trading unemployment -...

28
Trading Unemployment eline Carr` ere UNIGE and CEPR Marco Fuggazza UNCTAD Marcelo Olarreaga UNIGE and CEPR Fr´ ed´ eric Robert-Nicoud UNIGE and CEPR Arnoldshain XI University of Antwerp June 26-28, 2013

Upload: others

Post on 01-Sep-2019

2 views

Category:

Documents


0 download

TRANSCRIPT

Trading Unemployment

Celine CarrereUNIGE and CEPR

Marco FuggazzaUNCTAD

Marcelo OlarreagaUNIGE and CEPR

Frederic Robert-NicoudUNIGE and CEPR

Arnoldshain XIUniversity of Antwerp

June 26-28, 2013

Motivation:

I In policy circles trade is sometimes seen as a source ofunemployment

I ”One million jobs have been lost because of NAFTA”(Senator Obama, 2008)

I ”Unfair trade agreements, passed by both Republicans andDemocrats, have sent millions of jobs to other countries. Weneed to stop this hemorrhaging and find ways for Americanworkers to compete in the new market.”(Senator Feingold, 2004)

I ”Now, given the experience that we have had thus far, withour subsequent trade agreements with NAFTA and others, youwould think that with our experience of job loss that we havehad there that when you find yourself in a hole that you mightstop digging.”(Congressman Lynch, 2003)

I And sometimes seen as job-creatingI ”Every $1 billion in new exports of American goods supports

more than 6,000 additional jobs here at home.”(USITC website)

I ”The US-Korea trade agreement will support 70 thousandAmerican jobs”(President Obama, 2012)

I ”Free Trade with the EU will make unemployment disappear(President Saakashvili (Georgia), 2012)

I Who is right?I Does trade create jobs or unemployment?

I Nobody, according to (most) economists...I ”It should be possible to emphasize to students that the level

of employment is a macroeconomic issue depending in the longrun on the natural rate of unemployment, with microeconomicpolicies like tariffs having little net effect.”(Krugman, AER 1993 on What do undergrads need to knowabout trade)

I Y=C+I+G+X-MI ”In its analysis of the effects of reducing the automobile tariff

the Commission believes that in the long run, the aggregatesupply of labour is determined by factors unaffected by tariffpolicy changes.”(Productivity commission 1997)

I But in most models (full) employment is exogenously given.I When employment is endogenous, things get messy:

I Brecher (1974) HO model and impact depends on laborabundance

I Davis (1998): impact depends on which country has the morerigid labor market

I Davidson, Martin and Matusz (1999): unemployment as asource of comparative advantage

I Helpman and Itshkoki (2010) impact of trade onunemployment is ambiguous with trade leading to increases ordecreases in one or both countries depending on the structure.

I As put by Hoekman and Winters (2005): ”These are complexmodels with complex and ambiguous results, but at least theyadmit the possibility that trade reform could have long-runconsequences for employment.”

I And when theory gets messy, economist turn to empiricalevidence...

I But a lot of ambiguity there too....

I Revenga (1994): Tariff reductions led to employmentreductions in Mexico’s manufacturing

I Levinsohn (1999): Trade reforms contributed to the increasein unemployment in Chile in the 1990s

I Menezes et al. (2007): Trade liberalization in Brazil wasassociated with increases in unemployment

I Goldberg and Pavnick (2005): no impact of trade onunemployment in Colombia

I Hasan et al (2012): no relationship in India

I Dutt et al (2010) found openness reduces unemployment in across-section

I Milner and Wright (1998) found that openness reduceunemployment in Mauritius

I Lee (2005) trade growth reduced unemployment in China,India and Malaysia

Objectives

I Fill the gap by empirically testing a theory of trade andunemployment that can explain ambiguous results makingeveryone potentially right

I Move away from two sector models to better fit empiricaltesting with many goods

I We embed a DMP model a la Helpman and Itskhoki (2010) ina DFS Ricardian setup

I Predictions of the model are then empirically tested

I Why DMP’s search model?I GE model of search frictions with microeconomic foundationsI Key: w = w e + stuff, so that u > 0I Would obtain similar results with minimum wages, fair-wages

or centralized wage bargaining

I Why DFS Ricardian model with a continuum of goods?I Continuum of goods to better fit empirical testing with many

goodsI Key: sector-specific unemployment rate interacted with

comparative advantageI HO or even MC would lead to similar empirical predictions.

Bottom line

I Trade openness has an ambiguous effect on unemployment

I The sign depends on the correlation between sector level labormarket frictions and comparative advantage

I If positively correlated, then trade openness increasesunemployment

I If negatively correlated, then trade openness reducesunemployment

I Empirically we find strong support for these predictions

Rest of the talk

I DMP meets DFS

I Prediction

I Empirical tests

I Conclusion

When DMP meets DFS

I Preferences: U (Y ) = Y .I Y is non-traded, non-increasing returns to scale, perfect

competition.

I DFS-type Production:

lnY =

∫ 1

0lnX (z) dz , (1)

X (z) = a (z) n (z)−χ/(σ−1)

[∫ n(z)

0x (j)(σ−1)/σ

]σ/(σ−1)

,(2)

I a (z) is tfp in sector zI n (z) is the mass of firms entering sector zI χ is Lucas (1978) “span-of-control”

DMP-type Labor Markets

I Firms and workers first choose a sector

I Then search for workers and jobs.

I Labor mobility across sectors and risk neutral workers implyµ (z)w (z) = w

I where µ(z) is the probability of finding a job or theemployment rate in z

I µ(z) is obtained through the matching technology mapping:

µ (z) ≡ H (z)

L (z)=

[m (z)

V (z)

L (z)

]1/(1+α), (3)

I Economy-wide unemployment rate is then given by:

u =

∫ 1

0[1 − µ (z)]dz = 1 −

∫ 1

0µ (z)dz

= 1 −∫ 1

0

[w

v (z)

] 11+α

dz ,

I And finally

u = 1 −(

1

2βL

1−χσ−1

) 1−αα

exp (4)[1 − α

α

∫ 1

0ln(a (z) v (z)

−11+α

)dz

] ∫ 1

0ν (z)

−11+α dz

I In an open economy you would first need to determine whichare the goods that you produce domestically (the integral isnot over the entire continuum):

u = 1 − w

L[Lworld

∫ zh

0v (z)

−11+α dz (5)

+Ln

∫ zf

zh

v (z)−11+α dz ]

Prediction

Impact of openness on unemployment depends on ρ

I As countries open up to trade they export more goods andstop producing goods in which they do not have acomparative advantage

I The impact on unemployment will depend on the correlationbetween the degree of overall comparative advantage in eachsector A and v(z) (or u(z) the degree of labor marketfrictions).

I If ρ >> 0, as countries open up to trade, they reallocateworkers to sectors with more labor market frictions, andunemployment increase

I If ρ << 0, as countries open up to trade, they reallocateworkers to sectors with less labor market frictions andunemployment falls.

Empirical test

I This prediction can be tested by running:

uct = αc + αt + β1ρct + β2τct + β3 (τct × ρct) + µct (6)

where ρct is the covariance between a measure of comparativeadvantage and a measure of sector level labor marketfrictions, τct is a measure of trade costs of country c at timet, and µct is an i.i.d error term.

I The model predicts β3 < 0 (and β1 > 0) and

∂u

∂τ= β2 + β3 × ρ (7)

I Comparative advantage is computed using an extension of Vollrath’sextension of Balassa’s RCA

Rcs =r xcs

1 + r xcs− rmcs

1 + rmcs(8)

I Assuming labor market frictions are sector specific, but commonacross countries, we can compute as in Hausmann and Rodrik(2004):

us =∑c

r xcsuc (9)

Not exactly the same r x . In R it has to be comparable acrosssectors within a country. In u it has to be comparable acrosscountries within a sector.

Low labor-market frictionsHS2 Chapter ProduPrepr feathers & down; arti flower; articles human hair 3.87Tin and articles thereof. 4.41Silk. 4.94Manufactures of straw, esparto/other plaiting mat; etc 4.97Lac; gums, resins & other vegetable saps & extracts. 5.21Tobacco and manufactured tobacco substitutes 5.37Oil seed, oleagi fruits; miscell grain, seed, fruit etc 5.43

High labor-market frictionsHS2 Chapter ProduMineral fuels, oils & product of their distillation;etc 10.39Soap, organic surface-active agents, washing prep, etc 9.58Furskins and artificial fur; manufactures thereof. 9.43Railw/tramw locom, rolling-stock & parts thereof; etc 9.3Explosives; pyrotechnic prod; matches; pyrop alloy; etc 9.1Aircraft, spacecraft, and parts thereof. 9.08Beverages, spirits and vinegar. 9.04

Anecdotical evidence supporting our ranking of labor frictions:

Note that labor-unions in agriculture are weak, whereas in railway oraircraft they tend to be strong

Harder evidence:

See Robinson (CJE 1995)

High ρ countriesCountry ρSouth Africa 0.13Bulgaria 0.13Morocco 0.10Croatia 0.10Uruguay 0.09Poland 0.08Tunisia 0.08

Low ρ countriesCountry ρEthiopia -0.19Bhutan -0.16Mali -0.12Madagascar - 0.11Uganda -0.11Iceland -0.10Benin -0.10

Testing the empirical prediction

Average Collectedtariff duties

Labor force 0.00 0.00(0.00) (0.00)

τ -0.02 -0.10(0.05) (0.12)

ρ 32.76** 45.14**(8.59) (9.45)

ρ× τ -1.20** -2.18**(0.56) (0.75)

Obs. 729 648

Second prediction:Marginal effect of τ on unemployment

IV estimates:Marginal effect of τ on unemployment

Instruments are neighbors’s ρ and τ and own labor force (over-id passesp=0.64)

Concluding remarks

I We embedded a DMP labor market model into a DFS Ricardianmodel to show that analytically the impact of trade liberalization onunemployment is ambiguous. Trade liberalization leads

I to more unemployment in countries where comparativeadvantage is in sectors with strong labor market frictions

I to less unemployment in countries where comparativeadvantage is in sectors with weak labor market frictions

I The data supports the theoretical prediction which partly explainsthe ambiguity of results obtained in the previous empirical literaturethat did not control for initial conditions.

Let’s look at the results of the empirical literature mentioned earlier andcompare them to our results:

I Revenga (1994), Levinsohn (1999) and Menezes et al. (2007)estimate a negative impact of tariffs on unemployment in Mexicoand Chile, Brazil. In these three countries we found ∂u/∂τ < 0

I Lee (2005) predicts positive impact of tariffs on unemployment orChina, India and Malaysia. We found that ∂u/∂τ > 0 in thesecountries