Productivity Spillovers in the GVC. The Case of Poland and the New EU Member States

MotivationMethod

Conclusions

Productivity Spillovers in the GVCThe Case of Poland and the New EU Member States

Jan Hagemejer

Narodowy Bank Polski

University of Warsaw

September 10, 2016

The views presented here are those of the author and not necessarily of the National Bank of Poland. I greatly

acknowledge the �nancing by the National Science Centre, grant no: UMO- 2013/09/D/HS4/01519.

Hagemejer Productivity & GVC

MotivationMethod

Conclusions

Outline

1 MotivationIntroductionLiterature

2 MethodOutlineGVC measuresForeign ownership premiumSpillovers and GVC

3 Conclusions


MotivationMethod

Conclusions

IntroductionLiterature

Why?

Ongoing internationalization of New Member States economies dueto:

transitionEU integrationinvolvement in the GVC

Internationalization is believed to have important direct and indirecte�ects on �rm productivity

through selection e�ects (export related)through FDI hostingthrough FDI productivity spillovers

FDI & exports are already well established in the literature - but towhat extent participation in GVC and the position in the productionchain matters for productivity?


MotivationMethod

Conclusions


Why GVC?

Emerging economies compete for a good �placement� in the GVCs.This motivates �rms to restructure and reorganize.

Inclusion in GVC may involve:

adoption of high quality standardsadoption of modern technologyadoption of modern management techniques

The smile curve debate? Ye, Meng, and Wei (2015), Kowalski et al.(2015) or Cheng et al. (2015). Is the distribution of gains uniformalong the GVC? Is it good to be close to the �nal demand?


MotivationMethod

Conclusions


Literature

FDI spillover literature is already abundant.

Most studies follow the Sma»y«ska-Javorcik (2004) method basedon �rm-level data and input output tables. Other notable worksHaddad and Harrison (1993), Aitken and Harrison (1999), Djankovand Hoekman (2000) or Konings (2001).

Own sector e�ects, backward and forward e�ects.

Review can be found in Crespo, Fontoura, and Proenca (2009)

Irsova and Havranek (2013) analyse more than a 1000 of FDIspillovers in a large-scale meta-analysis showing that, NMS: theoverall evidence of FDI spillovers is heterogeneous.

Hagemejer and Kolasa (2011) show large spillovers from sectoralinternationalization (FDI, exporting, imports of intermediates).Spillovers are either horizontal of backward.


MotivationMethod

Conclusions

OutlineGVC measuresForeign ownership premiumSpillovers and GVC

What we do?

Use Amadeus database for the economies of the New Member States

Combine multiple waves of Amadeus to maximize the span of thesample: 1997-2011 for most countries

Merge �rm-level Amadeus database with the sector-level GVC andspillover measures computed using the WIOD database.

Augment the foreign productivity premia/spillover equations withGVC measures


MotivationMethod

Conclusions


GVC measures

We measure upstreamness according to the de�nition provided byAntras et al. (2012).

Ui = 1 · Xi

Yi+2 ·

∑Nij zijXj

Yi+3 ·

∑Nk=1∑

Nij zijzjk

Yi+ ... (1)

U is the distance from �nal demand measured in stages ofproduction computed for the WIOD database for a paper byHagemejer & Ghodsi (2015).

We measure foreign content of exports using Wang, Wei, and Zhu(2013) backward-based decomposition that is valid on the sectorallevel

FVA (foreign value added of exports) - from intermediate and �nalgoodsVS (vertical specialization) - overall foreign content of exports


GVC measures

MotivationMethod

Conclusions


Premia from foreign ownership

Is GVC participation associated with a lower productivity GAPbetween foreign and domestic �rms?

Similar to Bernard and Jensen (1997) exporter premia regressions

The following equation is estimated:

TFPit = β1foreignit + β2foreignit ·GVCit + β3GVCit + εit (2)

TFP: using Levinsohn and Petrin (2003) method using materials asa proxy for unobservables

Country-sector-clustered SE

Individual countries and full sample regressions


Baseline results: high foreign productivity premia

(1) (2) (3) (4) (5) (6) (7) (8)

BGR CZE EST HUN POL ROU SVK SVN

Foreign 0.418*** 0.614*** 0.648*** 0.651*** 0.374*** 0.385*** 0.530*** 0.394***

(0.0226) (0.0213) (0.0261) (0.0492) (0.0130) (0.0146) (0.0252) (0.0260)

Obs. 66,761 95,901 17,385 13,761 57,173 350,733 33,855 17,650

R2 0.626 0.546 0.306 0.383 0.459 0.292 0.494 0.431

Poland: foreign ownership premium drops with the foreign

content of intermediate goods

(1) (2) (3) (4) (5) (6)

VARIABLES All Mnfc All Mnfc Mnfc Mnfc

Foreign 0.721*** 0.497*** 0.612*** 0.531*** 0.455*** 0.420***

(0.0399) (0.0629) (0.0571) (0.0639) (0.0539) (0.0232)

Foreign * Upstreamness 0.0145 -0.0815 0.173** -0.148

(0.0483) (0.0602) (0.0752) (0.0987)

Upstreamness 0.301 0.230 0.366** 0.260

(0.184) (0.325) (0.172) (0.351)

Foreign * VS -1.277*** -0.228

(0.128) (0.202)

VS 3.130*** 2.544***

(0.355) (0.587)

Foreign * VS (final goods) -0.782*** -0.314 -0.210

(0.214) (0.213) (0.222)

VS (final goods) 4.051*** 2.681*** 2.459*** 2.395***

(0.514) (0.857) (0.744) (0.685)

Foreign * VS (intermediate

goods)-1.929*** 0.00648 -0.427* -0.365**

(0.233) (0.405) (0.244) (0.139)

VS (intermediate goods) 2.062*** 2.373*** 2.527*** 2.504***

(0.532) (0.622) (0.638) (0.545)

Observations 138,117 57,173 138,117 57,173 57,173 57,173

R-squared 0.425 0.502 0.426 0.502 0.501 0.501

Full NMS sample: point to heterogeneity of NMS

(1) (2) (3) (4) (5) (6)

VARIABLES All Mnfc All Mnfc Mnfc Mnfc

Foreign 0.280*** 0.294*** 0.236*** 0.288*** 0.352*** 0.360***

(0.0213) (0.0321) (0.0231) (0.0372) (0.0299) (0.0179)

Foreign * Upstreamness 0.218*** 0.135*** 0.307*** 0.150***

(0.0244) (0.0297) (0.0326) (0.0566)

Upstreamness 0.168 0.485*** -0.442*** 0.489***

(0.174) (0.119) (0.155) (0.141)

Foreign * VS 0.0814 0.161

(0.0628) (0.0987)

VS 0.615 1.678***

(0.424) (0.260)

Foreign * VS (final goods) 0.268*** 0.182 0.0434

(0.0917) (0.118) (0.113)

VS (final goods) 0.173 1.697*** 1.125*** 1.134***

(0.166) (0.376) (0.326) (0.326)

Foreign * VS (intermediate

goods)-0.326** 0.102 0.577*** 0.560***

(0.133) (0.201) (0.113) (0.104)

VS (intermediate goods) 3.670*** 1.668*** 1.548*** 1.552***

(0.445) (0.301) (0.311) (0.312)

Observations 2,172,952 654,105 2,172,952 654,105 654,105 654,105

R-squared 0.818 0.854 0.819 0.854 0.854 0.854

NMS: foreign ownership premium - VS in �nal goods

Table: Foreign �rms productivity premia in individual countries(NMS)-interaction with VS in �nal goods, manufacturing

(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES BGR CZE EST HUN POL ROU SVK SVN

Foreign 0.709*** 0.601*** 0.686*** 1.036*** 0.382*** 0.434*** 0.389*** 0.287***

(0.0605) (0.0483) (0.0517) (0.125) (0.0279) (0.024) (0.037) (0.0584)

Foreign -1.786*** 0.0443 -0.208 -1.952*** -0.0684 -0.365*** 0.832*** 0.649**

* VS (final goods) (0.317) (0.213) (0.234) (0.456) (0.197) (0.115) (0.208) (0.322)

VS (final goods) 0.398 4.517*** 0.726** 1.252** 1.730** -0.0175 -0.417 0.313

(0.855) (0.43) (0.349) (0.545) (0.726) (0.416) (0.647) (1.184)

Observations 66,761 95,901 17,385 13,761 57,173 350,731 33,855 17,651

R-squared 0.549 0.53 0.371 0.545 0.501 0.251 0.497 0.591

MotivationMethod

Conclusions


NMS: foreign ownership premium - VS in intermediate goods

Table: Foreign �rms productivity premia in individual countries (NMS) -interaction with VS in intermediate goods

(1) (2) (3) (4) (5) (6) (7) (8)


Foreign 0.295*** 0.749*** 0.616*** 0.987*** 0.421*** 0.350*** 0.555*** 0.455***

(0.0312) (0.0523) (0.0574) (0.0913) (0.0368) (0.0237) (0.0584) (0.0761)

Foreign * VS 0.885*** -0.741*** 0.162 -1.693*** -0.359** 0.369** -0.164 -0.366

(int. goods) (0.195) (0.239) (0.272) (0.4) (0.134) (0.168) (0.353) (0.414)

VS (int. goods) 0.603 3.745*** 0.881** -0.127 1.739*** 0.834* 3.960*** -0.309

(0.541) (0.679) (0.379) (0.637) (0.651) (0.466) (0.607) (1.036)

Observations 66,761 95,901 17,385 13,761 57,173 350,731 33,855 17,651

R-squared 0.549 0.528 0.371 0.544 0.501 0.251 0.498 0.591


MotivationMethod

Conclusions


Spillovers from GVC

Is GVC participation associated with a lower productivity GAPbetween foreign and domestic �rms?

The following equation is estimated for domestic �rms:

∆TFPijt = α0 + α1∆HZjt + α2∆BWjt + α3∆FWjt

+ α4∆GVCjt + α5∆EXPjt + εit (3)

∆TFPijt is a change of TFP in �rm i in sector j and time t. HZjt ,BWjt ,FWjt are the measures of horizontal, backward and forwardlinkages as de�ned originally by Smazynska-Javorcik (2004).

∆EXPjt is a change in export share of output at sectoral level toaccount for productivity e�ects related to exporting(learning-by-exporting or self selection).

Country-sector level e�ects, time dummies, sector-clustered SE

Individual countries and full sample regressions


Baseline results: not much FDI spillovers

(1) (2) (3) (4) (6) (7) (8) (9)

BGR CZE EST HUN POL ROU SVK SVN

Horizontal -0.825** 0.104 -0.1 0.00311 0.143 -0.106 0.162* 0.138

(0.385) (0.113) (0.0878) (0.0784) (0.103) (0.0786) (0.0904) (0.0971)

Forward 0.341 -0.867*** -0.0875 -0.721 -0.155 -0.12 -0.436* 0.245

(0.468) (0.303) (0.145) (0.462) (0.366) (0.199) (0.237) (0.775)

Backward 1.717 1.972*** 0.0563 -0.322 1.323*** 0.228 0.405* -0.908*

(1.174) (0.337) (0.19) (0.331) (0.39) (0.144) (0.238) (0.525)

Obs. 35,840 63,348 10,114 7,029 30,041 218,561 21,834 9,966

R2 0.108 0.083 0.036 0.046 0.1 0.082 0.051 0.093

Spillovers: Poland

(1) (2) (3) (4) (5) (6) (7)

VARIABLES All Mnfc Mnfc Mnfc Mnfc Mnfc Mnfc

Horizontal FDI -0.0894 -0.00942

(0.0799) (0.0900)

Forward FDI -0.903** -0.390

(0.351) (0.359)

Backward FDI 1.584*** 1.180*** 0.993*** 1.046*** 1.004*** 1.050*** 1.040***

(0.370) (0.350) (0.206) (0.218) (0.206) (0.218) (0.228)

Export share 0.710*** 0.834*** 0.861*** 1.184*** 1.233*** 1.071*** 0.995***

(0.160) (0.132) (0.142) (0.170) (0.168) (0.0993) (0.196)

VS 0.676 1.071*** 0.939***

(0.519) (0.327) (0.350)

Upstreamness -0.731*** -0.252* -0.233* -0.413** -0.408** -0.294** -0.409**

(0.170) (0.133) (0.135) (0.184) (0.191) (0.135) (0.165)

Foreign VA -0.790 -1.059

(final goods) (0.751) (0.737)

Foreign VA 0.769 1.078*

(intermediate goods) (0.577) (0.558)

VS 0.0652

(final goods) (0.718)

VS (intermediate 1.311***

goods) (0.354)

Observations 71,336 30,041 30,041 30,041 30,041 30,041 30,041

R-squared 0.113 0.122 0.122 0.121 0.121 0.121 0.122

Spillovers: NMS

(1) (2) (3) (4) (5) (6) ()

VARIABLES All Mnfc Mnfc Mnfc Mnfc Mnfc Mnfc

Horizontal FDI -0.185** -0.121

(0.0786) (0.0793)

Forward FDI -0.0592 -0.0226

(0.131) (0.164)

Backward FDI 0.345*** 0.466** 0.365** 0.337** 0.343** 0.346*** 0.334**

(0.132) (0.182) (0.142) (0.135) (0.139) (0.131) (0.139)

Export share 0.663*** 0.650*** 0.643*** 0.557*** 0.504*** 0.649*** 0.568***

(0.106) (0.159) (0.159) (0.154) (0.124) (0.0876) (0.167)

VS -0.335*** -0.116 -0.129

(0.117) (0.417) (0.417)

Upstreamness -0.204** -0.183 -0.189 0.0115 0.00729 -0.106 0.0613

(0.0891) (0.125) (0.127) (0.0882) (0.0879) (0.130) (0.0866)

Foreign VA 0.740 1.006*

(final goods) (0.686) (0.546)

Foreign VA -0.805 -1.148***

(intermediate goods) (0.606) (0.402)

VS (final goods) 0.805

(0.575)

VS (intermediate -0.885**

goods) (0.409)

Observations 1,295,481 397,210 397,210 397,210 397,210 397,210 397,210

R-squared 0.090 0.102 0.101 0.103 0.102 0.102 0.103

Spillovers - individual countries, manufacturing

(1) (2) (3) (4) (5) (6) (7) (8)


Horizontal FDI -0.820** 0.180* -0.116 0.00571 0.0122 -0.111 0.0862 0.127

(0.338) (0.105) (0.0831) (0.0701) (0.0852) (0.0747) (0.0803) (0.0921)

Forward FDI 0.123 -0.695** -0.0122 -1.420*** -0.393 -0.121 -0.518** -0.127

(0.414) (0.284) (0.15) (0.289) (0.357) (0.199) (0.23) (0.597)

Backward FDI 1.333** 0.907*** -0.0893 0.0704 1.204*** 0.212 0.11 -0.585

(0.61) (0.287) (0.132) (0.243) (0.364) (0.131) (0.218) (0.383)

Export share 1.673*** 0.441*** -0.0413 -1.152*** 0.965*** 0.282** 0.376** -0.141

(0.522) (0.13) (0.0627) (0.34) (0.183) (0.122) (0.189) (0.315)

VS (final -0.149 2.424*** 0.648** 1.484* 0.202 0.241 2.463*** 1.628

goods) (1.404) (0.441) (0.319) (0.789) (0.664) (0.426) (0.59) (1.159)

VS (int. -3.243*** -1.490*** -0.219 3.265*** 1.439*** -0.583** 1.464** 0.608

goods) (1.109) (0.541) (0.194) (0.547) (0.347) (0.249) (0.592) (0.609)

Upstreamness -0.0374 0.705*** 0.419*** -0.0785 -0.426** 0.0437 0.0132 -0.00488

(0.402) (0.16) (0.103) (0.214) (0.17) (0.0623) (0.233) (0.19)

Observations 35,840 63,348 10,114 7,029 30,041 218,561 21,834 9,966

R-squared 0.151 0.108 0.038 0.064 0.122 0.083 0.073 0.11

Conclusions

Poland: most of the GVC related productivity gains are inintermediate goods

this is where foreign content of exports is associated with lowerproductivity di�erences between domestic and foreign enterprises. Atthe same time productive �rms are, other things equal, located closeto the �nal demand.it pays of to be on close to the �nal consumer unless being furtheraway involves a high content of imported foreign value added inexported goods.

In most of the other countries (except Hungary where results aresimilar to that of Poland) where positive spillovers in the GVC exist,they tend to stem from production of �nal goods.

In Romania and Bulgaria the GVCs do not seem to bring to much ofproductivity improvement to domestic �rms

Results are similar when labour productivity is used instead of TFP.

Economy & Finance

Productivity Spillovers in the GVC. The Case of Poland and the New EU Member States