical change Basic idea: due to knowledge externalities, under laissez-faire firms will tend to innovate dirty. Hence a role for government intervention in redirecting technical change (carbon tax, research subsidies)

26 Empirical investigation We use patents to measure innovation EPO World Patent Statistical database (PATSTAT) We use data on fuel prices (including taxes) to proxy the effect of a carbon tax International Energy Agency

27 The data set 12,000 patents in clean technologies Electric vehicles, hybrid vehicles, fuel cells 36,000 patents in dirty technologies Regular combustion engines Filed by 7,000 patent holders Between 1978 and 2007

28 The variables The dependent variable: ratio between the number of Clean and Dirty Patents Explanatory variables: The fuel price (weighted by the geographical distribution of the firm s patent stock) The firm s stock of clean patents and stock of dirty patents 2 interaction variables to measure the effect of the fuel price conditional on the size of the clean/dirty patent stocks Control variables: country X year fixed effects (control for national technology-push policies and other economy-wide effects)

29 Results Variable Model 1 Fuel price Stock of clean patents Stock of dirty patents 0.590*** (0.150) 0.141*** (0.014) *** (0.014) An increase in the price of fuel favours clean innovation relatively to dirty innovation Firms build on their existing stock of technology-specific knowledge

30 Conclusion (1) Competition policy and single market must be pushed further In particular competition and entry limited by insufficient market finance and by insufficient investment in reallocation and retraining of workers

31 Conclusion (2) Yet, we need industrial policy Entry is limited by lack of finance and by lack of early customer (hence to role of public procurement) Need to redirect technical change (climate, finance) Industrial policy can be made competition friendly and even enhance competition

32 Conclusion (3) Industrial policy should select sectors, not particular firms Industrial policy should select sectors with large growth externalities (green innovations, health, nanotechnologies, numerique..) Importance of exit mechanisms and ex post evaluation Industrial policy should not prevent exit, hence the importance of not getting stuck with bad projects (role for cofinancing) Role of competition commission to evaluate competitive effects of sectoral aids ex post

33 Conclusion (4) EIB should put more emphasis on skill intensive sectors such as energy and information, not just on infrastructure and transports Competition Commission should move from ex ante legalistic to a more ex post evidence-based approach to sectoral state aids At the same time one should still push non-targeted policies such as subsidizing higher education, R&D, innovative inputs (lab equipments,..) and training..and one should actively pursue the achievement of a single market for goods and services!!