stop playing: the boycott of the moscow summer olympics ...econ.msu.edu/papers/cook/stop playing....

Stop Playing: The Boycott of the Moscow Summer Olympicsand Implications for Domestic Knowledge Creation ∗

Lisa D. Cook and Maksym Ivanyna

Corresponding Author:Lisa D. Cook

Department of Economics110 Marshall-Adams HallMichigan State University

486 W. Circle DriveEast Lansing, MI 48824

[email protected] (o)

September 3, 2018

∗The authors wish to thank Alberto Alesina, Susan Linz, Trevon Logan, Ilya Segal,Roald Sagdeev, Michael Schwarz, and Sergei Severinov for insightful conversations; sem-inar participants at Duke University, Harvard University, Michigan State University, theUniversity of Michigan, Moscow State University, and the National Bureau of EconomicResearch for helpful comments and suggestions; Mattan Alalouf for able research assis-tance; and staff at the Hoover Institution Library and Archives for their kind assistance.

1

Abstract

Recently, economists have examined how changes in foreign migra-tion change the domestic market for labor. Less attention has beenpaid to changes in other inputs, particularly concerning scientific dis-covery and technological spillovers. We use the escalation of the ColdWar in 1979-80 to identify a causal effect of scientific activity by for-eigners on the magnitude and direction of knowledge creation in theeconomy. Using data on over 165,000 U.S. and East German patents,we show that Soviet inventors who patent in the U.S. prior to 1979-80 both decrease their patent activity in the U.S. and increase theirpatent activity in countries not observing the boycott, namely EastGermany. We find that on average a 10-percent increase in inventiveactivity by Soviet scientists in East Germany increases domestic inven-tion by 2.5 percent in the long run. We also observe new collaborationbetween Soviet and East German inventors, which had been intention-ally limited during the Cold War, and increases in non-patent researchactivity, namely publications. Overall, our results suggest that foreignparticipation in scientific effort increases domestic scientific activity.

Keywords: innovation, inventive activity, technological spillovers,patents, Soviet Union, sport

JEL-classification: F22, F51, N44, N74, O31, O33, P16, Z20

2

1 Introduction

The Iron Curtain imposed constraints on international knowledge flows be-tween the West and Eastern Bloc countries during the Cold War due torestrictions on collaboration. Yet, the curtain was not impenetrable. Cook(2011) shows that many Soviet scientists did apply for patents in the U.S.1

For them, it was a principal means of being identified as the originator ofan idea and of disseminating research. We use a unique historical experi-ment arising from the boycott of the Moscow Summer Olympics in 1979-80to identify effects on scientific discovery and inventive activity. Using a noveldata set on patenting by Soviet inventors abroad, we uncover two new facts.First, international knowledge flows halted dramatically with greater defini-tion of the Western Bloc, namely following the escalation of the Cold War in1979-80, including the boycott of the Summer Olympics in Moscow in 1980and related boycotts. Second, scientific collaboration was not static but dy-namic during the Soviet period. We find that following the boycotts Sovietinventors substituted their lost links with the West by increasing collabora-tion with the Eastern-Bloc countries. East German scientific teams were theprimary beneficiaries of this development.

We use these facts to make two contributions to the literature. First,we provide a unique and the first systematic appraisal of the economic andscientific costs associated with the escalation of the Cold War and relatedboycotts. New micro data on patenting by Soviet residents abroad providedirect evidence that knowledge flows to the U.S. fell dramatically after theseevents. They correspondingly increased in East Germany. Cross-countrymacro data on patenting, that we also collect, corroborates our main findings.

Second, we extend the economic literature that has not settled the ques-tion of whether, how, and to what extent inventive activity by foreign scien-tists affects domestic knowledge production. Do foreigners introduce positiveexternalities for domestic scientists in the form of cooperation or generationof new ideas? Or do foreign scientists crowd out research funds or inventiveor innovative effort?2 On the one hand, Moser and Voena (2012) show that

1We use the terms ”scientists” and ”scientific teams” and ”inventors” and ”patentteams” interchangeably in this paper, because they were largely indistinguishable, unlikein the West, during the Soviet period. There is no evidence from the literature that Sovietscientific teams were reconfigured for the sake of patenting or registering their inventions,more generally.

2Unless otherwise stated, in this paper inventive activity will refer to activity related

3

domestic innovation increases in the long run in the case of a compulsory-licensing event (resulting from the Trade-with-the-Enemy Act), when thepool of foreign-owned knowledge becomes available at no cost for domesticresearchers, and the competition for new ideas becomes less intense. Onthe other hand, Borjas and Doran (2012) find no or negative spillovers ondomestic knowledge creation from the migration of Soviet mathematiciansto the U.S. when the Soviet Union disintegrated. The authors argue thatpositive human-capital externalities from knowledge creation by foreignersmay not materialize if the foreigners also compete for scarce research fundsor positions in universities.

Our paper explores the effect of foreign inventive activity on domesticinvention between these two extremes – foreign-owned knowledge becomingfree due to compulsory licensing and direct competition for research fundsresulting from scientists’ migration. Specifically, we ask whether increasedpatenting by foreign inventors affects domestic knowledge creation.

To identify the causal effect of patenting by foreigners on domestic inven-tive activity, we use the shock to geographical changes in foreign patenting bySoviet scientists, which happened exogenously as a result of the escalation ofthe Cold War starting in 1979, including the boycott of the Moscow SummerOlympics in 1980 and related events. After 1980, Soviet inventors continuedto respond to market-oriented incentives and desired to patent abroad toenhance their reputations, but they shifted their inventive efforts toward theclosest substitutes in the Eastern Bloc, such as East Germany. Given thatthe events of 1979-80 were driven by politics and not by developments in thescientific world, our identification strategy allows us to rule out alternativesources of changes in East German domestic knowledge creation, includingpatent class-specific trends or changes in Soviet-East German scientific co-operation.

We show that, as a result of the shock, knowledge spillovers increasedin East Germany, as measured by partnerships with Soviet scientists andpatenting by East German inventors. We find that on average a 10-percentincrease in inventive activity by Soviet scientists in East Germany increasesdomestic invention by 2.5 percent in the long run. We find that partnershipsbetween Soviet and East German inventors significantly increase immediatelyafter the shock. We also match our patent data to data on publications from

to process, product, or design invention. Innovative activity will refer to inventive activitythat has been commercialized.

4

Google Scholar and show that increased patenting by Soviet scientists mayalso have boosted other inventive activities, and hence have wider-reachingeffect on East German society.

2 Olympic and Other Boycotts

From 1952 when the Soviet Union rejoined the modern Olympic Games,these games served as a second front of the Cold War. The tradition ofthe Olympic Games as the focal point of international relations dates backto the Greeks. More recently, there are a few well-known examples. Hitlerused the 1936 Olympics in Berlin to legitimate the Nazi regime. In 1968 inMexico City, student protesters of the Olympics were killed by the Mexicangovernment, and African American medalists gave a ”Black Power” salute.During the 1972 Olympics in Munich, Palestinian terrorists kidnapped andkilled nine Israeli athletes, coaches, and officials. The 1976 Olympic Gamesin Montreal were boycotted by 32 countries, many from Sub-Saharan Africa,who protested New Zealand’s engagement in sports competitions with SouthAfrica. For 40 years, the U.S. and the U.S.S.R. competed vigorously to winmedal counts and point totals during the Olympic games.

Soviet troops were first deployed in Afghanistan in December 1979. In ad-dition to temporarily withdrawing the U.S. ambassador to Moscow, reducingthe number of passenger flights and cultural exchanges to the Soviet Union,and restricting grain and high-technology exports to the Soviet Union, onJanuary 23, 1980, President Jimmy Carter threatened a boycott of the XXIISummer Olympics in Moscow in 1980. Since the Soviets did not withdrawtheir troops by the mid-February deadline, President Carter announced aboycott of the Moscow Olympics on March 21, 1980. West Germany joinedthe boycott in May, as did Canada, Britain, Argentina, and Turkey. Is-lamic nations also decided to join the boycott, since Afghanistan was also aMuslim nation. Ultimately, a total of 65 countries refused to participate inthe games.3 Facing international embarrassment and economic and financiallosses, Soviet diplomats secretly urged West Germany and other countries toreverse their decisions to participate in the boycott. Despite these efforts, byJune 1980, 23 percent of available tickets remained unsold, at least 80,000

3Sixteen countries decided to allow athletes to participate under the Olympic flag ratherthan their national flag, including Britain, Ireland, France, Italy, Switzerland, and Den-mark.

5

Olympics-bound tourists had canceled their hotel reservations in the SovietUnion, and several Western sponsors, including Daimler-Benz, had droppedtheir sponsorship contracts.4

While the Olympic boycott is better known, another boycott immedi-ately preceded it. At the end of 1978, a group of American scientists createdthe Scientists for Sakharov, Orlov and Sharansky (SOS) Committee, whichwas designed to protest the detention and treatment of Soviet scientists whowere also political dissidents. By the second half of 1979, the Committeewas joined by 2,400 U.S. scientists. The list included Nobel prize winners,members of the National Academies of Science, and presidents of nationalscientific organizations. In 1980, after the exile of Andrei Sakharov, theCommittee expanded internationally, and roughly 8,000 scientists from 44countries joined the boycott. The SOS Committee committed to the suspen-sion of personal cooperation with the Soviet Union, including the transfer ofadvanced Western technology and cooperation with Soviet scientists, whichsupported the regime, until the three scientists were released, which onlyhappened in 1986. Members of the Committee also peacefully protested con-ferences that featured complying prominent Soviet scientists. Last but notleast, the SOS Committee lobbied the National Academies of Science in theUS to reduce the number of formal researchers’ exchanges with the Academyof Sciences of the USSR. As a result of the boycott – in sports, science, orpolitics – the number of short-term exchanges in science and technology fellfrom 296 in 1979 to 22 in 1980. The number of long-term exchanges declinedfrom 27 to 7.5

From Cook (2011), we know that Soviet inventors had strong incentivesto produce inventions with “international significance” that would ultimatelyimprove Soviet living standards. Was external inventive activity by Sovietinventors affected by the escalation of the Cold War?

2.1 Implications of the Boycotts

Relations between the Soviet Union and boycotting countries disintegrated.Most scholars studying the boycott note formal retaliation later through theSoviet boycott of the Los Angeles Olympics in 1984. However, this had animpact far beyond amateur sports. The U.S. State Department notes that

4Zhyrnov (2010).5Hoover Institution (1980); Ginsburgs and Slusser (1981); Ginsburgs (1987)

6

Figure 1: Patents by Soviet Residents in East Germany and US, 1975-1990,by Application Date

0100

200

300

Pate

nt count

1975h1 1980h1 1985h1 1990h1Half year

East Germany US

Cold War escalation (btw. red lines)

Source: East German patents - DEPATIS, German Patent and Trademark Office; USpatents - USPTO. Reliable data on East German patents starts in 1977. Half-yearlydata reported.

1979-80 marked the beginning of cooling relations between the Soviet Unionand the U.S.6

Bernard and Busse (2004) notwithstanding, little has been written abouteconomic factors and the Olympics. Neither the actions of the SOS Commit-tee and their consequences have properly been analyzed. Nonetheless, im-plications from such a disruption in technological spillovers have likely beenunderestimated. Unique new data from patents shed light on this. Despitelittle formal scientific cooperation between the U.S.S.R. and most Westerncountries during the Cold War, the significant technological flows via patentsobtained abroad by Soviet inventors between 1959 and 1991 were interrupted.In 1979-80, when the Cold War escalated, SOS Committee was created, andeventually the U.S. boycott of the Moscow Olympics was announced andmaterialized, there was a steep decline in the number of applications in theU.S. from Soviet inventors (Figure 1). Similar patterns with a small lag wereobserved in West Germany and Japan (Figure 2).

6U.S. Department of State, Office of the Historian (2012)

7

Figure 2: Patents of Soviet Origin in West Germany and Japan, 1970-1990,by Application Date

0200

400

600

pate

nt count

1970 1975 1980 1985 1990year

patent count Cold War escalation (btw. red lines)

a. West Germany

0100

200

300

400

pate

nt count

1970 1975 1980 1985 1990year

patent count Cold War escalation (btw. red lines)

b. Japan

Source: DEPATIS, German Patent and Trademark Office. Included are only patentswith Soviet Union priority numbers. Annual data reported.

What happened to the inventive effort that was externally oriented? Didit just come to a sudden halt? Figure 1 suggests otherwise. It shows thatflows from the Soviet Union to East Germany were orthogonal to flows tothe U.S. following the boycott. In 1978, there were 223 applications for EastGerman patents by Soviet scientists, compared to 450 in the U.S. In 1979the number of applications in East Germany was even smaller - 193. Butafter the boycott, the annual number of Soviet applications in East Germanyexploded, reaching its maximum of 561 in 1981, and then stabilizing at about300-350 applications per year until 1987.

Figures 3 and 4 demonstrate that the orthogonal dynamics of patentingby the Soviet residents in the U.S. and East Germany is not a simple co-incidence. A significant share of Soviet scientists did seem to shift in 1980from applying for a patent in the U.S. to East Germany. Figure 3 (solidline) shows applications for an East German patent by Soviet inventors whofiled at least one patent application in the U.S. in 1972-78. In 1980 thenumber of such applications increased by more than 100, which could ex-plain about a half the decline in the U.S. in 1979-80 and about a third theincrease in East Germany in 1981. Figure 4 demonstrates a similar patternfrom the other side: it shows U.S. patent applications (by year granted) bythose Soviet inventors, who filed for at least one patent in East Germanyin 1977-90. Again there is a sharp decline in the early 1980s.7 The two

7We collect data for these particular series from Google Scholar. The patent data is

8

Figure 3: Patents by Soviet Residents in East Germany, 1977-1990: InventorLevel Data

0200

400

600

050

100

150

200

250

1975 1980 1985 1990Year

applications, which include inv. with US patents from 1972−78

applications, which also include new inventors

all applications (rhs)

Source:East German patents - DEPATIS, German Patent and Trademark Office; USpatents - USPTO. Annual data. Inventors in DEPATIS and USPTO databases arematched by the surname and the first letter of the name. Data on East German patentsprior to 1977 are unavailable in electronic format.

figures do suggest that there was an actual flow of patent applications fromthe U.S. to East Germany in 1979-81 when the Cold War escalated and afterthe boycott. Another noticeable pattern on Figure 3 is that nearly all EastGerman patent applications by the former U.S. applicants were co-authoredwith “new” inventors, i.e. those, who did not apply for a patent in the U.S.in 1972-78. This suggests a significant increase in scientific cooperation as aresult of redirected patenting flows.

Below, we test whether the Cold War escalation in 1979-80 and the result-ing changes in the international knowledge flows affected domestic patentingby the East German inventors.

collected only by year when patent was granted. The patent applications should follow asimilar pattern, but shifted by few years earlier.

9

Figure 4: US Patents Granted to Soviet Residents with East German Patents,1970-1990, by Year Granted

0200

400

600

Pate

nt count

1970 1975 1980 1985 1990 1995Year

US patents Boycott announcement

Source: Google Scholar. Yearly data. Sample of scientists - Soviet residents who wheregranted at least one patent (including co-inventors) in East Germany in 1977-1990.

3 Data

Newly available data on East German patents are combined with U.S. patentdata to answer the questions above.

3.1 US Patent Data

Data on patents granted to Soviet citizens abroad were collected from theUSPTO. We know from Cook (2011) that Soviet inventors often bypassedtheir domestic patent office. Therefore, most patents obtained by Sovietinventors abroad would not refer to a Soviet patent or patent-like document.8

Hence, using the latter as a search criterion would result in an undercount ofpatents obtained by Soviet inventors abroad. We executed patent searchesusing further criteria – country of inventor and of assignee (applicant). Eachpatent record contains first initial or first name, middle initial or middle name

8Patents were very rarely awarded to Soviet inventors in the Soviet Union. Morecommon were inventor’s certificates, which offered inventors recognition of authorship butno control rights.

10

(patronymic), and surname of inventor(s); patent or IC number; applicationdate; grant date; USPTO class at the 6-digit level; applicant (assignee); anda brief description or title of the patent. There are 5,904 U.S. patents appliedfor by Soviet residents between 1963 and 1991 (and granted between 1971and 1997) in the data set.

3.2 East German Patent Data

We collected East German patent data from the German Patent and Trade-mark Office’s (GPTO) patent information system DEPATIS. Our data setcontains 160,814 patents records spanning 1977 to 1991.9 Each record con-tains first name, middle initial, and surname of inventor(s); patent number;application and grant dates; International Patent Classification (IPC) sec-tion, subclass and group; applicant; title of the patent.

In a separate exercise we use DEPATIS to find all East German patentsby Soviet residents. We search by country of priority (where it is first regis-tered), of inventor, and of assignee. In addition, we execute a manual checkof all records to identify Soviet and non-Soviet inventors on a patent due toproblems of classification in the data.10 Similarly to the U.S. data, using onlypatents with Soviet Union as the country of priority would result in under-count of patents by Soviet residents, as they ofter bypassed their domesticpatent office.

The system of intellectual property protection in East Germany waslargely similar to that of West Germany. Both countries retained their mem-bership in the Paris Convention for the Protection of Industrial Property,which Germany signed in 1903. The life of a patent in both countries was20 years, and in both countries the “first-to-file” rule was implemented. Themajor difference was that East Germany provided less protection for inven-tors. As in the Soviet Union, there were two types of patents in the GDR:the “exclusion patent” (Ausschliessungspatent) and the “economic patent”(Wirtschaftspatent). While the former type of patent granted an inventorthe exclusive right to use and license his or her invention, the latter typewas subject to compulsory licensing for a nominal fee. In 1991, 80 percentof active East German patents were of the “economic” type. Like Soviet in-

9We have data prior to 1977, but for most of the electronic records - 97% according toDEPATIS - from this time period there is no information about applicants and inventors,including their country of residence, which is necessary information for our research

10See Appendix for details of manual matching process.

11

ventor’s certificates, East German patents do not have backwards citations,since references to “prior art” were not required.

In total, our dataset contains 3,612 East German patents granted to So-viet residents. According to the International Patent Classification, 29 per-cent of patents are in performing operations, transportation; 22 percent arein metallurgy and chemistry; and 15 percent are in physics.

3.3 Other Data

We collect additional data for our auxilliary exercises. First, we use GoogleScholar to match our inventor-level patent data with the inventor-level dataon publications. We use the latter to test whether increased patenting alsoboosts other means of invention. See Section 5.5 for details. Second, we useDEPATIS to collect panel country-level data on patents with Soviet Unionas the country of priority. We find 20 countries, where Soviet inventors wereactively patenting in 1970-90. We use these data to check if our conjectureabout the link between international knowledge flows and the Cold Warescalation holds at the cross-country level. See Section 6.1 for details.

4 Empirical strategy

The main question we ask is whether changes in patenting by foreign appli-cants have any effect on changes in patenting patterns by domestic applicants.To answer this question we use a natural experiment - the escalation of theCold War and the consequent politically-driven change in patenting flows ofSoviet origin from the U.S. to East Germany. Variation in these flows acrosspatent subclasses allows to pin down the causal effect. Specifically, we wantto estimate the following regression:

< # GDR-origin patents in GDR >it= α+

+ β1 < # SU -origin patents in GDR >it +

+ β2 < # SU -origin patents in GDR and US >it +It + FEi + εit. (1)

< # X-origin patents in Y >it is the number of patents by inventors fromcountry X issued in country Y in patent subclass i over period t, by appli-cation date. GDR stands for East Germany. We explore the spillovers fromSoviet to East German inventors, both groups patenting in East Germany.

12

We also control for the general level of innovation by Soviet inventors, whichis proxied by the number of Soviet applications both in East Germany andthe U.S. FEi are subclass fixed effects, and It are period dummies.

Our interest is in β1. Estimating β1 by OLS (i.e., no FEi) or by fixedeffects would likely produce inconsistent results. First, < # GDR-originpatents in GDR > and < # SU -origin xvffpatents in GDR > may bedriven by common omitted technology- and innovation-specific factors. Forinstance, a major technology breakthrough in a class i would change bothdependent and independent variables in our regression, in particular if thisclass is an area of specialization of both Soviet and East German inventors.11

Second, the bias in β1 may appear due to a measurement error, which is verylikely to be present in our data.

The escalation of the Cold War in 1979-80 provides a source of variation inpatenting that is orthogonal to the invention process. During the escalation,applications for patents by Soviet residents shifted from countries in theWestern Bloc, in particular the U.S., to political allies like East Germany.As a result, technological fields in which Soviet inventors frequently patentedbefore the escalation may have experienced a concomitantly large inflow ofpatent applications by Soviet inventors in the GDR. As a result, the boycott-induced patent outflows from the U.S. could be used as an instrumentalvariable for < # SU -origin patents in GDR >it - East German patents ofSoviet origin.

To measure the boycott-induced patent outflows from the U.S. we proceedin two steps. In the first step, we use the pre-escalation (1972-78) data on theU.S. patenting by Soviet inventors to estimate “business-as-usual” projectedpatent counts in the U.S. in 1979-91. Our projection is a simple average ina subclass, or technological category, over the pre-escalation period, 1972 -78.12 In the second step, we define our IV as follows:

IV =

{0 in 1977-78

< # pats >predUS − < # pats >act

US in 1979-91,(2)

11Another example, that would bias β1, is an agreement on scientific cooperation be-tween the Soviet Union and East Germany, which would drive up patenting by both Sovietsand East Germans.

12We also use an alternative procedure to estimate “business-as-usual” patent counts.Instead of taking averages we fit an AR(2) process to each patent subclass. We find thatour main conclusions are robust to this alternative estimation procedure. The estimationresults are reported in Appendix, Table 6.

13

or the difference between predicted (< # pats >predUS ) and actual (< # pats >act

US)patent counts in the U.S. For example, suppose the average annual numberof patent applications in class X in the U.S. in 1972-78 is 5. Suppose, in1979-91 it is on average 2. Then our IV is 0 in 1977-78, and 3 in 1979-91.13

Defined like in (2) our variable is likely to be a valid IV. We test below inthe first stage whether it explains well the inflow of East German patents ofSoviet origin in 1979-91 (the answer is yes). Exclusion restriction is also likelyto hold. As we use only the U.S. data in defining the IV, the variable is likelyuncorrelated with any SU-GDR-specific trends in patent classes. Also ourIV is unlikely to be correlated with the measurement error in East Germanpatent counts. 14

Figure 5 demonstrates the workings of our IV. The left panel shows thedifference between patent counts by Soviet scientists in East Germany inthe subclasses, where the IV is larger than the sample median, and the restof the subclasses. The affected subclasses received higher inflows of patentapplications in 1979-91, and in 1979-83 the difference between affected andunaffected subclasses was substantially larger than during the pre-escalationperiod (1977-78) - a pattern which is unlikely to be the outcome of preexistingtrends.

The right panel shows the patent counts by East German scientists overthe same grouping of subclasses. In 1977-84 neither of the groups showsdifferent behavior before and during the Cold-War escalation. But in 1985 thedifference between the affected and unaffected subclasses sharply increasedand remained large until 1990. This suggests that the influx of Soviet patentsin the beginning of the 1980’s affected domestic invention in East Germanyhalf a decade later.

In specification (1) we assume two time periods. The pre-escalation periodextends from 1977 to 1978. The lower boundary reflects the availability ofthe data on East German patents. The upper boundary corresponds to theperiod when the SOS Committee began its activity, which also conformswith the pattern in Figure 1. The escalation and post-escalation periods aremerged in one and extend from 1979 to 1991.15 All data are averaged overboth periods, based on half-yearly frequency.

13We only need to define our IV for 1977-91, not earlier, since our East German datastarts in 1977

14The formal justification for our IV is provided in Appendix in a stylized model of EastGerman inventive activity.

15We also report regression results for other periods.

14

Figure 5: IV Estimation: First-Stage and Reduced-Form Models0

.1.2

.3.4

.5S

ovie

t in

ve

nto

rs p

ate

nts

in

GD

R,

ave

rag

e

1975h1 1980h1 1985h1 1990h1Half year

affected subclasses unaffected subclasses

First Stage

05

10

15

GD

R in

ve

nto

rs p

ate

nts

in

GD

R,

ave

rag

e

1975h1 1980h1 1985h1 1990h1Half year

affected subclasses unaffected subclasses

Reduced Form

Sources: East German patents - DEPATIS; U.S. patents - USPTO.Notes:“Affected subclasses” - if the IV is lower then the sample median; “unaffectedsubclasses” otherwise. Patent counts are by application date.

The patents in our specification are divided by the IPC four-symbol sub-classes. To compare with the U.S. patents by Soviet residents we do USPTO-to-IPC concordance using the Johnson and Evenson Patent Set.16 The con-cordance is described in detail in the Appendix. Overall, there are 644 IPCfour-symbol subclasses. So essentially our final dataset is a two-period panelof 644 patent subclasses.

In addition to technological spillovers to East German inventors, we assessthe benefits from patenting in East Germany on Soviet inventors themselvesand on international cooperation. We use the same specification (1) and theIV defined by (2), but the dependent variables are different. First, to measurepossible international spillovers we use number of patents by internationalteams of inventors, where an international team consists of at least one Sovietand one non-Soviet resident. Second, we look at the average number of Sovietinventors on patents in East Germany.

The statistical summary of dependent, independent and instrumentalvariables is given in Table 1. In regressions, all variables are in logarithms,e.g., xnew = log(1 + x).

16Accessed from http://acad.coloradocollege.edu/jeps/index.php on July 24, 2011

15

Table 1: Summary Statistics

Variable Mean St.dev Mean St.dev1977-78 1979-91

# GDR-origin patents in GDR 4.92 10.32 5.81 11.82# SU-origin patents in GDR 0.14 0.63 0.13 0.48# SU-origin patents in GDR and US 0.43 1.13 0.4 1.09IV (see Equation 2) 0 0 0.06 0.68# intl. teams on SU-origin patents in GDR 0.03 0.25 0.04 0.19# inventors on SU-origin patents in GDR 0.74 3.92 0.67 2.89# IPC four-symbol subclasses 644 644

Sources: GDR patents - DEPATIS; U.S. patents - USPTO. Inventors’ team is considered international ifit contains at least one Soviet (SU) resident and at least one non-Soviet resident. All variables areaveraged by half-year and by IPC subclasses.

5 Results

The main estimation results are reported in Table 2.

5.1 Did Soviet inventors shift activity Eastward afterthe boycotts?

Column (8) in Table 2 shows the first stage of the IV estimation. The IVwe use is fairly strong. The difference between the actual number of Sovietapplications for the U.S. patents in 1979-91 and its predicted value basedon 1972-78 is a very good predictor of how many patent applications Sovietscientists submitted in East Germany in 1979-1991. The point estimate forthe IV is 0.41. That is, a lack of patenting by Soviet scientists in the U.S. ismatched by an increase in patenting in East Germany by almost a half.

5.2 Did the boycott-induced Soviet presence in EastGermany affect patterns of invention in East Ger-many?

Columns (1-3) of Table 2 report results for technological spillovers to EastGerman inventors. Estimates are reported for OLS, FE, and IV estimation.We find a positive effect of the influx of Soviet patents on patenting by East

16

Table 2: IV estimation: Spillovers to East German inventors1979-91 1979 1979-82 1979-85 1979-88 First stage

(1) (2) (3) (4) (5) (6) (7) (8)OLS FE IV IV IV IV IV FE

# SU-origin 0.57∗∗∗ 0.72∗∗∗ 0.89∗∗∗ 0.01 0.40∗∗∗ 0.67∗∗∗ 0.88∗∗∗

patents in GDR (0.19) (0.12) (0.18) (0.18) (0.15) (0.14) (0.16)# SU-origin 1.67∗∗∗ -0.27∗∗∗ -0.36∗∗∗ 0.04 -0.12 -0.26∗∗ -0.35∗∗∗ 0.87∗∗∗

patents in US and GDR (0.12) (0.09) (0.12) (0.13) (0.13) (0.11) (0.12) (0.02)=1 if 1977-78 -0.40∗∗∗ -0.29∗∗∗ -0.28∗∗∗ -0.08∗∗∗ -0.19∗∗∗ -0.28∗∗∗ -0.32∗∗∗ -0.03∗∗∗

(0.04) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.00)IV 0.44∗∗∗

(0.02)Constant 1.12∗∗∗ 1.50∗∗∗ 1.50∗∗∗ 1.28∗∗∗ 1.39∗∗∗ 1.49∗∗∗ 1.54∗∗∗ -0.09∗∗∗

(0.03) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.01)Observations 1288 1288 1288 1288 1288 1288 1288 1288R-squared 0.51 0.40 0.68R-squared overall 0.11 0.11 0.11 0.12 0.12 0.14 0.78

Note Dependent variable: columns (1)-(7) - # GDR-origin patents in GDR, column (8) - #SU-origin patents in GDR. ∗ -significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. See definitions of variables in Table 1.Robust standard errors are reported in parentheses. All variables are in logarithms: xnew = log(1 + x)

German inventors in the same subclasses in 1979-91.17 The magnitudes of ourIV estimates are not trivial. In case of IV, for subclasses with a large numberof applications, the elasticity of East German applications with respect toSoviet ones is about 0.45. On average, we find that a 10-percent increase inSoviet patent activity in East Germany increases domestic inventions by 2.5percent.18

The positive spillover from Soviet to East German scientists does notmaterialize immediately. Columns (4-7) of Table 2 show its evolution overtime. In the first year after the escalation of the Cold War, the effect isnegative, but not statistically different from zero. This is consistent withthe finding in Borjas and Doran (2012) that a sudden influx of competition(for research ideas, funding) may reduce innovation. With time the spilloverbecomes positive, and in 1985 it is statistically significant at 10% and remainsso until 1991.19

We also find that the boycott and the Cold War escalation had a long-lasting effect on Soviet inventors’ patenting in East Germany. First, anincreased number of Soviet patents in East Germany led to significantly morepatents by international teams in the same classes (Table 3). Second, more

17OLS and FE estimates are higher than the ones from IV, which is consistent with themodel described in Section 4 and in Appendix.

18Given the functional form that we use in the regressions, the exact effect of x on y is:∆y = ∆x(1 + y)/(1 + x).

19This is consistent with the view in the literature that the positive innovation spilloverstake time to materialize. For example, Moser and Voena (2012) shows that the significantbenefits from compulsory licensing appeared only seven years after it was introduced.

17

Table 3: IV estimation: Spillovers to Soviet inventors - Number of Interna-tional Teams

1979-91 1979 1979-82 1979-85 1979-88(1) (2) (3) (4) (5) (6) (7)

OLS FE IV IV IV IV IV# SU-origin 0.42∗∗∗ 0.34∗∗∗ 0.32∗∗∗ 0.43∗∗∗ 0.26∗∗∗ 0.27∗∗∗ 0.31∗∗∗

patents in GDR (0.02) (0.02) (0.04) (0.05) (0.04) (0.03) (0.03)# SU-origin -0.05∗∗∗ -0.04∗∗ -0.03 -0.09∗∗∗ -0.01 -0.01 -0.03patents in US and GDR (0.01) (0.02) (0.02) (0.03) (0.03) (0.03) (0.02)=1 if 1977-78 0.01∗ 0.00 0.00 -0.00 0.01 0.00 0.00

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)Constant -0.01∗∗ 0.00 0.00 0.01∗∗ -0.00 -0.00 0.00

(0.00) (0.00) (0.00) (0.01) (0.01) (0.00) (0.00)Observations 1288 1288 1288 1288 1288 1288 1288R-squared 0.51 0.32R-squared overall 0.51 0.51 0.45 0.47 0.47 0.51

Note ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. See definitions of variables inTable 1. Robust standard errors are reported in parentheses. All variables are in logarithms: xnew = log(1 + x)

inventors are recorded on patents in the affected classes (Table 4). Unlikespillovers to German inventors, spillovers to Soviet inventors are positiveand significant regardless of which escalation/post-escalation period we use(Columns 4-7 in the respective tables). In fact, the effect of more Sovietpatent applications on number of inventors on a patent team and number ofinternational teams is higher in the beginning of the escalation period (seeColumn 4 vs. Column 3 in Table 3). This suggests that, especially in thebeginning of the escalation, Soviet inventors found scientific contacts, in EastGermany and among themselves, that would help them “switch” from theU.S. to East German patent office.

Below we present a case study to illustrate how boycotts led to a chain ofevents that would force Soviet scientists to leave the U.S. Patent Office andthat would in turn increase domestic patenting in other countries.

5.3 A Case Study from Molecular Macrocompounds

Increased networking and international cooperation in the beginning of theescalation period could serve as a channel for increased domestic innovationin East Germany five years later. A case study of East German chemistHeinz Raubach demonstrates how this channel worked.

In 1980, inventors from the Soviet Union Nina Bekasova and Vasily Ko-rshak renewed their cooperation with Raubach with whom they patentedonce before.20 Together with other East German scientists they filed for a

20Raubach and Korshak cooperated before, but their last joint application for a patent

18

Table 4: IV estimation: Spillovers to Soviet inventors - Number of Inventorson a Patent

1979-91 1979 1979-82 1979-85 1979-88(1) (2) (3) (4) (5) (6) (7)

OLS FE IV IV IV IV IV# SU-origin 2.32∗∗∗ 2.40∗∗∗ 2.58∗∗∗ 2.62∗∗∗ 2.13∗∗∗ 2.21∗∗∗ 2.35∗∗∗

patents in GDR (0.05) (0.06) (0.09) (0.11) (0.10) (0.08) (0.08)# SU-origin 0.10∗∗∗ -0.03 -0.13∗∗ -0.10 0.12 0.10 0.00patents in US and GDR (0.03) (0.05) (0.06) (0.08) (0.09) (0.07) (0.06)=1 if 1977-78 -0.05∗∗∗ -0.04∗∗∗ -0.03∗∗ 0.02∗ -0.03∗∗ -0.04∗∗∗ -0.04∗∗∗

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Constant 0.09∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.04∗∗∗ 0.08∗∗∗ 0.08∗∗∗ 0.10∗∗∗

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Observations 1288 1288 1288 1288 1288 1288 1288R-squared 0.91 0.83R-squared overall 0.91 0.90 0.90 0.90 0.91 0.91

Note ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. See definitions of variables inTable 1. Robust standard errors are reported in parentheses. All variables are in logarithms: xnew = log(1 + x)

patent in one of the subclasses of molecular macrocompounds (C08G). Afterthis patent application Raubach continued active work in subclass C08G.Together with his East German colleagues he successfully applied for (andobtained) six patents in this subclass between 1981 and 1986, and then fornine more between 1987 and 1989. The Raubach’s example is suggestivethat it was more than general interest in macrocompounds that led to aflurry of patent activity between him and his partners after 1980. The tim-ing is consistent with incremental additions to knowledge being augmentedby collaboration with a Soviet inventor, which is also consistent with ourempirical analysis showing that the inflow of Soviet inventors in the early1980’s led to increased domestic inventive activity in East Germany.

5.4 Other Technological Spillovers

The direct effect of increased cooperation between Soviet and East Germanscientists, while easy to trace empirically, is not the only channel that con-nects foreign patents to domestic invention. Networks and research ideascould spread beyond initial ties. For example, Raubach worked with manyother East German chemists in the 1980’s, who could then apply for a patentin subclass C08G on their own or as a part of other research teams. Informa-tion about the Soviet inventions became “closer” to East German science inthe 1980’s: it was more wide spread in East Germany, and it was translated

before 1980 was in 1976. No applications in subclass C08G were made by Raubach between1976 and 1980.

19

into German. Finally, patenting in one subclass could lead to more inventiveactivity in other subclasses. For example, at the end of the 1980s, Raubachpatented extensively in subclass C08L (also in the group of molecular macro-compounds). In this sense, we likely underestimate the true effect of foreignpatenting on domestic inventive activity, and our estimates should be inter-preted as a lower bound on technological spillovers in domestic invention.

5.5 Linkages with Alternative Means of Invention

Is patenting by the Soviet inventors in East Germany associated with othermeans of invention? We collect Google Scholar data on publications by theseinventors in Soviet and foreign journals to check whether inventors with morepatents in GDR tend to publish more.21 We also check whether being a coin-ventor on a GDR patent is associated with being a coauthor on a publication,and whether this relationship also holds for international teams of inventors.In light of the results in Table 3, the latter would be direct evidence ofspillovers from foreign to domestic invention not only though patents butalso through publications. 22

The results are presented in Table 5. Overall, we found 24,362 publi-cations (papers, chapters, or books), dated 1976 to 1995, by 10,805 Sovietinventors, who filed at least one application for a patent in GDR between1976 and 1992. Of those, 1,427 publications are coauthored by the pairs ofinventors, who are also coinventors on a patent (39,489 such pairs). Twohundred sixty-four publications are by the international (Soviet plus non-Soviet inventor) pairs of inventors (9,044 pairs). Column 1 in Table 5 showsthat inventors with more patents also tend to publish more. Columns 2-4demonstrate that coinventing on a patent is positively associated with thecoauthorship on a publication. This holds both for Soviet and internationalpairs of inventors, although for Soviet inventors (Column 3) the effect isstronger. These results are robust to the functional specification of regres-sions.23

21Note that in general the positive relationship between patents and publications doesnot necessarily hold. It may hold in case of Soviet inventors, as majority of them workedin research institutes and were driven by reputation motives.

22At this point we test only the most basic preliminary specifications: no time dynamics,no division by patent classes or types of publication, and no publications by East Germaninventors (except those in coauthorship with Soviet inventors). These extensions are left

20

Table 5: SU Patent Applications in GDR and Number of Publications# publ. by inventor, 1976-92 # publ. by pair of inventors, 1976-92

(1) (2) (3) (4)SU inventors SU+foreign SU foreign

# GDR patents by 0.38∗∗∗

inventor, 1976-92 (0.02)# GDR patents by 0.02∗∗∗ 0.02∗∗∗ 0.01∗∗

pair of inventors, 1976-92 (0.00) (0.00) (0.00)Constant 0.05∗∗ -0.00 -0.00 -0.00

(0.02) (0.00) (0.00) (0.00)Observations 10805 38761 29717 9044R-squared 0.02 0.00 0.00 0.00

Note: ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. Sample: column (1) - allinventors with at least one GDR patent application in 1976-92; columns (2)-(4) - all pairs of inventors, which co-authorat least one GDR patent application in 1976-92. Dependent variable - publications by the inventor (column 1) or by thepair of inventors (columns 2-4) in 1976-92, as registered by the Google Scholar. Inventors are identified by the last nameand the first letter of the first name. Columns (1) and (3) includes only Soviet inventors; column (4) includes only pairsof inventors with at least one foreign (non-Soviet) inventor; column (2) includes both Soviet, foreign and mixed pairs ofinventors. All variables are in logarithms: xnew = log(1 + x). Data on publications: Google Scholar (accessed in August2011).

6 Robustness checks

We subject our results to a number of robustness checks. The results of thesetests are reported in the Appendix.

First, we use an alternative IV definition. Instead of long-run averages,now our “business-as-usual” patent applications are estimated by an AR(2)process in each subclass.24 The results are reported in Table 6. Our mainresults hold.

Second, we check if our results hold if we use alternative functional spec-ifications. Table 7 reports the results when patent counts are included inlevels as opposed to logarithms in the main specification. Table 8 reportsresults of Poisson and negative binomial regressions. Our main conclusionshold in all specifications.

Third, we rerun our main specification counting only patents withoutpatent families, i.e. those patents which were filed for only in East Ger-many. This is to address a potential concern that the patent applications bySoviet inventors could be driven by a commercial interest rather than rep-utational motives. The results are reported in Table 9. They still suggestthat (a) patent applications by Soviet inventors shifted from the U.S. to EastGermany in the end of the 1970s, and (b) this shift has positively affected do-

for further research.23See Table 11. Like patentees in other countries, most inventors do not also have

publications. In our sample, 80 percent of inventors do not have any publications.24See definition in Section 4

21

mestic patenting in East Germany, as well as number of international teamsfiling for a patent. Compared to the main specification, the coefficients inTable 9 are smaller, and in some specifications they lose statistical signifi-cance because of that. A possible explanation is that the patents with nopatent families are likely to be less important or influential, and hence lesslikely to spur further research and invention.

6.1 Were Soviet Patent Flows in 1979-80 Driven Bythe Escalation of the Cold War?

We use cross-country data on patent applications with Soviet priority num-bers to demonstrate that the SU patent flows reversal in the U.S. and EastGermany in 1979-90 was not a coincidence, but was indeed driven by theescalation of the Cold War. We check how patent applications with SU pri-ority numbers in a given country responded in the beginning of the 1980s tothe country’s adherence to the Moscow Olympics boycott. While we do notwant to overemphasize the importance of the boycott itself, the adherencelikely serves a good proxy of allies and opponents of the Soviet Union duringthe Cold War escalation.25 The results are presented in Table 10. As esca-lation of the Cold War developed gradually in the end of the 1970s, we usethree definitions of pre- and post-escalation periods. We also average oversufficiently long periods of time (5 or 10 years) to reduce the role of outliers.In all our specifications, both strong and weak (allowing athletes to competeunder the UN flag) adherence to the boycott is associated with the fall inpatent grants with SU priority numbers in post- vs. pre-escalation periods.

7 Conclusion

We use newly available data from the U.S. and East German patent offices toexamine the effects of foreign participation in domestic knowledge creation.We exploit the abrupt, unanticipated rupture of East-West scientific ties re-sulting from the Cold War escalation in 1979-80, which culminated in aninternational boycott of the Moscow Summer Olympics in 1980. Using data

25Neither is patent applications with SU priority numbers the best measure of Sovietpatent applications. For example, in East Germany and the U.S. it would almost entirelyoverlook the patent flow reversal. Nevertheless, it is the only proxy at the cross-countrylevel which can be obtained at a reasonable cost of time and effort

22

on patents obtained by Soviet inventors in the U.S. and East Germany, weshow statistically significant changes in the level and direction of domesticinventive activity following increased participation by Soviet scientists. Tothe extent that inventions were used by state enterprises and that such usesraised living standards, e.g., by increasing employment or increasing vari-eties of consumption goods, such an appreciable change in innovative activitycould result in higher living standards, as they would in Western economies.More broadly, like Borjas and Doran (2012) and Moser and Voena (2012),our research provides further evidence of changes in domestic knowledge pro-duction due to foreign participation, albeit from different sources and typesof participation.

At least one policy implication of this research is clear. Policymakersinterested in encouraging domestic scientific discovery may want to encourageforeign participation in domestic scientific activities, as well. This may comein the form of easing regulation or reducing other barriers to entry for foreignparticipants. More broadly, Cook (2011) suggests that broader participationin invention and innovation could raise innovative output.

On the way to demonstrating these effects, we also present for the firsttime a systematic analysis of the economic and broader effects of the boycottsassociated with intensification of the Cold War. Cook (2011) showed thatthese international knowledge flows existed outside the Soviet Union. Thisresearch extends this by answering the question of what happened to theseflows when sanctions were imposed.

23

A Appendix

A.1 Search for East German Patents by Soviet Inven-tors

We take all records, which satisfy at least one of the following criteria:

• A patent has SU priority numbers;

• At least one inventor on a patent has a patronymic letter in his/hername (indicating he/she is from Soviet Union);

• At least one inventor on a patent indicates SU as his/her country oforigin.

We did manual check of all records to identify Soviet/non-Soviet inventorson a patent. This is due to some problems with the original data: countryof origin is necessarily indicated only starting from 1983; there are obviousflaws with the country of origin - surely German inventors are sometimesrecorded as SU; some Soviet inventors are recorded without patronymic.

We do the following procedure to identify patents by Soviet inventorsfrom all records that we find. First, we fill in the list of Soviet names. Theseare the names of those inventors who are marked as Soviet in DEPATISdatabase (only after 1983). Second, we clean the list from obviously Germannames of inventors, which were apparently marked as Soviet by mistake, -such names as Matthias, Gunter, or Sylvia. We also clean the list from“suspicious” names like Leon or Sofia. Third, we run through all recordsand mark an inventor Soviet if she/he has a Soviet name from our list. Therest of inventors are marked as international. Fourth, we manually checkall records with at least one international inventor, which may also includeSoviet inventors with “suspicious” names.

As a result, we should have all Soviet inventors, and all patents with SUpriority numbers. The only omission is Soviet inventors recorded withoutpatronymic, and using non-SU priority numbers. This is unlikely to be a biggroup but it may introduce downward bias in the number of internationalteams of inventors.

24

A.2 USPTO-to-IPC Concordance

Using the Wellesley Technology Concordance (WTC) probability tables, wetransform all patent classes into IPC (International Patent Classification)notation. We do it for two reasons. First, the datasets of all East Germanpatents and East German patents by Soviet residents classify patents usingonly IPC notation. Therefore, we only need to transform the dataset of USpatents by Soviet residents. Second, the WTC probability tables only allowtransformation from USPTO subclasses (six-digit codes) to IPC subclasses(four-symbol codes), therefore reverse concordance, which is possible to con-struct, would be less precise.

USPTO-to-IPC concordance goes the following way. Probability of eachUSPTO subclass falling into domain of IPC subclass X are estimated basedon patent families assignments by WIPO and USPTO in 1976-1994. Forexample, if of 100 patents assigned to subclass 100/001 by USPTO 50 wereassigned to, say, A01B and 50 were assigned to A01D, then correspondingprobabilities for A01B and A01D are 0.5 and 0.5. The probabilities for therest of IPC subclasses are zero. Accordingly, in our dataset a patent assignedUSPTO subclass 100/001 counts as 0.5 patents with A01B IPC subclass,and 0.5 patents with A01D IPC subclass. According to WTC, one USPTOsubclass concords with on average 2.4 IPC subclasses, although the numbervaries from 1 to 91. The average maximal probability of a USPTO subclassbeing concorded with a IPC subclass is 0.85, it varies from 0.11 to 1. Lessthan 50 percent of USPTO subclasses concord with IPC subclass exactly (i.e.maximal probability is one), while for 25 percent of USPTO subclasses themaximal probability is less than 0.5.

A.3 A Brief Guide to IPC Notation

IPC assigns each patent a section, class, subclass, and group. Section is thehighest level of classification’s hierarchy. There are eight of them: A - Hu-man Necessities; B - Performing Operations, Transporting; C - Chemistry,Metallurgy; D - Textiles, Paper; E - Fixed Constructions; F - MechanicalEngineering, Lighting, Heating, Weapons, Blasting; G - Physics; H - Elec-tricity. Sections are divided into classes, which are represented by two digits,and then classes are divided into subclasses, which are denoted by a Latinletter. Subclasses are further divided into groups, however we are only inter-ested in subclasses given the difficulties with USPTO-to-IPC concordance.

25

Figure 6: Example of IPC Classification Symbol

Note Copied from IPC Guide(http : //www.wipo.int/classifications/ipc/en/guide/guide ipc.pdf) - retrieved onMarch 24, 2012.

An example of IPC classification symbol is presented on Figure 6.Our merged dataset of East German (all) and US (only by Soviet resi-

dents) patents contains representatives of all eight IPC sections, 256 classes,and 1110 subclasses.

A.4 “Scientists for SOS” Logo

Figure 7: “Scientists for SOS” Logo

Source: Hoover Institution, SOS Archives.

26

A.5 A stylized model of East German domestic inno-vation

The following small stylized model of East German domestic innovation ex-plains why our IVs are likely to be valid in the specification (1). Our as-sumption is that the number of East German patent applications by domes-tic scientists in subclass i in period t, < # DD pats in DD >it depends onthe following:

< # DD pats in DD >it= α∗ + β1 < # SU pats in DD >it +

+ β∗2 < World >it +β∗

3 < DD − SU >it +It + FEi + ξit, (3)

where < World >it are world-wide factors or shocks to innovation in sub-class i, < DD− SU >it are factors related to bilateral scientific cooperationbetween East Germany and a country in focus, in our case the Soviet Union,ξit includes factors that are not related to < # SU pats in DD >, e.g. EastGerman domestic R& D policy. Let sit be the share of the Soviet patentapplications in East Germany in the total number of Soviet applications inthe U.S. and East Germany. Then

< # SU pats in DD >it= sit∗ < # SU pats in DD and US >it . (4)

The regression error εit in the specification (1) contains < World >it

and < DD − SU >it, and our concern is that < # SU pats in DD >it iscorrelated with these two variables, which makes OLS or FE estimator of β1biased.

Controlling for < # SU pats in DD and US >it, correlation between< # SU pats in DD >it and < World >it is likely to be zero:

E(< # SU pats in DD >it ∗ < World >it | < # SU pats in DD and US >it) =

=< # SU pats in DD and US >it ∗∗ E(sit∗ < World >it | < # SU pats in DD and US >it) = 0, (5)

since the worldwide innovation shocks affect the total innovation effort in theSoviet Union, but not how much of it is patented in a given country.

The situation with < DD − SU >it is different, as scientific cooperationbetween the Soviet Union and East Germany can affect not only the totalinnovation effort by Soviet scientists, < # SU pats in DD and US >it, but

27

also the share of inventions that are patented in East Germany, because theinventions resulting from the cooperation are more likely to be patented inEast Germany than in the U.S.:

E(sit∗ < DD − SU >it | < # SU pats in DD and US >it) 6= 0. (6)

The proposed instrumental variables help resolve the situation. First, ifSoviet scientists do shift their patent applications from the U.S. to East Ger-many due to the boycotts, then the difference between actual and “business-as-usual” U.S. patent applications by Soviet scientists should affect patentapplications in East Germany:

E(< # SU pats in DD >it ∗IVj| < # SU pats in DD and US >it) 6= 0, j = 1, 2.(7)

We test this assumption formally. Second, controlling for the total numberof Soviet applications in the U.S. and East Germany, our IVs should notbe correlated with εit, i.e. with < World >it or < DD − SU >it. Theargument for the correlation with the < World >it is the same as in (5).< DD − SU >it could be correlated with the patent applications by Sovietscientists in the U.S. but only if the DD-SU cooperation involved dimensions,which are not related to the innovation process, e.g. reduction of barriersto patent in East Germany or boycott-related issues. These factors affectdomestic innovation in East Germany only through the innovation effort bySoviet scientists.

28

A.6 Alternative Specifications

Table 6: Spillovers to East German inventors: Alternative IV1979-91 1979 1979-82 1979-85 1979-88 First stage


# SU-origin 0.57∗∗∗ 0.72∗∗∗ 0.79∗∗∗ 0.01 0.39∗∗ 0.64∗∗∗ 0.81∗∗∗

patents in GDR (0.19) (0.12) (0.23) (0.22) (0.17) (0.15) (0.18)# SU-origin 1.67∗∗∗ -0.27∗∗∗ -0.31∗∗ 0.04 -0.11 -0.24∗∗ -0.31∗∗ 0.77∗∗∗


(0.04) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.01)IV 0.27∗∗∗

(0.02)Constant 1.12∗∗∗ 1.50∗∗∗ 1.50∗∗∗ 1.28∗∗∗ 1.39∗∗∗ 1.49∗∗∗ 1.54∗∗∗ -0.05∗∗∗


Note Dependent variable: columns (1)-(7) - # GDR-origin pat’s in GDR, column (8) - #SU-origin pat’s in GDR. ∗ -significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. IV is defined by equation (2), where

< # pats >pred is estimated by fitting an AR(2) process on pre-escallation data (1972-1978) for each subclass. Seedefinitions of other variables in Table 1. Robust standard errors are reported in parentheses. All variables are inlogarithms: xnew = log(1 + x)

29

Table 7: Spillovers to East German inventors: Specification in Levels1979-91 1979 1979-82 1979-85 1979-88 First stage


# SU-origin 13.17∗∗∗ 10.25∗∗∗ 15.10∗∗∗ 0.09 6.90∗∗∗ 13.00∗∗∗ 17.32∗∗∗

patents in GDR (1.34) (0.95) (1.44) (0.74) (0.60) (0.88) (1.28)# SU-origin 4.16∗∗∗ -5.24∗∗∗ -6.69∗∗∗ -0.43 -3.49∗∗∗ -6.14∗∗∗ -7.81∗∗∗ 0.63∗∗∗


(0.59) (0.29) (0.31) (0.14) (0.17) (0.27) (0.35) (0.01)IV 0.60∗∗∗

(0.03)Constant 4.80∗∗∗ 8.27∗∗∗ 7.79∗∗∗ 6.36∗∗∗ 6.90∗∗∗ 7.60∗∗∗ 8.07∗∗∗ -0.12∗∗∗


Note Dependent variable: columns (1)-(7) - # GDR-origin pat’s in GDR, column (8) - #SU-origin pat’s in GDR. ∗ -significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. Robust standard errors are reported inparentheses. All variables are in levels (as opposed to logarithms in the main specification)

Table 8: Spillovers to East German inventors: Alternative Functional FormsPoisson Neg. binomial

(1) (2) (3) (4)OLS FE OLS FE

dv1# SU-origin 0.46∗∗∗ 0.06∗ 0.92∗∗∗ 0.06patents in GDR (0.02) (0.03) (0.19) (0.04)# SU-origin 0.24∗∗∗ 0.03∗∗ 0.69∗∗∗ 0.04∗

patents in US and GDR (0.01) (0.02) (0.09) (0.02)=1 if 1977-78 -0.69∗∗∗ -0.54∗∗∗ -0.51∗∗∗ -0.54∗∗∗

(0.02) (0.03) (0.07) (0.03)Constant 1.87∗∗∗ 1.36∗∗∗ 4.66∗∗∗

(0.01) (0.05) (0.49)lnalphaConstant 0.06

(0.05)Observations 1288 1250 1288 1250Pseudo R-squared 0.33 0.09

Note Dependent variable: # GDR-origin pat’s in GDR. ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ -significant at 1% level. Robust standard errors are reported in parentheses. All variables are in levels. For negativebinomial mean dispersion is assumed.

30

Table 9: Spillovers to East German inventors: Patents without Patent Fam-ilies

DD patents Int. teams Num. inv. First stage(1) (2) (3) (4) (5) (6) (7) (8)log lev log lev log lev log lev

# SU-origin 0.23 10.61∗∗∗ 0.80∗∗∗ 0.38∗∗∗ 0.36 0.67patents in GDR, no pat. family (0.30) (3.76) (0.31) (0.14) (0.48) (0.68)# SU-origin 0.25∗∗ -2.02∗ -0.26∗∗ -0.08∗∗ 0.01 -0.10 0.51∗∗∗ 0.50∗∗∗

patents in US and GDR (0.12) (1.09) (0.12) (0.04) (0.19) (0.20) (0.09) (0.09)=1 if 1977-78 -0.54∗∗∗ -5.71∗∗∗ 0.08 0.09 0.03 0.50 -0.13∗∗∗ -0.19∗∗

(0.07) (2.14) (0.07) (0.08) (0.11) (0.39) (0.03) (0.09)IV 0.31∗∗∗ 0.69∗∗∗

(0.07) (0.13)Constant 2.06∗∗∗ 2.85 -0.33 -0.13 1.25∗∗∗ 3.22∗∗∗ 0.58∗∗∗ 0.85∗∗∗

(0.24) (5.32) (0.24) (0.20) (0.38) (0.97) (0.05) (0.11)Observations 455 455 455 455 455 455 455 455R-squared 0.31 0.27R-squared overall 0.15 0.01 0.03 0.12 0.00 0.00 0.30 0.24

Note Dependent variable: columns (1)-(7) - # GDR-origin pat’s in GDR, column (8) - #SU-origin pat’s in GDR withoutpatent families. ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. See definitions ofvariables in Table 1. Robust standard errors are reported in parentheses. All variables are in logarithms:xnew = log(1 + x)

Table 10: Boycott Adherence and SU Patent Flows: Results from Cross-Country Data

1975-80 vs. 1981-85 1974-79 vs. 1980-84 1970-80 vs. 1981-90(1) (2) (3) (4) (5) (6)∆ %∆ ∆ %∆ ∆ %∆

strong adherent -22.72∗∗ -1.05∗∗∗ -10.50∗∗ -0.52∗∗∗ -14.89 -0.94∗∗

(10.52) (0.31) (4.11) (0.14) (10.62) (0.33)weak adherent -10.94∗∗ -0.82∗∗ -4.79 -0.32 -10.82∗ -0.92∗∗

(4.94) (0.35) (4.07) (0.21) (5.61) (0.35)Constant 3.28∗ 0.36 1.90 0.16 3.33∗ 0.50∗

(1.61) (0.30) (1.28) (0.13) (1.76) (0.26)Observations 19 16 19 16 19 16R-squared 0.24 0.47 0.19 0.27 0.11 0.37

Note: ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. Dependent variable in eachregression - average over period quarterly number of patents with Soviet Union priority numbers, ∆ - absolute differencebetween post-escalation and pre-escalation, %∆ - percent difference. Different pre- and post-escalation periodsconsidered: 1975-80 vs. 1981-85; 1974-79 vs. 1980-84; 1970-80 vs. 1981-90. Strong adherents: Canada, China, Japan,South Korea, West Germany, United States. Weak adherents: Australia, Switzerland, Spain, France, U.K., Italy,Netherlands, Sweden. Non-adherents: Czechoslovakia, East Germany, Finland, Hungary, Yugoslavia. Poland is excludedin all regressions as a likely outlier (due to anomalously high application counts in the second half of 1977, which resultsin a very high and negative post- to pre-escalation percentage change in Soviet patent applications).

31

Table 11: SU Patent Applications in GDR and Number of PublicationsOLS, levels Negative binomial Poisson

(1) (2) (3) (4) (5) (6)SU+foreign foreign SU+foreign foreign SU+foreign foreign

main# GDR patents by 0.02∗∗∗ 0.00 0.26∗∗ 0.29 0.15∗∗∗ 0.11∗

pair of inventors, 1976-92 (0.00) (0.00) (0.11) (0.22) (0.01) (0.06)Constant 0.01∗∗ 0.00∗∗ -3.82∗∗∗ -5.40∗∗∗ -3.67∗∗∗ -5.16∗∗∗

(0.00) (0.00) (0.16) (0.36) (0.03) (0.16)lnalphaConstant 5.52∗∗∗ 5.15∗∗∗

(0.07) (0.29)Observations 38761 9044 38761 9044 38761 9044R-squared 0.00 0.00

Note: ∗ - significant at 10% level, ∗∗ - significant at 5% level, ∗∗∗ - significant at 1% level. Sample: all pairs ofinventors, which co-author at least one GDR patent application in 1976-92. Dependent variable - publications by thepair of inventors in 1976-92, as registered by the Google Scholar. Inventors are identified by the last name and the firstletter of the first name. Columns (2), (4) and (6) include only pairs of inventors with at least one foreign (non-Soviet)inventor; columns (1), (3), and (5) include both Soviet, foreign and mixed pairs of inventors. Data on publications:Google Scholar (accessed in August 2011).

References

Bernard, A. and M. Busse (2004). Who Wins the Olympic Games? EconomicResources and Medal Totals. Review of Economics and Statistics 6 (1):413–417.

Borjas, G. and K. Doran (2012). How high-skill migration affects science:Evidence from the collapse of the USSR. Quarterly Journal of Economics127 (3): 1143–1203.

Cook, L. (2011). A green light for red patents: New evidence from Sovietinnovation abroad, 1933 to 1991, Michigan State University working paper.

Ginsburgs, G. (1987). A Calendar of Soviet Treaties, 1974-1980 (MartinusNijhoff Publishers, Dordrect, The Netherlands, 1987).

Ginsburgs, G. and R. Slusser (1981). A Calendar of Soviet Treaties, 1958-1973 (Martinus Nijhoff Publishers, Dordrect, The Netherlands, 1981).

Hoover Institution (1980). Scientists for Sakharov, Orlov, and SharanskyPapers, Hoover Institution Archives, Stanford Univeristy, accessed January2014.

Moser, P. and A. Voena (2012). Compulsory licensing: Evidence from Tradewith the Enemy Act. American Economic Review 102 (1): 396–427.

32

U.S. Department of State, Office of the Historian (2012). The Olympic Boy-cott, 1980 (http : //history.state.gov/milestones/1977 − 1980/Olympic,2012).

Zhyrnov, Y. (2010). No other Olympics was so unprofitable. KommersantVlast 21 (875).

33

stop playing: the boycott of the moscow summer olympics ...econ.msu.edu/papers/cook/stop playing....

Documents