1 s2.0-s0048733309000262-main

DE

SR

a

ARRAA

JOOOP

KNTIIF

1

cbagiri

ueStBaEab

0d

Research Policy 38 (2009) 845–860

Contents lists available at ScienceDirect

Research Policy

journa l homepage: www.e lsev ier .com/ locate / respol

rivers of national innovation in transition: Evidence from a panel of Easternuropean countries

orin M.S. Krammer ∗

ensselaer Polytechnic Institute, Economics Department, 110 8th Street, Troy, NY 12180, USA

r t i c l e i n f o

rticle history:eceived 11 September 2008eceived in revised form 20 January 2009ccepted 20 January 2009vailable online 14 March 2009

EL classification:304757

a b s t r a c t

Innovation plays a crucial role in determining today’s economic growth patterns. But what enables somecountries to innovate more than others? This study attempts to answer this question by analyzing inpremiere a panel of sixteen Eastern European transition countries. It provides a detailed description ofinnovation identifying regional differences in terms of historical heritage, technological specialization,commitments and main actors involved in this process, before and after the fall of communism. Sec-ondly, it explores empirically the main drivers of their innovative output, proxied by patents, using avariety of econometric techniques and control variables. The results confirm the crucial role of univer-sities and existing national knowledge base complemented by R&D commitments from both public andprivate sources. Policy measures, such as intellectual property rights protection or a favorable business

20

eywords:ational innovationransition economiesntellectual property rightsnternational trade

climate, increase significantly the propensity to patent, while measures of transitional downturn andindustrial restructuring diminish it. Finally, globalization contributes to developing new innovations inthese countries through inflows of foreign investment and trade.

© 2009 Elsevier B.V. All rights reserved.

oreign direct investment

. Introduction

In today’s dematerialized global economy, the ability of aountry to develop, adapt and harness its innovative potential isecoming critical for its long run economic performance. This fact,cknowledged by the endogenous growth literature, is starting toenerate concrete policy results as well; the two recent Europeannnovation strategies of Lisbon in 2000 and Barcelona in 2002eflect this trend, aiming to reduce the gap between EU and USn innovation, productivity and ultimately, economic growth.

This study builds on the previous literature and provides anpdated and comprehensive overview of innovation in transitionconomies from Eastern Europe and the former Soviet Union (EECs).ince the early 1990s these countries have experienced a painfulransition from a closed centralized economy to a free market one.esides their economic performance, this difficult conversion has
lso impeded their innovative capacities. As a result, the existingast–West technological divide in Europe as well as differencesmong the EECs have increased during the 1990s. In terms ofoth inputs (R&D investment, researchers, business and govern-
∗ Tel.: +1 518 276 7918; fax: +1 518 276 2235.E-mail address: [email protected].

048-7333/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.respol.2009.01.022

ment involvement, multinational’s presence) and output (numberof patents, technological strengths) the EECs are quite diverse.Such significant heterogeneity, reintegration in the global eco-nomic structures and the unique effects of the transition processprovide many interesting questions for research on innovation inthis region. However, significant challenges reside in successfullyaddressing the multitude of relevant factors and data requirements.

Secondly, this work provides an econometric analysis of thedeterminants of innovation in transition economies. So far thecross-country literature has focused mostly on developed coun-tries, usually OECD ones. Using newly constructed panel data onpatents, R&D inputs, policy variables, innovation infrastructure,transition and globalization effects, I employ an eclectic but consis-tent approach to explain the evolution of innovative output (patentsissued in USA and EU) in Eastern Europe. The robustness of themodel is tested using a variety of econometric techniques andadditional control variables such as human capital, education pol-icy and population. Our results show that, despite low investmentand skewed technological specialization, the existing knowledge
stock complemented by current R&D efforts increases the amountof international patents from these countries. The governmentovertakes businesses as the most important source of funds forinnovation, consistent with previous work on latecomers. Regard-less of their limited involvement, universities retain a crucial role in
http://www.sciencedirect.com/science/journal/00487333

http://www.elsevier.com/locate/respol

mailto:[email protected]

dx.doi.org/10.1016/j.respol.2009.01.022

8 rch Po

lvg(o

seadeaiwiRkotet

tpistt5cc

2

lqsiwlTtw

eacipatLwaaptwcg

sH

are important in determining EECs’ patenting rates, their tech-nological specialization is very much path dependent, and thatthere is significant country heterogeneity in how R&D is conducted.Marinova (2000)5 emphasizes the sharp transitional decline in

46 S.M.S. Krammer / Resea

inking financial resources to the national infrastructure for inno-ation. Furthermore, strong intellectual property rights regime andlobal flows (trade and FDI) increase patenting while transitionoutput drop, industrial distortions) and certain policy factors (costf doing business) significantly decrease it.

Finally, the paper aims to provide some relevant policy conclu-ions for transition and developing countries. Throughout history,xtraordinary “outsiders” like Ireland, Finland, Israel, South Koreand Taiwan, have become major global technological players in justecades, all due to the right mix of policies and investments. East-rn Europe needs to take better advantage of its own comparativedvantage, in terms of existing human capital, competencies andnnovative heritage, and build solid innovative capacities which

ill ensure economic growth in the region. Future challengesnclude stimulating the business community to undertake more&D, especially the small firms, encouraging the presence andnowledge transfer from multinationals, supporting developmentf managerial and technical skills though tertiary education andraining activities, facilitating the flow of capital for innovativentrepreneurs as well as providing comprehensive national policieshat will sustain an innovative culture in these countries.

The paper is organized as follows. Next section discusses theheoretical background of innovation at the national level androvides an overview of the empirical work dealing with these

ssues in a cross-country dimension. Section 3 describes thetatus quo of Eastern European innovation providing both his-orical and inter-regional comparisons. Section 4 introduces theheoretical approach and the dataset employed while Section

presents the empirical analysis and results. Finally, Section 6oncludes, discussing findings, policy implications and possibleaveats.

. Perspectives on national innovation

Since the relationship between economic growth and techno-ogical development has been postulated in the literature,1 theuestion of analyzing the determinants of this capacity of firms,ectors and countries to generate new flows of knowledge has beennvestigated by numerous scholars. A commonly employed frame-

ork is the knowledge production function (KPF) approach thatinks inputs and outputs associated with the process of innovation.his section reviews the main findings of this literature focusing onhe cross-country studies and presents the existing work dealingith Eastern Europe on which this paper builds on.

In terms of inputs to innovation, usually some measures of R&Dxpenditure, there are only few empirical investigations lookingcross countries at their determinants. Keeping in mind their short-omings, specifically endogeneity and unobserved heterogeneityssues, this literature provides some interesting conclusions. Patentrotection, openness and cultural factors (Varsakelis, 2001) as wells government expenditure (Bebczuk, 2002) have a close rela-ionship with R&D investment around the world. Furthermore,ederman and Maloney (2003) notice that R&D intensity increasesith development level and that the main reasons behind this surge

re the protection of intellectual property, government investmentnd quality of research institutions. They also examine the under-erformance of Latin America in comparison to the impressive R&D
akeoffs of outstanding outliers like Finland, Israel, Korea and Tai-an. Both their analysis and the follow-up by Bosch et al. (2005)
over industrial and developing countries and control for endo-eneity issues.

1 Several canonical models assume constant returns to knowledge and largepillovers that lead to increases in growth rates (Romer, 1990; Grossman andelpman, 1991; Aghion and Howitt, 1992; Jones, 1995).

licy 38 (2009) 845–860

In order to gain better insights on the KPF, the quest for goodquantitative proxies for innovation led to a significant and increas-ingly sophisticated literature that uses patents and patent relatedmeasures. As a result, patents have become a common mea-sure for innovational output and a good way to track flows ofknowledge across firms, sectors and countries. Moreover, the lastdecades record a tremendous increase in the number of patentsissued worldwide reflecting the increasing importance of intel-lectual property in today’s knowledge based economy.2 Like anyother proxy, patents present both advantages and disadvantages(Acs et al., 2002). However, they remain the best available sourcefor assessing technological change and innovation since “nothingelse comes close in quantity of available data, accessibility andthe potential industrial organizational and technological details”(Griliches, 1990).

Similarly to the studies on inputs, the evidence on what deter-mines innovative output (usually patents) at the country levelemploys a variety of explanatory factors but is mostly confinedto developed countries. Assuming that innovation stems withina national framework of institutions, Varsakelis (2006) incorpo-rates in his analysis specific measures of governance (civil liberties,political rights, free press and corruption) and education (math-ematics and science mean scores). When exploring the role ofpolitical institutions persistence, findings show that the institu-tional system tenure, regardless of the type, increases US patentapplications from the foreigners in the case of several Latin Ameri-can and Caribbean countries (Waguespack et al., 2005). Rather thanfocusing on few specific factors, the concept of national innova-tive capacity (NIC) investigates the overall sources of innovationsystems at the country level (Furman et al., 2002, henceforth FPS).Thus, the NIC concept, defined as the ability of a country to produceand commercialize a flow of innovative technology over the longrun, converges three main sets of ideas: (a) the KPF from endoge-nous growth theory (Romer, 1990; Jones, 1995); (b) Porter’s (1990)interaction between the private sector and the national industrialclusters and (c) the national innovation systems (NIS) literatureemphasizing institutional interactions within a complex nationalsystem at work3(Freeman, 1982; Lundvall, 1992; Nelson, 1993). FPStest successfully this conceptual framework using 17 OECD coun-tries, while the subsequent extensions analyzing 29 OECD countries(Furman and Hayes, 2004) or the case of East Asian “tigers” (Hu andMathews, 2005) validate their conclusions.

With respect to Eastern Europe, the existing literature islimited to descriptive analyses and case studies.4Radosevic andAuriol (1999) depict an overall picture of innovation in six Cen-tral and Eastern European countries; they conclude that despitetheir “downward shift in terms of stocks of R&D spending andemployment, EECs have managed to maintain and intermedi-ate position between developed and less developed OECD/EUeconomies”, failing however to transform these stocks in sourcesof growth. Radosevic and Kutlaca (1999) claim that income levels

2 Approximately 110,000 applications were filled at the European Patent Office(EPO) and almost 315,000 were registered by the US Patent and Trademark Office(USPTO) in 2000, compared with nearly 60,000 and 108,000, respectively, in 1991.

3 Such as university systems, intellectual property, historical industrial organiza-tion, R&D labor division, private industry structure and governmental support.

4 For a recent collection of such case studies see the volume edited by Piech andRadosevic (2006).

5 However, this study has two important drawbacks. First, the use of aggregatedformer entities (e.g. Yugoslavia, USSR, Czechoslovakia) after their official secessionis hard to justify and interpret. Secondly, the USPTO methodological inconsistencies

S.M.S. Krammer / Research Policy 38 (2009) 845–860 847

e USA

pRit

3

btwaUawtcs

3

(ncp

Wsafatrot(s

(d

s

Fig. 1. Historical trends in patenting in th

atents revealing regional and country specific strengths, whileadosevic (2004) using recent data on NIS variables confirms the

ntermediate position of some EECs at the lower innovative spec-rum of the enlarged European Union.

. Stylized facts about Eastern European innovation

At the beginning of the 1990s a colossal natural experimentegan when all Eastern Europe suddenly embarked on its wayowards a free market economic system. This deep transformationas quite painful with important macro-economic disturbances6

nd institutional collapse following the regime shift. The resulting-shaped response of GDP, with a sharp initial decline followed byrecovery in the late 1990s, left its mark on innovation output asell. This section explores the progress Eastern European innova-

ion during this transition by looking in detail at its inputs (R&Dommitments) and output (US patents) and attempting to identifyome of its drivers and their trends (Table 1).

.1. East–West historical divide

The difference between Eastern (EECs) and Western EuropeWECs) in producing new technologies and products is quite sig-ificant. Moreover, this holds also in an historical context whenomparing their international patenting numbers (see Fig. 1) oratent intensity, defined as the stock of patents per capita (Table 2).

This fact has multiple causes which are hard to disentangle.hile the WECs increased their R&D commitments and their sub-

equent development in commercial innovation was driven by anctive business environment, the socialist EECs were rather isolatedrom the world’s flows of trade and ideas, and failed to diversifynd keep the pace with the latest technologies (Murrel, 1990). Inerms of patent intensity, even the most successful EECs are cur-
ently far away from the frontier, between the southern peripheryf Western Europe (Spain, Portugal and Greece) and its core, whilehe emerging ones still struggle at the bottom of this classificationsee Table 2). With regards to patent assignees, things have sub-tantially changed in transition, mainly due to the recognition of
coding of countries) for the EECs, not corrected until 2006, yielded systematicallyownward biased counts, especially for the newly formed countries.6 Capital stock shrinkages, labor force movements, trade reorientation, significant

tructural changes.

. Eastern–Western European comparison.

private and intellectual property.7 Thus, the assigned percentage ofpatents to foreign entities experienced a significant increase after1990. A more detailed analysis of this transformation is given inSection 3.5.

3.2. Regional heterogeneity

There are also significant differences between and withinregions. The usual North–South division applies to Western Europe,while for Eastern Europe the picture is more diverse (see Fig. 2).The overall innovative leader in the communist block was the for-mer Soviet Union which had the vast majority of granted patentsin the US between 1975 and 1995.8 Its heir, Russia, remained ontop after 1991 and is still responsible for about half of the USPTOpatents from Eastern Europe with 3 695 patents granted. The restof the countries in our study can be grouped in innovative termsinto three categories.9First tier innovators average patent stocksbetween 400 and 1300 between 1990 and 2007. Hungary (1208)is the most consistent one but in a slight regression compared withthe prior period, Czech Republic (663) and especially Poland (669)have shown remarkable improvements while Ukraine (480) justrebounded after a sharp drop in the early 1990s. The second tieraveraged between 100 and 400 first inventor patents in this inter-val and can be divided into two subgroups: improvers like Slovenia(308), Croatia (199), Romania (142) and a surprising Lithuania (105),all with lower starting points followed by significant growth andlaggards such as Bulgaria (152), Belarus (118) or Slovakia (127)exhibiting higher initial starting points but an overall stagnating orregressing trend. Finally, the third tier is formed from small coun-tries with few USPTO patents that seem not to have improved muchover time: Latvia (51), Estonia (62), Georgia (34) and Serbia andMontenegro (100), a shadow in innovative terms of the formerYugoslavia.

3.3. Specialization patterns

The decades of communist isolationism no doubt influenced theEECs towards certain technologies. To compare their technologi-

7 An assignee is the holder of the rights to use the patent for commercial purposes.8 About 66% of the total EEC patents in the 1970s and 45% in the 1980s.9 15 EECs that have a total of 30 or more first inventor patents in this 18 year

interval; the rest are two small to be taken into account and have many of zerocounts. See Appendix A for the list of countries.

848 S.M.S. Krammer / Research Policy 38 (2009) 845–860

Tab

le1

East

ern

Euro

pea

nfa

ctsh

eet

(20

06or

the

late

stye

arav

aila

ble.

)

Cou

ntr

yPo

pu

lati

on(m

illi

ons)

GD

Pp

erca

pit

a(U

S$20

06)

GD

Pp

erca

pit

a(U

S$PP

P)

Rec

ent

real

GD

Pgr

owth

(20

01–2

005

)

USP

TOp

aten

ts(9

0–06

)

Res

earc

her

s(t

hou

san

ds)

GER

D(%

GD

P)IP

Rin

dex

Trad

eop

enn

ess

Cos

tbu

sin

ess

(day

s)

Ind

ust

rial

dis

tort

ion

sin

dex

Ou

tpu

td

iffe

ren

tial

(20

00–

1990

)

Edu

cati

onsp

end

ing

(%G

DP)

Publ

icR

&D

(%G

ERD

)

Priv

ate

R&

D(%

GER

D)

Un

iver

sity

R&

D(%

per

f)

Bu

lgar

ia7.

73,

683

9,97

54.

9%14

110

.45

0.4

9%5.

1414

1.47

320.

296

−9.1

2%2.

40%

71.4

1%18

.52%

10.0

4%B

elar

us

10.3

3,55

28,

688

6.9%

101

18.5

60.

62%

3.19

154.

4777

0.52

4−9

.72%

3.78

%32

.60%

50.9

7%16

.43%

Cze

chR

epu

blic

10.2

13,0

3520

,563

3.2%

598

30.6

41.

22%

9.57

154.

1336

0.27

43.

31%

3.45

%22

.97%

61.0

8%15

.63%

Esto

nia

1.3

9,88

217

,672

6.2%

505.

090.

75%

5.72

178.

0447

0.26

15.

52%

3.18

%16

.97%

30.6

2%47

.82%

Geo

rgia

4.7

986

3,27

77.

2%29

12.0

00.

29%

387

.822

0.18

4−5

6.60

%1.

85%

––

16.2

7%C

roat

ia4.

58,

422

13,1

864.

3%17

811

.14

1.12

%3.

7111

5.39

48

0.16

7−6

.49%

3.25

%22

.26%

42.6

6%35

.08%

Hu

nga

ry10

.011

,885

17,7

333.

6%1,

129

29.7

61.

02%

11.2

516

4.03

450.

266

12.5

2%3.

54%

32.8

5%35

.47%

25.1

6%Li

thu

ania

3.6

7,34

215

,46

47.

3%77

9.52

0.67

%2.

5712

2.26

260.

195

−24.

77%

3.15

%33

.39%

16.8

6%4

9.75

%La

tvia

2.3

7,17

513

,938

7.4%

44

6.10

0.42

%5.

7699

.71

160.

242

−16.

51%

3.45

%18

.99%

40.8

8%40

.12%

Pola

nd

38.5

8,60

214

,137

3.0%

586

90.8

40.

58%

9.69

66.4

631

0.14

341

.14%

3.42

%45

.46%

20.3

4%33

.92%

Rom

ania

22.3

3,98

58,

602

6.0%

118

24.6

40.

38%

2.71

92.9

320

0.20

8−1

2.92

%2.

80%

24.1

7%60

.26%

15.5

6%R

uss

ia14

2.9

6,14

312

,142

6.1%

3,38

341

4.6

81.

25%

6.08

75.2

335

0.42

5−3

1.78

%1.

68%

24.4

6%69

.88%

5.42

%Se

rbia

and

Mon

ten

egro

8.0

3,38

35,

549

4.8%

9110

.86

1.17

%2.

2039

.50

350.

598

−37.

86%

1.85

%4

4.19

%3.

95%

51.8

6%

Slov

akia

5.4

10,3

2617

,266

4.6%

111

15.3

90.

57%

9.57

172.

5151

0.33

01.

32%

2.92

%26

.56%

64.

33%

9.10

%Sl

oven

ia2.

018

,816

23,1

023.

4%27

44.

64

1.53

%10

.56

123.

4760

0.10

819

.77%

3.24

%23

.06%

59.6

8%15

.55%

Ukr

ain

e46

.72,

245

7,80

28.

6%43

785

.21

0.95

%6.

0811

9.32

350.

520

−54.

15%

3.20

%40

.00%

54.5

9%5.

41%

Table 2Patent stock and intensity in Western and Eastern Europe (as of 2007).

Country Patent stock a Patent intensity b

Switzerland 53,236 7,108.21Sweden 39,517 4,389.91Germany 304,161 3,689.87Finland 13,466 2,577.99Netherlands 36,817 2,243.91United Kingdom 126,663 2,095.63France 113,969 1,878.93Denmark 9,948 1,831.26Austria 14,512 1,773.07Belgium 16,004 1,544.13Norway 5,564 1,211.40Iceland 260 876.20Italy 44,990 774.31Ireland 2,329 579.98Hungary 2,805 281.02Spain 5,852 145.06Slovenia 199 98.99Bulgaria 505 68.38Greece 506 47.43Croatia 141 31.37Czech Republic 312 30.48Portugal 267 25.27Estonia 32 24.16Poland 843 21.88Romania 421 18.88Russian Federation 2,233 15.63Lithuania 31 8.64Slovakia 44 8.09Latvia 17 7.47Georgia 26 5.58Belarus 57 5.54Ukraine 231 4.95Serbia and Montenegro 6 0.75

Source: Own calculations based on Table A1-1a, Breakout by country of origin num-ber of patents granted as distributed by year of patent grant. USPTO Patent StatisticsReports 2008.For the newly formed countries this statistic only incorporates the patents granted
from that year onwards (e.g. here the stock of Russian patents in the USA does notinclude the one of the former USSR).
a USPTO utility patents granted between 1963 and 2007.b Number of patents per million people.

cal paths with those of Western market economies, I employ theNBER US Patent dataset by Hall et al. (2001) and their classifica-tion into six broad technological categories: Chemical, Computersand Communication, Drugs and Medical, Electrical and Electronic,Mechanical and Other. These broad results confirm a significantdecrease in EECs’ innovational output over the last 30 years andimportant changes in their innovational mix during the 1990s (seeFig. 3).

When using a 14-industry level of detail, both similarities anddifferences between East and West are evident (see Fig. 4). Bothare strong in heavy industries, textiles, chemicals, food and homeproducts. Seemingly, EECs have a comparative advantage in drugsand medicine, metallurgy and energy, while their Western neigh-bors are better, as expected, in new industries with a higher valueadded like Engines and Vehicles, Communications, Computers andMiscellaneous Structures. These different strengths indicate alsocomplementarity of technological specialization within a widerEurope specifically beneficial from a trade perspective.

3.4. Commitments to innovation

There is a strong correlation between the level of income and
national commitments to innovation supporting the conclusionsof endogenous growth theory (see Fig. 5a). However, the numberof researchers in the labor force does not depend necessarily on thegross domestic expenditure on R&D (GERD) as shown in Fig. 5b.At the country level, the evolution of researchers employed is very


stern

dUcaPacs

Fig. 2. Regional heterogeneity in Ea

iverse compared to 1990: some heavy reductions (Bulgaria—70%;kraine—64%; Georgia—53%; Romania—45% and Russia—41%),onstancy (Hungary, Serbia and Montenegro, Slovakia, Slovenia
nd the Baltic states) and even increases (Czech Republic 34% andoland 22%). Nevertheless, in GERD terms, both within EECs andmong them and the WECs, the differences are notable even afterontrolling for purchasing power parity (PPP) disparities. As for theource of funding, government remains a major player in Eastern
Fig. 3. Historical technological specializ

European innovation (1990–2007).

European R&D while the involvement of businesses and highereducation is still quite limited (see Fig. 6a and b).

3.5. Old versus New players

When looking at the distribution of main EEC patent holderssome interesting facts can be identified (see Table 3). While virtu-ally there was no foreign presence prior to mid 1990s, except the

ation: East vs. West (1967–1999).


Fig. 4. Detailed technological specialization map using USPTO patent stocks (1999).

Fig. 5. (a and b) Relationship between R&D intensity, income levels and researchers in Eastern Europe.

Table 3Main Eastern European patent holders. Breakout by organizations.a

1990–1994 1995–1999 2000–2004

I F N I F N I F N

Bulgaria 4 0 7 5 0 0 13 0 0Belarus 1 0 1 5 0 5 8 0 1Czech Republic – – – 10 5 10 32 22 8Estonia 0 0 0 0 0 0 0 5 0Georgia 1 0 0 3 0 0 2 8 0Croatia 0 0 0 16 0 18 15 13 11Hungary 49 0 237 49 27 70 66 59 48Latvia 0 0 0 2 0 0 4 0 0Poland 8 2 9 12 4 5 13 3 2Romania 0 0 0 7 0 0 14 0 0Russia 10 3 0 160 14 2 216 28 3Slovakia – – – 3 0 0 11 0 0Slovenia 4 0 0 17 1 8 0 8 23Ukraine 4 0 0 24 2 0 19 19 0

Source: Basedon a report from Patent Technology Monitoring Branch (PTMB) “Count of 1969–2005 Utility Patent Grants by calendar year of grant”.a Organizations receiving 5 or more utility patents during 1969–2004; I—individuals; F—foreign entities (firms, universities); N—domestic entities (research institutes,

institutions, firms).


Fig. 6. (a and b) Government versus business R&D in the EECs and GERD by sector of performance (GOV—Governmental, BUS—Business, HED—Higher Education).

Table 4Main organizations holding Eastern European patents in the communism a(1969–1989).

No. Code Name of the organization Country Patents Main field of activity

1 N CESKOSLOVENSKA AKADEMIE VED Czechoslovakia 425 Education2 N RICHTER GEDEON VEGYESZETI GYAR RT Hungary 267 Pharmaceutical3 N CHINOIN GYOGYSZER ES VEGYESZETI TERMEKEK GYARA RT. Hungary 224 Pharmaceutical4 N ELITEX ZAVODY TEXTILNIHO STROJIRENSTVI GENERALNI REDITELSTVI Czechoslovakia 121 Textiles5 N VYZKUMNY USTAV BAVLNARSKY Czechoslovakia 105 Textiles6 N EGYT GYOGYSZERVEGYESZETI GYAR Hungary 69 Pharmaceutical7 N SPOFA, UNITED PHARMACEUTICAL WORKS Czechoslovakia 52 Pharmaceutical8 N ADAMOVSKE STROJIRNY, NARODNI PODNIK Czechoslovakia 50 Polygraphic presses9 N VYZKUMNY A VYVOJOVY USTAV ZAVODU VSEOBECNEHO STROJIRENSTVI Czechoslovakia 50 Metallurgy

10 N INSTITUT ELEKTROSVARKI IMENI E.O. PATONA AKADEMII NAUK UKRAI USSR 48 Metallurgy11 N ELITEX, KONCERN TEXTILNIHO STROJIRENSTVI Czechoslovakia 36 Textiles12 N MEDICOR MUVEK Hungary 30 Medical equipment13 N ESZAKMAGYARORSZAGI VEGYIMUVEK Hungary 29 Chemical14 N INSTITUT GORNOGO DELA SIBIRSKOGO OIDELENIA AKADEMII NAUK SSS USSR 25 Metallurgy15 N INSTITUTE PO METALOZNANIE I TECHNOLOGIA NA METALITE Bulgaria 24 Metallurgy16 N MINISTERUL INDUSTRIEI CONSTRUCTIILOR DE MASINI Romania 23 Mechanical17 N VSESOJUZNY NAUCHNO-ISSLEDOVATELSKY I PROEKTNO-KONSTURKTORSKY USSR 22 Constructions18 N POLITECHNIKA GDANSKA INSTYTUT CHEMII I TECHNOLOGII ORGANICZN Poland 19 Education19 N POLITECHNIKA WARSZAWSKA Poland 19 Education2 N.SOL

S t of 19ntitie

fediaiMth(mcttt

3

icwtac

specification augmented by additional factors that impact innova-tion identified in the national innovation systems (NIS) literature:

0 N LEK TOVARNA FARMACEVTSKIH IN KEMICNIH IZDELKOV,

ource: Based on a report from Patent Technology Monitoring Branch (PTMB) “Couna Organizations receiving 5 or more utility patents during 1969–2004; F—foreign e

ormer Yugoslavia, this has dramatically changed in the last years,specially for the first tier of innovators in the region. Prior to 1990,omestic individual holders and entities (firms, governmental bod-

es or research institutions) were dominant (see Table 4). After 1990,significant difference is the emergence of global players in the EEC

nnovation arena such as General Electric, Samsung Electronics, Sunicrosystems, Ericsson and Bosch Siemens (see Table 5). Moreover,

he dominant orientation of these patents has shifted away fromeavy and labour intensive industries towards today’s “hot” fieldspharmaceutical and biotech, computers and semiconductors, com-

unications). Despite this positive trend, foreign assignees are stillonfined to a handful of countries, with Hungary and Russia leadinghe way in absolute numbers, Czech Republic, Slovenia and Croatiarailing behind, while the others have little or no foreign assigneeso their international patents.

.6. Summary

Historically, the EECs’ innovative productivity has been declin-ng since the late 1970s associated with growing inefficiencies of the
ommunist regime. Moreover, the transitional shock made thingsorse by inducing significant reductions in the total commitments
o R&D due to the lack of funds. Despite these facts, EECs man-ged to retain an intermediate position between the core Europeanountries and the less developed peripheral EU states. Their legacy

.O. Yugoslavia 17 Pharmaceutical

69–2005 Utility Patent Grants by calendar year of grant”.s (firms, universities); N—domestic entities (research institutes, institutions, firms).

in some key fields (e.g. chemicals, pharmaceuticals) and trainedhuman resources provides opportunities for a successful revivalof innovation in Eastern Europe in which some of the global R&Dplayers are already involved.

4. Measuring the determinants of innovation

4.1. Theoretical framework

The theoretical specification departs from an endogenousgrowth model and includes additional controls for a small openeconomy. Since this paper has an empirical objective, a full specifi-cation will not be included here.10 This generic economy has threesectors producing: a consumption good (output), a human capitalgood (experience, education or skill) and new varieties of capitalgoods (ideas, innovations). For our purpose we are interested only inthe production function of new ideas. This builds on Jones’s (1995)

At = ıhˇt LAtA

�t ˝t (4.1)

10 For details please see the original papers.


Table 5Main organizations holding Eastern European patents after the fall of communism a(1990–2005).

No. Code Name of the organization Country Patents Main field of activity

1 N RICHTER GEDEON VEGYESZETI GYAR RT Hungary 79 Pharmaceutical2 N CHINOIN GYOGYSZER ES VEGYESZETI TERMEKEK GYARA RT. Hungary 74 Pharmaceutical3 N EGIS GYOGYSZERGYAR Hungary 63 Pharmaceutical4 F GENERAL ELECTRIC COMPANY Hungary 51 Various5 F SAMSUNG ELECTRONICS CO., LTD. Russia 36 Electronics6 N ELBRUS INTERNATIONAL LTD. Russia 32 Computer Technology7 F LSI LOGIC CORPORATION Russia 31 Communications; Semiconductors8 F GENERAL ELECTRIC COMPANY Russia 30 Various9 F SUN MICROSYSTEMS, INC. Russia 30 Computer Technology

10 N BIOGAL GYOGYSZERGYAR RT. Hungary 26 Chemicals11 N PLIVA FARMACEUTSKA, KEMIJSKA, PREHRAMBENA I KOZMETICKA Croatia 24 Pharmaceutical; Cosmetics12 N LEK PHARMACEUTICAL AND CHEMICAL COMPANY D.D. Slovenia 24 Pharmaceutical;Chemicals13 F AJINOMOTO COMPANY INCORPORATED Russia 21 Food14 F CERAM OPTEC INDUSTRIES, INC. Russia 21 Optical fiber; Lasers15 N OTKRYTOE AKTSIONERNOE OBSCHESTVO

“NAUCHNO-PROIZVODSTVENNOE OBIEDINENIE “ENERGOMASH”IMONI AKADEMIKA KAKSOLMIKA V.P. GLUSHKO”

Russia 20 Energy; Engines

16 F R-AMTECH INTERNATIONAL, INC. Russia 19 Emerging Technologies17 F TELEFONAKTIEBOLAGET LM ERICSSON Hungary 18 Telecommunications18 N TUNGSRAM RESZVENYTARSASAG Hungary 18 Lighting19 F BOSCH SIEMENS HAUSGERATE GMBH Slovenia 16 Household appliances20 F UNIVERSITY OF CHICAGO Russia 16 Education21 F SEMICONDUCTOR COMPONENTS INDUSTRIES, LLC Czech Republic 15 Semiconductors22 N GYOGYSZERKUTATO INTEZET KFT Hungary 15 Pharmaceutical

S t of 19ntitie

AotakmsicfFctfc

•

•

•

•

•

ource: Based on a report from Patent Technology Monitoring Branch (PTMB) “Couna Organizations receiving 5 or more utility patents during 1969–2004; F—foreign e

˙ represents the flow of new ideas at time t, h is the average skillf labor while ı is the productivity of skilled adjusted unit of laborhat increases with the existing stock of ideas A. Consistent with

“standing on the shoulders of giants” effect of inter-temporalnowledge spillovers I expect � > 0. LA is the effort (total invest-ent or in the original model, labor input) devoted to the “ideas”

ector. ˝ is a vector that comprises other factors that impact thedea production process within both national (institutions, poli-ies, resources) and international (trade, foreign direct investment)rameworks. This approach builds on the NIC concept developed byPS, the rich descriptive work on NIS and previous findings in theross-sectional literature. Moreover, ˝ comprises additional fac-ors specific to transition countries which I expect to be relevantor their innovative output. All these variables are grouped in fiveategories:

National infrastructure for innovation (XINFR) comprising resourcesavailable for innovation, such as sources of finance for R&D,human capital available, investment in education and basic sci-ence, as well as policies like protection of intellectual property,ease of doing business and tax regimes.Cluster specific environment for innovation (YCLUS) emphasizing themicro-economic decisions of firms to undertake R&D aggregatedat the country level in the total business expenditures on R&D;this factor is recognized as the main driver of commercial inno-vation in industrialized countries.Linkages between the infrastructure and industrial clusters (ZLINK )are activities of various organizations like universities, researchinstitutes and public think tanks that provide a link between var-ious elements of the first two groups.Transition specific factors (TTRAN) are extremely important for East-ern Europe and their disruptive influence should also impactinnovation via economic mechanisms.
Globalization related factors (WGLOB) such as inflows of trade, for-eign investment and migration (brain drain) play and increasinglysignificant role in today’s global economy giving opportunities forlearning, imitating and building new technologies, and reducingthe duplication of R&D efforts among countries.
69–2005 Utility Patent Grants by calendar year of grant”.s (firms, universities); N—domestic entities (research institutes, institutions, firms).

Thus, ˝ = {XINFR, YCLUS, ZLINK , TTRAN, WGLOB} incorporates awide set of resources, policies and economic variables that shouldinfluence the intensity of innovation in a country according toprevious findings in the literature. Moreover, while this set-upimplies complementarity among its components, it also raises someeconometric problems due to low variance in certain variables andpossible collinearity of regressors. For the estimations, I opt for loglinear specification of (4.1), except for the qualitative and percent-age variables, which makes it less sensitive to outliers and easier tointerpret as elasticities. Hence, the flow of new ideas is specified asa production function where all the factors described above enteron the right hand side and � is an error term:

log A = ˇ log h + � log A + ı log LA + ıI log XINFR + ıC log YCLUS

+ ıL log ZLINK + ıT log TTRANS + ıI log WGLOB + � (4.2)

4.2. Research hypothesis

Building on the strains of literature presented in Section 2 thisstudy will also explore some pertinent research questions in thecontext of developing and transitional economies. These hypothe-ses are presented below.

Hypothesis 1. A stronger and more effective intellectual propertyrights (IPR) regime increases the number of “new-to-the-world” inno-vations produced in a country. There is an ongoing debate whetherdeveloping countries should increase their legislative measures andenforce more vigorously in order to develop faster. One argumentis that a strong IPR policy increases the incentives for produc-ing local innovations that are patentable (Aghion et al., 2001) andalso attracts FDI with higher technological potential for spillovers(Smarzynska Javorcik, 2002; Kanwar and Evenson, 2001). However,multinationals may invest only in labor intensive segments abroad
while upstream activities (such as R&D) will remain exclusively intheir home office. Moreover, since IPR is applied equally over allsectors, the gains from attracting FDI in one industry may be offsetby losses from the others that have benefitted through imitation(Léger, 2005; Glass, 2004).


Table 6Descriptive statistics and panel unit root tests of the variables employed.

Variable N Mean Std. Dev. Min Max LL IPS

log (US patents)t+2 257 2.48 1.36 0.00 5.84 −6.23∗∗∗ −4.17 ∗ ∗∗log(R&D total) 229 12.15 1.38 8.77 15.82 −9.33∗∗∗ −5.73∗∗∗

log(US Patent Stock) 288 3.81 1.50 0.29 7.41 −5.84∗∗∗ −3.90∗∗∗

IPR index 288 6.05 3.04 2.20 11.25 – –Cost of doing business 288 38.45 15.27 16.00 76.50 – –log(R&D Government) 162 11.61 1.16 9.52 15.76 −26.27∗∗∗ −5.02∗∗∗

log(R&D Business) 164 11.41 1.46 7.70 14.66 −10.73∗∗∗ −1.55∗

R&D performed university 178 0.20 0.14 0.02 0.56 −8.19∗∗∗ −0.89log(industrial distortions) 236 −1.20 0.42 −2.25 0.20 −4.08∗∗∗ −2.55∗∗∗

Cumulative output drop 288 −11.02 25.28 −56.60 41.14 – –log(FDI inflows) 252 6.37 1.73 1.39 10.27 −7.86∗∗∗ −4.63∗∗∗

log(trade intensity) 232 −7.19 0.58 −9.42 −5.96 −4.37∗∗∗ −2.93∗∗∗

log(human capital) 255 2.19 0.15 1.79 2.35 0.06 1.12Education share (%GDP) 196 0.03 0.01 0.01 0.05 −4.53∗∗∗ −1.40∗

log(population) 256 9.10 1.20 7.18 11.90 −3.52∗∗∗ 1.35log (EPO patents) 202 2.47 1.55 0 5.47 −11.88∗∗∗ −8.93∗∗∗

l 7

N ; variai ** and

HpTricole

HtoetltH

HotihsntE

4

eatertden

mcsf

through regulations, which in turn affects the overall inventivenessof firms at the national level. A bureaucratic country will be less suc-cessful both in attracting foreign innovative firms and encouragingdomestic entrepreneurs. The data shows that the number of proce-

t+2

og(EPO patent stock) 192 3.55 1.7

ote: All panel unit root tests include individual effects and individual linear trendsn our data set; The null hypothesis for these tests is non-stationarity (unit root); *,

ypothesis 2. EECs’ ability to produce innovations relies more onresent R&D commitments than on the stock of prior knowledge.hus, by this assumption, the current human capital and financialesources employed should have a greater impact on commercialnnovations rather than the amount of previous knowledge. In thease of the EECs, these knowledge stocks are expected to be mostlyutdated and concentrated in mature industries with a presentow propensity to patent, another negative legacy of centralizedconomic systems (Radosevic and Kutlaca, 1999).

ypothesis 3. Transitional countries will rely more on public ratherhan business R&D expenditure, identified usually as the main driverf innovation in developed economies. Since the whole marketconomy system is still relatively new for the EECs, one could expecthat the main push in innovative activities will come from pub-icly funded research institutions rather than private businesseshat require time for building up competitiveness (Suarez-Villa andasnath, 1993; Hu and Mathews, 2005).

ypothesis 4. Globalization related factors have a positive effectn national innovation. This view is consistent with the belief thatrade and foreign direct investment are accompanied by signif-cant positive spillovers on productivity and innovation of theost country (Coe and Helpman, 1995; Saggi, 2002). The recenturge in trade, FDI and outsourcing activities may give rise toew niches of innovation production for developing countrieshat possess a certain level of absorptive capacity such as theECs.

.3. Variables

In this section I will briefly describe the variables used tostimate (4.2). Further details on data sources and constructionre provided in Appendix A. The data covers 16 EECs duringhe period 1990 to 2007. Means and standard deviations of themployed variables are reported in Table 6, while pair-wise cor-elations appear in Table 7. As expected, the main challenge washe availability and quality of data, since some of these countriesid not collect this type of data prior to mid 1990s while oth-rs have adopted quite late the international classifications andorms.

As a proxy for “new-to-the-world” innovation, I employ iny analysis the number of patents at the USPTO. This variable

onstitutes a good measure of technologically and commerciallyignificant innovations at the world’s frontier, especially usefulor cross-country studies since it avoids comparability issues for

−0.67 7.07 1.69 −1.26∗

bles for which a value for this test is not available do not possess a time dimension*** indicate parameters that are significant at the 10%, 5% and respectively 1%.

national granted patents such as differences in standards, costs,protection offered and commercial benefits.11

In order to control for the previous knowledge stock of a coun-try (A), FPS and subsequent studies have used GDP per capita as aproxy for what they call “technological sophistication”. However,this variable includes a multitude of influences beside the histori-cal and country-specific component of technological progress. Thus,for this purpose I opt for a more straightforward measure of knowl-edge stock by constructing a patent stock variable using yearly flowsof patents. The computational details are presented in Appendix A.

Consistent with the previous micro evidence, I include the grossdomestic expenditure on R&D as a raw measure of national inno-vation effort. However, empirical evidence suggests that it mightbe important to distinguish between the contribution of privatebusinesses to R&D finance (Bassanini and Ekkehard, 2002) and thegovernment’s support (Hu and Mathews, 2005), since the latter maybe particularly important in the case of emerging innovators suchas the EECs.

Skilled human capital complements the financial R&D effortsto produce innovations (Griffith et al., 2004). Furthermore, uni-versities represent a vital link in this process that provides boththe human resources needed and the basis for research yieldingspillovers back to the industry (Jaffe et al., 1993; Adams et al., 2001).In a systemic view, the R&D performed by universities successfullylinks the available national infrastructure for innovation with thespecific business efforts.

From the policy perspective, a key variable is the IPR regime.This issue has become increasingly important in the last years withits embedment in international trade agreements such as TRIPSand its controversial impacts on developing nations. The IPR indexemployed in this paper takes into account both dimensions of thisissue: the formal one, represented by legal commitments to IPtreaties (Park and Wagh, 2002) and the informal one that considersthe actual enforcement of these laws (Smarzynska Javorcik, 2002).

The cost of doing business variable is used as a proxy for thecountry’s ability to create and stimulate the business environment

11 Usually, the benefits are proportional to the size of the market where the patentis granted. In 2005 USA has attracted the largest patent applications (417,508) andgrants (157,717) worldwide (USPTO Patent Statistics Chart, 2006).


Table 7Pair-wise correlations.

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

1 log (US patents)t+2 1.002 log (R&D Total) 0.78 1.003 log(US Patent Stock) 0.89 0.81 1.004 IPR index 0.50 0.47 0.48 1.005 Cost of doing business 0.08 0.09 0.20 0.26 1.006 log (R&D Government) 0.81 0.93 0.77 0.38 −0.02 1.007 log (R&D Business) 0.76 0.97 0.75 0.43 −0.02 0.88 1.008 R&D performed university −0.28 −0.41 −0.32 −0.16 −0.10 −0.29 −0.49 1.009 log (industrial distortions) −0.02 0.05 0.05 −0.23 0.12 −0.06 −0.02 −0.28 1.00

10 cumulative output drop 0.19 0.28 0.23 0.67 0.36 0.19 0.12 0.13 −0.55 1.0011 log (FDI inflows) 0.66 0.49 0.66 0.40 −0.13 0.56 0.46 −0.04 −0.22 0.26 1.0012 log (Trade intensity) 0.18 −0.04 0.31 0.27 0.44 −0.25 −0.28 0.11 −0.09 0.23 0.39 1.0013 log (Human capital) 0.18 0.14 0.17 0.05 −0.18 0.07 0.02 −0.29 0.40 −0.14 0.21 −0.04 1.00

−0.40.60.70.5

dssgi

flsaitnana

sad(Ttde

bitaeit

5

5

ap

tdo

Most of the variables enter in log form, yielding useful resultsin terms of subsequent interpretation (elasticities) and minimiz-ing the influence of possible outliers.16 In order to take advantage

14 Education share (% GDP) −0.44 −0.47 −0.43 0.02 0.20 −0.4115 log (Population) 0.61 0.72 0.65 0.03 −0.17 0.7116 log (EPO patents)t+2 0.88 0.76 0.90 0.46 0.27 0.7317 log (EPO patent stock) 0.80 0.55 0.86 0.39 0.20 0.57

ures required to start a firm and the costs associated with it varyignificantly worldwide.12 Such high entry costs are associated withignificant corruption, larger black market and low quality publicoods (Djankov et al., 2002) and are expected to have a negativempact on innovative output.

Prior to 1990s, Eastern Europe was virtually closed to foreignows of capital and goods. Since then, EECs have accomplishedignificant improvements changes and today they represent anttractive destination for foreign direct investment. As a result, anncreasing number of multinationals come in this region seeking toake advantage of the cheap yet skillful labor force. This suggestsew opportunities for local innovation production via FDI (Athreyend Cantwell, 2007). Moreover, previous studies document that sig-ificant knowledge spillovers arise from trade flows giving indirectccess to the R&D performed abroad (Coe and Helpman, 1995).

It is widely known that in the early 1990s the EECs facederious challenges in reallocating resources, a result of a rigidnd closed communist economic system.13 These distortions haveisruptive effects both on economic and innovation mechanismsSrholec, 2007) and need to be accounted in our regressions.herefore, I employ an industrial distortion index that monitorshe progress of a country over time in terms of reducing theseistortions compared with an international benchmark of marketconomies.

The cumulative output decline shows the percent differenceetween the end of transitional downturn in the region (2000) and

nitial (1990) levels of GDP and is a proxy for the harshness of theransition affecting also the resource allocation towards innovativectivities. Although most of the countries have surpassed the lev-ls of development by 2004, exceptions can be found afterwardsn some Commonwealth of Independent States (CIS) countries andhe war haunted Serbia and Montenegro.

. Empirical analysis and results

.1. Methodology

The estimation of the relationship between innovation’s outputnd inputs is not straightforward for several reasons. First, out-ut (patent) data are discrete, non-negative but potentially with

12 From 2 business days in Australia to 521 in Madagascar in 2005.13 Two distinctive features were a very small service sector and an oversized indus-rial one as a result of the development strategy of the 1960s. Most of the countriesealt with this legacy in the 1990s through privatization, restructuring or liquidationf such industrial mammoths.

3 0.47 −0.01 0.08 −0.35 0.08 −0.01 1.008 −0.51 0.33 −0.20 0.39 −0.23 0.45 −0.45 1.000 −0.24 −0.01 0.23 0.70 0.23 −0.01 −0.40 0.58 1.003 −0.13 0.00 0.19 0.69 0.28 0.06 −0.38 0.46 0.81 1.00

many zeros, suggesting the use of count models that assume aPoisson or a negative binomial distribution (Hausman et al., 1984;Blundell et al., 2002). However, in our case the low number of zeroobservations (5.83% of the total) reduces somewhat this concern.14

Secondly, patenting involves some lag between inputs and output,while its length is still subject of research (Hall et al., 1986). Thispaper attempts to address all these issues by exploring the datausing several approaches and perform additional robustness checksto validate the results.

With respect to the empirical implementation of (4.2), the flowof new innovations, A(j, t), is proxied by the number of patentsgranted in year t to country j. However, in practice, we observe alag between a patent application and a grant at the USPTO whichis on average two years (Furman and Hayes, 2004). Hence, (4.2)becomes:

log PATj,t+2 = ˇ log hj,t + � log PATSj,t + ı log LAj,t + ıI log XINFRj,t

+ ıC log YCLUSj,t + ıL log ZLINK

j,t + ıT log TTRANSj,t

+ ıI log WGLOBj,t + �j,t (5.1)

where j is the country subscript, t is the time subscript, PAT isthe flow of USPTO patents, h is a measure of human capital, PATSrepresents the computed stock of knowledge and LA is the totalR&D expenditure. The other variables included in the empiricalspecification are elements of ˝ described in Section 4.1. Thus, allregressors are lagged by two years with respect to the left handside variable.15 This structure also addresses some endogeneityissues since lagged regressors are predetermined with respect tothe dependent variable.

5.2. Estimation

of three key variables that lack the time dimension (IPR index,

14 Moreover, recent aggregated studies treat patents as continuous (Botazzi andPeri, 2003; FPS and subsequent work on NIC; Bosch et al., 2005) while employingcount models can provide additional robustness for our estimates.

15 We also test the robustness of our results to a one year lag and no lag specifica-tions, not reported here.

16 The exceptions are given by percentage and qualitative variables for which alogarithmic specification wouldn’t make sense: the IPR index, cumulative outputdrop, cost of doing business, the percentage of R&D performed by universities andthe percentage of GDP spent on secondary and tertiary education.

rch Pol

cIrsrfw(tcs

thcdrtffitt

nttcfipaft(pecgitcarTwtnnIe

arrsne

t

evi

a

S.M.S. Krammer / Resea

ost of doing business and cumulative output drop in transition),opt for a GLS estimator and include various controls (year and

egional dummies) to capture as much as possible of the unob-erved heterogeneity (Wooldridge, 2002).17 To make sure that myegressions are not spurious from a time series perspective, I per-orm the most common two panel unit root tests: Levin et al. (2002)hich assumes a cross-sectional common unit root and Im et al.

2003) that allows for individual unit root processes across sec-ions. The outcomes of these tests are presented in the last twoolumns of Table 6 confirming that the variables employed aretationary.

In addition, I run a couple of diagnostic tests to make sure thathe proposed estimates are efficient. Using a likelihood ratio test,omoskedasticity is firmly rejected in all models. Beyond this, serialorrelation could also be biasing the estimates so I perform the testescribed by Wooldridge (2002). However, the null of no serial cor-elation is strongly rejected in all models except Robust2, provinghat this is an issue for my estimations. Due to these concerns, I use aeasible generalized least squares (FGLS) estimator that is robust torst-order panel-specific autocorrelation and panel heteroscedas-icity. Table 8 presents the panel estimations including additionalime fixed effects and dummies for the CIS and Baltic countries.

Model 1 estimates a simple knowledge production function forational innovation that includes the current resources devoted tohe R&D sector (R&D TOTAL) and the stock of previous knowledge inerms of patents (US PATENT STOCK). Both terms are highly statisti-ally significant with estimated coefficient in the range of previousndings. Another estimation, not reported here to preserve com-arability with the other models, yields similar results using anlternative measure of LA (number of researchers) and controllingor R&D intensity of a country.18Model 2 explores in detail the fac-ors underlying a country’s national infrastructure for innovationXINFR) in terms of both commitments (R&D GOVERNMENT) andolicies (IPR regime, COST OF DOING BUSINESS) that complementxisting knowledge (US PATENT STOCK). Model 3 incorporates theontribution of industrial clusters (YCLUS) within a nation aggre-ated in R&D BUSINESS and linkages between it and the nationalnnovative capacities (ZLINK ) emphasizing the role of universi-ies (UNIVERSITY R&D PERFORMANCE) in facilitating technologicalommunication, sharing and development. Model 4 incorporatesdditional controls for the negative effects of transition (TTRAN) viaestructuring costs in the economic sectors (INDUSTRIAL DISTOR-IONS) and overall collapse of the economy (OUTPUT DECLINE)hich prove to be very important for Eastern European innova-

ion as well. Finally, Model 5 or the “Full Model” goes beyond to theational dimensions of innovative performance by including sig-ificant external forces (WGLOB) such as FDI INFLOWS and TRADE

NTENSITY that cannot be ignored in today’s global and interlinkedconomy.

The first point to highlight is that regardless of the numbernd variety of regressors in these estimations, the results are quiteobust. Keeping in mind that these are transition countries, theesults are remarkable. Secondly, the estimations reveal a con-istent picture with the postulated production function (5.1) for
ew-to-the-world innovations proxied by US patents, despite itsclectic nature.
In the following I am going to briefly review and summarizehe results of the FGLS estimations. For model selection, I employ

17 Besides those that have only one value for each cross-section, human capital,ducational expenditure as a percentage of GDP and the R&D statistics present a lowariance over time. A fixed effects estimation will discard the information containedn these variables.18 The elasticities are 0.23 for log(researchers), 0.25 for log(R&D per researcher)nd 0.66 for log(US Patent Stock), all significant at p < 0.000.

icy 38 (2009) 845–860 855

the two most common information criteria (Akaike and SchwarzBayesian) reported in the lower part of Table 8. Both rank Model 5or the “full” one as preferred from an econometric perspective so itsfindings will be emphasized here. The previous stock of knowledgeappears to be the most important national source for the streamof new innovations with an estimated elasticity of 0.37. This isin line with previous findings on developed countries but contra-dicts our second hypothesis formally tested in Model 1. Governmentinvestment in research and development (0.13) outplays the corre-spondent business expenditure (0.11), consistent with Hypothesis3. As expected, both exercise a positive and highly significant impacton the stream of new patents for the EECs. The link between indus-try and government provided by universities and laboratories hasa crucial impact (1.53) while the policy choices represented by theregulatory burden on businesses (−0.01) and effective IPR regime(0.13) are also very important. In the case of EECs, one cannotignore the adversity of transition process seized by the cumulativedecline of output (−0.01) and the disruptive process of realloca-tion and restructuring captured by an industrial distortion index(−0.68). Finally, globalization plays a significant role in the processof knowledge spillovers via trade (0.37) and foreign direct invest-ment flows (0.14). The effects of these two variables are similaror even higher than the contribution of domestic R&D (businessand government) to patenting in the EECs, consistent with previ-ous findings in the literature on R&D spillovers. All variables entersignificantly and additively in these regressions suggesting theimportance of these diverse factors to the innovative capacity of acountry.

5.3. Robustness

5.3.1. Additional controls: human capital, human capitalenhancing policies, population and dummies for possible outliers

Human capital (HK) plays both a direct role in impacting theproductivity and innovative capacity of a country (Engelbrecht,2002) as well as an indirect one, by determining the efficiency ofits absorptive capacity (Nelson and Phelps, 1966). Thus, I wouldlike to control for these cross-country differences in HK andsee their impact on innovation output (Robust 1). The skills ofhuman capital available in these countries is proxied by the widelyemployed variable from Barro and Lee (1996) and its updated2000 version. This index reports the average years of secondaryschooling in male population over 25 years old over five-yearperiods.

Moreover, in line with the literature (Acs et al., 2002; Varsakelis,2006; FPS) that emphasizes the role of education in stimulat-ing national innovation, I include also a control for the effectof this policy (Robust 2). Specifically, I use the expenditure ontertiary and secondary education as a percentage of GDP inmy regressions, under the assumption that a high educatedlabor force increases the amount of innovation undertaken in acountry.

Finally, theoretical considerations suggest that market size(proxied here by population) also matters by providing incentivesfor innovations. Furthermore, using patent counts as dependentvariable may raise some concerns regarding a scale effect since big-ger countries would put up bigger number. One option would be tonormalize all relevant variables by population and obtain per capitavalues, while an alternative would be to include it in the regressionsand test whether scale effects have an independent influence onpatenting activity. For convenience, I opt for the latter technique in
Robust 3.
When included in the preferred model, all three variables pre-sented above appear with the expected signs and high magnitudecoefficients but no statistical significance, suggesting that the pre-ferred model is well specified while keeping in mind also the low


Table 8Determinants of new Eastern European innovations (FGLS estimation a, 1990–2007).

Models Model 1 Model 2 Model 3 Model 4 Model 5 Robust 1 Robust 2 Robust 3 Robust 4Jones simpleprod.function

Nationalcommit-ments

+ Ind. clustersand linkages

+ Transitionfactors

+ Globalizationfactors (Full)

Full + humancapital

Full + educpolicy

Full +population

Full with EPOPatents

Variables

LA log R&D TOTAL 0.242 ***

(0.063)

A log US Patent Stock 0.664 *** 0.595 *** 0.446 *** 0.441 *** 0.372 *** 0.368 *** 0.376 *** 0.350 ***

(0.070) (0.061) (0.077) (0.086) (0.079) (0.079) (0.080) (0.082)

XINF IPR index (legis+enforce) 0.051 ** 0.091 *** 0.144 *** 0.135 *** 0.146 *** 0.136 *** 0.144 *** 0.159 ***

(0.023) (0.020) (0.033) (0.037) (0.038) (0.040) (0.038) (0.027)

XINF cost doing business −0.006 * −0.012 *** −0.014 ** −0.016 *** −0.005 −0.016 *** −0.000 −0.013 *

(0.003) (0.003) (0.005) (0.005) (0.010) (0.006) (0.011) (0.007)

XINF log R&D Government 0.274 *** 0.153 ** 0.164 ** 0.131 * 0.153 ** 0.149 ** 0.130 ** −0.050(0.056) (0.072) (0.074) (0.067) (0.069) (0.075) (0.065) (0.058)

YCLUS log R&D Business 0.247 *** 0.107 0.118 * 0.133 * 0.117 0.095 0.238 ***

(0.080) (0.083) (0.069) (0.070) (0.074) (0.071) (0.064)

ZLINK University R&D Performed 1.389 ** 1.573 *** 1.535 *** 2.308 *** 1.648 *** 1.407 ** 0.228(0.579) (0.609) (0.569) (0.791) (0.613) (0.560) (0.652)

TTRANS log industr distortions −0.335 ** −0.681 *** −0.886 *** −0.680 *** −0.829 *** −0.418 **

(0.137) (0.174) (0.229) (0.184) (0.191) (0.183)

TTRANS output decline −0.012 *** −0.013 *** −0.023 *** −0.014 *** −0.019 *** 0.001(0.004) (0.004) (0.008) (0.004) (0.005) (0.007)

WGLOB log FDI inflows 0.137 *** 0.119 *** 0.134 *** 0.109 ** 0.110 ***

(0.041) (0.042) (0.042) (0.043) (0.033)

WGLOB log trade intensity 0.368 * 0.240 0.384 * 0.422 ** 0.488 ***

(0.205) (0.230) (0.213) (0.208) (0.246)

ControlsXINF log human capital 1.201

(0.939)

XINF Education share 4.158(6.999)

XINF log population 0.263(0.168)

A log EPO patent stock 0.231 ****(0.052)

Constant −3.439 *** −2.546 *** −3.972 *** −3.604 *** −0.668 −5.482 −0.897 −2.997 0.977(0.739) (0.764) (0.844) (0.919) (1.814) (4.241) (1.940) (2.417) (2.317)

CIS dummy −0.128 0.060 0.597 *** 1.005 *** 1.345 *** 0.931 ** 1.351 *** 0.708 1.731 ***

(0.127) (0.153) (0.159) (0.299) (0.283) (0.442) (0.293) (0.501) (0.530)

Baltics dummy 0.178 0.092 −0.019 −0.250 −0.343 −0.591 * −0.360 0.018 −0.504 *

(0.289) (0.209) (0.267) (0.324) (0.312) (0.353) (0.331) (0.371) (0.293)

Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes

Wald Chi square 634.60 678.53 903.74 949.83 1,436.70 1,568.33 1,432.67 1,480.57 1,602.71AIC 249.96 150.18 109.87 103.83 96.74 96.24 100.23 95.97 64.97BIC 317.93 217.83 181.32 180.86 179.70 182.17 181.37 181.89 144.39N 221 160 145 143 143 143 143 143 126

LR test: H0 no heteroskedasticity 147.51 *** 71.02 *** 123.86 *** 120.35 *** 114.05 *** 116.55 *** 105.88 *** 90.21 *** 87.42 ***

W test: H0 no serial correlation 4.251 * 6.724 ** 4.226 * 7.360 ** 12.88 *** 15.67 *** 11.84 *** 11.90 *** 2.90

a GLS estimation robust to heteroskedasticity and group specific autocorrelation of order one.

vcmpi

sC

* P < 0.1.** P < 0.05.

*** P < 0.01 (standard errors in parenthesis).

ariance of these controls. Also, the results of the three robustnesshecks point out the importance of skills available (1.20) and invest-ent in higher education (4.15) while some scale effects (0.26) are

resent in our estimations. However, Model 5 remains robust to thenclusion of all these variables.

Moreover, multiple reasons such as economies of scale, mas-ive R&D capabilities and defense expenditures inherited from theold War era, suggest that the Russian Federation could be an out-

lier that influences our results. However, when employing a singledummy variable for Russia the results are very similar to Robust 3,emphasizing the role of previous stock of knowledge, governmental
support and FDI, while the cost of doing business, openness to tradeand private R&D are not statistically significant. Moreover, the AIC(106.11) and BIC (192.04) are higher than before, suggesting thatthis model (Model 5) remains the preferred one. These results areavailable upon request.


Table 9Determinants of new Eastern European innovations (additional estimations, 1990–2007).

Variables Model 5 Estimation (full model with log USPATt+2)

FGLS OLS (N-W) a Poisson NegBin (ML) NegBin (QML)

A log US Patent Stock 0.372 *** 0.421 *** 0.536 *** 0.438 *** 0.501 ***

(0.079) (0.080) (0.040) (0.059) (0.067)XINF IPR index (legis+enfor) 0.135 *** 0.082 ** 0.046 *** 0.066 *** 0.051 *

(0.037) (0.036) (0.018) (0.022) (0.029)XINF cost doing business −0.016 *** −0.007 −0.012 *** −0.007 −0.010 *

(0.005) (0.008) (0.003) (0.006) (0.005)XINF log R&D Government 0.131 * 0.269 ** 0.213 *** 0.227 *** 0.211 **

(0.067) (0.124) (0.054) (0.085) (0.087)YCLUS log R&D Business 0.118 * 0.069 0.125 ** 0.130 * 0.137 *

(0.069) (0.100) (0.050) (0.078) (0.081)ZLINK University R&D Performed 1.535 *** 1.074 0.775 * 0.883 0.805

(0.569) (0.869) (0.418) (0.661) (0.679)TTRANS log industr distortions −0.681 *** −0.733 *** −0.736 *** −0.654 *** −0.710 ***

(0.174) (0.226) (0.107) (0.163) (0.176)TTRANS output decline −0.013 *** −0.012 ** −0.008 *** −0.010 *** −0.009 **

(0.004) (0.005) (0.002) (0.003) (0.004)WGLOB log FDI inflows 0.137 *** 0.191 *** 0.082 *** 0.149 *** 0.111 **

(0.041) (0.055) (0.027) (0.044) (0.046)WGLOB log trade intensity 0.368 * 0.520 ** 0.650 *** 0.548 ** 0.613 ***

(0.205) (0.211) (0.104) (0.221) (0.186)

CIS dummy 1.345 *** 1.030 *** 0.942 *** 0.905 *** 0.919 ***

(0.283) (0.338) (0.154) (0.298) (0.267)Baltics dummy −0.343 0.025 0.279 0.177 0.268

(0.312) (0.374) (0.178) (0.263) (0.291)Constant −0.668 −0.930 −0.967 −2.008 −1.402

(1.814) (2.031) (0.862) (1.639) (1.489)Year fixed effects yes yes yes yes yes

CT test overdispersion – – 0.005 *** –Woolridge test overdispersion – – 0.007 ** –

AIC 96.74 1.47 7.29 6.98 7.11BIC 179.70 2.05 7.87 7.58 7.69R2 – 0.90 0.98 0.95 0.97N 143 143 143 143 143

a Regression with Newey-West standard errors that uses autocorrelations up to m = 4 to compute the standard errors. For the truncation parameter m, I employ the usualrule of thumb and compute it as (0.75 × (N ∧ 1/3)) which equals 3.92, rounded up to 4.

5

pteptPsaacgiswRsh

5

mttb

W

* P < 0.1.** P < 0.05.

*** P < 0.01 (standard errors in parenthesis).

.3.2. Different proxies for innovation at the technological frontierUSPTO patents have been used extensively in the literature as a

roxy for new-to-the-world innovations. However, one might arguehat factors such as trade intensity and affinity (historical, political,conomic) to the USA may induce a bias the amount of innovationsatented here. To test the results of our preferred model againsthis issue, I perform a similar analysis of (5.1) using EPO (Europeanatent Office) patents and patent stocks (Robust 4). For the EECs, thiseems even a more reasonable choice due to their recent economicnd political reintegration in the European space. Using EPO patentpplications has two potential disadvantages. First, in theory, appli-ations are a weaker proxy for innovation since not all of them areranted patents. Secondly, the data is limited to 1990–2006 affect-ng to some extent the accuracy of our computations of EPO patenttocks. Despite these limitations, the model performs reasonablyell. While the coefficients for GOVERNMENT R&D, UNIVERSITY&D PERFORMANCE and OUTPUT DECLINE loose their statisticalignificance, the other estimates remain in the previous range andighly significant.

.3.3. Estimation techniques
As mentioned in the beginning of this section, the choice of esti-
ation techniques is an important issue in this literature. Whilehe low number of zero patents in our sample justifies treatinghis variable as continuous, I would like to check these findingsy employing also count models estimations.

Table 9 presents the results of these additional estimations. Thefirst column reports the FGLS results with the preferred M odel 5. Inthe second column, I present a simple OLS regression with Newey-

est standard errors (OLS NW) which are both heteroskedasticityand autocorrelation consistent. The majority of the coefficientsremain robust to this estimator compared to the benchmark (FGLS).Next, I estimate a Poisson regression model (PRM) which assumesequidispersion or Var(PAT) = E(PAT) and conditional zero mean oferrors but allows for heteroskedasticity. The coefficients remainhighly significant and with expected signs while R squared isextremely high (0.98). However, in practice, PRM rarely fits the datadue to overdispersion. Estimates of a PRM for overdispersed dataalthough unbiased are inefficient since usually the standard errorsare biased downward. To test this restriction I use two methods:the one described by Cameron and Trivedi (1990) and Wooldridge(1997) approach. The former uses fitted values of the dependentvariable to regress (PAT − PAT f )2 − PAT on (PAT f )2, while the lat-ter regresses the standardized residuals on predicted values ofpatents. In our case, both t-statistics are highly significant leadingus to reject the Poisson restriction. Moreover, the significance of theestimated coefficients indicates overdispersion in the residuals. As
expected, these results suggest that the negative binomial distribu-tion is preferred over the Poisson one. Column 3 reports the negativebinomial maximum likelihood (NegBin ML) estimation with esti-mated coefficient in the line with previous results, except the COSTOF DOING BUSINESS and UNIVERSITY R&D that lose their statistical

8 rch Po

sob(aWatt

6

tlebmeetai

ipima(titc(R

spiaolatc

lnceitoag

wlao

p(2


ignificance just like in the OLS NW estimation. Given the evidencef overdispersion, I re-estimate the model using a two-step negativeinomial quasi-generalized pseudo maximum likelihood estimatorNegBin QML) with generalized linear model (GLM) standard errorsnd covariance and using the determined variance from the aboveooldridge test. The results are robust to inclusion of GLM covari-

nces and standard errors proving that our initial conjectures abouthis preferred model are confirmed regardless of the estimationechnique.

. Discussion and conclusions

This study contributes to the cross-country empirical litera-ure by investigating the determinants of innovation at a nationalevel in the transitional Eastern Europe while the analysis could bextended for other developing countries. The approach undertakenuilds on the theoretical grounds of endogenous growth comple-ented with elements from the national innovation systems in an

clectic approach similar to FPS. The results are broadly robust tostimation technique and imply, despite the great numbers of fac-ors considered and endogeneity concerns, that they contributedditively to a country’s innovative performance as proxied bynternational patenting.

From the results a number of interesting observations and pol-cy recommendations emerge. The existing infrastructure (XINFR)rovides the basis for future innovations. Overall, EECs’ variation

n patenting rates depends significantly on private R&D commit-ents similar to OECD studies (FPS; Furman and Hayes, 2004) but

lso on governmental efforts, like in the case of Asian latecomersHu and Mathews, 2005).19 Moreover, government’s contribu-ion outperforms the business R&D investment (YCLUS), usuallydentified as the main driver of innovation in developed coun-ries. Associated with the latter’s continuous downward trend, thisalls for significant measures to provide incentives and supportin form of tax breaks or subsidies) to private firms performing&D.20

The infrastructure comprises also some relevant policy mea-ures. Reducing the bureaucratic burden in these countries has alsoositive impacts over the medium/long term evolution of patent-

ng. Furthermore, stronger intellectual property protection sendssignal to both domestic inventors and multinationals increasingverall the number of patents produced by a country. The postu-ated role of human capital and policies meant to enhance it, suchs investment in tertiary and secondary education, have a posi-ive effect although not statistically significant, probably due toollinearity issues.

Our results also emphasize the role of universities (ZLINK ) ininking the innovative infrastructure with the R&D effort of busi-esses. Unfortunately, the EECs are lagging behind their Westernounterparts in this aspect and have not been improving so far. Asxpected, the macroeconomic transitional forces (TTRAN) play anmportant negative role. The overall fall in GDP and industrial dis-
ortions exhibited by former communist economies took their tolln patenting, especially in the 1990s when the hard adjustmentsnd reforms were taking place. Opposite to their disruptive effect,lobalization (WGLOB) has opened the channels for new sources of
19 These papers use similar empirical specifications to the one employed by FPS,hile the present study has a different specification to accommodate the particu-

arities of transition countries. Thus, qualitative inferences between their findingsnd our results are still acceptable, while any comparison between the magnitudesf various coefficients is not valid.20 With the exception of Slovakia and Czech Republic, all the listed EECs performoorly in terms of business R&D averaging under 0.5% of GDP while the global leadersSweden, Japan, USA) invest between 2.4 and 1.91% (European Innovation Scoreboard006).

licy 38 (2009) 845–860

knowledge via trade, FDI, communication or migration. However,despite the generalized increase of FDI in the EECs between 1990and 2007, the bulk of it remains confined to a handful of countriesthat provide the right mix of policies and infrastructure. The restneed to improve and catch-up also in this perspective.

Given the methodological and data availability issues, the resultsof this paper should be treated with some caution. While we believeinternational patents to be a good proxy for commercially relevantnew technologies, they do not reflect the entire spectrum of inno-vative activities in a country, especially a developing one. Thesecountries are not very successful in producing inventions at thefrontier, due to the prominent requirements of this process; nev-ertheless, they can successfully adopt and diffuse it. Thus, anotherway to map innovation would be to include such additional mea-sures in the analysis. Moreover, there are significant differencesbetween countries in terms of R&D distribution (private, public),type of research undertaken (basic or applied science), industrialstructure (propensity to patent varies across sectors) and even cul-ture towards patenting, all affecting the overall amount of patentsoriginating from a country. A more detailed level of analysis (sector,firm) could shed light on these issues.

In conclusion, this analysis finds that both innovation-orientedand business friendly policies along a balanced innovation invest-ment mix are prerequisites to develop the EEC’s national innovationcapacities and boost their competitiveness on international mar-kets. These countries have proven that innovation takes placedespite inefficiencies and austere conditions while the need for sus-tained growth in the region requires more efforts to support anddevelop it. This fact becomes even more important now, when theinitial drivers of growth (reallocation of resources, benefits fromrestructuring, comparative advantage in labor intensive industries)are slowly petering out through economic integration within awider Europe.

Acknowledgements

The author is grateful for advice on earlier drafts from TheophileAzomahou, Bengt-Åke Lundvall, Loet Leydesdorff, Bianca Poti, Ken-neth Simons and two anonymous referees. I would also like to thankthe participants of the PRIME Winter School at CERIS-CNR Rome,Globelics Academy 2008 in Tampere and 4th PRIME Ph.D. confer-ence in Budapest for useful remarks.

Appendix A. Data: construction and sources

A.1. Sample countries

The 16 EECs considered for this study are: Belarus, Bulgaria,Croatia, Czech Republic, Estonia, Georgia, Hungary, Latvia, Lithua-nia, Poland, Romania, Russian Federation, Serbia and Montenegro,Slovakia, Slovenia and Ukraine. The term EEC refers generically toall countries from the Eastern Europe and the former Soviet Union(nowadays CIS).

A.2. Patent counts

Patent counts are drawn from the USPTO Full Text Database(http://patft.uspto.gov/) covering all patents issued in the USA froman EEC inventor. I use this measure instead of the usual “firstinventor” utility patent measure for two reasons: first, the at the
beginning of the 1990s these countries have a very low numberof patents and by using an “all inventor” approach the number ofzero counts is lowered; secondly, this measures allows for collabo-rative inventions that have at least one Eastern European inventoron board. In any case, the two counts are strongly correlated (0.99)
http://patft.uspto.gov/

rch Policy 38 (2009) 845–860 859

aafttcto

A

uflp(fiacS

A

RbY2trtp

A

T

A

Dpp

A

(gippdsb

ostadslo

Table A.1IP2 Index—effective enforcement of IPR regimes.

Score Description

1 Close to adequate IPR legislation present by the end of

S.M.S. Krammer / Resea

nd by this I gain a bit in terms of variance while taking into accountll innovative output with EEC origins. For accuracy, the numbersor the early 1990s have been corrected in the case of countrieshat have broken up by individual patent assessment. Those emit-ed after the breakup were reassigned towards one of the formeromponent countries using as criteria the geographic location ofhe inventor. All null observations have been converted to 0.001 inrder to make it possible to take logs.

.3. Patent stocks

For computations of the patent stocks in year t, I use the perpet-al inventory model and take into account the number of patentsows between the initial year 1963 and year (t − 1). The initialatent stock is computed using the method developed by Griliches1990): S0 = P0/(g + ı), where g is the average growth rate for therst 10 years of available data and ı is the depreciation rate, sett 15%, the most commonly used rate in the literature for dis-ounting patent flows. The subsequent stocks are computed ast = (1 − ı)St−1 + Pt , where Pt is the patent flow at year t.

.4. R&D variables

Data on gross, business and government expenditure on R&D,&D performed by universities (percentage of the total) and num-er of researchers comes from UIS S&T database, UNESCO Statisticalearbooks and OECD Main Science and Technology Indicators007 supplemented with compatible national statistics data andhen converted to comparable and constant US$. In the case ofesearchers, due to reduced availability of some indicators, desireo “let the data speak” and avoid extrapolations, I use the mostrevalent indicator in the case of the EECs (the head-count).

.5. Foreign direct investment

The FDI statistics on aggregated FDI inflows originate from UNC-AD’s FDI Online database.

.6. Trade intensity

Data comes on imports and exports are taken from the IMFirection of Trade Database 2007 (DOTS). Trade intensity is com-uted as the total trade amount (imports plus exports) as aercentage of GDP.

.7. Intellectual property rights (IPR) index

A widely used index of patent rights updated by Park and Wagh2002) places Hungary and Croatia on top, while Lithuania and Bul-aria are on the opposite side of the spectrum. This index (IP1)s constructed by taking into account five dimensions of patentrotection: the extent of coverage, membership in internationalatent agreements, protection from restrictions on patent rights,uration and mechanisms of enforcement, all broken down intoub-categories with assigned weights yielding final index valuesetween 0 and 5, with 5 being the highest degree of IPR protection.

However, in the case of the EECs as well as in many other devel-ping countries, the “official” enforcement standard often differsignificantly from the actual one. To control for this, a measure ofhe actual enforcement implemented in these countries is more
ppropriate (Radosevic, 2004). Thus, an IPR index like the oneeveloped by Smarzynska Javorcik (2002) serves better our analy-is (see Table A.1 below). This simple index (IP2) captures both theegislative and actual enforcement of the IPR regime and is basedn the description of IPR regimes by the International Intellectual
1995; some enforcement efforts undertaken2 Close to adequate IPR legislation present by the end of

1995; no enforcement efforts undertaken3 Lack of adequate IPR legislation at the end of 1995

Property Alliance for the US Special 301 Watch list, paying closeattention to trademark and copyright laws.

The IPR index employed in our study is a combination of theabove two dimensions: legislative protection and degree of enforce-ment. The values are obtained by multiplying IP1 and IP2 and wereextended (based on the guidelines presented in Table A.1) to coveralso the former Yugoslav republics.

A.8. Cost of doing business

This variable indicates the regulatory cost of business in a coun-try and was computed using the average duration for starting up anew business between 2003 and 2006. World Bank’s Doing Busi-ness database provides objective measures of business regulationsand their enforcement in comparable terms across 175 economies.The limitation of this dataset is given by its rather short time span(2003–2006).

A.9. Industrial distortion index

The distortion index is based on the predicted (benchmark)employment percentages (S∗

s ) in four broad sectors (agriculture,industry, market-oriented services and non-market-oriented ser-vices) in 50 other market based economies and after controlling forlevels of GDP, following a similar approach to that of Raiser et al.(2004). However, their index is defined differently as the sum ofabsolute values of (s − s∗) divided by 2, where s is the actual shareof employment in sector j, and s∗ is the benchmark in the samesector. This computational choice is a bit hard to justify so I optfor a classical distance measure given by an Euclidian formula asfollows:

DISTj,t =√

(SA − S∗A)2 + (SI − S∗

I )2 + (SMS − S∗MS)2 + (SNMS − S∗

NMS)2

(1)

where Si represents the actual share of employment in sector i andS∗

irepresents the benchmark share in the same sector. The index

is a measure of the overall distance of a transition economy froma market economy with the same per capita income. The valuescome from my own computations using employment data avail-able through LABORSTA (the International Labor Office statisticsdatabase) and GDP per capita levels from the World Bank’s Devel-opment Indicators 2007.

A.10. Education expenditure

The public national expenditure on education is collected mainlyfrom the World Bank’s EdStats database and completed with datafrom the national statistics offices. The tertiary and secondaryexpenditures as a percentage of GDP vary between 0.42 and 0.73among countries but very little over time within the same nationalentity.

A.11. Population

Population is drawn from the World Development Indicators2007. The mean is about 20 million people while the size of coun-

8 rch Po

tm5P1

A

t

R

A

A

A

A

A

B

B

B

B

B

B

C

C

D

E

F

F

F

G

G

G

G

H

H


ries in our sample varies a lot, from a really small one around 2illion (Estonia, Latvia and Slovenia) to a medium one betweenand 10 million (most of the countries) and above it (Romania,

oland and Ukraine with 20 to 45 million) while Russia (around50 million) is significantly bigger.

.12. EPO patents

Are extracted from EPO statistics on European patent applica-ions 1990–2006 (www.epo.org) complemented with Eurostat.

eferences

cs, Z., Anselin, L., Varga, A., 2002. Patents and innovation counts as measuresof regional production of new knowledge. Research Policy, Elsevier 31 (7),1069–1085.

dams, J.A., Chiang, E., Starkey, K., 2001. Industry-University Cooperative ResearchCenters. The Journal of Technology Transfer 26 (1–2), 73–86.

ghion, P., Howitt, P., 1992. A model of growth through creative destruction. Econo-metrica 60 (2), 323–351.

ghion, P., Christopher, H., Howitt, P., Vickers, J., 2001. Competition, imitationand growth with step-by-step innovation. Review of Economic Studies 68 (3),467–492.

threye, S., Cantwell, J., 2007. Creating competition?: globalisation and the emer-gence of new technology producers. Research Policy 36 (2), 209–226.

arro, R.J., Wha, L.J., 1996. International measures of schooling years and schoolingquality. American Economic Review 86 (2), 218–223.

assanini, A., Ekkehard, E., 2002. Labour market regulation, industrial relations andtechnological regimes: a tale of comparative advantage. Industrial and CorporateChange 11 (3), 391–426.

ebczuk, N., 2002. R&D expenditures and the role of government around the world.Estudios de Economia 29 (1), 109–121.

otazzi, L., Peri, G., 2003. Innovation and spillovers in regions: evidence from Euro-pean patent data. European Economic Review 47, 687–710.

osch, M., Lederman, D., Maloney, W., 2005. Patenting and research and develop-ment: a global view. World Bank Policy Research Working Paper 3739.

lundell, R., Griffith, R., Windmeijer, F., 2002. Individual effects and dynamics incount data models. Journal of Econometrics 108, 113–131.

ameron, A., Trivedi, P., 1990. Regression-based tests for overdispersion in the Pois-son model. Journal of Econometrics 46, 347–364.

oe, D., Helpman, E., 1995. International R&D spillovers. European Economic Review39 (5), 859–887.

jankov, S., La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 2002. The regulation ofentry. The Quarterly Journal of Economics 117 (1), 1–37.

ngelbrecht, H.J., 2002. Human capital and international knowledge spillovers inTFP growth of a sample of developing countries: an exploration of alternativeapproaches. Applied Economics 34, 831–841.

reeman, C., 1982. Technological infrastructure and international competitiveness.Draft Paper Submitted to the OECD Ad Hoc Group on Science, Technology andCompetitiveness (Mimeo).

urman, J., Porter, M., Stern, S., 2002. The determinants of national innovative capac-ity. Research Policy 31, 899–933.

urman, J., Hayes, R., 2004. Catching up or standing still? national innovative pro-ductivity among “Follower” Nations, 1978–1999. Research Policy 33, 1329–1354.

lass, A.J., 2004. Intellectual property protection and international technol-ogy diffusion. Working Paper Economics Department Texas A&M. Available:http://econweb.tamu.edu/aglass/ippitd.pdf.

riffith, R., Redding, S., Van Reenen, J., 2004. Mapping the two faces of R&D: pro-ductivity growth in a panel of OECD Industries. The Review of Economics andStatistics 86 (4), 883–895.

riliches, Z., 1990. Patent statistics as economic indicators: a survey. Journal of Eco-nomic Literature 28 (4), 1661–1707.

rossman, G., Helpman, E., 1991. Innovation and Growth in the Global Economy. MITPress, Cambridge, MA.

all, B., Hausman, J., Griliches, Z., 1986. Patents and R&D: is there a lag? InternationalEconomic Review 27 (2), 265–283.

all, B., Jaffe A., Trajtenberg, M., 2001. The NBER patent citation data file: lessons,insights and methodological tools. NBER Working Paper 8498.

licy 38 (2009) 845–860

Hausman, J., Hall, B., Griliches, Z., 1984. Econometric models for count data withapplication to the patents-R&D relationship. Econometrica 52 (4), 909–938.

Hu, M.C., Mathews, J.A., 2005. National innovative capacity in East Asia. ResearchPolicy 34, 1322–1349.

Im, K., Pesaran, M., Shin, Y., 2003. Testing for unit root in heterogeneous panels.Journal of Econometrics 115 (1), 53–74.

Jaffe, A.B., Trajtenberg, M., Henderson, R., 1993. Geographic localization of knowledgespillovers as evidenced by patent citations. The Quarterly Journal of Economics108 (3), 577–598.

Jones, C., 1995. R&D-based models of economic growth. Journal of Political Economy103 (4), 759–784.

Kanwar, S., Evenson, R., 2001. Does intellectual property protection spur technolog-ical change? Working Papers 831, Economic Growth Center, Yale University.

Lederman, D., Maloney, W.F., 2003. Research and Development (R&D) and Develop-ment. World Bank Policy Research Working Paper 3024, Washington, DC.

Léger, A., 2005. Intellectual property rights in developing countries: do they play arole? World Development 33 (11), 1865–1880.

Levin, A., Lin, C., Chu, C., 2002. Unit root tests in panel data: asymptotic and finitesample properties. Journal of Econometrics 108, 1–24.

Lundvall, B.A., 1992. Introduction, National Systems of Innovation: towards a theoryof innovation and interactive learning. Pinter Publishers, London.

Marinova, D., 2000. Eastern European patenting activity in the USA. Technovation21, 571–584.

Murrel, P., 1990. The Nature of Socialist Economies. Lessons from Eastern EuropeanForeign Trade. Princeton University Press, Princeton, NJ.

Nelson, R., Phelps, E., 1966. Investment in humans, technology diffusion and eco-nomic growth. American Economic Review 56, 66–75.

Nelson, R., 1993. National Innovation Systems: A Comparative Analysis. Oxford Uni-versity Press, New York, Oxford.

Park, W., Wagh, S., 2002. Index of patent rights. In: J. Gwartney (Ed.), EconomicFreedom of the World: 2002 Annual Report. The Fraser Institute, Vancouver.

Piech, K., Radosevic, S. (Eds.), 2006. The Knowledge-Based Economy in Central andEast European Countries and Industries in a Process of Change. Palgrave Macmil-lan.

Porter, M., 1990. The Competitive Advantage of Nations. Free Press, New York.Radosevic, S., Auriol, L., 1999. Patterns of restructuring in research, development

and innovation activities in central and eastern European countries: an analysisbased on S&T indicators. Research Policy 28, 351–376.

Radosevic, S., Kutlaca, D., 1999. Technological “catching-up” potential of Central andEastern Europe: an analysis based on US foreign patenting data. TechnologyAnalysis and Strategic Management 11, 95–111.

Radosevic, S., 2004. A two-tier or multi-tier Europe? assessing the innovation capac-ities of central and east european countries in the enlarged EU. Journal ofCommon Market Studies 42 (3), 641–666.

Raiser, M., Schaffer, M., Schuchhardt, J., 2004. Benchmarking structural change intransition. Structural Change and Economic Dynamics 15, 47–81.

Romer, P., 1990. Endogenous technological change. Journal of Political Economy 98(5), 71–102.

Saggi, K., 2002. Trade, foreign direct investment, and international technology trans-fer: a Survey. World Bank Research Observer 17 (2), 191–235.

Srholec, M., 2007. A multilevel approach to geography of innovation. Working Paperson Innovation Studies 20071010, Centre for Technology, Innovation and Culture,University of Oslo.

Smarzynska Javorcik, B., 2002. The composition of foreign direct investment andprotection of intellectual property rights: evidence from transition economies.European Economic Review 48 (1), 39–62.

Suarez-Villa, L., Hasnath, S.A., 1993. The effect of infrastructure on innovation:innovative capacity and the dynamics of public construction investment. Tech-nological Forecasting and Social Change 44, 333–358.

Varsakelis, N., 2001. The impact of patent protection, economy openness and nationalculture on R&D investment: a cross-country empirical investigation. ResearchPolicy 30, 1059–1068.

Varsakelis, N., 2006. Education, political institutions and innovative activity: a cross-country empirical investigation. Research Policy 35, 1083–1090.

Waguespack, D.M., Birnir, J.K., Scroeder, J., 2005. Technological development andpolitical stability: patenting in Latin America and the Caribbean. Research Policy
34, 1570–1590.
Wooldridge, J., 1997. Quasi-likelihood methods for count data. In: Pesaran, M.,Schmidt, P. (Eds.), In: Handbook of Applied Econometrics, vol. 2. Blackwell Pub-lishing, Oxford, pp. 352–406.

Wooldridge, J.F., 2002. Econometric Analysis of Cross-section and Panel-data. MITPress, Cambridge, MA.

http://www.epo.org

http://econweb.tamu.edu/aglass/ippitd.pdf

1 s2.0-s0048733309000262-main

Documents

Transcript of 1 s2.0-s0048733309000262-main