Trade Liberalisation and Poverty What have we learned

Trade Liberalisation and Poverty: What have we learned in fifteen years? L Alan Winters Professor of Economics, University of Sussex CEPR, IZA, GDN

Trade Liberalisation and Poverty: What have we I learned in fifteen years? L Alan Winters Professor of Economics, University of Sussex CEPR, IZA, GDN

What is the issue? 25 th June 2013 Antwerp 3

Conclusion from 1999 Trade Liberalisation • generally stimulates growth • and through it poverty alleviation BUT • it creates losers • some of whom may be, or may become, poor

Therefore Public Policy should: • • proceed with liberalisation, predict the poverty impacts, possibly pre-empt them, and protect the poor with general anti-poverty policies 25 th June 2013 Antwerp 5

Trade and Growth: Levels vs. Changes • Income – Higher vs. growing more rapidly – Permanent vs. transitory growth effects Y” Log (y) Y’ 6 Y Event time

Empirical challenges 1. Defining and measuring openness Binary (Sachs/Warner); (X=M)/GDP Averages – weights, Anderson-Neary 2. Establishing causation Liberalisation → growth or vice versa? 3. Separating openness from other policies the attribution problem 25 th June 2013 Antwerp 7

What have we learned (I) Case studies • No closed economy has developed post WW 2 – Nineteenth century and 1930 s low relevance • Five common features of successful growers: – – – Macro stability High savings and investment Use markets to allocate resources Committed, credible, capable government Fully exploited world economy Growth Commission (2008) 25 th June 2013 Antwerp 8

What have we learned? (II) Instruments • Frankel and Romer (AER, 1999) gravity – Many followers; not always successful • Noguer and Siscart (2005, JIE) – More countries in gravity instruments → clarity • But see Bazzi and Clemens (AEJ-MAC, 2013) – Strong critique of weak instruments – Country size is treacherous - little time variance – Be serious about the exclusion restriction – This includes GMM and system GMM too! 25 th June 2013 Antwerp 9

Exclusion • Deaton (2010, JEL) – instrumentation requires a narrative • Bazzi and Clemens – Many studies take the form: growth = f(x, W); x=g(Z) – If you estimate: growth = h(y, W*); y=m(Z) – You strictly have to reject every one of them! 25 th June 2013 Antwerp 10

Other instruments • Feyrer (2009, NBER WP) – Time varying instrument for trade – Importance of difference in sea distance and air distance becomes more significant as air travel cheapens • Romalis (2007, NBER WP) – US tariffs (also time varying), but • Commodity structure; • Relates trade policy level to output growth 25 th June 2013 Antwerp 11

What have we learned (III) time series • Mixed results • Gries, Kraft & Meierrieks (2009, WD) – GMM – Individual African economies – Granger-Hsaio causation • Wacziarg & Welch (2008, WBER) – 13 country statistical case studies • Kneller, Morgan & Kanchanahatakij (‘ 08, WE) – Panel of event studies 25 th June 2013 Antwerp 12

What have we learned (IV) Conditions • Heterogeneity of country studies • E. g. Chang, Kaltani & Loayza (2009, JDE) and Bolaky and Freund (JDE, 2008) – Interactions of openness with Property Rights Financial Depth Lab. market flexibility Rule of law Education Firm entry Governance Telecoms Firm exit 25 th June 2013 Antwerp 13

Income, Openness and Regulation (Bolaky-Freund; JDE 2008) Less regulated half 14 More regulated half

Property Rights Rule of law Governance Financial Depth Education Telecoms Lab. market flexibility Firm entry Firm exit • Weaker effects of liberalisation in poor countries • Benefits for low-income Africa uncertain • Linear interaction – where is it identified from in the sample space? • Conditions are highly correlated, tested 1 -by-1 – Is it just catching Africa in 1980 s/1990 s – Is Africa different in 2000 s/2010 s? 25 th June 2013 Antwerp 15

What have we learned (V) Productivity • Selectivity • Imported inputs (Amiti & Koenings (2007, AER) and AER P&P 2009 • Learning by Exporting (Fernandes & Isgut, 2005, WB; Blalock & Gertler, 2004, JDE) • Learning to Export (Iacovone, 2009, WB) 25 th June 2013 Antwerp 16

What does this mean for policy? • Vast majority of growth policy is not trade policy; try to make trade policy simple and unobtrusive • Treat as decision, not a hypothesis test 25 th June 2013 Antwerp 17

Policy-Relevant Growth Econometrics g = Xα + Zγ + βt X ‘maintained’ variables; Z ‘optional’ variables t is tariff level • Hypothesis testing H 0 : β=0 against H 1 : β≠ 0 • Trade policy ‘affects’ growth if t-statistic on β > 2 • Issues: (a) uncertainty about β (specification as well as sampling ), (b) relative size of g and βt 25 th June 2013 Antwerp 18

Decision Approach • • Have to decide Balance of evidence and priors Costs of different errors Cost of uncertainty 25 th June 2013 Antwerp 19

Distribution of growth increments Kneller et al (2008) Sample of 47 events 25 th June 2013 Antwerp 20

Cost of error • ‘We know of no credible evidence … that suggests that trade restrictions are systematically associated with higher growth rates’ Rodriguez and Rodrik (2001, p. 317) 25 th June 2013 Antwerp 21

Growth and Poverty • Large literature – not dealt with here • One influential study 25 th June 2013 Antwerp 22

Shares in Long-run Poverty Reduction (Kraay, JDE 2006, cross-section) 23 23

The Microeconomics/Distribution • Conclusion from 1999 • Trade Liberalisation – creates losers – some of whom may be, or may become, poor • Much is uncertain, but not all 25 th June 2013 Antwerp 24

Conceptual Framework 25 Winters World Economy (2002) 25

Two approaches • CGE Modelling supplemented with modules or satellite accounts for personal incomes – Hertel, Tarr, Cockburn – Clear, quantitative, perfect attribution, necessary ex ante – But not strongly grounded in data and outcomes • Frontier is ex post studies – Based on actual outcomes, messy, difficult • Start with a hybrid to point up the issues 25 th June 2013 Antwerp 26

Porto (JIE, 2006) • Uses the framework on Mercosur – Welfare = f(prices, incomes); – SOE – traded prices changed by tariffs etc – Non-traded prices and wages = f(traded prices) • Estimates based on theory, then simulates 25 th June 2013 Antwerp 27

25 th June 2013 Antwerp 28

…and sons • Nicita (JDE, 2009) • Mexico – Expenditure main effect • Marchand (JDE, 2012) • India – Consumption dominates 25 th June 2013 Antwerp 29

Empirical Use of the Framework • Identity, definition; LHS unobservable • Observable equivalent Δ(real consumption) Δrcj = …. . β (Σi wji dlnτi )+ uj – Second order effects – Omitted variables – Parameter heterogeneity – Errors of observation Δrcj = …. . β (exposure) + uj • Partial studies: pass-through; wage effects, ε 25 th June 2013 Antwerp 30

Topalova (AEJ-AE, 2010) • India, 1990 s reforms • Real consumption or poverty rate by region (77) and district (450) • Regressed on employment-weighted tariffs and fixed effects; hence, – Import-based – D-i-D – only relative effects • Greater exposure → worse poverty outcomes 25 th June 2013 Antwerp 31

Refinements • Standard robustness tests • Agricultural wages/returns main channel • Harm is associated with lack of mobility – regional and sectoral = big issue for the poor – i. e alternatives to agriculture weakened • Inflexible labour laws are a key factor • Export exposure → liberalisation reduces poverty (Topalova, 2007) 25 th June 2013 Antwerp 32

Off-shoots • Hasan et al (REc. Stat, 2007) state level data and add data on NTBs: reverse result • Krishna et al (NBER, 2010): – reverse result for ‘leading’ regions; strong in laggards • Castilho (WD, 2012) Brazil: – using same exposure variable, same result – using exports → lib. helps; imports → lib. hurts • Mc. Caig (JIE, 2011) Vietnam – Exposure to US tariff cuts reduces poverty 25 th June 2013 Antwerp 33

Wages – skills gap • Large literature – generally suggests liberalisation increases the skills premium • Amiti and Cameron (JIE, 2012): Indonesia – Unskilled abundant – Firm-level data – estimate within-firm effects; – Tariffs on inputs and outputs – Tariff cuts for outputs – insignificant effects – Tariff cuts for inputs – closes skills gap – Stolper-Samuelson type of effect? 25 th June 2013 Antwerp 34

Conclusion from 1999 2013 Trade Liberalisation • generally stimulates growth • and through it poverty alleviation • Generally more secure, but maybe with a few more caveats for low income countries

Conclusion from 1999 2013 • BUT, it creates losers • some of whom may be, or may become, poor • Conclusion re-inforced • High-quality empirics – more precise and focussed, but all partial in one way or another • Role of segmentation strongly evident

Therefore Public Policy should: • proceed with liberalisation, • predict the poverty impacts, • possibly pre-empt them, especially by increasing mobility and flexibility, and • protect the poor with general anti-poverty policies 25 th June 2013 Antwerp 37

Thank you 25 th June 2013 Antwerp 38

Households and Markets • Do border price shocks get transmitted to poor households? • Are markets created or destroyed? • How well do households respond? • Do the spillovers benefit the poor? • Does trade liberalisation increase vulnerability? Winters, Mc. Culloch and Mc. Kay JEL (2004) 39 39

Wages and Employment • Does liberalisation raise wages or employment? • Is transitional unemployment concentrated on the poor? Government Revenue and Spending • Does liberalisation actually cut government revenue? • Do falling tariff revenues hurt the poor? 40 40