Zhao Chen # , Sandra Poncet * , Ruixiang Xiong #

very preliminary

Comparative Advantage and the Effects of Place-Based Policies: Evidence from

China’s Export Processing Zones

Zhao Chen#, Sandra Poncet*, Ruixiang Xiong#

#China Center of Economic Studies, Fudan University

*Paris School of Economics (University of Paris 1) and CEPII23/4/22

I. Introduction

• Industry policies, place-based policies and Economic Zones in China

• No conclusive conclusions about the effectiveness of place-based policies (Moretti, 2010, Busso et al., 2013)

• Difficulties in evaluation of industry policies (Krugman, 1983)

– How to measure industry policy– How to identify the causality

23/4/22

I. Introduction

• In this paper– The effects of export-processing zones

(EPZs)• Clear policy purpose• The role of comparative advantage

– DID estimation using a quasi-experiment of EPZs in China.

23/4/22

II. Brief literature review

• The effect of industry policies (Cai, Harrison and Lin, 2011)

– Tariff policy has positive impact on TFP of industries with comparative advantage

– Comments:• Tariff policy & TFP• Comparative advantage: exporting firms• Policy of protection vs. policy of promotion

II. Brief literature review• Policy evaluation of Economic Zones• City-level data (Wei, 1995; Wang, 2013; Alder et al., 2013)

• Firm-level data (Head and Ries, 1996; Schminke and Van Biesebroeck, 2013)

• Comments:– policy at city-level, no within city difference– Few concern about the heterogeneous impact

• This paper:– Policy difference at city-industry level– Comparative advantage

III. Background of China’s EPZs

• Aim: promote exports by preferential policies

• Establishment :

23/4/22

Establishment year Number of EPZs

2000 15

2001 3

2002 8

2003 13

2005 18

total 57

III. Background of China’s EPZs

• Only some industries chosen as key industries could enjoy preferential policies

• Preferential policies in EPZs ： free VAT, free trade for imported

components; facilitate firm’s exporting outside ： tax reimbursement when providing

firms in EPZs with intermediate goods

23/4/22

IV. Data

• Data source– China annual survey of manufacturing

firms from 1998 to 2007

• Sample– To make the cities more comparable, we

only include the cities having EPZs by 2005

23/4/22

3. Data

• Data cleaning Basic cleaning (Brandt et al., 2012, Nie et al.,

2012) Industry classification adjustment (Yang et al.,

2013) Administrative division adjustment (Bao et al.,

2013) Price deflator

Exporting firms (1998-2007) Matching key industries

23/4/22

4. Empirical ResultsOLS：Ycijt=αc+βi+δt+ψTc+θKci+ρTc· Kci+Xjtλ+εcijt

FE：Ycijt=γj+δt+ψTc+θKci+ρTc· Kci+Xjtλ+εcijt Y: the natural logarithm of export value;

c，city；i，3 digits industry；j, firm，t, year；

Tc：before EPZs’ establishment equals 0, otherwise 1；

Kci：not key industry equals 0，otherwise 1；

Xjt：firm and city level control variables；

εcijt：random error term；

ρ: the effects of place-based policy.

23/4/22

4. Empirical Results

• How to define comparative advantage (CA):– Qci = 1, if location entropy for industry i in city c > 1 before

EPZ establishment, otherwise 0

• Regression– Full sample– Subsample with CA– Subsample without CA

– Triple-interaction term with Qci

23/4/22

4.1 Basic model　 (1) (2) (3) (4) (5) (6)

Full-sampleWithout

CAWith CA

Full-sample

WithoutCA

With CA

　 OLS OLS OLS FE FE FE

T -0.0477 -0.0106 -0.0547

-0.0594*

** -0.0317

-0.0640**

*

(0.0326) (0.0344) (0.0360)(0.0191

) (0.0291) (0.0217)K -0.0332 0.00541 -0.0414

(0.0337) (0.0364) (0.0394)

T*K0.0663** 0.00903

0.0830***

0.104*** 0.0367 0.123***

(0.0255) (0.0299) (0.0283)(0.0157

) (0.0303) (0.0169)

Constant -0.909***-

1.388*** -0.404 -0.247 -0.328* -0.229(0.232) (0.143) (0.283) (0.243) (0.169) (0.313)

Cluster-City YES YES YES YES YES YESObs 338,211 97,966 240,245 338,211 97,966 240,245

R-squared 0.912 0.912 0.913 0.919 0.916 0.921Number of

panelid 　 　 　 75,197 22,464 52,733

23/4/22

4.2 long-run effects

Reference group: n = - 5 [-7, -6, -5]

(n=-4) * Kci

(n=-3) * Kci ……(n= 4) * Kci (n= 5) * Kci

4.2 long-run effects: full sample

23/4/22

-.2

0.2

.4R

eg

ressio

n c

oe

ffic

ien

ts:

% c

ha

ng

e in

exp

ort

va

lue

-4 -3 -2 -1 0 1 2 3 4 5years since(to) EPZs' establishment

Estimates 95% upper bound

95% lower bound

4.2 long-run effects: sub-sample w/o CA

23/4/22

-.1

0.1

.2.3

Re

gre

ssio

n c

oe

ffic

ien

ts:

% c

ha

ng

e in

exp

ort

va

lue

-4 -3 -2 -1 0 1 2 3 4 5years since(to) EPZs' establishment

Estimates 95% upper bound

95% lower bound

4.2 long-run effects: sub-sample with CA

23/4/22

-.2

0.2

.4R

eg

ressio

n c

oe

ffic

ien

ts:

% c

ha

ng

e in

exp

ort

va

lue

-4 -3 -2 -1 0 1 2 3 4 5years since(to) EPZs' establishment

Estimates 95% upper bound

95% lower bound

5. Robustness checks

Ycijt =γj+δt+φ Tc+θ Kci+ω Qci + ρ Tc·Kci+ϑ Tc · Qci+π Kci · Qci + τ Tc·Kci · Qci+𝐗jt∆+εcij

23/4/22

5.1 Using interaction terms （ FE ）

T -0.0185(0.0325)

T*K 0.0451(0.0319)

T*Q -0.0549*(0.0330)

T*K*Q 0.0811**(0.0333)

Constant -0.241(0.240)

Cluster-city YESObservations 338,211

R-squared 0.919Number of panelid 75,197

23/4/22

5.2 Considering firm’s relocation

23/4/22

(1) (2) (3)

full-sample Without CA With CA

T -0.0481** -0.0192 -0.0539**

(0.0194) (0.0282) (0.0225)

T*K 0.108*** 0.0376 0.128***

(0.0159) (0.0287) (0.0172)

Constant -0.265 -0.356** -0.253

(0.217) (0.162) (0.276)

Cluster-City YES YES YES

Observations 277,610 79,517 198,093

R-squared 0.919 0.916 0.921

Number of panelid 56,090 16,555 39,535

5.3 Omitted governance abilities （ FE ）

(1) (2) (3)

Full-sample Without CA With CA

T -0.0169 -0.0115 -0.0149

(0.0347) (0.0441) (0.0378)

T*K0.0873*** 0.00393 0.108***

(0.0281) (0.0434) (0.0311)

Constant -0.0469 -0.144 -0.0499

(0.269) (0.147) (0.364)

Cluster-City YES YES YES

Observations 158,340 45,687 112,653

R-squared 0.920 0.918 0.921

Number of panelid 34,569 10,233 24,336

23/4/22

5.4 Higher export intensity firms （ FE ） Prologis （ 2008 ）

　 (1) (2) (3)

Full-sample Without CA With CA

T -0.0524** -0.00726 -0.0652**

(0.0256) (0.0354) (0.0266)

T*K0.110*** 0.0346 0.130***

(0.0215) (0.0457) (0.0231)

Constant -0.539 -0.570** -0.546

(0.436) (0.225) (0.554)

Cluster-City YES YES YES

Observations 166,809 47,361 119,448

R-squared 0.930 0.932 0.930

Number of panelid 38,766 11,512 27,254

23/4/22

6. Conclusions and implications

• Conclusions Average effects

Overall: 10.4%Industries with CA ： 12.3%; otherwise no

effectsLong-run effects

Industries with CA ： from 9.8% to 24.4%Otherwise no effects

• Policy implication ：local initial conditions are important when

making place-based policies23/4/22

Thank You ！

23/4/22