Title: Location Choices of the Pharmaceutical Industry in Europe after 1992
1Location Choices of the Pharmaceutical Industry
in Europe after 1992
- Prof. Frances P. Ruane
- The Economic and Social Research Institute,
Ireland and The Institute for International
Integration Studies, Trinity College Dublin,
Ireland - Xiaoheng Zhang
- The Institute for International Integration
Studies and the Department of Economics, Trinity
College Dublin, Ireland
2Context
- The Single Market Programme (1st, January, 1993)
removed the non-tariff barriers between EU Member
States to allow free movement of goods, capital,
people and services. - Multinationals rationalize their production by
consolidating production facilities within a
country or across countries to fully utilize
economies of scale. - Pharmaceuticals become a focus because it is an
important industry to the EU economy and European
people. - We expect its significant response to the Single
Market because this industry is able to benefit a
lot from rationlization due to high increasing
returns to scale.
3Geographic Concentration Trend
- Measure the geographic concentration of
pharmaceutical production across 14 EU countries
using Theil Index and Location Gini coefficient
(1993 to 2002) - - Data source OECD STructural ANalysis
(STAN) database - Pharmaceutical
production and employment at country level -
where is the production in country i
in the country set that under investigation, and
n is the number of countries. Location Gini
coefficient of concentration is defined as the
area between the Lorenz curve and 45 degree line
in a space where , the pharmaceutical
production share of country i in the data set
that under investigation, is cumulated on the
Y-axis and the number of countries cumulated on
the X-axis with equal interval of width 1/N.
Countries are ranked by .
4- Theil Indices of Geographic Concentration of
Pharmaceutical Production EU15 - (OECD STAN data, EU 15, exclude Luxembourg)
5- Location Gini coefficient
- (OECD STAN data, EU 15, exclude Luxembourg)
6Theoretical explanations of the
agglomeration/dispersion
- New Economic Geography (NEG) theories
- Theories to analyse the spatial
distribution of the economic activities between
two or more regions. The subject is increasing
returns to scale industry. - Krugman (1991), Venables (1996), Baldwin
(1997), Baldwin (2002), Puga (1999) - Two different predictions on the relationship
between trade costs and agglomeration.
7- Krugman (1991) Monotonic relationship
Puga (1999) Bell-shaped curve - X-axis trades costs level
- Y-axis share of the industry in each of two
regions
8Implication to the EU Single Market and the
Pharmaceutical industry
- Dispersion trend in the pharmaceutical industry
and low trade cost imply that the agglomeration
process of this industry may be at the left half
of the Bell-shaped curve high wages and
congestion in the agglomerated region drive the
industry to the less agglomerated regions. - Empirical question
- - What are the determinants of
pharmaceutical multinationals location choice? - - Main focuses
- Country level agglomeration
Corporate tax rate Market size
9- A Discrete-choice Framework (I)
- Multinationals choose a country from a set of
alternative countries to expand their production
or build up new facilities. Selected country is
supposed to be able to maximize the
multinationals profit. Profit depends on the
observable attribute of the alternative
countries. - The Conditional Logit Model (CLM) McFadden
(1974) -
-
-
-
10- The problems with CLM
- - Simple but restrictive assumption on error
term - ?The ratio of probabilities of any
two alternatives being - chosen is independent on any
other alternatives. This is - called Independence from
Irrelevant Alternatives (IIA). -
- ? Individual taste behaves as an
individual effect ? - correlation between error
terms of alternatives ? violation - of IIA and inconsistent ML
estimation - - Not able to accommodate complicated
individual structure in our - case several location choices made by
the same MNE
11- A discrete-choice framework (II)
- The Mixed Logit Model (MXL)
Train (2003) - Rabe-Hesketh et al. (2004)
-
-
- - Coefficient follows a normal
distribution (random effect) - - Control for MNE parent-Subsidiaries
hierarchy -
12Data
- Subjects Pharmaceutical MNEs subsidiaries in
11 out of EU15 countries - Data source Amadeus data Collection of
European firms accounts - Samples High-performance sample 224 existing
- pharmaceutical firms
experiencing - above-median expansion of
turnover b/w 1995 - and 2003
- New firms sample 119 firms
that were - established after 1993
13Major Explanatory Variables
14Empirical Models
- High-performance Sample
- New-firm Sample
- For both CLM and MXL. Only agglomeration
variables, tax rate and market size are treated
as random-effect variables.
15Results High-performance Sample
- Effective tax rate, market size, agglomeration of
the - pharmaceutical industry matter.
- Firm heterogeneity shows up through interaction
terms. - Hausman test rejects IIA for Germany, Portugal,
Spain and Sweden if they are excluded. - CLM and MXL show
- similar results.
-
-
16Results New-firm Sample
- Only distance to Brussels and familiarity matter.
- Firm heterogeneity isnt found.
- Hausman test cannot rejects IIA
- CLM and MXL show
- similar results.
-
-
-
17Future Improvement
- Endogeneity in estimation of High-performance
sample - Petrin and Train (2002) a control
function approach - Lewbel (2004) very
exogenous variable approach - Adding variables to the models to test the
assumptions of NEG models - Krugmans assumption inter-region labour
mobility ? use skilled pharma workers in
neighbouring countries to proxy potential labour
flow - Venerable/Pugas assumption intra-region
labour mobility ? use workers in Chemical
industry in the same country to proxy potential
intra-region labour flow -
18Conclusions
- Evidence is found to support Puga and Venables
models of a non-monotonic relationship between
industrial agglomeration and trade costs. - The expansion in production at existing plants in
Europe may contribute to Europe-level geographic
dispersion of pharmaceutical production. - The use of the conditional logit model in this
research is justified by comparing its
performance with those of the mixed logit models.
-
- Any comments and critiques are welcome!