Upload
gervais-mills
View
218
Download
0
Tags:
Embed Size (px)
Citation preview
Foundation for the Advancement of Economics andFoundation for the Advancement of Economics and Friedrich Ebert Stiftung (Belgrade) lecture series Friedrich Ebert Stiftung (Belgrade) lecture series
The impact of European The impact of European FDI FDI in the EU 'neighbourhood': in the EU 'neighbourhood':
relative size, geographical scale and relative size, geographical scale and spatial selectivityspatial selectivity
Vassilis Monastiriotis (with Mireia Borrell)Vassilis Monastiriotis (with Mireia Borrell) LSEE – LSE Research on South Eastern Europe European Institute, London School of Economics (Contact: [email protected])
Faculty of Economics, University of Belgrade, 10 September 2013
IntroductionIntroduction
General motivation / context
EU “approximation” a ‘policy distortion’ Preferential trade, trade re-orientation, institutional convergence
But is the “EU influence” good?
Often, these advantages are assumed (‘given’), even in the context of ‘conditionality without accession’ (!)
Here, we examine one only aspect of the ‘economic’: FDI spillovers One aspect but not a side-issue: impacts on domestic productivity
and thus on industrial structure and international competitiveness Because of FDI’s contribution to political transition (Grabbe, 2006; Bevan/Estrin, 2004) Because of its contribution to economic development (Markusen/Venables,1999)
Why should we care? Iff the ‘EU influence’ is not unidirectional, then there is a role
(and responsibility) for the EU to correct and/or compensate
Institutional(capacities)
Political(democratisation)
Economic(modernisation/growth)
IntroductionIntroduction
FDI in ENP (vs in CEE)
In CEE, prospect of accession paramount in mobilising foreign investments
Part of changing geographical organisation of production (spatial DoL)
Part of a wider process of restructuring for the European industry
Deeper integration process, stronger / more organic linkages => stronger technology transfers to domestic firms
In ENP (and SEE?) less FDI and with different motives
Less integration, more market capture & resource ‘expropriation’
Still, pref’l framework a motive for more organic / less speculative FDI
IntroductionIntroduction
Research questions Does EU FDI have positive productivity spillovers?
To justify the FDI attraction policies Does EU FDI have a spillover advantage over FDI of other origins?
To justify the ‘approximation’ policies Does (EU) FDI have different impacts in ENP than in CEE?
Would suggest need to combine openness w/ deeper/political integration… Does (EU) FDI create spatial distortions despite its overall +ve effects?
Would necessitate corrective policy interventions
Empirical analysis Firm-level data (10k+) from BEEPS (28 countries, 2002-2009) Production-function approach to estimate direct and intra-industry
productivity spillovers, looking in particular at:
How these vary across country groups How they vary by origin The geographical scale of spillovers (nat’l/sectoral v reg’l/localised) The spatial selectivity of spillovers (capitals v the rest)
IntroductionIntroduction
Structure
Introduction
Considerations for the analysisEmpirics – theoretical considerations – data & model
Empirical analysis I: impact of FDI by origin and recipient groupEuropean vs non-European spillovers in ENP, SEE & CEE
Empirical analysis II: scale and location of FDI impactsExamining localisation of spillovers and capital-city effects
Conclusions and policy discussion
Considerations for the analysisConsiderations for the analysis
What we know (empirics) Vertical / inter-industry spillovers / stronger / more positive
Damijan et al, 2003; Gorodnichenko et al, 2007; Nicolini/Resmini, 2010 Horizontal often non-significant or negative (also in LDCs)
Konings, 2001; Javorcik, 2004; Sabirianova et al, 2005 Horizontal (and vertical) conditional / contextual
Firm/sector characteristics (firm size – Pojar, 2012; absorptive capacity / technological
distance – Halpern/Murakozy, 2007; sector/region – Monastiriotis/Jordaan, 2010)
Recipient country characteristics (level of development, corruption, political regime –
Tytell/Yodaeva, 2005) Characteristics related to the foreign investors
(origin, extent of ownership, export-orientation – Javorcik/Spatareanou, 2011)
Significant selection (Jordaan, 2013) & hysterisis (Monastiriotis/Alegria, 2011)Damijan et al, 2013: an
extensive study of effects and conditioning factors
Here, interest is in the geography of spillovers, within a context of political-institutional-economic approximation
Considerations for the analysisConsiderations for the analysis
On origin and geography
OriginExisting research: different processes and conceptions of ‘proximity’
J&S: distance => local sourcing => vertical spillovers M&A: cultural-technological proximity => more ‘embeddedness’,
greater scope for / more absorbable spillovers (vertical – horizontal)Here: political-inst’l proximity (‘approximation’); more productive due to:
FDI part of ‘strategy’ to strengthen local capacities/markets local links/synergies Local firms more keen to ‘engage’/compete with EU investors more ‘learning’ Institutional approximation (transposition) facilitating/forging such links
larger / more absorbable spillovers
Scale and spaceFDI literature not particularly ‘theoretical’ on this – nor too ‘empirical’ either
Focus on sectors and the nature of spillovers (eg, pecuniary – technological)‘Spatial’ theory directly from ‘economics of agglomeration’ literature
Knowledge spillovers are localised; MAR & Jacobs externalities / economiesSome work on impact of agglomeration (‘local capacities’ / absorption)
Driffield/Hughes, 2003; Haskel et al, 2007; Girma/Wakelin, 2009 (UK); Mullen/Williams, 2007 (USA); Sgard, 2001 (HU); Jordaan, 2008 (MEX); Monastiriotis/Jordaan, 2010 (GR)
Considerations for the analysisConsiderations for the analysis
Data
Business Environment and Enterprise Performance Survey (BEEPS) EBRD/World Bank survey on firms / business environment Unbalanced panel from three waves – 2002, 2005 and 2009 Survey contains 28k obs (22k/10k after cleaning); 28 transition countries Sample sizes b/w 82 (MG) and 1,000+ (BG, UK, RU, TK, PL, RO, CR) Info on sales, employment, fixed assets, share & nationality of foreign
ownership, share of exports, sector (2-digit NACE), region, etc (info on foreign presence/nationality in cases projected from x-sectional survey)
Independent variable – Foreign share within sector (horizontal) Based on firm-level foreign shares: share of output attributable to
foreign ownership (within sector-year or sector-region-year) Output-based measure, favoured over E-based due to pvity diffs Using 10% threshold for decision not to include to group of
domestic firms – other thresholds (min, maj, full) similar results
Descriptive statistics (next slide)
Considerations for the analysisConsiderations for the analysis
Considerations for the analysisConsiderations for the analysis
Table 2. Sample sizes by country (group) and year
Considerations for the analysisConsiderations for the analysis
Table 3. Foreign ownership shares by country (group) and year
Considerations for the analysisConsiderations for the analysis
Table 4. FDI inflows as share of GDP (UNDP data)
Considerations for the analysisConsiderations for the analysis
Method
Production-function approach – total factor productivity effects Sales (Y) a function of labour (E) and capital (FA) Policy variable (spillovers): share of foreign presence (sector/region) Controls (selection etc): sector and country dummies; firm FEs
Estimating model
where y is output (sales) of firm i in sector s of country c in year t; k is capital (fixed assets); l is employment; h is our measure of horizontal FDI; Dc and Dt are vectors of binary dummies for countries and years; e is a normally distributed error; and the b’s are parameters to be estimated
Estimation considerations Clustered standard errors, OLS and FE (within) estimations FEs reduce sample and heterogeneity: only indicatively here No correction for selectivity of capital – little sectoral differences
Aggregate effects
Empirical resultsEmpirical results
Dependent: log(sales) (1) (2) (3) (4) (5) (6) Employment (log) 0.544*** 0.333*** 0.831*** 0.853*** 0.566*** 0.329*** (0.023) (0.056) (0.013) (0.013) (0.023) (0.061) Fixed assets (log) 0.503*** 0.635*** 0.189*** 0.182*** 0.508*** 0.625*** (0.017) (0.017) (0.019) (0.018) (0.021) (0.019) Foreign share (own) 0.198*** 0.181
(0.047) (0.284) Foreign share (sector) -0.0085 -0.213 -0.206** -0.126** 0.694* 2.328*** (0.088) (0.201) (0.074) (0.056) (0.385) (0.659) Foreign share squared (sector)
-0.955* -3.631*** (0.461) (0.884)
Constant 9.813*** 5.248*** 21.06*** 20.87*** 9.171*** 4.774*** (0.435) (0.307) (0.642) (0.569) (0.416) (0.412) Fixed effects
Country, Year
Firm, Year
Country (x) Year
Country (x) Year, Sector
Country, Year,
Sector
Firm, Year,
Sector Observations 10,696 10,696 9,313 9,313 9,313 9,313 R-squared 0.868 0.598 0.922 0.929 0.872 0.634
Direct (own) effect strong and positiveBut not when controlling for selectionEither way, external (horizontal) effect negative and non-significantBut significant with interactive country-year FEs (national trends / country-specific temporal effects)Effect smaller with sector-FEs, suggesting, if anything, selection into low-productivity sectors
But significant non-linearity (also with firm-FEs) hump-shaped effect, positive for low FDI concentrations, negative for high ones
Non-linearity
Empirical resultsEmpirical resultsPredicted total effects on firm productivity (vertical axis) for different values of sectoral shares of foreign presence (horizontal axis) based on alternative estimation equations
Col. 5
Col. 6
Stock-taking – aggregate analysis
Overall, in our sample countries, foreign investments offer little benefits to the sectors in which they locate (in terms of spillovers)
There is a direct positive effect for the firms that receive the foreign investment – although, still, this may be subject to self-selection
Some evidence of positive spillovers in sectors where foreign presence is (positive but) at rather low levels – although country differences in temporal dynamics may exaggerate this effect
Firms in sectors that are dominated by foreign-owned production (well above 50%) show in fact lower productivity
When restricting the sample to surviving domestic firms only(FE estimation) the effect becomes more negative
This suggests that this effect is not compositional (e.g., creaming-off via take-overs or crowding-out via exit) but a pure negative externality
Empirical resultsEmpirical results
Impact of origin / destination
Empirical resultsEmpirical results Hypothesis 1: origin matters / EU advantageHypothesis 2: destination matters
H1 confirmed: EU FDI significantly more advantageousAlthough insignificant with country-specific time effects, difference from non-EU FDI remains (and is significant)
Difference remains also in FE & non-linear specifications (EU FDI has hump-shaped effect, non-EU linear negative) Impact negative insignificant in CEE: benefits exhausted?
But very strongly hump-shaped (essentially always positive) in SEE: lack of domestic capital/technology base?
In ENP the effect highly insignificant: suggesting perhaps a capacity (absorption) issue?
(1) (2) (3) (4) (5) (6) (7) (8)
ALL ALL CEE SEE ENP CEE SEE ENP
Employment 0.566*** 0.853*** 0.793*** 0.397*** 0.607*** 0.793*** 0.384*** 0.609***
(0.009) (0.013) (0.026) (0.045) (0.042) (0.027) (0.044) (0.043) Fixed assets 0.508*** 0.182*** 0.253*** 0.704*** 0.477*** 0.254*** 0.710*** 0.476***
(0.006) (0.018) (0.027) (0.059) (0.029) (0.028) (0.060) (0.030) Horizontal (total)
-0.0461 3.130*** -0.135
(0.599) (1.000) (0.566)
Horizontal squared (total)
-0.0309 -3.437*** 0.0590
(0.689) (1.052) (0.682) EU horizontal 0.362*** 0.0006
-0.0721 1.308** 0.567*
(0.093) (0.047)
(0.131) (0.576) (0.282) Non-EU horizontal
-0.267* -0.295***
-0.335** 1.003 -0.0427
(0.139) (0.094) (0.153) (1.490) (0.331) Constant 9.179*** 20.84*** 8.265*** 5.171*** 8.234*** 8.271*** 5.266*** 8.179*** (0.155) (0.575) (0.313) (1.272) (0.456) (0.309) (1.277) (0.429) Fixed effects Country,
Year, Sector
Country (x) Year, Sector
Country, Year,
Sector
Country, Year,
Sector
Country, Year,
Sector
Country, Year,
Sector
Country, Year,
Sector
Country, Year,
Sector Observations 9,313 9,313 4,233 1,581 3,499 4,233 1,581 3,499 R-squared 0.872 0.929 0.913 0.876 0.867 0.913 0.876 0.867
Hypothesis 3: EU role by destination
In all cases productivity advantage for EU FDI: - significantly positive in the SEE and ENP, where
non-EU is non-significant (but +ve in SEE) - EU negative non-significant in CEE, compares favourably to non-EU (significant negative)
Stock-taking – role of European FDI
Overall, the results offer support to the underlying hypothesis that the origin of FDI matters, not only in itself but also in relation to the recipient country
In the CEECs, where levels of development are comparatively high and where economic openness happened earlier / faster, the benefits
from FDI, in an intra-industry sense, seem to have been exhausted – for both EU and non-EU FDI
Instead, in the SEECs, where the EU has been playing a pivotal role for political-economic stabilisation and development, the productivity
effects of European FDI are exceptionally strong and much more significant than spillovers from FDI of other origins
Finally, in the ENP region, where EU’s involvement is also preferential but much less significant, or influential, European FDI produces smaller spillovers, albeit still more beneficial ones compared to FDI from other regions
Empirical resultsEmpirical results
Impact of geography / scale
Empirical resultsEmpirical results (1) (2) (3) (4) (5) (6) (7) Employment 0.853*** 0.853*** 0.853*** 0.853*** 0.853*** 0.853*** 0.853*** 0.854***
(0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Fixed assets 0.182*** 0.182*** 0.182*** 0.182*** 0.182*** 0.182*** 0.182*** 0.182***
(0.018) (0.018) (0.018) (0.018) (0.018) (0.018) (0.018) (0.018) FP (region-sector) 0.0197 0.515***
(0.067) (0.166) FP^2 (region-sector) -0.727***
(0.196) EU FP (region-sector) 0.0962
(0.075) Non-EU FP (region-sector)
-0.157 (0.117)
Interactions Capital regions (x) …
FP (sector) 0.0900 0.813*** (0.121) (0.274)
FP (sector-region) 0.193* 1.002*** (0.102) (0.187)
FP^2 (sector) -1.154*** (0.317)
FP^2 (sector-region) -1.233*** (0.244)
EU FP (region-sector) 0.215** (0.091)
Non-EU FP (region-sector)
-0.097 (0.152)
Other regions (x) … FP (sector) -0.212*** -0.140
(0.061) (0.209) FP (sector-region) -0.098 0.172
(0.073) (0.203) FP^2 (sector) -0.051
(0.260) FP^2 (sector-region) -0.369
(0.250) EU FP (region-sector) 0.0094
(0.094) Non-EU FP (region-
sector) -0.194 (0.136)
Constant 20.82*** 20.77*** 20.81*** 20.88*** 20.86*** 20.87*** 20.83*** 20.84*** (0.583) (0.586) (0.580) (0.564) (0.575) (0.564) (0.579) (0.579)
Observations 9,313 9,313 9,313 9,313 9,313 9,313 9,313 9,313 R-squared 0.929 0.929 0.929 0.929 0.929 0.929 0.929 0.929
Is the absence of strong results due to our focus at the national-sectoral level?
Are spillovers (more) localised?
As before, strong evidence of non-linearity (across specifications): concentration matters (adversely)
But the effect is flatter, suggesting little localisation and probably evidence of spatial heterogeneity
Impact of geography / scale
Counter to expectations, no localised effects either
Effect (at least) positive, but insignificant(also with firm-FEs, non-interactive dummies, etc)
The hump-shaped effect is also consistently flatter for the three country blocks (and only significant in SEE)
But note that the region-wide measure (not region-sector) returns strong positive spillovers
Spillovers are inter-industry (region-wide): urbanisation, not localisation, effects
EU FDI still more beneficial, but effect non-significant: by all evidence, intra-industry (horizontal) spillovers are not localised
But for intra-industry at least, the effects appear to be heterogeneous in space: the negative effect is driven by domestic firms in the periphery
EU FDI still more beneficial, but effect non-significant: by all evidence, intra-industry (horizontal) spillovers are not localised
In the case of the region-specific measure, FDI returns a positive spillover in capital regions
Spillover effect more positive everywhere, suggesting that localisation matters, but rather differently in different types of regions
The picture is similar when we examine non-linearities:
For capital regions, strong hump-shaped spillovers and stronger with co-location (spillovers are localised)
The picture is similar when we examine non-linearities:
For capital regions, strong hump-shaped spillovers and stronger with co-location (spillovers are localised)
For other regions, spillovers are insignificant / negative (although still less so with co-location)
These patterns appear to be strongest in CEE and weakest in SEE (not shown):
Especially in the CEECs, negative spillovers / exhaustion of benefits comes from the peripheral regions
Concerning the origin distinction:
EU FDI more advantageous in both types of regions
FDI of any origin more advantageous in capital regions
EU FDI in capital regions produces clear positive spillovers
(1) (2) (3) (4) (5) (6) (7) Employment 0.853*** 0.853*** 0.853*** 0.853*** 0.853*** 0.853*** 0.853*** 0.854***
(0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Fixed assets 0.182*** 0.182*** 0.182*** 0.182*** 0.182*** 0.182*** 0.182*** 0.182***
(0.018) (0.018) (0.018) (0.018) (0.018) (0.018) (0.018) (0.018) FP (region-sector) 0.0197 0.515***
(0.067) (0.166) FP^2 (region-sector) -0.727***
(0.196) EU FP (region-sector) 0.0962
(0.075) Non-EU FP (region-sector)
-0.157 (0.117)
Interactions Capital regions (x) …
FP (sector) 0.0900 0.813*** (0.121) (0.274)
FP (sector-region) 0.193* 1.002*** (0.102) (0.187)
FP^2 (sector) -1.154*** (0.317)
FP^2 (sector-region) -1.233*** (0.244)
EU FP (region-sector) 0.215** (0.091)
Non-EU FP (region-sector)
-0.097 (0.152)
Other regions (x) … FP (sector) -0.212*** -0.140
(0.061) (0.209) FP (sector-region) -0.098 0.172
(0.073) (0.203) FP^2 (sector) -0.051
(0.260) FP^2 (sector-region) -0.369
(0.250) EU FP (region-sector) 0.0094
(0.094) Non-EU FP (region-
sector) -0.194 (0.136)
Constant 20.82*** 20.77*** 20.81*** 20.88*** 20.86*** 20.87*** 20.83*** 20.84*** (0.583) (0.586) (0.580) (0.564) (0.575) (0.564) (0.579) (0.579)
Observations 9,313 9,313 9,313 9,313 9,313 9,313 9,313 9,313 R-squared 0.929 0.929 0.929 0.929 0.929 0.929 0.929 0.929
Stock-taking – geography of spillovers
Evidence of localised spillovers particularly weak – e.g., in quadratic model hump-shaped curve much flatter than in sectoral analysis
Region-wide FDI, however, produces strong positive spillovers (vertical / inter-industry spillovers; urbanisation effects)
But these effects are conditional on space / heterogenous: capital regions consistently benefit more / suffer less (esp. in CEE) hump-shaped non-linearity essentially only in capital regionsconditioning on space, localisation effect becomes larger
Implication: even when overall effect is positive (moderate concentration, European origin, etc), foreign presence has a detrimental spatial
effect as it tends to increase productivity differences b/w capital-city regions (which already possess development/agglomeration advantages) and the rest
Effect not due to concentration of FDI in capitals (see result stability in cols.6 & 7)But given this concentration, the spillover advantage observed in capital regions
implies necessarily that foreign-frim presence magnifies regional disparities in the host economies (at least those of our sample)
Empirical resultsEmpirical results
Summary and conclusionsSummary and conclusions
Starting points Knowledge/empirics of FDI spillovers extensive/consolidated – but little
emphasis on supra- and sub-national geographies and political contexts Here, examined the ‘origin & geography’ issue within context of EU external
relations (neigh/hood policy): inst’l approximation – enhanced econ flows Research questions: Is it good? Is it worth it? Is it cost-free (or distortive)?
Main findings / key highlights (more, in the paper…) EU FDI has ‘productivity advantage’ in ‘neighbourhood’ (more +ve / less –ve)
If EU MNCs not systematically more productive, policy a likely source of advantage FDI spillovers in ENP not maximised (countries not sufficiently ‘developed’?)
More ‘approximation’ can bring benefits as in CEE (past) and SEE (today) Spillovers not too localised but stronger in capital regions
Agglomeration matters; but spatially distortive / enhancing regional disparities Conclusions for policy
Approximation process and ‘gravitational pull’ create an inadvertent reorientation of trade and economic / production structures With it, come costs: productivity spillovers that are often negative; spatial
distortions producing more inequality / geographical differentiation Role for (EU) policy to correct/compensate : existing policies and assistance
instruments to obtain spatial focus; new ones to address spatial imbalances
About LSEE . Research unit at the European Institute, LSE
bringing together LSE’s expertise on SEE, complementing the work of the HO and the
Contemporary Turkish Studies Programme
Aims to provide a platform for high quality, independent research & facilitate research collaborations and public dialogue on SEE, including via academic visits & public events
LSEE has developed an extensive network of research and institutional collaborations with a range of academic, governmental and international bodies, including the EC and the RCC.
LSEE researchers and research affiliates are engaged in a broad range of in-house and externally funded projects in the region, along three broad themes:
Social Cohesion (social policy, regional policy, labour markets)Macroeconomy (European integration, institutional reform, growth)Security/Minorities (international relations, state-building, civil society)
http://www.lse.ac.uk/LSEE