You are on page 1of 16

 

FIRM DYNAMICS AND


REGIONAL DEVELOPMENT:
AN EMPIRICAL STUDY ON PRODUCTIVITY GROWTH
FOR EU 28 REGIONS
Supervisor Student

Prof. Fabio Pieri Francesco Androni


Index
• Setting the scene
• Regional trends
• Classical determinants of regional growth
• Firm dynamics

• Empirical analysis
• Data
• Empirical specification and methodology
• Regression analysis

• Conclusions
• Main findings
• Policy implications

• Annex

2
“Persistent poverty, economic decay, and lack of opportunities are at the root of
considerable discontent in declining and lagging-behind areas the world over. Poor
development prospects and an increasing belief that these places have ‘no future’ –
as economic dynamism has been posited to be increasingly dependent on
agglomeration economies – have led many of these so-called ‘places that don’t
matter’ to revolt against the status quo.” (Rodríguez-Pose)

3
Setting the scene | Regional trends (1)

Averages of top 10% (frontier), bottom (10% lagging) regional GDP per worker.
Across countries Within countries Total inequality Across countries Within countries Total inequality

Theil index for GDP per capita Theil index for GDP per worker
0.1 0.1

0.09 0.09 • If countries have


0.08 0.08
experierienced a process of
0.07

0.06
0.07

0.06
convregence for low
0.05 0.05 income countries, within-
0.04 0.04 country inequalities are
0.03 0.03
increasing in time.
0.02 0.02

0.01 0.01

0 0
95 96 9 7 9 8 99 00 01 0 2 0 3 0 4 05 06 07 0 8 0 9 1 0 11 12 13 1 4 1 5 95 96 9 7 98 99 00 0 1 0 2 0 3 04 0 5 06 07 08 0 9 1 0 1 1 12 13 14
19 19 19 19 19 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 19 19 19 19 19 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20

Source: Productivity and Jobs in a Globalised World, OECD, 2018.

4
Setting the scene | Regional trends (2)

Average productivity growth of top 10% (frontier), remaining 75% and lowest
10% for NUTS 2 regions.

• Frontier regions are also the


ones showing highest
productivity growth,
increasing inequalities.

Fr on tier re gio ns La gg in g r eg io n s 75 % o f re gio ns


USD PPP pe r e mp loye e

1.6% per year

1.3% per year

1.3% per year

Source: Productivity and Jobs in a Globalised World, OECD, 2018.

5
Setting the scene | Regional trends (3)

Composition of spatial inequalities in productivity, Theil Index, 2015, TL2 regions.


Between TL2 regions Within TL2 regions
Theil index across TL3 regions
0.15;0.2
0.05 1

0.04

0.03

0.02

0.01

0.00
ESP ITA IRL FRA POL NO HU SVN GR PRT AUT OEC BEL CZE DNK NLD GBR SW DEU KOR FIN SVK
(11) (18) (2) (22) (16) R N (2) C (3) (8) D (2) (5) (4) (10) (12) E (7) (14) (5) (3) (3)
(7) (7) (13)
Country (number of TL2
regions)

Source: OECD Regions and Cities at a Glance 2018 (OECD, 2018).

6
Setting the scene | Classical determinants of
regional growth
Three streams of literature:

• Neoclassical theory – accumulation of physical and human capital.

• Endogenous growth theory – stock of knowledge of a region.

• New economic geography – economies of scale (localization or urbanization


economies).

7
Setting the scene | Firm dynamics

• Not all firms that enter the market produce the


described results.

• High rates of failure produce a «revolving


door» dynamic.

• The quality of entrants is important in


fostering such effects: necessity vs opportunity
entrepreneurship (Vivarelli, 2013).

[Fritsch,
2008]
8
Empirical analysis | Data
• OECD Regional Database.
• 101 regions belonging to 10 EU 28 countries, T=6: 2010-2015.
• A total of 516 observations.

Year
Country
2010 2011 2012 2013 2014 2015
AT 9 9 9 9 9 9
DK 5 5 5 5 5
ES 16 17 17 17
FI 5 5 5 5 5 5
FR 19 21 19 20 21 21
HU 6 6 6 6 6
IT 21 21 21 21 21
PT 5 5 5 5 5 3
RO 8 7 7 7 7
SK 4 4 4 4 4

9
Empirical analysis | Empirical specification

• 

Where:
• Firm dynamics variables

And the control variables:


• Capital deepening
• Convergence effect
• Technological capital () and human capital ();
• Employment share per macrosector () and Herfindahl–Hirschman Index ();

10
Empirical analysis | Empirical specification – firm
dynamics

  𝑦 𝑖𝑗𝑡 =𝛿 ∆ 𝑘 𝑖𝑗𝑡 + 𝛼 𝑙𝑝 𝑖𝑗𝑡 − 1+ 𝛾 ′ 𝐼𝑁𝑁𝑂𝑖𝑗𝑡 −1 + 𝜏 ′ 𝑆𝑇𝑅 𝑖𝑗𝑡 − 1+ 𝜷 ′ 𝑭𝑫 𝒊𝒋𝒕 − 𝟏+𝜂 𝑖𝑗 +𝜂𝑡 + 𝜀 𝑖𝑗𝑡 −1

Firm dynamics variables Definitions from literature


Net entry - Market opportunities and increased competitiveness (Fritsch,
(A curvinear relationship is also tested) 2008).

Turnover rate -> turbulence - Trial and error (Jovanovic, 1982);


- Necessity entrepreneurship (Vivarelli, 2013).

- Experimentation (Klepper, 1997);


- Market cleansing (Caballero and Hammour, 1994).

Firm size (controlling for startup size) - Willingness to growth (Haltiwanger, 2011).

11
Empirical analysis | Methodology

• Two-way fixed effects: considering region (or country)-invariant and time-invariant components.

• Pooled OLS with country and year dummies.


• Fixed effects within group estimator with year dummies.
• Quantile regression with country and year dummies.

• Random effects estimator appears incostintent.

• Cluster robust standard errors at regional level.

12
(1) Pooled (2) FE (3) RE (4) Pooled (5) FE (6) RE
Variable Dependant variable: ∆𝑦𝑗𝑖𝑡 = growth rate (%)
∆𝑘𝑖𝑗𝑡 0.5687*** 0.5529*** 0.5116*** 0.5624*** 0.5522*** 0.5005***
(0.1144) (0.1295) (0.0985) (0.1129) (0.1291) (0.0997)
𝐿𝑃𝑖𝑗𝑡 −1 -0.0471** -0.4430*** -0.0109 -0.0444** -0.4274*** -0.0158
(0.0199) (0.0671) (0.0143) (0.0198) (0.0697) (0.0138)
𝑇𝐾𝑖𝑗𝑡 −1 0.0051** 0.0568*** 0.0047** 0.0047** 0.0490*** 0.0009
(0.0020) (0.0156) (0.0019) (0.0020) (0.0159) (0.0020)
ℎ𝑘𝑖𝑗𝑡 −1 0.0146 -0.0457 -0.0107 0.0157 -0.0442 0.0004
(0.0226) (0.0496) (0.0114) (0.0213) (0.0483) (0.0130)
ℎℎ𝑖𝑖𝑗𝑡 −1 0.1717*** 0.1748 0.1203** 0.1790*** 0.1254 0.1404**
(0.0633) (0.5881) (0.0571) (0.0593) (0.5566) (0.0645)
𝑠𝑒𝑐
𝑠ℎ𝑎𝑟𝑒𝑖𝑗𝑡 −1 0.0882** -0.5076 0.0324 0.0847** -0.6142 0.0351
(0.0412) (0.4267) (0.0298) (0.0405) (0.3771) (0.0280)
𝑡𝑒𝑟𝑡
𝑠ℎ𝑎𝑟𝑒𝑖𝑗𝑡 −11 0.0887** 0.0819 0.0323 0.0791* -0.1238 0.0389
(0.0444) (0.4301) (0.0352) (0.0410) (0.3718) (0.0327)
𝑛𝑒𝑡𝑖𝑗𝑡 −1 0.1930* 0.2029*** 0.0961 0.2682** 0.2531*** 0.1711***
(0.1080) (0.0762) (0.0626) (0.1032) (0.0727) (0.0633)
𝑡𝑢𝑟𝑏𝑖𝑗𝑡 −1 -0.0336 -0.0952** 0.0621*
(0.0421) (0.0430) (0.0353) Net entry rate
2 [Birth(t)-Death(t)]/Active(t-1)
𝑛𝑒𝑡𝑖𝑗𝑡 −1 -3.0547*** -2.4051** -2.3688** (.0426667,.0608]
(.01412,.0426667]
(1.1585) (1.0837) (1.0413) (.0052333,.01412]
(-.0073,.0052333]
𝑎𝑐𝑡𝑖𝑣𝑒
𝑆𝐼𝑍𝐸𝑖𝑗𝑡 −1 -0.0078 -0.0477* -0.0196*** -0.0058 -0.0432 -0.0116* (-.01066,-.0073]
(-.01674,-.01066]
(0.0060) (0.0260) (0.0065) (0.0058) (0.0261) (0.0068) [-.0299,-.01674]
No data
𝑏𝑜𝑟𝑛
𝑆𝐼𝑍𝐸𝑖𝑗𝑡 −1 -0.0035 0.0099 0.0114* -0.0064 0.0089 0.0064
(0.0095) (0.0114) (0.0059) (0.0084) (0.0110) (0.0058)
Constant 0.3892** 4.7180*** 0.0622 0.3627** 4.7448*** 0.1261
(0.1852) (0.8434) (0.1399) (0.1803) (0.8484) (0.1370)
:  peak reached at 4.53% rate
N

516
0.3706
516
0.5142
516 516
0.3964
516
0.5252
516
(pooled) and 5.26% (FE).
Adj-R² 0.3385 0.4986 0.3657 0.5100
It includes respectively 15% and
Notes:
-
- Cluster robust standard errors are given in the parentheses
Significance levels: * p<0.10, ** p<0.05, *** p<0.01 10% of all observations.
- CAPITAL LETTER VARIABLE: log-transformed variable.
- Country dummies and year dummies are included for the pooled model FE and RE model include year
dummies.
- The Hausman specification test run for both FE-RE models shows the RE model to be inconsistent.
𝜎𝜂2
- 𝜌= ; 𝜌 = 0.95 in column 2, 𝜌 = 0.94 in column 5: most of the variance is due to regional effects.
𝜎𝜂2 +𝜎𝜀2
- F-test rejects the null hypothesis that 𝜂𝑗𝑖 are conjunctly 0.

Notes: The Hausman specification test run for both FE-RE models shows the RE model to be inconsistent.

13
(1) (2) (3) (4) (5)
10 pctile 25 pctile 50 pctile 75 pctile 90 pctile
Variable Dependant variable: ∆𝑦𝑗𝑖𝑡 = growth rate (%)
∆𝑘𝑖𝑗𝑡 0.3359** 0.3762*** 0.3965*** 0.4429*** 0.5542***
(0.1665) (0.0758) (0.0710) (0.0798) (0.1169)
𝐿𝑃𝑖𝑗𝑡 −1 -0.0352 -0.0306* -0.0422** -0.0257 0.0160
(0.0241) (0.0163) (0.0170) (0.0181) (0.0293)
𝑇𝐾𝑖𝑗𝑡 −1 0.0018 0.0037* 0.0051** 0.0062*** 0.0017
(0.0039) (0.0021) (0.0021) (0.0016) (0.0037)
ℎ𝑘𝑖𝑗𝑡 −1 -0.0211 -0.0514** 0.0002 0.0126 -0.0044
(0.0399) (0.0249) (0.0236) (0.0193) (0.0436)
ℎℎ𝑖𝑖𝑗𝑡 −1 0.1419 0.0648 0.1103 0.2515*** 0.1944*
(0.0866) (0.0638) (0.0761) (0.0740) (0.1009)
𝑠𝑒𝑐
𝑠ℎ𝑎𝑟𝑒𝑖𝑗𝑡 −1 0.0979* 0.0973*** 0.0406 -0.0440 -0.0553
(0.0536) (0.0350) (0.0429) (0.0485) (0.0672)
𝑡𝑒𝑟𝑡
𝑠ℎ𝑎𝑟𝑒𝑖𝑗𝑡 −11 0.0969* 0.1066*** 0.0473 -0.0423 -0.0880
(0.0571) (0.0368) (0.0438) (0.0529) (0.0754)
𝑛𝑒𝑡𝑖𝑗𝑡 −1 0.2065* 0.1883* 0.1271* -0.0366 -0.0909
(0.1178) (0.1057) (0.0725) (0.1415) (0.1004)
𝑡𝑢𝑟𝑏𝑖𝑗𝑡 −1 -0.0581 -0.0478 -0.0300 0.0540 0.0831*
(0.0936) (0.0497) (0.0432) (0.0478) (0.0448)
𝑎𝑐𝑡𝑖𝑣𝑒
𝑆𝐼𝑍𝐸𝑖𝑗𝑡 −1 0.0052 0.0014 0.0011 -0.0246*** -0.0186**
(0.0282) (0.0296) (0.0097) (0.0068) (0.0087)
Constant 0.2525 0.2130 0.3649** 0.2165 -0.1561
(0.2355) (0.1536) (0.1631) (0.1827) (0.2495)
N 516 516 516 516 516
R² 0.1774 0.2160 0.3415 0.2446 0.2097
Notes: - Cluster robust standard errors are given in the parentheses
- Significance levels: * p<0.10, ** p<0.05, *** p<0.01
- CAPITAL LETTER VARIABLE: log-transformed variable.
- Country and year dummies are included.

14
Conclusions | Main findings
• Net entrance is clearly associated with increased market opportunities, positively related to
following year’s productivity growth.
• Too much high net entry is detrimental to productivity growth: above a certain threshold too
much «enthusiasm» for market opportunities leads to the entrance of productivity-detrimental
firms (similarly to what said for turbulence).
• Turbulence is generally associated with negative productivity change. Processes of trial and
error and the entrance of «low quality» entrepreneurship burn resources rather than being the
outcome of experimentation of innovative processes.
• Productivity growth is associated in regions with negative productivity change with an
exploitation of market opportunities, while the ones growing the most show to be driven by
increase in turbulence, that can be ultimately identified as a process of experimentation that leads
creative destruction.

15
Conclusions | Policy implications
• Fostering entrepreneurial activity should be considered by policy-makers for reducing regional
disparities and increasing productivity growth.
• The quality of entrepreneurship must be considered, in order not to produce negative turbulence.
• Following a place-based approach (Iammarino et al. 2017) that values territorial capital, such
policies should focus on building capabilities, infrastructures and innovation systems that are
able to stimulate a positive entrepreneurial spirit (Audretsch and Keilbach, 2004).
• A further analysis on French, Romanian and Slovakian firm dynamics should provide better
insights for high net entry rates.

16

You might also like