FDI in Germany

7/21/2019 FDI in Germany

http://slidepdf.com/reader/full/fdi-in-germany 1/41

FDI of German companies during globalization and deglobalization

Gerhard Kling,a Joerg Baten b and Kirsten Labuskec

a University of Southampton, b University of Tuebingen, c The onference Board

THIS IS NOT THE FINAL !OST"#E$IE%& $E#SION

'O( FIND THE FINAL $E#SION HE#E)

Kling, G!, Baten, J! and K! Labuske "#$%%& '() of German companies during

globali*ation and deglobali*ation, Open Economies Review ##"#&, #+-#$!

Based on micro-level data of German companies from %./ to %0#, 1e identified hori*ontal

and vertical '() applying a Kno1ledge-apital model and analy*ed individual '() decisions!

2ur K model revealed that market-driven '() predominated3 ho1ever, 1age gaps and

differences in human capital stimulated cost-driven '() flo1s, 1hich accounted for up to

%$4 of total '()! 2n an individual level, large companies 1ith high profitability conducted

more '()! 5igher tariffs after 66) enhanced '(), as companies could circumvent trade

barriers 7 but declining openness reduced '()! )n spite of disintegration after 66), the

propensity to invest increased due to higher market concentration and firm specific

investment patterns - albeit industry agglomeration effects 1ere of minor importance!

J8L codes9 '#%, :/, :+

Key 1ords9 '()3 globali*ation3 protection3 Germany

;ckno1ledgements9 6e thank the anonymous referee for her or his comments that helped us

to improve our paper significantly! 6e thank 2livier ;ccominotti, <arc 'landreau, <ichael

lemens, <ar =ubio for providing crucial data on protection, and participants of the

8conomic 5istory Society ;nnual onference #$$>, the Second onference on German

liometrics #$$>, seminars at the Universities of Barcelona and Tuebingen, the Tuebingen

economic history research group for their comments on earlier versions!

%



*+ Introduction

The end of the %0th century and the beginning of the #$th century marked a period of

e?panding demand, gro1th in productive capacity and rising e?ports in Germany! Germany

became one of the leading industrial countries and took a pioneering role especially in the

chemical, electric, and engineering industry until 66) "5enderson, %0@, p!%/&! )nventions

such as the dynamo and the electric bulb 1ere as important to the electric industry as

synthetic dyes to the chemical industry or the development of steam engines to the

engineering industry! :e1ly founded companies, like ;8G or Siemens, achieved commercial

success and entered the global stage! 'or instance, the fastest-gro1ing electro-technical firm

;8G invested / times abroad from %./ to %0#! ;lready in %.0#, ;8G conducted their first

'() in the UK, soon after 8mil =athenau acAuired a license for 8disons patents on lamps and

the foundation of ;8G in %.. "see Cohl, %0..&! Gro1ing competition, especially in ne1ly

emerging industries 1ith high gro1th potential, reAuired entering ne1 markets and reducing

production costs, 1hich triggered more and more '()! ;fter a period of enhanced economic

and financial integration, 66) marked a turning point in international relations and a phase

of protection and Ddeglobali*ation started, 1hich allegedly contributed to the economic and

social breakdo1n in the %0/$s "hase, #$$+&!

2ur paper focused on '() during the period %./-%0# undertaken by German

companies3 hence, by covering periods of integration and disintegration, 1e had the uniAue

opportunity to assess the impact of Ddeglobali*ation on '() streams bet1een countries and

individual investment decisions! )n a first step, 1e aggregated individual '() to a panel of

country level investment streams! 6e applied an e?tended Kno1ledge-apital model "K& to

identify country characteristics that attracted '() and to distinguish bet1een market and cost-

driven factors! The second step e?ploited our micro-level dataset on 0+. individual '()

transactions of / Eoint stock companies, as 1ell as a control group of @@> Eoint stock

companies 1ithout '(), as it allo1ed us to reveal company characteristics that stimulated

#



entering foreign markets! 'inally, 1e investigated the endogeneity of '() decisions using the

heterogeneity of our panel data! )n particular, 1e addressed firm and industry-specific

investment patterns and therefore considered agglomeration effects and past individual

investments!

Theoretical models of hori*ontal "market-driven& '() focused on the trade-off

bet1een firm-level economies of scale and transportation costs, 1hich suggests that in the

absence of transportation costs firms prefer producing in one factory and e?porting goods to

foreign markets!% ;s transportation costs 1ere relatively high from %./ to %0#, 1e 1ould

e?pect a strong incentive to conduct hori*ontal '(), and therefore companies developed

production and distribution net1orks in host countries to meet local demand! )n contrast,

differences in factor intensities and factor prices cause vertical "cost-driven& '(), 1hich

contends that companies shift parts of their production process or their entire production

"leaving only the headAuarter services in the home country&, 1hich are not skill-intensive, into

countries 1ith lo1 1ages for unskilled labor!# )f countries 1ith a high 1age gap, namely

lo1er real 1ages compared to Germany, and lo1 relative skill levels "measured by the

number of patents and primary school enrolment rates& attracted '(), 1e could regard these

investments as being primarily cost-driven! There are many terms used to describe different

forms of vertical '() like Dslicing up the value chain coined by Krugman "%00>&,

fragmentation "see Jones and Kier*ko1ski, %00$& or outsourcing3 ho1ever, these forms 1ere

not technologically feasible in the first phase of globali*ation! :evertheless, shifting the entire

production and not Eust parts of it to a different country for the sake of lo1ering production

costs 1as possible from a technological point of vie1! ;ccordingly, the research Auestions

arise 1hether 1e can find vertical '() in the period %./-%0# and 1hat type of vertical '()

1as used! arr, <arkusen and <askus "#$$%& combined both, market and cost-driven

% 6e refer to <arkusen "%0.+& and <arkusen and Fenables "%00.&!

# 5elpman "%0.+& and 5elpman and Krugman "%0.@& developed the first theoretical models that e?plain vertical

'()!

/



theoretical models, into the Kno1ledge-apital "K& model, 1hich 1as empirically tested

and modified by Braconier, :orbaeck, and Urban "#$$@& and (avies "#$$+&! By aggregating

our firm level '() data, 1e applied a modified K model to e?plain '() across countries and

to assess the relative importance of market and cost-driven '()!

2ur second approach disentangles individual '() decisions and controls for firm and

industry-specific effects! '() decisions on the firm level are likely to be influenced by firm

heterogeneity in terms of productivity "5elpman, <elit* and eaple, #$$+3 Bernard and

Jensen, %00@& and agglomeration economies, 1hich refer to positive e?ternalities of an

industry cluster in a particular host country that stimulate investment of firms in the same

industry 7 see Greena1ay and Kneller "#$$& for a revie1 of studies on agglomeration

effects&!

2ur paper is organi*ed as follo1s9 the literature revie1 stresses the current debate

concerning hori*ontal and vertical '() on an aggregated and micro-level and e?plores sources

of heterogeneity bet1een firms and industries "e!g! agglomeration effects&! To assess the

relevance of our study, the historical conte?t is essential, as our period combines the

e?perience of the first phase of globali*ation and the subseAuent disintegration after 66)3 the

third section highlights our data collection efforts and sho1s descriptive findings3 the fourth

section presents our empirical results along the line of the follo1ing three Auestions9 "%& 1ere

'() flo1s bet1een countries primarily market or cost-driven3 "#& 1hich factors influenced

individual '() decisions taking into account firm characteristics, industry-specific effects and

the time pattern of '()3 "/& did protection and disintegration change investment incentives!

'inally, 1e conclude and discuss our findings!

,+ Literature re-ie.

+



The traditional vie1 of capital flo1s based on a simple neoclassical model is that investments

from capital abundant countries should flo1 to economies that have lo1 relative capital

endo1ment and are HrichI in other factors such as unskilled labor or natural resources! / )n

contrast, most empirical studies found that the largest share of '() usually flo1s from rich,

high-1age countries to other rich, high-1age countries! 5ence, another strand of theoretical

models 7 labeled :e1 Trade Theory 7 evolved that 1as better able to e?plain the empirical

facts! 5o1ever, the models that e?plained hori*ontal "market-driven& activities "<arkusen,

%0.+& failed to e?plain vertical "cost-driven& investments "5elpman, %0.+& and vice versa!

The Kno1ledge-apital model "arr, <arkusen, and <askus, #$$%&, in contrast, allo1ed

hori*ontal and vertical activities and tried to e?plain international activities by country-

specific factors! These factors include the Eoint market si*e of the home and host country, a

dispersion factor "sAuared difference in real G(C&, a measure for skill differences, and trade

costs! The K model assumes that the assets of kno1ledge-based firms can be used in many

types of economies, including rich and human capital-abundant countries! )t comprises as

special cases the hori*ontal "market-driven& and the vertical "cost-driven& strategy9 the

hori*ontal strategy means that production processes are placed via '() in countries that are

very similar in human capital intensity to the headAuarter economy "home market&! The main

motive is to gain market access more easily, by moving production into the pro?imity of

foreign consumers, 1hich lo1ers transport costs and circumvents other trade barriers "e!g!

tariffs&! )n empirical studies, G(C of the target economy should be a strong driver for

hori*ontal '(), as it indicates market potential! )n addition, similarities in terms of market

si*e and endo1ment favor hori*ontal activities!

uite contrary, the vertical strategy of '() follo1s the idea that the stages of the

production process are sliced up vertically, and each stage of production takes place 1here the

factor costs are lo1est9 simple stages of production are relocated to lo1-1age, lo1-human

/ Lucas "%00$& discussed the neoclassical prediction and its limitations!

@



capital countries, and human capital intensive processes take place in high-1age countries, for

e?ample headAuarter services! The empirical implication 1ould be that G(C of the host

country should not matter much, and dissimilarity in terms of endo1ment and skill levels

resulting in factor price differences "e!g! 1ages& should stimulate vertical activities! The

modern form of slicing up the production process vertically reAuires a considerable amount of

technology "e!g! information technology&, 1hich 1as not feasible in the pre-%0# period!

ost-driven '() decisions and the shift of total production facilities3 ho1ever, 1as possible

"e!g! mining and access to ra1 material, factories in host countries&, 1hich the follo1ing

section discusses in detail!

The K model nests both models, and it predicts that companies employ the vertical

or hori*ontal strategy, 1hichever might be most suitable for a given situation! <ost empirical

studies for the last fe1 decades tended to confirm that market-driven '() is the predominant

strategy, but there is also some evidence for cost-driven '(), especially from intervie1s 1ith

entral 8uropean firms that aim at using the 1age differential bet1een entral and 8astern

8urope to slice up their production chain! )n sum, most of the recent literature stressed that

market access is the strongest motive for '(), although vertical "cost-driven& motivations play

a role in some situations "especially 1here factor price differences are only separated by

relatively open borders, such as 1ithin the 8U, or bet1een the U!S! and <e?ico&!

;ccordingly, our estimation approach is motivated by arr, <arkusen and <askus "#$$%&,

<arkusen and <askus "#$$%, #$$#&, and Blonigen, (avies, and 5ead "#$$/& that applied the

K model to macro-level data on '()! )n contrast to arr, <arkusen and <askus "#$$%&, 1ho

used sales of affiliates as dependent variable, 1e tried to e?plain '() flo1s bet1een Germany

and host countries using country-specific factors "total market si*e, dispersion of real G(C,

distance, difference in human capital measured by patent data and primary school enrolment

rates& as e?planatory variables! To assess the changing legal and political environment after

>



66), 1e included t1o measures, namely tariff barriers and an openness indicator "defined as

value of merchandise e?ports relative to G(C&!

Besides analy*ing aggregated '() streams, firm heterogeneity and the endogeneity of

'() have received much attention in the theoretical and empirical literature "Greena1ay and

Kneller, #$$3 5elpman, <elit* and eaple, #$$+3 Bernard and Jensen, %00@&! )n particular,

firm characteristics "e!g! productivity, si*e& might influence '() decisions 7 but also industry

structure seems to play a role! )n specific, the agglomeration effect that refers to positive

e?ternalities e?perienced by firms that move into industry clusters might e?plain '() patterns

"the DSilicon Falley 8ffect&! 'or e?ample, Japanese firms tend to cluster together in U!S!

regions "5ead, =ies, and S1enson, %00@&! Those agglomerations provide information-

processing advantages and 7 in the case of similar skills demanded on the labor market 7

industry clusters might be beneficial!

;s only a fe1 papers on '() "for e?ample Buch et al!, #$$@3 6agner and Schnabel,

%00+& analy*ed recent micro-level data, 1e stress that this is the first approach to investigate

German micro-level '() data in the first globali*ation period and the %0#$s!+ <ost studies so

far used aggregated data, 1hereas firm level studies have been only conducted for the U!S!,

S1eden, and Germany using data of the last t1o decades! )n particular, long run studies that

could make use of the special feature of the K model 7 that behavior should be determined

by varying economic environments 7 are lacking so far! 2ne interesting recent study on

Germany compared results at the firm level and at the aggregated level "Buch et al!, #$$@&!

The authors mobili*ed a very large panel dataset recorded by the Bundesbank and compared

aggregated and individual '() to target countrys G(C and similarity of G(C! 2n the

aggregated level, an additional percent of G(C results in almost % percent additional '()!

This can be decomposed into "a& the increase of the number of affiliates and "b& a higher

+ Buch et al! "#$$@& pointed at the limitations of studies that only focus on macroeconomic or aggregated data, as

they did not allo1 assessing firm-specific characteristics and incentives for '()!



investment per affiliate! Buch et al! "#$$@& found that the investment per affiliate accounts for

about one third of additional '() volume, 1hereas additional affiliates account for the

remaining t1o thirds "assuming that there is no omitted variable and measurement error bias

in their regressions&! The authors also uncovered that similarity of G(C of the host country

compared to Germany, 1hich is the only headAuarter economy in this study, has a positive

influence on the si*e of the investment and sales per affiliate, 1hereas protection has a

negative influence!

The economic and political conte?t during our investigation period might affect

investment patterns and hence is a central aspect of our study, as it allo1s assessing the

impact of globali*ation and disintegration on aggregated '() flo1s and on individual '()

decisions! The economic history of foreign investment 1as multi-facetted and can be outlined

here only briefly! 6ilkins "%0$, %0+, %0.0, and #$$+& described '() of U!S! firms abroad

and of foreign firms in the U!S! in a series of monographs! Unfortunately, for other countries

including Germany, a systematic data collection and evaluation of '() is yet missing 7 see

5agen "%00& on German investments in the UK and Schaefer "%00@& on portfolio

investments! Broadly, the period can be divided into an early phase of globali*ation before

66), and a post-1ar environment! )n the pre-1ar period, '() took place in a liberal 1orld, in

1hich trade blocs and similar constraints 1ere largely unkno1n! 2n a regional basis,

municipalities restricted foreign investment, since they 1ere influenced by local employers

afraid of rising 1ages due to increased competition on the labor market "Baten, %00/, p! +&!

et if one municipality generated administrative obstacles, the investing firm could easily

target another location! Besides the lack of barriers for '(), any subsidies "ta? breaks etc!&

1ere also absent! (uring the %0#$s, a large number of countries tried actively to attract

foreign investment, although ta? breaks as a systematic strategy 1as still not common!

)nstead, more rapid administrative procedures 1ere offered! To highlight the influence of

66) and the disintegration phase, Table % reports the largest recipients of German '() in the

.



pre-%0%+ and post-%0%+ period! The UK and ;ustria had by far the largest share before 66),

1hereas the U!S!, )taly, =ussia, and 'rance 1ere apparently also interesting markets! )n the

post-1ar period, there 1as a certain change! German '() 1ent more often to neighboring

countries, such as post-1ar Coland and especially the city of (an*ig, 1hich 1as under the

League of :ation administration after Germany lost 66), but economic ties 1ith Germany

1ere still important! :eighboring S1it*erland 1as the third most important destination,

1hereas it had not appeared on the pre-1ar top %$ list! ;ustria remained in the lead after

66), and *echoslovakia "no1adays *ech and Slovak =epublic& became more important!

But still, the US attracted about 4 of '() before and after 66) 1ith no apparent difference!

=ussia disappeared from the list of potential host countries after %0% and the Soviet Union

did not replace it, as the communist country 1ould not have accepted German '() "at least

officially&!

")nsert Table %&

The '() to1ards the U!S! and other 1ar enemies of 66) 1as influenced by the 1ar

policies of sei*ing foreign property! 'or e?ample, in the U!S!, German '() 1as de facto

e?propriated in %0%, although a number of companies found either ;merican stooges or

people living in neutral countries "such as S1it*erland& to continue the activities of their

foreign plants and sales units "6ilkins #$$+, pp! %%/-%%+&! 'or those investors 1ho 1ere not

able or 1illing to initiate such a solution, the debates about returning Halien trustsI 1ere long

and complicated! 2nly small investments 1orth less than US %$,$$$ 1ere returned by the

la1 of %0#/, i!e! small sales units might have been given back! 2ne of the large foreign

affiliates o1ned previously by the <etallgesellschaft, even caused a la1suit about bribery

"6ilkins #$$+&! The complicated situation in some of the previous opponents of 1ar countries

might have stimulated additional '(), as the plants and sales units, 1hich 1ere not restored

fast enough, needed to be replaced by ne1 '()! :evertheless, investments in former enemy

0



countries might have been considered more risky than before! )n sum, those push and pull-

factors might to a certain e?tent have offset each other, as the U!S!, the UK and 'rance 1ere

still among the top %$ host countries!

;nother issue 1hich had changed bet1een the pre-1ar and post-1ar 1orld 1as the

emergence of trade blocs! )t became even more important during the %0/$s 1hen Germany

aimed at increasing trade 1ith 8ast and Southeast 8urope, and it might be that '() follo1ed

trade relations! 5o1ever, already during the %0#$s, the Sterling and the 'ranc trade blocs

became more consistent and closed than before 66)! ;gain, there might have been counter-

acting forces! ; trade bloc 1ith its non-tariff restrictions against non-members "indirect

barriers to trade& had a similar effect as a tariff9 trade 1as limited and property rights might

have been more difficult to enforce, but the missing trade might also have initiated additional

'() to substitute trade! )t is difficult to reveal empirically the effect of the Sterling and 'ranc

trade blocs for German '(), as those 1ere dominated by former German 1ar enemies! 5ence,

the decline of German '() in the UK might have been caused by the Sterling *one as 1ell as

the problematic relationship 1ith former 1ar enemies! )n sum, 1hile the economic and

political environment bet1een the pre- and the post-1ar period has changed Auite

substantially, it is not clear 1hether there 1as a pronounced effect on German '() because of

counter-acting forces! Looking at the list of the most freAuently chosen target countries, the

UK clearly lost its dominance to S1it*erland, but other1ise the list of top host countries

hardly changed after 66)!

/+ Data and construction of -ariables

2ur analysis included all Eoint stock companies listed on German stock e?changes and

documented in the 5andbMcher der deutschen ;ktiengesellschaften from %./ to %0# that

undertook '()! onseAuently, / Eoint stock companies conducted in total 0+. '()

%$



transactions during the investigation period! To account for companies that did not undertake

any '(), 1e randomly selected @@> Eoint stock companies as a control group! Generally, 1e

have information about the year, the destination "to1n and country&, the industry, and the

amount of investment of the respective '() transaction! 2ur dataset provides information on

'() of German companies in @@ countries and / industries! The %$ most important countries

1ere by far ;ustria-5ungary "current borders, after 66)9 ;ustria and 5ungary& and the UK,

follo1ed by 'rance, US;, )taly, =ussia "SU after 66)&, Coland "before 66)9 part of

neighbor empires&, S1it*erland, :etherlands, and *echoslovakia "before 66)9 part of

;ustria-5ungary&! 'ig! % plots the relative importance of specific industries regarding '()!

6e observe a particularly high number of '() in the electric and machinery industry 1ith a

share of about +$4 of total '(), follo1ed by the chemical and metal processing industry!

:one of the remaining industries accounted for more than >4 of total '()!@

")nsert 'ig! %&

'ocusing on the type of '() provided by the N5andbMcher der deutschen

;ktiengesellschaften, 1e can identify to some e?tent the motive for '() "access to ra1

material, production, sales& and the mode of market entry "eAuity stake, merger, Greenfield

investment&! (ue to data limitations, 1e cannot identify the type of '() for all '()

transactions3 nevertheless, 1e obtain an overvie1 of the relative importance of different

motives and modes of entry! Table # sho1s that Greenfield investments accounted for #4 of

total '() and cross-border mergers 1ere of marginal importance3 ho1ever, #>4 of '() 1as

based on eAuity stakes! Credominantly, '() referred to starting production and sales

operations in a host country, but #4 of '() focused on sales "e!g! sales agencies& and %/4 of

'() 1as due to production! )n the mining and steel industry, '() 1as important to obtain

access to ra1 material "e!g! mines, forests&! onseAuently, some forms of '() "e!g! shift of

@ 6e account for the four most active industries concerning '() by including dummy variables into our

regression models!

%%



entire production, mining& 7 certainly different from modern forms "e!g! outsourcing of

finance function& 7 1ere observable and given the level technical progress feasible!

")nsert Table #&

'ollo1ing arr, <arkusen and <askus "#$$%&, 1e estimated the K model to

disentangle vertical and hori*ontal investment activities! ;ccordingly, our regression models

refer to the follo1ing country-specific variables9 "%& the sum of real G(C, "#& the sAuared

difference in real G(C of Germany and the respective host country "dispersion factor&, "/& the

distance bet1een Germany and the respective host country, and "+& the skill difference

bet1een home and host country! 'rom a theoretical perspective, large and similar countries in

terms of factor endo1ments and market si*e should attract hori*ontal activities, as market

access seems to be the most important motive! )n contrast, skill differences should result in

factor price differences and hence could trigger vertical '() flo1s, as countries can shift

production to benefit from factor price differentials! Based on <addisons "%00@& G(C time

series 1ith constant prices, 1e derive the sum of real G(C of Germany and the host country

" sum_gdp) and calculate the sAuared difference in real G(C, 1hich indicates the degree of

similarity in terms of market si*e "dispersion)! To account for transportation costs and as a

common pro?y for cultural differences, 1e include the distance in km bet1een the capitals of

the home and host country "distance&! ;dEacent countries tend to have closer economic

linkages due to lo1 transportation costs, similar culture and familiar regulatory frame1orks

"e!g! legal system&! 8?tending the idea of cultural similarity, 1e controlled for the official

language of the host country "language)! To account for skill differences, 1e used t1o

indicators, namely the number of patents of foreign firms "located in the host country& in

Germany " patent & and primary school enrolment rates!0 Catent data provide an output measure

for skills in an economy compared to input-oriented measures like the number of scientists

> To ensure that patents are of high importance for the respective company, 1e follo1ed Streb, Baten and in

"#$$>& 1ho distinguished bet1een lo1 and high-value patents!

%#



per %$$$ 1orkers or primary school enrollment rates "arr, <arkusen, and <askus, #$$%&! ;s

Germany can be regarded as skill abundant country, a high skill level in a host country

suggests that the host country is more similar compared to Germany! Using the number of

patents " patents& as an indicator for the host countrys skill level e?hibits inherent limitations3

in particular, '() flo1s might lead to patents to protect ne1ly entered markets, 1hich results

in an endogeneity bias! onseAuently, primary school enrollment rates serve as a more robust

indicator of skill levels, albeit schooling is an input measure! To measure the skill difference

in terms of primary school enrollment rates, 1e calculated the difference of the host countrys

enrollment rate and the top #@4 of all host countries "enrollment &3 thus, a negative value

indicates that the respective host country has lo1er enrollment rates compared to the leading

#@4 of host countries!

2ur investigation period covers the first phase of globali*ation and a period of

disintegration and protection follo1ing 66)3 thus, 1e attempted to measure the impact of

protection and disintegration on '() streams! 2n the one hand, 1e 1ould e?pect more '()

bet1een 1ell-integrated markets, if the costs of production are substantially different "vertical

activities& or the pro?imity of production to consumer markets plays a large role "hori*ontal

activities&! 2n the other hand, '() 1as often used as a substitute for trade, 1hen tariff barriers

1ere high! )n particular, if intangible assets 1ere an incremental part of the firms product 7

for instance special e?pertise, reputation, and brand 7 capital could be moved behind tariff

1alls by setting up a production facility in the selected host country! 'or e?ample, the Singer

se1ing machine company 1as a famous U!S! multinational, 1hich often moved behind tariff

1alls and even pretended to become a Dnative company of respective host countries "6ilkins,

%0.>3 2=ourke and 6illiamson, %000, p!#%.&!

6e used the ratio of custom revenues to total imports for a given country and time to

construct a measure of protection " protect &! 2ne maEor source of data 1as ;ccominotti and

'landreau "#$$>&! They collected protection rates for ;ustria-5ungary, Belgium, S1it*erland,

%/



Spain, 'rance, the UK, )taly, :etherlands, =ussia, and the U!S! for the period %.+$-%.0$!

;nother important source 1as =ubios "#$$>& data for Latin ;merican countries3 thus, 1e

1ere able to fill gaps until %0%$ for some countries! 'inally, 1e used average tariff rates from

the Auite comprehensive lemens and 6illiamson "#$$+& dataset for the %0#$s! 5o1ever, our

protection variable might be biased if there are very high tariffs for some goods among others

1ith lo1er protection! ;s the imported volume of a good 1ith a high tariff might decrease, the

1eight of a highly protected good reduces in the calculated overall measure of protection!

This effect might understate the degree of protection of the respective host country! Therefore,

1e added an additional indicator that Auantifies the degree of openness " openness&, 1hich is

based on the value of merchandise e?ports relative to G(C! 'inally, to Auantify factor price

differences directly, our regression model incorporated the difference in real 1ages based on

6illiamson "%00@& "wagediff &!

To address firm heterogeneity and the endogeneity of '(), 1e analy*ed in a second

step individual '() decisions and accounted for the time pattern of investments! Besides our

measures for protection and openness that might influence individual '() decisions "e!g! tariff

barriers might increase '()&,. 1e follo1ed 5elpman, <elit* and eaple "#$$+& by analy*ing

the relationship bet1een firm productivity and '()! They emphasi*ed that firms conducting

'() are not only larger, but also more efficient and productive than firms that produce for the

home market or choose to e?port! )n contrast to 5elpman, <elit* and eaple "#$$+& that used

the firm si*e dispersion as a measure of firm productivity, 1e measured productivity directly

by calculating the return on eAuity " ROE &!

8conomies of scale on the firm and plant level could influence '() decisions, as high

economies of scale on the firm level relative to the plant level make hori*ontal '() more

Based on <addison "#$$/&!

. )n the case of our control group that contains companies 1ithout '(), 1e use the median of the level of

protection and openness based on the '() transactions in the respective year!

%+



likely "Brainard, %00/3 <arkusen and Fenables, %00.&30 hence, 1e take the log of total eAuity

as pro?y of firm si*e to control for si*e advantages " size&!%$ Besides the interrelation of firm

si*e and economies of scale, larger companies have easier access to capital markets! Thus,

'() is easier to finance, and cost of capital is lo1er, 1hich makes '() more attractive! %%

<oreover, the year of the establishment of a company " foundation& controls for the degree of

firms maturity! To control for market concentration in Germany, the 5erfindahl )nde? " HI &

served as a pro?y! The inde? refers to the sum of sAuared market shares of firms 1ithin the

same industry3 a high 5erfindahl )nde? indicates a high level of concentration! Typically, the

5erfindahl inde? is standardi*ed to create a range of inde? values bet1een $ and %$$ "lo1 to

high level of concentration&! The definitions of variables and sources used in the empirical

part of the paper are summari*ed the appendi?!

To obtain an overvie1 regarding the firm characteristics of companies that conducted

'() and companies that did not, Table / reports means, standard deviations, @ and 0@-

percentiles for firm si*e, value of '() relative to total assets, return on eAuity and year of

establishment! Based on group-1ise descriptive statistics, Table / underlines that companies

1ith '() activities tended to be larger, and more profitable indicated by a higher return on

eAuity "=28&! 'urthermore, there seem to be industry-specific effects that need to be analy*ed

further3 particularly, the si*e of '() relative to total assets 1as larger in the metal industry

compared to the electric, machinery and chemical industry!

")nsert Table /&

0 Clant level of scale economies cannot be measured due to a lack of data!

%$ :ote that other pro?ies like the number of employees 1ould reduce the number of observations considerably

due to missing data! )n addition, alternative measures are highly correlated 1ith our pro?y!

%% Tilly "%0.#& argued that the companies la1s of %..+ and the ne1 e?change la1 established %.0> favored

larger companies! 'or instance, the la1 reAuired that the minimum issue volume had to e?ceed one million <ark!

5ence, a larger company had advantages to finance e?pansion by issuing ne1 shares! The companies la1 and

the ne1 e?change la1 mainly determined the legal frame1ork in the pre-%0%+ period!

%@



1+ Empirical results

4! "as #$I market or cost%driven&

'ollo1ing arr, <arkusen and <askus "#$$%&, <arkusen and <askus "#$$%, #$$#& and

Blonigen, (avies, and 5ead "#$$/&, 1e specified K models to e?plain cross-country '()

flo1s and to identify 1hether '() 1as driven by market access "hori*ontal activities& or cost

savings "vertical activities&! ;ccordingly, 1e aggregated individual '() flo1s for every year

and country for 1hich 1e could detect '() flo1s!

Therefore, log investment streams "aggregated to country '() streams based on

individual '()& served as dependent variable! Using 5offmanns "%0>@& price inde? for

Germany, 1e deflated the value of '()! ;s the value of '() 1as not al1ays reported, 1e have

only /> underlying observations that 1e aggregated to %.$ country level '() streams! *, To

describe the global geography of German '()s during this period, 'ig! # highlights the

residual '() per billion G(C after accounting for distance and common language3 hence, the

residual value is adEusted for spatial and cultural pro?imity! )t is apparent that Scandinavia

received much less German '() per billion G(C, after accounting for its pro?imity! ;lso

Southeastern 8urope, Turkey, and 8gypt did receive only modest amounts, and the same

applies to anada, Cortugal and Spain, 1hereas the rest of the 1orld received above-average

'() streams! ;ccordingly, spatial and cultural pro?imity alone did not e?plain investment

streams sufficiently, and a more elaborate K model is needed to distinguish bet1een

different types of '()! The follo1ing basic regression eAuation is used to analy*e cross-

country '() streams! 5ost countries are inde?ed f and time is inde?ed t!

ft ft ft u+++++= f +f /#%ft languagedistancedispersionsumOG(C&log"invest β β β β α "%&

%# Balance sheet information could help to overcome this data limitation3 ho1ever, 1e cannot distinguish

bet1een ne1 '() and old stakes in foreign enterprises! )n spite of @@# balance sheet observations regarding the

foreign activities of a company "minority stakes, foreign subsidiaries etc!&, 1e cannot rely on this information!

%>



<odel ; is our basic model that includes the gravity component " sum_'$( &, the

dispersion term "dispersion& that highlights the degree of dissimilarity bet1een Germany and

the host country in terms of market si*e, distance as a pro?y for transportation cost and

cultural differences "distance& and the use of the German language as additional cultural

component "language&! 6e e?tended the basic model ; by inserting 1age differences

"wagediff &, as lo1er 1ages in host countries might trigger cost-driven '(), for firms can

reduce costs by shifting their labor-intensive production to a labor abundant country!%/ 6age

differences can be the result of differences in skill endo1ment, and thus model B considers

the number of patents " patent & as human capital indicator! (ue to an alleged endogeneity bias,

since patents could depend on '(), models to 8 use primary school enrollment rates

"enrollment & as human capital measure!

<odel ( 1as e?tended further by incorporating a measure for protection " protect & to

control for differences 1ith regard to trade policies! ;s our protection measure "custom

revenues divided by imports& might not indicate the degree of disintegration and might be

biased "as discussed in the previous section&, model 8 embeds the value of merchandise

e?ports relative to countrys G(C as an indicator of openness "openness&!

")nsert Table +&

Table + reports the results of the K models 1ith different measures for skill

differences and market integration! The signs of the coefficients are interesting3 as it sho1s

that cost-driven '() 1as important at least in the case of some host countries 1ith lo1 skill

levels measured by primary school enrollment rates "negative values indicate enrollment rates

belo1 the top #@4 of all host countries& and a high 1age gap compared to Germany "positive

values indicate that the real 1age in Germany e?ceeds the real 1age in the host country&!

The standard gravity model ; 1ith the sum of G(C, dispersion factor, distance, and

common language dummy has e?planatory po1er in our country panel! 5enceforth, 1e

%/ 6e collected real 1ages for host countries in the respective year of '() inflo1s!

%



confirmed the presence of '() motivated by hori*ontal strategies9 large host countries

attracted more '() due to their large markets, and the dispersion factor has a strong negative

impact on '() flo1s, 1hich sho1s that countries of similar si*e 1ith regard to their G(C

e?hibited higher '() flo1s! (istance did not have the alleged negative impact on '()3 hence,

it is not a robust predictor for cultural and institutional similarities! ommon language has an

insignificant effect3 thus, there 1as no apparent language barrier! oncerning differences in

human capital and real 1ages, Table + reveals that cost-driven '() 1as relevant for some

countries that had lo1er 1ages compared to Germany "see model B to 8&! )nterestingly, the

number of patents " patent & did not seem to be a robust variable to e?plain skill differences

"see model B&! Catents might be driven by '() flo1s, as firms tried to obtain patents to secure

their markets and to deter market entry! onseAuently, model to 8 replaced patents as

human capital measure by primary school enrollment rates, 1hich e?hibited a significantly

negative impact on '() streams! ountries 1ith lo1 primary school enrollment allo1ed

ceteris paribus complementarities to German high-skilled production factor and hence

attracted vertical activities in some cases!%+

")nsert Table @&

To illustrate our findings, Table @ sho1s partial effects predicted by model 8 "see

Table +& for all countries, lo1 skill countries that e?hibit primary school enrollment rates

belo1 the #@-percentile, lo1 1age countries 1ith the highest 1age gap "top #@-percentile&

and selected countries 1ith the lo1est skill level ")ndia&, highest 1age difference "8gypt& and

high degree of similarity in terms of market si*e "'rance&! )ndia e?hibited the lo1est primary

school enrollment rates, and hence model 8 predicts that 0!/4 of '() should be driven by

skill "+!.4& and 1age differences "+!@4&! 8gypt attracted >!/4 of '() due to the highest

1age gap compared to Germany! Therefore, vertical drivers of '() did not e?plain a

%+ Crimary school enrollment rates also seem to be a good general indicator for the level of development and

industriali*ation, as correlation bet1een enrollment rates and real G(C per capita 1as $!.!

%.



substantial part of investment streams! ountries more similar to Germany, such as 'rance,

sho1ed a lo1 level of '() due to 1age or skill differences, namely $!@4! ;ccordingly, 1e

confirm '() 1as mainly market-driven, 1hich is in line 1ith findings for recent investigation

periods3 ho1ever, to some e?tent 1age and skill differences mattered for '() decisions! )n

addition, protection due to tariff barriers did not influence aggregated '() streams 7 but

market openness "openness& e?hibited a significantly positive effect, albeit the magnitude of

impact 1as negligible "see Table @&! 'urthermore, trade barriers and openness might influence

individual investment decisions, 1hich the follo1ing section analy*es! ompared to arr,

<arkusen and <askus "#$$%&, our K model e?hibited rather lo1 levels of e?planatory

po1er 1ith an adEusted =-sAuared bet1een #$4 and /$43 hence, 1e tested for a potential

omitted variable bias using the =amsey =8S8T test and confirmed that our model

specifications did not suffer from an omitted variable bias! The lo1 level of e?planatory

po1er might be due to using '() streams as dependent variable and not sales of affiliates

"arr, <arkusen and <askus, #$$%&, 1hich are less volatile!

4 "*ic* companies conducted #$I&

;fter analy*ing aggregated '() flo1s, 1e modeled the decision Hto invest or not to investI

from a business perspective3 accordingly, 1e used firm-level data for the 1hole period from

%./ to %0# and tried to e?plain individual '() decisions! Besides tariff barriers " protect &

and market openness "openness) used in the K models "see Table +&, 1e incorporated firm-

specific variables to Auantify firm profitability " ROE &, firm si*e " size& and firms maturity

measured by the year of establishment " foundation&! 6e also considered industry effects

"denoted E& by using conditional "fi?ed-effects& logit models! ;ccounting for the most active

industries in terms of '(), 1e used four main categories "chemical, electric, machinery, and

metal&! Besides these maEor industries, 1e could distinguish / different sub-industries!

%0



onseAuently, 1e run the follo1ing logit models that address the binary decision concerning

individual '()!

it

ft

u

openness

+⋅

+⋅+⋅+⋅+⋅+=

i+

it/it#%ft$ti

foundation

=28si*e protectfdi

β

β β β β α "#&

Table > sho1s the regression output for the basic model 1ithout industry effects "model '&,

the conditional logit model 1ith / industries "model G& and a specification 1ith the four

maEor industries "model 5&! <odel ) also embedded the openness indicator to illustrate the

effect of deglobali*ation in the %0#$s leading to lo1er levels of market integration and the

pro?y for tariff barriers " protect &!

")nsert Table >&

'irm si*e stimulated '() indicating that economies of scale "on the company level&,

easy access to capital markets and other si*e advantages like market po1er 1ere relevant for

conducting '()! ;ccordingly, our findings confirm Brainard "%00/& and <arkusen and

Fenables "%00.& 1ho argue that firm level scale economies drive '()! Crofitability made '()

more likely3 hence, profitable firms could use their factor advantages to enter foreign markets!

Therefore, firm heterogeneity regarding profitability e?plained the propensity to invest

abroad, 1hich is in line 1ith 5elpman, <elit* and eaple "#$$+&! The year of foundation,

sho1s a negative but not significant impact "after controlling for industry effects& on the

probability to conduct '()! Crotection stimulated '() flo1s3 ho1ever, the magnitude of

impact is rather limited and needs further investigation! Therefore, 1e can conclude that

increasing tariffs created incentives for companies to circumvent trade barriers by shifting

their production facilities! et disintegration of markets measured by our openness indicators

seemed to reduce the probability of '(), since openness declined after 66)! To determine

the magnitude of impact of the identified factors and to e?plore firm heterogeneity and

investment patterns, the follo1ing section e?pands our model further!

4+ Heterogeneit, and endogeneit, of #$I decisions- . firm and industr,%specific view

#$



To e?ploit path dependency of individual '(), 1e included an indicator variable " past_fdi&

that takes the value one if the same company conducted '() before "1ithin a ten-year

1indo1& and *ero if the firm did not undertake any '()! 2ut of 0+. '() transactions, +@0

firms conducted '() previously and %#$ firms invested in the same country in the past ten

years! ; positive coefficient of past_fdi 1ould indicate that investments e?hibited path

dependency3 therefore, past '() triggered more '() in the future! ;part from this firm-

specific investment pattern, industry-1ide clustering might affect location and investment

choice! 'ollo1ing the literature on firm heterogeneity and agglomeration effects "Greena1ay

and Kneller, #$$&, the logit model in eAuation / considers industry clusters in the four main

industry categories "metal, machinery, chemical and electric& by counting the number of '()

of firms 1ithin the same industry in the respective host country during the last decade! Table

sho1s our findings and highlights that agglomeration effects did not influence individual '()

decisions 7 but past investments on the individual level made subseAuent investments more

likely!

it

/

/

ft

u+⋅+⋅+⋅+⋅

+⋅+⋅+⋅+⋅+=

∑=

E

+

%

l>it@i+

it/it#%ft$it

ionagglomerat5) pastOfdifoundation

=28si*eopenness protectfdi

χ β β β

β β β β α "/&

<odels J and K e?panded our previous logit model "see eAuation #& by adding the

time pattern of individual '() " past_fdi& or agglomeration effects "agglomeration& for the

four maEor industries "labeled E&! <odel L and < also included the standardi*ed 5erfindahl

inde?, as a measure for market concentration, for the / sub-industries "labeled l&!

onseAuently, model L and < captured individual patterns of '(), agglomeration effects and

industry concentration! The difference bet1een model L and < is that a logit model produced

the results of model L, 1hereas model < 1as based on a probit model! The logit model

allo1s modeling more e?treme probability distributions and hence has thicker tails than the

probit distribution! ;s there is no theoretical argument for preferring either the logit or probit

model, 1e ran both specifications and obtained similar results concerning the significance of

#%



the e?planatory variables! This underlines that our results are robust! ;part from statistical

significance, the directions of influence are the same3 ho1ever, the magnitudes of impact

differ, 1hich is due to the different underlying probability distributions!

")nsert Table &

'irm si*e and profitability had a significant and positive impact on '()! )nterestingly,

the impact of tariff barriers " protect & on '() activity is significant on the individual level 7 but

not on the aggregated level as sho1n in our K models! 5igher tariff barriers stimulated '(),

as firms could circumvent trade barriers, 1hereas lo1er levels of market openness hindered

'()3 therefore, the impact of Ddeglobali*ation on '() decisions is not straightfor1ard to

assess! To illustrate the magnitude of impact of the different firm, industry and country-

specific factors, 1e calculated marginal effects and used the changes in dependent variables to

determine the partial impact on the propensity to invest based on model <! Table . sho1s that

the propensity for conducting '() increased by +!@4 from the pre-%0%+ to the post-%0%+

period, 1hich 1as mainly due to a pronounced increase of industry concentration! 5igher

tariff barriers increased the investment probability by $!/4, 1hereas the lo1er degree of

market openness reduced the investment probability by $!>4!

")nsert Table .&

2+ 3onclusion

6e analy*ed a ne1 database on German '() conducted by Eoint stock companies bet1een

%./ and %0#! 2ur first approach investigated the destination choice of German companies

going abroad by aggregating individual investment streams by host countries and by the time

of investment, 1hich resulted in an unbalanced panel of cross-country investment streams!

'ollo1ing arr, <arkusen and <askus "#$$%&, <arkusen and <askus "#$$%, #$$#& and

Blonigen, (avies, and 5ead "#$$/&, our K model confirmed the predominant role of market-

driven investment strategies already in the late %0th and early #$th century! )nterestingly, 1e

##



also found some evidence for cost-driven '(), as 1age differentials caused '() flo1s to lo1

1age countries! )n addition, differences in human capital measured by primary school

enrollment rates triggered cost-driven '(), since German firms brought their proprietary

kno1ledge 1ith them and hence did not have to rely on human capital of host countries, but

could use complementary advantages! 6e also assessed the effect of distance, both

geographic and cultural, the latter captured by the pro?y common language! et, after

accounting for gravity "in terms of G(C& and dispersion "differences in G(C&, spatial and

cultural pro?imity did not seem to have a pronounced impact on '() streams! 2n the

aggregated level, the increase in trade barriers during the %0#$s, 1hich started a

Ddeglobali*ation period, had no impact on '() streams 7 but lo1er levels of integration

measured by our openness indicator deterred foreign investments!

The market environment concerning tariffs and market openness, ho1ever, influenced

individual investment decisions significantly, although the impact of tariffs seemed to be

positive declining market openness hampered '()! )n total, the effect of tariffs and openness

is minor compared to firm and industry-specific variables! ontrarily, Buch et al! "#$$@&

found that the level of market protection has a negative impact on '()! (ue to the increase in

market concentration in the German home market 7 measured by the 5erfindahl )nde? 7

companies conducted more '()! 'urthermore, our logit and probit models emphasi*ed the

importance of firm si*e and efficiency for '() in the late %0th and early #$th century, 1hich is

in line 1ith 5elpman, <elit* and eaple "#$$+&! Cast firm-specific investment decisions

influenced subseAuent '()3 thus, 1e confirm path-dependency on the firm level, although our

models reEect path-dependency on the industry level! ;s a conseAuence, industry

agglomeration effects did not seem to matter in the period %./-%0#, 1hich differs from

empirical evidence for later periods!

#/



#eferences

;ccominotti, 2! and <! 'landreau "#$$>& (oes bilateralism promote tradeP :ineteenth

century liberali*ation revisited! 8C= (iscussion Caper @+#/!

Baten, J! "%00/& (ie 1irtschaftliche 8nt1icklung der Stadt Lahr im #$! Jahrhundert, in9 Stadt

Lahr, Kaufmann, 8! "ed!&, Geschichte der Stadt Lahr im Sch1ar*1ald /9 >>-0+!

Benavot, ;! and C! =iddle "%0..& The e?pansion of primary education, %.$-%0+$9 trends and

issues! Sociology of 8ducation >>9 %0%-%#$!

Bernard, ;! and J!B! Jensen "%00@& 8?porters, Eobs and 1ages in US manufacturing, %0>7

%0., Brookings Capers on 8conomic ;ctivity! <icroeconomics9 >7%%0!

Blonigen, B!;!, (avies, =!B! and K! 5ead "#$$/& 8stimating the Kno1ledge-apital of the

multinational enterprise3 comment! ;merican 8conomic =evie1 0/9 0.$-00+!

Braconier, 5!, :orbaeck, C!J! and (! Urban "#$$@& =econciling the evidence on the

Kno1ledge-apital <odel! =evie1 of )nternational 8conomics %/9 $-.>!

Brainard, S!L! "%00/& ; simple theory of multinational corporations and trade 1ith a trade-

off bet1een pro?imity and concentration! :B8= 6orking Caper :o! +#>0!

Buch, !<!, Kleinert, J!, Lipponer, ;! and '! Toubal "#$$@& (eterminants and effects of

foreign direct investment9 evidence from German firm-level data! 8conomic Colicy

#$9 @/-%%$!

arr, (!L!, <arkusen, J!=! and K!8! <askus "#$$%& 8stimating the Kno1ledge-apital model

of the multinational enterprise! ;merican 8conomic =evie1 0%9 >0/-$.!

hase, K!;! "#$$+& )mperial protection and strategic trade policy in the inter1ar period!

=evie1 of )nternational Colitical 8conomy %%9 %-#$/!

lemens, < and J! 6illiamson "#$$+& 6hy did the tariff-gro1th correlation change after

%0@$P Journal of 8conomic Gro1th 09 @-+>!

#+

http://web27.epnet.com/searchpost.asp?tb=1&_ug=sid+FA20CAD3-9C96-4559-AB9A-11A6DB19A8E5@sessionmgr4+dbs+ecn+cp+1+DF4F&_us=frn+1+hd+False+hs+False+or+Date+fh+False+ss+SO+sm+ES+sl+-1+dstb+ES+mh+1+ri+KAAACBXC00023654+4267&_uso=hd+False+tg%5B2+-+tg%5B1+-+tg%5B0+-+st%5B2+-+st%5B1+-urban+st%5B0+-braconier+db%5B0+-ecn+op%5B2+-And+op%5B1+-And+op%5B0+-+5DB5&ss=JN%20%22Review%20of%20International%20Economics%22&fscan=Sub&lfr=Lateral&





(avies, =!B! "#$$+& Ta? treaties and foreign direct investment9 potential versus performance!

)nternational Ta? and Cublic 'inance %%9 @-.$#!

Greena1ay and Kneller "#$$& 'irm heterogeneity, e?porting and foreign direct investment!

The 8conomic Journal %%9 %/+-%>%!

5agen, ;! "%00& (eutsche (irektinvestitionen in Grossbritannien, %.%-%0%.! Stuttgart9

Steiner!

5andbuch der deutschen ;ktiengesellschaften "%.0-%0$%, %0%#,%0#&9 issues %.0>Q0, vol )-

))3 %.0Q0., vol )-))3 %.0.Q00, vol )-)) %.00Q%0$$, vol )-))3 %0$$Q$%, vol )-)))3 %0%%Q%#,

vol! ) - ))3 %0#, vol! )-)F, Berlin9 5oppenstedt!

5ead, K!, =ies, J! and (! S1enson "%00@& ;gglomeration benefits and location choice9

evidence from Japanese manufacturing investments in the United States! Journal of

)nternational 8conomics /.9 ##/-#+!

5elpman, 8! "%0.+& ; simple theory of international trade 1ith multinational corporations!

Journal of Colitical 8conomy 0#9 +@% - +%!

5elpman, 8! and C!=! Krugman "%0.@& <arket structure and foreign trade, ambridge, <)T

Cress!

5elpman, 8!, <elit*, <!J! and S!=! eaple "#$$+& 8?port versus '() 1ith heterogeneous

firms! ;merican 8conomic =evie1 0+9 /$$-/%>!

5enderson, 6!2! "%0@& The rise of German )ndustrial Co1er %./+ 7 %0%+! London9 Temple

Smith!

5offmann, 6!G! "%0>@&9 (as 6achstum der (eutschen 6irtschaft seit der <itte des %0!

Jahrhunderts! Berlin9 Springer Ferlag!

Jones, =!6! and 5! Kier*ko1ski "%00$&9 The role of services in production and international

trade9 a theoretical frame1ork, in Jones, =!6! and ;!2! Krueger "eds!&, The political

#@

http://web27.epnet.com/searchpost.asp?tb=1&_ug=sid+FA20CAD3-9C96-4559-AB9A-11A6DB19A8E5@sessionmgr4+dbs+ecn+cp+1+DF4F&_us=hd+False+hs+False+or+Date+fh+False+ss+SO+sm+ES+sl+-1+ri+KAAACBXC00024018+dstb+ES+mh+1+frn+1+95D9&_uso=hd+False+tg%5B2+-+tg%5B1+-+tg%5B0+-+st%5B2+-+st%5B1+-+st%5B0+-davies%2C++r*+db%5B0+-ecn+op%5B2+-And+op%5B1+-And+op%5B0+-+3F1E&ss=JN%20%22International%20Tax%20and%20Public%20Finance%22&fscan=Sub&lfr=Lateral&

http://web27.epnet.com/searchpost.asp?tb=1&_ug=sid+FA20CAD3-9C96-4559-AB9A-11A6DB19A8E5@sessionmgr4+dbs+ecn+cp+1+DF4F&_us=hd+False+hs+False+or+Date+fh+False+ss+SO+sm+ES+sl+-1+ri+KAAACBXC00024456+dstb+ES+mh+1+frn+1+60C2&_uso=hd+False+tg%5B2+-+tg%5B1+-+tg%5B0+-+st%5B2+-swenson+st%5B1+-ries+st%5B0+-head+db%5B0+-ecn+op%5B2+-And+op%5B1+-And+op%5B0+-+9FD4&ss=JN%20%22Journal%20of%20International%20Economics%22&fscan=Sub&lfr=Lateral&


http://web27.epnet.com/searchpost.asp?tb=1&_ug=sid+FA20CAD3-9C96-4559-AB9A-11A6DB19A8E5@sessionmgr4+dbs+ecn+cp+1+DF4F&_us=hd+False+hs+False+or+Date+fh+False+ss+SO+sm+ES+sl+-1+ri+KAAACBXC00024018+dstb+ES+mh+1+frn+1+95D9&_uso=hd+False+tg%5B2+-+tg%5B1+-+tg%5B0+-+st%5B2+-+st%5B1+-+st%5B0+-davies%2C++r*+db%5B0+-ecn+op%5B2+-And+op%5B1+-And+op%5B0+-+3F1E&ss=JN%20%22International%20Tax%20and%20Public%20Finance%22&fscan=Sub&lfr=Lateral&





economy of international trade9 essays in honor of =obert 8! Bald1in, ambridge,

<;9 Black1ell, /%-+.!

Kaiserliches Catentamt "%.@-%0#&, Fer*eichnis der von dem kaiserlichen Catentamt im

Jahre R!!! erteilten Catente, Berlin9 arl 5enmanns Ferlag!

Krugman, C!=! "%00>& (oes Third 6orld gro1th hurt first 1orld prosperityP 5arvard Business

=evie1 #9 %%/-%#%!

Lindert, C! "#$$+& Gro1ing public! ambridge University Cress, ambridge!

Lucas, =!8! "%00$& 6hy doesnt capital flo1 from rich to poor countriesP ;merican

8conomic =evie1 .$9 0#-0>!

<addison, ;! "%00@& <onitoring the 1orld economy %0#$-%00#! Caris %00@!

<addison, ;! "#$$/& The 1orld economy9 historical statistics! (evelopment Studies entre,

28(!

<arkusen, J!=! "%0.+& <ultinationals, multi-plant economies, and the gains from trade!

Journal of )nternational 8conomics /-+9 #$@-##>!

<arkusen, J!=! and ;!J! Fenables "%00.& <ultinational firms and the :e1 Trade Theory!

Journal of )nternational 8conomics +>9 %./-#$/!

<arkusen, J!=! and K!8! <askus "#$$%& ; unified approach to intra-industry trade and foreign

direct investment! :B8= 6orking Caper :o! .//@!

<arkusen, J!=! and K!8! <askus "#$$#& (iscriminating among alternative theories of the

multinational enterprise! =evie1 of )nternational 8conomics %$9 >0+-$!

2=ourke, K! and J!G! 6illiamson "%000& Globali*ation and history9 the evolution of a

nineteenth-century ;tlantic economy! ambridge9 <)T Cress!

Cohl, <! "%0..& 8mil =athenau und die ;8G! <ain*9 v! 5ase u! Koehler!

#>

http://web27.epnet.com/searchpost.asp?tb=1&_ug=sid+FA20CAD3-9C96-4559-AB9A-11A6DB19A8E5@sessionmgr4+dbs+ecn+cp+1+DF4F&_us=hd+False+hs+False+or+Date+fh+False+ss+SO+sm+ES+sl+-1+ri+KAAACBXC00025091+dstb+ES+mh+1+frn+131+0D34&_uso=hd+False+tg%5B2+-+tg%5B1+-+tg%5B0+-+st%5B2+-+st%5B1+-+st%5B0+-markusen%2C++james+db%5B0+-ecn+op%5B2+-And+op%5B1+-And+op%5B0+-+FD66&ss=JN%20%22Journal%20of%20International%20Economics%22&fscan=Sub&lfr=Lateral&


http://web27.epnet.com/searchpost.asp?tb=1&_ug=sid+FA20CAD3-9C96-4559-AB9A-11A6DB19A8E5@sessionmgr4+dbs+ecn+cp+1+DF4F&_us=hd+False+hs+False+or+Date+fh+False+ss+SO+sm+ES+sl+-1+ri+KAAACBXC00025429+dstb+ES+mh+1+frn+1+537E&_uso=hd+False+tg%5B2+-+tg%5B1+-+tg%5B0+-+st%5B2+-+st%5B1+-venables+st%5B0+-markusen%2C++james+db%5B0+-ecn+op%5B2+-And+op%5B1+-And+op%5B0+-+2C3D&ss=JN%20%22Journal%20of%20International%20Economics%22&fscan=Sub&lfr=Lateral&






=ubio, <! "#$$>& Crotectionist but globalisedP Latin ;merican custom duties and trade during

the pre-%0%+ belle poAue! 8conomics 6orking Capers, Universitat Compeu 'abra!

Schaefer, K!! "%00@& (eutsche Cortfolioinvestitionen im ;usland %.$-%0%+! Banken,

Kapitalmaerkte und 6ertpapierhandel im Veitalter des )mperialismus! <uenster9

<uensteraner Beitraege!

Streb, J!, J! Baten and S! in "#$$>& Technological and geographical kno1ledge spillover in

the German 8mpire, %.-%0%.! 8conomic 5istory =evie1 @09 /+-//!

Tilly, =! "%0.#& <ergers, e?ternal gro1th, and finance in the development of large-scale

enterprise in Germany, %..$-%0%/! Journal of 8conomic 5istory +#9 >#0->@.!

6agner, J! and ! Schnabel "%00+& (eterminants of German foreign direct investment9

evidence from micro data! Veitschrift fMr 6irtschafts- und So*ial1issenschaften %%+9

%.@-%0%!

6ilkins, <! "%0.>&9 8uropean multinationals in the United States9 %.@-%0%+, in9

<ultinational enterprise in historical perspective, edited by Teichova, ;!, Levy-

Leboyer, <! and 5! :ussbaum, ambridge, :e1 ork, and <elbourne9 ambridge

University Cress, Caris9 8ditions de la <aison des Sciences de l5omme, @@->+!

6ilkins, < "%0$& The emergence of multinational enterprise9 ;merican business abroad

from the colonial era to %0%+! ambridge9 5arvard UC!

6ilkins, < "%0+& The maturing of multinational enterprise9 ;merican business abroad from

%0%+ to %0$! ambridge9 5arvard UC!

6ilkins, < "%0.0& The history of foreign investment in the United States to %0%+! ambridge9

5arvard UC!

6ilkins, < "#$$+& The history of foreign investment in the United States, %0%+-+@!

ambridge9 5arvard UC!

#



6illiamson, J!G! "%00@& The evolution of global labor markets since %./$9 background

evidence and hypotheses! 8?plorations in 8conomic 5istory /#9 %+%-0>!

#.



Table * Top %$ host countries and their '() share in the pre and post-%0%+ period

pre-%0%+

post-%0%+

UK %+!04 ;ustria %$!/4;ustria %+!+4 Coland 0!%4'rance .!#4 S1it*erland 0!%4US; !$4 *echoslovakia !+4)taly >!>4 US; >!04=ussia >!@4 UK >!/45ungary +!.4 :etherlands >!/4Coland +!$4 )taly @!%4

:etherlands /!%4 Bra*il /!+4Spain /!%4 Spain #!04

#0



Table , Types of '() and modes of entry

<odes of entry Type of '()

:umber Cercent :umber Cercent

Greenfield >.% #4 Sale W production #@0 #48Auity stake #@$ #>4 Sale #@. #4<erger %$ %4 Croduction %# %/4Unspecified %4 =a1 material / +4

Unspecified #> #.4

Total 0+. 0+.

/$



Table / (escriptive statistics

2bs! <ean Std! (ev! Bottom @4 Top @4

Canel ;9 ;ll firms

6ithout '() si*e @0@ %+!@ %!/ %#!% %>!>

=28 @0@ @!. #0! -/%!/ /$!.

foundation @0+ %.0@!+ %>!/ %.>. %0##6ith '() '()Qassets ##% %#!> +$!% $!# />!/

si*e 0+. %@!. %!@ %/! %.!>

=28 0+. %$!% ##!+ -#+!+ /@!>

foundation 0+. %.0+ %>!@ %.$ %0##

Canel B9 hemical industry

6ithout '() si*e $ %+!% %!# %#!# %@!.

=28 $ %%!. %.!0 -%#!@ /!>

foundation >0 %.0%!. %0!/ %.@@ %0#$

6ith '() '()Qassets #$ .!+ %/!@ $!% /0!.

si*e %/% %@!/ %!% %/!. %>!0

=28 %/% %%!. #+!+ -#+!+ /+!#

foundation %/% %.0+!% %>!% %.>@ %0%/

Canel 9 8lectric industry

6ithout '() si*e ./ %+! %!+ %#!. %>!

=28 ./ %$!@ .!0 -#!# #/!

foundation ./ %0$$!# 0 %..+ %0%$

6ith '() '()Qassets /0 @!% 0!# $!% #%!@

si*e #/# %!# %!+ %+!> %.!

=28 #/# %#!@ 0 %!/ #$!>

foundation #/# %.0!@ %/!% %../ %0#

Canel (9 <achinery industry

6ithout '() si*e %$> %+!/ %!+ %%!@ %>!+

=28 %$> #!@ /+!# ->/!/ #.!0foundation %$> %.0!@ %+!. %.% %0##

6ith '() '()Q assets #> @!+ >!. $!# %

si*e %0> %@!+ % %/! %>!.

=28 %0> $!# /#!/ ->/!% /.!>

foundation %0> %.0.!@ %!. %.# %0##

Canel 89 <etal industry

6ithout '() si*e 0> %+!@ %!# %#!# %>!@

=28 0> $! /#!. -+>!/ #+!

foundation 0> %.0.!0 %+!/ %.% %0##

6ith '() '()Qassets #+ %$!# %+!0 $! #>!@

si*e .. %@! %!. %/!/ %.!0

=28 .. %@!@ /$! -%!/ ++!%foundation .. %.0#!. %#!> %.# %0$0

=28 and '() relative to total assets is in percentage points! Si*e refers to the natural

logarithm of total eAuity and is denominated in German current and deflated! The

appendi? contains more details on the construction of variables and data sources!

/%



Table 1 2LS estimation of gravity models on country level

<odel ; <odel B <odel <odel ( <odel 8

countryOfdi countryOfdi countryOfdi countryOfdi countryOfdi

sumOgdp .!0@$XXX %#!#+/@XXX %!%>#0XXX //!>/@XXX #$!0/+XXX

"+!.& "@!/0& "@!/& "+!%>& "+!0%&dispersion $!$$/. -$!/.>#XXX -$!@00$XXX -%!/>+/XXX -$!>@/XXX"$!$+& "-/!>+&XX "-+!#/& "-/!$>& "-/!@@&

distance -$!$$%%XX -$!$$$% -$!$$# -$!$$$. $!$$$+"-#!#>& "-$!/@& "-$!@%& "-%!$.& "%!$.&

language %!#+/% $!#>// $!%%% 0!+$#$ $!@>#"$!#.& "$!$!%%& "$!$>& "%!@@& "$!%>&

patents - $!>/+/ - - -"$!.&

enrollment -$!##%XXX -$!+%+$XXX -$!#.%#XXX"-#!0%& "-/!@>& "-#!.0&

1agediff - %%!/+##XXX %$!##.XXX #.!/@$>XX %!>+%$XXX"#!.#& "#!>& "#!%>& "/!+>&

protect - - - $!#@ -"$!@0&

openness - - - - $!/#%XXX"#!.+&

constant -#$$!#+XXX -#@@!>%/XXX -/>+!$$.XXX -##!#>+XXX -+>@!@.>/XXX"-@!$+& "-@!/& "-@!#& "-+!#+& "-+!0/&

observations %.$ %@$ %+0 %$# %/>adE! =-sAuared $!#%/ $!#>. $!#>/ $!# $!/$$=amsey =8S8T +!%$

"$!$$.&

%$!

"$!$$$&

!/

"$!$$$&

+!+

"$!$$>&

!/>

"$!$$$&=obust t statistics in parenthesesXXX pY$!$%, XX pY$!$@, X pY$!%

/#



Table 2 )llustration of partial effects

;llcountries

Lo1skill

Lo11age )ndia 8gypt 'rance

enrollment %!$4 /!%4 /!$4 +!.4 +!@4 -$!%4

1agediff $!%4 %!>4 #!%4 +!@4 >!/4 $!>4openness %!.4 $!04 $!4 $!/4 $!$4 %!#4

The partial effects refer to e?pected '() streams forecasted by model 8 in Table +! To

illustrate the relative importance of selected e?planatory variables, the partial effects are

e?pressed in percent of the total predicted '() flo1! The assessment is based on average

values of e?planatory variables for all countries, a sub-sample of countries or individual

countries!

//



Table 0 Logit and conditional "fi?ed-effect& logit models

<odel ' <odel G <odel 5 <odel )

Basic model )ndustry-effects

<aEor industries

2pennessmeasure

fdi fdi fdi fdi protect $!$##XX $!$#$X $!$#$XX $!$>>XXX"#!%.& "%!.0& "#!$$& "@!+%&

openness - - - $!%/0XXX">!./&

si*e $!@%XXX $!.+.XXX $!%XXX $!.$$XXX"%#!/@& "%#!$#& "%%!0/& "%%!$&

=28 $!.%XX $!.%@XX $!.@XX %!%.0XXX"#!/0& "#!%$& "#!+$& "/!$+&

foundation -$!$%$X -$!$$+ -$!$$0 -$!$$@"-%!.& "-$!>+& "-%!@& "-$!.$&

chemical - - $!>$0XXX $!@#$XX"#!>@& "#!%.&

electric - - $!$$# -$!$>."$!$%& "-$!/$&

machinery - - $!+#+XX $!/@#X"#!$>& "%!>+&

metal - - -$!>/XX -$!>>0XX"-#!@#& "-#!@/&

constant >!>+# - +! -@!>@/"$!>/& "$!+/& "-$!@$&

observations %$#/ 0. %$#/ 00#

pseudo = # $!%0# $!#$0 $!#$0 $!#+>* statistics in parentheses

XXX pY$!$%, XX pY$!$@, X pY$!%Table > reports coefficients based on logit models!

/+



Table 4 'irm heterogeneity and agglomeration effects

* statistics in parenthesesXXX pY$!$%, XX pY$!$@, X pY$!%

Table reports coefficients based on logit or probit models!

<odel J <odel K <odel L <odel <

fdi fdi fdi fdi

protect $!$>0XXX $!$>@+XXX $!$>##XXX $!$+$>XXX

"@!$& "@!+#& "/!+$& "+!>@&openness $!%0>>XXX $!%/##XXX $!%/..XXX $!$0+>XXX">!.$& ">!+>& "+!$>& ">!$@&

si*e $!+>+%XXX $!.%>XXX $!#+0+XX $!%.0XXX"@!#& "%%!@& "#!@#& +!%/

=28 %!>//0XXX %!%$0XXX %!$$/. $!0#>XXX"/!+.& "#!.& "%!>$& "/!//&

foundation -$!$%@@XX -$!$$>% -$!$%@/ -$!$$0"-#!%& "-%!$>& "-%!>#& "-%!.$&

pastOfdi +!>$+%XXX %!0.00XXX %!>%0XXX"%$!>.& "+!%+& ".!//&

5) $!>.+XXX $!$>>>XXX"%%!$$& "!/%&

;gglomeration effects

chemical $!%$%@XX $!$/%. $!$/#/"#!>#& "$!@$& "%!%#&

electric -$!$$./ -$!$>@. -$!$/0>"-$!/%& "-%!%0& "-%!@&

machinery $!$/#+ -$!$#/ -$!$$+"$!.>& "-$!/0& "-$!#&

metal -$!%0.XX -$!%$>+ -$!$/>"-#!>0& "-%!$$& "-%!@+&

constant %.!+0./ -#!+> #$!0>%# 0!0@/"%!/#& "-$!##& "%!%+& "%!%>&

observations 0#. 00+ 0#. 0#. p-seudo = $!++@ $!#+> $!# $!+0>

/@



Table 5 The impact of protection and openness on individual '() decision

Sensitivity

Falue pre-%0%+

Falue post-%0%+

hange )mpact

protect $!$$%> %#!+. %+!@> #!$ $!/4open $!$$+$ 0!@@ !0. -%!@ -$!>4si*e $!$$. %@!%0 %+! -$!+# -$!+4roe $!$+/$ $!%/ -$!%% -$!#+ -%!$4foundation $!$$$+ %.0/!%0 %0%>!.+ #/!>@ $!.4

pastOfdi $!$@.$ $!/ $!+ $!%$ $!>45) $!$$/$ >!+> #/!@$ %!$+ @!#4

;gglomeration effects

chemical $!$$%/ $!. $!%# -$!>> -$!%4electirc $!$$#$ %!/ $!// -%!$+ -$!#4machinery $!$$$+ $!./ $!@0 -$!#+ $!$4

metal $!$$/# $!+/ $!#% -$!## -$!%4Total impact on propensity to invest +!@4

Table . decomposes the change in the propensity to invest based on the probit model < in

Table into partial impacts! Sensitivities refer to marginal effects3 hence, the sensitivity

indicates the change in the propensity to invest due to a change of the independent

variable by one unit! ;s sensitivities change 1ith the level of the respective independent

variable, one has to regard the figures in Table as an appro?imation to illustrate the

magnitude of partial impacts!

/>



Fig+* The most important industries and their share in '(), %./-%0#

Share in %

/





Appendi6) $ariable definitions and data sources

Endogenous -ariables9

log

0invest /t )

:atural logarithm of aggregated '() flo1s bet1een Germany and the respective host

country "inde?ed E& at time t! 6e convert '() flo1s into German currency and deflateusing 5offmanns price inde?3 thus, '() is e?pressed in real terms!(ata source9 ;mount of investment 1as deflated, using 5offmanns price inde?"%0>@, p!>$%, col!%@&!

fdiit 6e measured the binary choice of individual '()3 hence, the dummy variable takesthe value one if firm "inde?ed i& conducted '() in year t!

E6planator7 -ariables for t8e 93 model&)

sum_gdp 6e calculated the sum of Germanys and the host countrys G(C9 ln"G(C f &Zln"G(C5ome&! (ata source9 <addison "%00@&!

dispersion To indicate the degree of similarity in terms of market si*e, 1e defined the dispersioninde? as follo1s9 "ln"G(Cf &-ln"G(C5ome&&#! (ata source9 <addison "%00@&!

distance (istance bet1een Berlin and the host countrys capital in km(ata source9http9QQ111!macalester!eduQresearchQeconomicsQC;G8Q5;F8<;:QTrade!=esourcesQ(ataQGravityQdist!t?t

language The dummy variable takes the value of if German is the official language of the hostcountry and *ero other1ise!

patents The variable is defined as the number of foreign patents, 1hich have been e?tended"so-called high-value patents&, of companies of the host country in Germany! This

indictor should highlight the degree of technological progress of the host county! (atasource9 Kaiserliches Catentamt "%.@-%0#&!

enrollment 6e used primary school enrollment rates in percent! To measure the skill difference,1e calculated the difference of the host countrys enrollment rate and the top #@4 ofall host countries "enrollment &3 thus, a negative value indicates that the respectivehost country has lo1er enrollment rates compared to the leading #@4 of hostcountries! (ata source9 Benavot and =iddle "%0..& and Lindert "#$$+&

wagediff 6e determined the difference in real 1ages of the host country compared toGermany3 ln"1age5ome&-ln"1agef &! 5ence, a positive value indicates that the hostcountry has lo1er real 1ages than Germany! (ata source9 6illiamson "%00@&!

protect 6e measured protection by the average tariff rate, 1hich is defined as the ratio ofcustom revenues to the total value of imports! (ata source9 ;ccominotti and'landreau "#$$>&, lemens and 6illiamson "#$$+& and =ubio "#$$>&!

openness 6e determined the ratio of the value of merchandise e?ports relative to G(C!Standard practice is to add imports and e?ports and divide by G(C 7 but 1e do nothave reliable import data for %./-%0#! 6e use <addisons "#$$/& data for %.$,%0%/ and %0#0 on merchandise e?ports to construct the openness indicator "e?portsQG(C&! (ata source9 <addison "#$$/&!

Additional e6planator7 -ariables for logit:probit models&) size 6e used the natural logarithm of total eAuity as measure for firm si*e

/0

/0



ROE =eturn on eAuity as a measure of firms profitability 1as defined as net income"distributable to common shareholders& divided by total common eAuity "e?cluding

preferred eAuity&! foundation ear of establishment 1hen Eoint stock company entered share register!

past_fdi )ndicator variable 1hich takes the value one if the firm invested abroad during thelast %$ years and *ero other1ise!

HI The standardised 5erfindahl )nde? sho1s the level of market concentration in therespective industry! 6e determined the sum of sAuared market shares "SS<S& basedon total eAuity "as industry revenue figures are not available& and standardi*ed asfollo1s9 "SS<S-min"SS<S&&Q"ma?"SS<S&-min"SS<S&&X%$$!

c*emical1

electric1

mac*iner,

1

metal

Used as industry dummy variables, e?cept in table as 1e determined industryagglomeration effects defined as the number of '() conducted by the industry 1ithinthe last decade in the respective host country!

The individual industries listed in the source on 1hich have companies 1ere9

asphalt

baths

banks

real estate

bre1eries

cement

chemicals

iron and coalelectrotechnical

dyes

other finance

gas

grain mills

glass

other ra1 material

rubber

colonial goods

other food

machineryQenginerring

metal processing

musical instruments

food processing

oil millsoptical instruments

+$

+$



paper and printing

petrol

porcelain

gun po1der

chocolate

other

toys

starch

te?tiles

rail1ays and urban commuting

clocks

other transportsinsurance

<ain data source "e?cept mentioned in the table above&95andbuch der deutschen ;ktiengesellschaften "%.0-%0$%, %0%#, %0#&!

+%

FDI in Germany

Documents