Trade and frictional unemployment in the global economy

Transcription

1 Trade and frictional unemployment in the global economy Céline Carrère, Anja Grujovic Frédéric Robert-Nicoud,, Geneva School of Economics and Management HEC Lausanne CEPR FIW, Vienna, 27 June 2017

2 What we do Develop a theory displaying rich effects of trade on unemployment: Multi-country, multi-sector gravity model with trade frictions (multi-sector EK as in Costinot, Donaldson and Komunjer, 2012) Labour market frictions and equilibrium unemployment ( static DMP as in Helpman and Itskhoki, 2010) Provide a structural estimation of key parameters of the model Quantify welfare and employment effects in counterfactual scenarii Abrogate NAFTA TTIP TPP Balanced trade

3 What we find (Theory) 1 Trade unemployment through expansion and reallocation effects 1 u i ( ) l i ω 1 λ [ = i 1 Cov (s ik s ik, µ ] ik) 1 u i l i ω i µ i 2 Trade liberalisation may result in higher unemployment rate 3 acr real wage gains from trade formula generalizes to ω i ω i = ( π ii π ii ) 1 λ θ i and encompasses Ossa (2015) and Heid and Larch (2016) 4 Welfare real wages when society is averse to inequality unemployment also matters

4 What we find (Quantitative) 1 Sector-specific labor market frictions Heterogeneous Robust across time and oecd countries 2 Unemployment rises for some members of ttip and tpp even in presence of gains from trade 3 Repeal of nafta leads to an increase in unemployment of 20% in Mexico of 6% in the us

5 Motivation Significant and persistent differences in unemployment rates across US manufacturing sectors

6 Related literatures Labor market outcomes of trade: Brecher (1974); Davidson, Martin, and Matusz (1988, 1999); Davis (1998); Costinot (2009); Helpman and Itskhoki (2010); Helpman, Itskhoki, and Redding (2013); Autor, Dorn, and Hanson (2013); Carrère, Fugazza, Olarreaga, and Robert-Nicoud (2014); Heid and Larch (2016). Labor market frictions as source of comparative advantage: Cuñat and Melitz (2012); Davidson, Martin, and Matusz (1988); and Helpman and Itskhoki (2010). Economic effects of TTIP and TPP:Bertelsmann (2013); Aichele, Felbermayr, and Heiland (2014); Egger, Francois, Manchin, and Nelson (2015); Felbermayr, Heid, Larch, and Yalcin (2015); Fontagné, Gourdon, and Jean (2013); Francois, Manchin, Norberg, Pindyuk, and Tomberger (2013); Petri, Plummer and Zhai (2012). Gains from trade: Atkeson and Burstein (2010); Arkolakis, Costinot, and Rodríguez- Clare (2012); Head, Mayer, and Thoenig (2014); Melitz and Redding (2015); Edmond, Midrigan, and Xu (2015); Ossa (2015; Heid and Larch (2016). Welfare criterion in a trade context: Antràs, de Gortari, and Itskhoki (2015); Galle, Rodríguez-Clare, and Yi (2015); Porto (2015).

7 Plan 1 Introduction 2 Model 3 Estimation 4 Counterfactuals 5 Summary

8 EK Consumption, production, and trade Environment One factor, Labor K sectors I countries

9 EK Consumption, production, and trade Environment One factor, Labor K sectors I countries Preferences Cobb-Douglas across sectors ces across varieties within sectors

10 EK Consumption, production, and trade Environment One factor, Labor K sectors I countries Preferences Cobb-Douglas across sectors ces across varieties within sectors Technology and Ricardian comparative advantage { ( ) } θk z tfp: P {z ik (x) > z} = 1 exp θ k governs the scope for comparative advantage z ik

11 EK Consumption, production, and trade Environment One factor, Labor K sectors I countries Preferences Cobb-Douglas across sectors ces across varieties within sectors Technology and Ricardian comparative advantage { ( ) } θk z tfp: P {z ik (x) > z} = 1 exp θ k governs the scope for comparative advantage Trade costs Iceberg: τ ijk 1 and τ iik = 1 z ik

12 DMP Search and matching frictions in labour market Timing: 1 Firms open vacancies and workers choose a sector Matching function: H ik = µ ik V 1 λ ik L λ ik, Matching efficiency adjusted for vacancy costs: µ ik µ ik ν 1 λ ik 2 Job matches are realized and sector-specific wages w ik are determined

13 DMP Search and matching frictions in labour market Timing: 1 Firms open vacancies and workers choose a sector Matching function: H ik = µ ik V 1 λ ik L λ ik, Matching efficiency adjusted for vacancy costs: µ ik µ ik ν 1 λ ik 2 Job matches are realized and sector-specific wages w ik are determined Equilibrium: Ex-ante labour mobility and risk neutral workers together imply: Employment rate in sector k: l ik w ik = w i l ik = ω 1 λ i µ ik

14 EK + DMP Equilibrium Gravity equation: E ijk = π ijk α jk Y j, where π ijk = ( ) θ tijk T jk where and remoteness: t ijk τ ijk wi λ P 1 λ i z ik µ ik ( ) 1/θ T jk i =1 t θ i jk

15 EK + DMP Equilibrium (cont.) Generalized ACR (2012) formula: ω i ω i = ( πii π ii ) 1 λ θ i, where π ii and θi ( ) αik K θ i θ π k iik (0, 1) k=1 ( K k=1 α ik θ 1 k ) 1 Encompasses generalizations to multiple sectors (Ossa, 2015) to endogenous employment (Heid and Larch, 2016)

16 EK + DMP Equilibrium (cont.) Trade shock and employment effects: Two effects: Expansion effect: ˆl i = (1 λ)ˆω i + 1 µ i K k=1 ŝ ik (µ ik µ i ) } {{ } =Cov(ŝ ik,µ ik ) Reallocation effect: (1 λ)ˆω i > 0 = ˆl i Cov(ŝik,µ ik )=0> 0 Cov (ŝ ik, µ ik ) < 0 = ˆl i ˆωi =0< 0

17 EK + DMP Equilibrium (cont.) Social welfare function and inequality aversion (Atkinson, 1970): W(η) [ K k=1 s k l k ω 1 η k = ω (1 u) + ub1 η }{{} between inequality + (bω) 1 η u ] 1 1 η + E ( ω 1 η) (Eω) 1 η ω 1 η } {{ } within inequality 1 1 η, using agi correction.

18 EK + DMP Equilibrium (cont.) Social welfare function and inequality aversion (Atkinson, 1970): W(η) [ K k=1 s k l k ω 1 η k = ω (1 u) + ub1 η }{{} between inequality + (bω) 1 η u ] 1 1 η + E ( ω 1 η) (Eω) 1 η ω 1 η } {{ } within inequality 1 1 η, using agi correction. Setting η = 2 and b 0: W W 1 = ω /u ω/u 1 ω ω u u

19 Plan 1 Introduction 2 Model 3 Estimation 4 Counterfactuals 5 Summary

20 Estimation Gravity Standard gravity regression (cross section, 2008): ln E ijk = β tariff k Data: ln(1 + tariff ijk ) + βk rta RTA ij + β D k ln Distance ij + β cont k Contiguity ij + β lang k Common Language ij + βk colon Common Colonial Empire ij + βk curr Common Currency ij + FE ik + FE jk + ɛ ijk Trade flows: CEPII (BACI database) Tariffs: UNCTAD (TRAINS database) RTA dummy: Jeffrey Bergstrand Sectoral production: OECD (ISICRev3 database)

21 Estimation Labor-market matching efficiencies, µ k From l ikt = ω 1 λ i µ ikt, assuming µ ikt = µ it µ k : ln l ikt = ln ( µ it ω 1 λ it ) + ln µk, If employment rate data is available (15 aggregate sectors k ): ln l ikt = FE it + FE k + ν it If unavailable (21 disaggregated manufacturing sectors m ): µ manuf = m s imt µ m One step estimation: ( 21 ) ln l ikt = FE it + (1 I k ) ln µ k + I k ln s imt µ m + error ikt, where I k = 1 if k = manuf, otherwise 0. m=1

22 Estimation Labor-market matching efficiencies, µ k (cont.) Data (35 sectors, 25 oecd countries, ): Yearly unemployment rates by country and sector: ILO (KILM database) Unemployment rates by manufacturing sector (USA): BLS Examples examples

23 Estimation Labor-market matching efficiencies, µ a (15 aggregate sectors) 90% confidence intervals for estimates of 15 µ ik s with available employment rate data

24 Estimation Labor-market matching efficiencies, µ m (21 manufacturing sectors) 90% confidence intervals for estimates of 21 disaggregated manufacturing sectors

25 Correlation between µ m and us employment rates (manufacturing sectors), ρ (0.56, 0.74)

26 Correlation between µ m and sector-specific volatility (Cuñat and Melitz, 2011), ρ = 0.29

27 Estimation Labor-market matching efficiencies, µ k (cont.) Robustness checks Stability of µ k estimates across countries and time µ k Out-of-sample prediction of employment rates out

28 Plan 1 Introduction 2 Model 3 Estimation 4 Counterfactuals 5 Summary

29 Calibration results: ttip ( u i 2008 k sttip ) ( ik µ k u TTIP ) ( k s i µ ik k u i ω TTIP ) ( ω W TTIP ) W TTIP Austria members Belgium Czech Republic Denmark Estonia Finland France Germany Greece Hungary Ireland Italy Netherlands Poland Portugal Slovenia Spain Sweden United Kingdom EU-19 Average United States Other Canada OECD Iceland Israel Japan Mexico New Zealand Norway Switzerland

30 Calibration results: ttip (cont.)

31 Calibration results: wall scenario ( k swall ) ( ik µ k u WALL ) ( k s i µ ik k u i ω WALL ) ( ω W WALL ) W NAFTA United States members Mexico Canada EU-19 Austria Belgium Czech Republic Denmark Estonia Finland France Germany Greece Hungary Ireland Italy Netherlands Poland Portugal Slovenia Spain Sweden United Kingdom EU-19 Average Other Iceland OECD Israel Japan New Zealand Norway Switzerland

32 Calibration results: wall scenario (cont.)

33 Plan 1 Introduction 2 Model 3 Estimation 4 Counterfactuals 5 Summary

34 Summary and Conclusion A multi-country, multi-sector Ricardian/Gravity trade model Labour market frictions and equilibrium unemployment Inequality or risk aversion Unemployment matters Comparative advantage matters for unemployment Estimate sector-specific labour-market frictions Counterfactual exercises Repeal of nafta increase in unemployment: 21% in Mexico and 6% in us ttip increase in real wages (0.2% in eu, 0.3% in us) ttip unemployment: 0.85%-increase in us, 1.2%-reduction in eu

35 Appendix

36 Estimates of sector-specific matching efficiencies High matching efficiencies ISIC Rev 3 Sector description Matching eff. 26 Other non-metallic mineral products Printing and publishing Textiles Leather, leather products and footwear Fabricated metal products, except machinery and equipment Pulp, paper and paper products Low matching efficiencies ISIC Rev 3 Sector description Matching eff. 55 Hotels and restaurants Food, beverages and tobacco products Radio, television and communication equipment Coke, refined petroleum products and nuclear fuel Other transport equipment Other miscellaneous manufacturing BACK

37 Robustness tests Stability of µ k estimates across OECD countries and time Correlation between lower and higher income OECD countries: 0.8 Correlation between pre-2005 and post-2005 income OECD countries: 0.7 Out-of-sample validity check l i,2011 = l ik,2011 k s ik,2011µ k l i,2008 l ik,2008 k s k. ik,2008µ k BACK

38 Robustness test: Out-of-sample prediction for year 2011 BACK