Volume 9 • 2022 • Number transnational corporations investment and development

Table 3a. Correlation matrix for key variables, 1997–2007
Variable
All manufacturing industries
Home welfare 
spending
Firm size
Labour 
intensity
Unit labour 
costs
Host welfare 
spending
Home welfare spending
1
-
-
-
-
Firm size
0.0848
1
-
-
-
Labour intensity
0.0008
-0.0006
1
-
-
Unit labour costs
0.0249
-0.0951
0.0003
1
-
Host welfare spending
0.0259
0.0107
0.0026
-0.0025
1
Source
: Authors’ calculations.
Note
: The detailed variable definitions are provided in section 3.2.

---

14
TRANSNATIONAL CORPORATIONS 
Volume 29, 2022, Number 2
Table 3b. Correlation matrix for key variables, 2013–2019
Variable
All manufacturing industries
Home welfare 
spending
Firm size
Labour 
intensity
Unit labour 
costs
Host welfare 
spending
Home welfare spending
1
-
-
-
-
Firm size
-0.2748
1
-
-
-
Labour intensity
0.0131
-0.2174
1
-
-
Unit labour costs
0.0316
-0.3287
0.3305
1
-
Host welfare spending
-0.0023
0.0045
-0.0001
-0.0014
1
Source
: Authors’ calculations.
4. Results
The relationship between welfare spending and location is presented in table 4. The 
table present the regression results for the whole manufacturing sector during the 
periods of 1997–2007 and 2013–2019. The negative coefficient on 
home welfare 
spending
indicates that MNEs are less likely to relocate when the home country’s 
welfare state is well developed. While the coefficient is only statistically significant 
for the 1997–2007 period, it is also negative – although less precisely estimated – 
over this period. Overall, this result does not support the conventional wisdom that 
welfare state expenditure pushes MNEs to invest more abroad at the detriment of 
expanding at home. At the same time, the coefficient on 
host welfare spending
is 
positive (and statistically significant in both cases), indicating that MNEs are more 
likely to relocate to host countries with generous welfare state provisions. 
Taking these results together, we can confirm our first and second hypothesis, 
namely that welfare spending tends to support MNEs and that firms are both 
attracted and retained by welfare spending. This suggests that while one can 
interpret welfare spending as an institution, one could also extend it to the 
importance of welfare spending to labour markets voids which would otherwise 
deter FDI. 
The subsequent estimates, reported in table 5, distinguish between technology 
levels and offer a test of hypothesis 3, which states that welfare spending may 
be more important for relocation decisions in high-tech industries. The results for 
the two periods are in line with this hypothesis when considering 
home welfare 
spending
. While home welfare spending matters for relocations in both high- 
and low-tech manufacturing industries in the 1997–2007 period, the estimated 
coefficient size for the high-tech industries is almost twice that of the low-tech 
industries. In the 2013–2019 period, we find that home welfare spending only 
returns the expected negative coefficient for the high-tech industries. 

---

15
Multinational enterprises and the welfare state
In these high-tech manufacturing industries, the focus on the “war for talent” is 
particularly fierce (Beechler and Woodward, 2009). In such contexts, labour market 
voids created through the absence of welfare support deter high-tech firms and 
encourage their relocation. While the issue of skill shortages among high-tech firms 
has been known for some time, no one appears to have considered it in the context 
of welfare spending and FDI. The results further suggest that welfare expenditure 
reduces the likelihood of relocation away from a country. Hence, it seems that firms 
attach value to a home country’s welfare state.
8
1
8 
One may argue that, if the main point of hypothesis 3 is about the “war for talent”, then the main variable 
of interest should be public expenditures in education and R&D. While this appears reasonable, we 
should stress that t
h
e “war for talent” is only one aspect of hypothesis 3, the other important point 
being the avoidance of labour market risks for high-tech firms (see section 2.1). 
Source
: Authors’ calculations. 
Note
: Average marginal effects from Probit Model estimation of equation (1) are reported. Explanatory variables are lagged one year. All 
specifications include a full set of country, industry and year dummies. Standard errors (clustered at the country level) in parentheses, 
*** p<0.01, ** p<0.05, * p<0.10.