Monopolistic Competition and International Trade: Reconsidering the Evidence David Hummels and Peter James Levinsohn QJE, 1995

Jan 28, 2009

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

1 / 15

Helpman (1987) All trade is intraindustry trade monopolistic competition, each producing a differentiated good with IRS technology, zero profit each good is produced in only one country countries have identical and homothetic preferences and trade is balanced, no trade cost Two countries and two goods: V T = s(P X ∗ + Y ∗ ) + s∗ (P X + Y ) = sGDP ∗ + s∗ GDP = 2sGDP ∗ = 2ss∗ GDP A group ofP countries A (region): GDP A = i∈A GDP i , ejA = GDP j /GDP A , eA = GDP A /GDP within-group of trade: P volume P P P V T A = j∈A k∈A,k6=j sj GDP k = j k∈A,k6=j sj ekA GDP A = P P P GDP A j sj (1−ejA ) = GDP A j ejA eA (1−ejA ) = GDP A eA [1− j (ejA )2 ] Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

2 / 15

Helpman (1987) Testable hypothesis: X j V TA = e [1 − (eA )2 ] A GDP A j∈A

P 1 − j∈A (ejA )2 size dispersion index: mamized when countries of the same size Predictions: countries exchange a larger fraction of output when a. they become mores similar in terms of size b. their total size as a group increases (i.e. produce more varieties) Gravity equation (coherent theory of why product differentiation occurs)

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

3 / 15

Helpman (1987)

Helpman: group A - OECD countries (more likely intraindustry trade), time series study Hummels and Levinsohn: I I

91 country-pair observations, 1962-1983, panel study other factors: border trade, seasonal trade, trade restrictions that are country-pair specific

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

4 / 15

First Look 806

QUARTERLY JOURNAL OF ECONOMICS

- 1,51288

1125079

Helpman's Size Dispersion Index

There are several reasons, though, why the fixed-effects Testing be Monopolistic Competition and Tradewe are treating Jan 28, 2009 (2) might misspecified. For example,

Hummels and Levinsohn (1995) estimate of

5 / 15

Estimation There are several reasons, though, why the fixed-effects estimate of (2) might be1misspecified. For example, we are treating 2 )2 )] + v +  ln V Tit = αthe − (eitin)2fact − (e pi's as fixed they i ln[GDP it (1 when it The second it may be irandom. of Table I gives the estimates of (2) when a random-effects instrument:column country factor endowment estimator is employed, and it makes no difference to the results.

TABLE I EQUATION (2) ESTIMATES OECD DATA(1962-1983) Fixed effects

Random effects

Fixed effects (instrumental variables)

Fixed effects (detrended data)

1.405 110.8 ,865 ,981 2002

1.403 110.8 ,860

1.268 47.7 -

2002

2002

1.094 33.7 .373 ,980 2002

OLS (detrended data)

-

a1

t-statistic R2 (wldummies) No. of obs.

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

1.170 44.8 .501 2002

Jan 28, 2009

6 / 15

Non-OECD countries Brazil, Cameroon, Colombia, Congo, Greece, Ivory Coast, South Korea, Nigeria, Norway, Pakistan, paraguay, Peru, Philippines, and Thailand Non-OECD data does not explain as much of the variation in the volume of MONOPOLISTIC COMPETITION AND TRADE 809 trade as OECD data TABLE I1 (2) ESTIMATES EQUATION NOECD DATA(1962-1977)

a1

t-statistic R (wldummies) No. of obs.

Fixed effects

Random effects

Fixed effects (instrumental variables)

Fixed effects (detrended data)

OLS (detrended data)

1.573-3 24.40 ,304 ,671 1456

1.523-3 24.93 ,299

1.543-3 22.03 -

1.18E-3 21.60 .242

1456

1456

1.623-3 19.40 .215 ,654 1456

1456

All estimates are in levels.

do not vary much.15 While using the NOECD data set does not as was Jan the28, 2009

explain(1995) as much Testing of theMonopolistic variation Competition in the volume of trade Hummels and Levinsohn and Trade

7 / 15

Non-OECD countries – Debaere (2005)

Non-OECD countries are not appropriately characterized by differentiated goods trade or by identical demands Debaere (2005) JIE ln(

V TtA ) = αij + γ ln(sit + sjt ) + β ln[1 − (e1it )2 − (e2it )2 ] A GDPt

Evidence contradicts with Hummels and Levinsohn: Helpman’s formula is strongly supported for hte OECD countries, but receive little or no support for the non-OECD countries.

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

8 / 15

A More General Test Some trade is intraindustry (motivated by relative factor abundance), and some interindustry (motivated by the exchange of varieties of differentiated goods). Suppose X- differentiated good (IRS), Y - homogeneous good (CRS) V T = s∗ P X + sP X ∗ + sY − Y Grubel-Lloyd index of intraindustry trade between countries jand k in industry i: 2 min(sP X ∗ ,s∗ P X) IN T RA IIT = IN T RA+IN T ER = s∗ P X+sP X ∗ +sY −Y I I

IIT = 1, if K/L ratios across countries are identical, ie. sY − Y = 0 IIT decreases as |Kj /Lj − Kk /Lk | increases 2

P

min(X

,X

)

ijk ikj IITjk = Pi (Xik +X ikj ) i – Xijk exports of industry i from country j to country k.

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

9 / 15

A More General Test 832

QUARTERLY JOURNAL OF ECONOMICS

HOME COUNTRY A

ST-Y

s* p X

A spx*

v FOREIGN COUNTRY

intraindustry trade share equals one. Factor reallocations that 28, 2009 relativeJan size

Hummels and Levinsohn (1995) Testing Monopolistic Competition Trade widen capital-to-labor differences withoutand changing

10 / 15

A More General Test

j

k

GDP Helpman: IITjk = α0 + α1 log | GDP Nj − Nk | + α2 min(log GDP j , log GDP k ) + α3 max(log GDP j , log GDP k ) + jk j

I

I

k

GDP | GDP N j − N k | not a good proxy for factor composition – conditional on two factors and all goods are traded – demand effect: countries with similar per capita income with have similar demand and produce and export similar goods α1 < 0, α2 > 0, α3 < 0 j

k

j

k

T K T IITjk = α0 + α1 log | K Lj − Lk | + α2 log | Lj − Lk | + j k α3 min(log GDP , log GDP ) + α4 max(log GDP j , log GDP k ) + jk

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

11 / 15

A More GeneralMONOPOLISTIC Test COMPETITION AND TRADE TABLE I11

(3) ESTIMATES WITH GDP PER WORKER INSTEAD OF GDP EQUATION (1962-1983)

815 PER

CAPITA

Year

YDZF

MZNGDP

MAXGDP

R2

1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

-0.275** -0.316** -0.227** -0.255** -0.290** -0.217** -0.234** -0.206** -0.156* -0.196* -0.222** -0.233* -0.109 -0.189 -0.159 -0.260** -0.214* -0.126 -0.074 -0.078 -0.064 -0.091

0.389** 0.377** 0.371** 0.366** 0.334** 0.349** 0.309** 0.332** 0.355** 0.335** 0.336** 0.253** 0.235** 0.226* 0.181* 0.152 0.111 0.130 0.127 0.116 0.103 0.054

-0.012 -0.009 -0.013 0.022 0.011 0.023 0.007 -0.009 -0.062 -0.085 -0.040 -0.031 -0.058 -0.045 -0.050 -0.043 0.004 -0.057 -0.090 -0.106 -0.105 -0.064

,209 .239 .216 .239 .242 ,230 .213 ,214 .I71 ,176 ,188 .I47 ,074 .089 ,056 .095 .067 .028 ,008 .011 .011 .003

The estimated regression is

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

12 / 15

816

QUARTERLY JOURNAL OF ECONOMICS

A More General Test

TABLE IV

EQUATION (3) OLS ESTIMATES WITH

CAPITAL-TO-LABOR AND bND-TO-LABOR

RATIOS ( 1 9 6 2 - 1 9 8 3 )

Year

KLDIF

TLDZF

MINGDP

MAXGDP

R2

The estimated regression is

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

13 / 15

Panel Regression Land to labor ratio may highly collinear with country-pair fixed effect Country-pair dummies seem to explain a tremendous proportion of the variation in our IIT MONOPOLISTIC COMPETITION AND TRADE 819 TABLE V

EQUATION (7) ESTIMATES (1962-1983)

Variable

No fixed effects

YDZF

Fixed effects

Random effects

-0.194 (- 10.900)

KLDIF MZNGDP MAXGDP R2 (wldummies)

0.281 (14.038) -0.009 (-0.488) .I64

The estimated regression is

ZIT,,,

=a .

+ a, log

I X : 1' L:

-

+ a, log

L;

'1 1' T'

Li

-

Tk

+ ma mm (log GDP:, log GDP:)

L;

+ a, max (log GDP:, log GDP:) + u,, Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

+

Jan 28, 2009

14 / 15

Geography An increase in distance may have a larger negative effect on intraindustry trade, if elasticity of substitution between varieties of a differentiated product is greater MONOPOLISTIC COMPETITION AND TRADE 827 than between homogeneous goods TABLE X

EQUATION (7) OLS ESTIMATES

WITH DISTANCE, CAPITAL-TO-LABOR, AND LAND-TO-LABOR RATIOS (1962-1983) Year

KLDIF

TLDIF

MZNGDP

MAXGDP

DIST

R2

1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

-0.086 -0.124 -0.097 -0.059 -0.086 -0.071 -0.072 -0.059 -0.040 -0.030 -0.028 0.011 0.037 0.029 0.071 0.057 0.064 0.078 0.094* 0.116** 0.127** 0.116**

-0.056 -0.073* -0.103** -0.095** -0.104** -0.100** -0.115** -0.110** -0.118** -0.111** -0.102** -0.080** -0.108** -0.075* -0.061* -0.065* -0.067* -0.036 -0.035 -0.030 -0.021 -0.043

0.464** 0.456** 0.412** 0.408** 0.372** 0.373** 0.337** 0.358** 0.354** 0.334** 0.341** 0.279** 0.232** 0.245** 0.211** 0.200** 0.151* 0.169** 0.161** 0.152** 0.146** 0.094

0.160* 0.124 0.123 0.165* 0.140* 0.178** 0.137* 0.141* 0.126* 0.118* 0.154* 0.167** 0.166** 0.197** 0.206** 0.199** 0.213** 0.200** 0.190** 0.190** 0.175** 0.203**

-0.601** -0.527** -0.427** -0.461** -0.447** -0.443** -0.428** -0.418** -0.428** -0.458** -0.463** -0.478** -0.441** -0.509** -0.546** -0.558** -0.489** -0.544** -0.546** -0.570** -0.541** -0.515**

,647 .630 .558 .593 ,624 ,631 ,650 ,633 .642 ,668 .641 .66 .669 ,654 ,668 ,652 ,559 .642 .644 .697 .709 .653

The estimated regression is

Hummels and Levinsohn (1995)

Testing Monopolistic Competition and Trade

Jan 28, 2009

15 / 15