Microsoft Word - ea 2019 1-2 ver 4.docx


DOI:	10.28934/ea.19.52.12.pp1‐22	
	
	
ORIGINAL		SCIENTIFIC	PAPER	
	
	

Business	Sector	Innovation	and	Economic	Growth:	a	
Comparative	Analysis	between	EU	Countries		

Cláudia	Caseiro1			|			Marta	Simões1*	
1	CeBER	and	Faculty	of	Economics,	University	of	Coimbra,	Portugal		

	
	
ABSTRACT	
This	paper	examines	the	relationship	between	innovation	carried	out	by	the	business	sector	and	
economic	growth	in	the	28	member	states	of	the	European	Union,	divided	into	two	groups	(1	and	2)	
according	 to	 their	 innovation	 performance.	 We	 use	 fixed	 effects	 panel	 data	 methods	 to	 test	 the	
hypothesis	that	business	sector	innovation	plays	a	relevant	role	in	explaining	the	behaviour	of	real	
GDP	per	capita,	estimating	two	growth	regressions	according	to	data	availability	(1990‐2015	and	
2008‐2015;	 unrestricted/restricted	 sample).	 The	 results	 indicate	 that	 the	 role	 of	 business	 sector	
innovation	in	economic	growth	not	only	varies	according	to	the	sample	of	countries	and	the	period	
under	 analysis	 but	 also	 the	 proxy	 for	 innovation	 used.	 In	 group	 1	 (above	 average	 innovation	
performance)	the	innovation	indicators	statistically	significant	in	explaining	growth	also	present	a	
positive	sign	(with	a	few	exceptions).	In	group	2	(below	average	innovation	performance)	on	the	
other	hand,	the	statistically	significant	business	sector	innovation	indicators	present	a	negative	sign.	
One	possible	justification	for	these	signs	are	differences	in	absorptive	capacity	so	that	the	growth	
benefits	of	innovation	activities	depend	on	aspects	such	as	human	capital	availability,	accumulated	
knowledge,	 technological	 and	 financial	 support.	 Since	 group	 1	 includes	 countries	 with	 higher	
absorptive	capacity,	business	sector	innovation	is	effectively	translated	into	faster	economic	growth.	
In	 group	 2,	 innovation	 activities	 do	 not	 translate	 into	 productivity	 increases	 due	 to	 a	 lack	 of	
absorptive	capacity.	Additionally,	resources	used	in	innovation	activities	might	compete	with	other	
activities	more	relevant	in	terms	of	the	stage	of	the	growth	process	these	countries	are	in	so	that	
innovation	saps	growth	

	
Key	words:	business	sector	innovation,	economic	growth,	EU	28,	fixed	effects		
	
JEL	Classification:	C23,	O30,	O47,	O52		

	

INTRODUCTION	

There	 seems	 to	 be	 a	 consensus	 in	 the	 literature	 that	 innovation	 drives	 economic	 growth,	
supported	both	by	the	neoclassical	theory	of	exogenous	growth,	but	especially	by	endogenous	
growth	models,	pointing	to	technological	progress	and	thus	innovation	as	drivers	of	economic	
growth,	albeit	from	different	perspectives	((Solow,	1956);	(Lucas,	1988);	(Romer,	1990);	(Jones,	
1995);	 (Jones,	 2005)).	 At	 the	 firm	 level,	 several	 studies	 examine	 the	 relationship	 between	
innovation	 carried	 out	 by	 firms	 with	 the	 respective	 productivity	 levels	 and	 growth	 ((OECD,	
2009);	(Mohnen	and	Hall,	2013)).	Eurostat	data	show	that	the	intensity	of	R&D	expenditures	in	
the	28	EU	Member	States	(EU)	increased	between	2005	and	2014,	with	expenditure	carried	out	
by	the	business	sector	increasing	from	1.10	%	of	GDP	in	2005	to	1.30%	in	2014.	Innovation	

																																																													
*	Corresponding	author,	e‐mail:	mcsimoes@fe.uc.pt	



2
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

through	R&D	and	investments	in	technology	are	considered	essential	to	ensure	competitiveness	
and	increase	productivity	and	in	this	way,	sustain	economic	growth	(Pece,	2015).	
Innovation	is	however	difficult	to	measure	since	it	encompasses	technological	aspects,	such	as	

new	or	significantly	improved	products/	services,	but	also	includes	non‐technological	aspects,	
such	as	new	organisational	models,	new	marketing	strategies,	or	new	processes	(Mohnen	and	
Hall,	 2013).	 The	 aim	 of	 this	 paper	 is	 to	 relate	 the	 different	 aspects	 of	 innovation	 by	 firms	
(technological	 and	 non‐technological)	 to	 a	 country's	 long‐term	 macroeconomic	 performance	
(economic	 growth),	 grouping	 the	 different	 countries	 under	 analysis	 according	 to	 different	
characteristics/indicators	of	their	business	sector	innovation	activity.	In	particular,	we	aim	at	
better	understanding	the	relationship	between	inputs	and	outputs	of	business	innovation	and	
the	behaviour	of	output	in	different	groups	of	countries.	Following	the	identification	of	country	
groups	based	on	the	performance	of	 indicators	of	 inputs	and	outputs	of	 innovation	activities	
relative	 to	 the	 EU28	 average,	 we	 estimate	 two	 empirical	 models	 that	 correspond	 to	 growth	
regressions	where	the	different	innovation	indicators	are	the	main	explanatory	variables	taken	
alongside	other	relevant	growth	determinants.	Depending	on	data	availability,	the	period	under	
analysis	is	1990‐2015	or	2008‐2015,	for	unrestricted	and	restricted	country	samples.	
The	remainder	of	the	paper	is	organized	as	follows:	after	the	introduction,	the	second	section	

contains	a	brief	review	of	the	literature	on	the	relationship	between	innovation	and	growth.	In	
the	third	section,	we	analyse	the	average	performance	of	firms’	innovation	activity	in	terms	of	
inputs	and	outputs,	grouping	the	countries	under	analysis	according	to	their	position	relative	to	
the	EU28	average.	The	next	section	contains	the	methodology	and	results.	In	the	final	section,	we	
present	the	main	conclusions.	

INNOVATION	AND	GROWTH:	THEORY	AND	EVIDENCE	

In	the	1930s,	Joseph	Schumpeter	highlighted	that	economic	growth	is	caused	by	innovation,	in	
particular,	creative	destruction,	with	business	innovation	fundamental	for	this	change.	(Solow,	
1956)	Also	stresses	that	technological	progress	is	the	driver	of	growth,	but	does	not	provide	an	
explanation	how	the	decisions	by	economic	agents	in	terms	of	 input	allocation	between	final	
goods	production	and	innovation	activities	can	influence	the	pace	of	economic	growth.	In	the	
mid‐1980s,	 endogenous	 growth	 theories	 assign	 knowledge	 a	 fundamental	 role	 in	 the	
explanation	 of	 technological	 progress.	 First‐generation	 endogenous	 growth	 models	 consider	
innovation	as	an	unintentional	by‐product	of	 inputs	accumulation,	 for	 instance	through	from	
learning	by	doing	or	learning	by	studying	((Romer,	1986);	(Lucas,	1988)).	Second‐generation	
endogenous	growth	models,	pose	that	decisions	by	firms	concerning	R&D	activities	increase	the	
respective	productivity	levels	and	also	generate	positive	externalities	in	terms	of	productivity	
improvements	in	other	firms,	and,	consequently,	faster	growth	((Romer,	1990);	(Jones,	1995)).	
There	 thus	 seems	 to	 be	 a	 consensus	 in	 the	 literature	 that	 technological	 progress	 and	

innovation	 are	 at	 the	 basis	 of	 economic	 growth,	 with	 many	 empirical	 studies	 presenting	
evidence	 on	 this	 link.	 For	 instance,	 (Ulku,	 2004),	 using	 panel	 data	 to	 study	 the	 relationship	
between	economic	growth,	R&D	expenditures	and	innovation	in	20	OECD	member	states	and	10	
non‐member	 countries,	 concludes	 that	 innovations	 (measured	 by	 patents)	 have	 a	 positive	
growth	impact,	although	only	developed	OECD	countries	seem	to	be	able	to	increase	the	level	of	
innovation	 through	R&D	expenditures.	 (Pradhan	et	al.,	2016)	studied	18	Eurozone	countries	
over	the	period	1961‐2013,	using	time	series	data.	The	dependent	variable	considered	was	the	
growth	rate	of	real	GDP	per	capita	and	the	independent	variables	correspond	to	five	indicators	
of	innovation	(number	of	patents	per	resident	per	thousand	inhabitants,	number	of	patents	per	
non‐residents	per	thousand	inhabitants,	number	of	patents	per	resident	and	non‐resident	per	
thousand	 inhabitants,	 real	 R&D	 expenditures	 as	 a	 percentage	 of	 real	 GDP	 and	 researchers	
involved	in	R&D	activities	per	million	inhabitants)	and	eight	indicators	of	financial	development.	
The	authors	conclude	that	financial	development	and	innovation	are	important	in	determining	



	 		
Cláudia	Caseiro,	Marta	Simões	 3	

economic	growth	and	that	long‐term	growth	depends	on	countries'	ability	to	innovate	in	order	
to	remain	competitive,	which	requires	adequate	(financial)	resources	devoted	to	R&D	activities.	
In	a	recent	study,	(Maradana	et	al.,	2017)	examined	19	EU	member	states	and	concluded	that	
there	is	a	long‐run	equilibrium	relationship	between	the	innovation	indicators	used	(resident	
patents,	 non‐resident	 patents,	 R&D	 expenditures,	 R&D	 researchers,	 high‐tech	 exports,	 and	
scientific	and	technical	journal	articles)	and	growth.	
However,	despite	the	consensus	on	the	importance	of	innovation	for	economic	growth,	many	

authors	recognise	that	it	is	difficult	to	measure	this	concept,	which	hampers	the	results	from	
empirical	studies.	Measuring	 innovation	 in	a	precise	and	rigorous	way	 is	critical	 to	correctly	
assess	its	effect	on	growth	(Hasan	and	Tucci,	2010).	Often	business	innovation	is	not	included	in	
the	 analysis	 due	 to	 data	 availability	 and	 accessibility	 (microdata)	 issues	 (OECD,	 2009).	 In	
addition,	there	is	no	unique	and	perfect	measure	of	innovation.	
According	to	(Hong	et	al.,	2012),	in	practice,	companies	innovate	in	different	ways:	investing	

in	R&D,	actively	patenting,	cooperating	with	external	partners,	acquiring	technology	externally	
via	licensing,	through	design,	marketing	and	staff	training.	It	is,	therefore,	crucial	to	choose	the	
adequate	innovation	proxies,	although	previous	studies	do	not	all	point	in	the	same	direction	
concerning	this	choice.	Several	authors	consider	R&D	expenditures	as	a	proxy	for	innovation,	
e.g.	(Goel	and	Ram,	1994)	and,	more	recently,	(Freimane	and	Bāliņa,	2016).	But,	according	to	
(Wang,	2013),	since	(Schmookler,	1966)	demonstrated	the	adequateness	of	patent	data	as	an	
indicator	of	 innovation,	 these	have	often	been	the	preferred	proxy	 for	 innovation.	Similar	to	
other	authors,	 (Wang,	2013)	considers	 that	R&D	expenditures	 are	only	 inputs	 to	 innovation	
activities	 and	 do	 not	 measure	 innovation	 results/outputs.	 The	 literature	 also	 distinguishes	
between	two	types	of	measurement	of	innovation:	indirect	and	direct	(Hong,	Oxley	and	McCann,	
2012).	 According	 to	 the	 authors,	 R&D	 expenditures	 are	 indirect	 measures	 since	 they	 only	
measure	 inputs	 to	 innovation	 activities,	 while	 patents	 focus	 on	 the	 successful	 commercial	
application	 of	 innovations.	 However,	 (Hong,	 Oxley	 and	 McCann,	 2012)	 consider	 that	 for	
econometric	analysis	the	best	option	is	to	use	direct	measures,	which	may	be	objective	(when	it	
comes	to	the	number	of	innovations	or	"count	of	innovations")	or	subjective.	
The	 OECD	 Oslo	 Manual	 (OECD,	 2005)	 identifies	 the	 existence	 of	 four	 different	 types	 of	

innovation:	 product,	process,	 marketing	 and	 organisational.	 Product	 innovation	 refers	 to	 the	
"introduction	 of	 new	 or	 significantly	 improved	 goods	 or	 services	 concerning	 its	 intended	
characteristics	or	uses"	((OECD,	2005),	p.57).	Process	innovation	refers	to	the	"implementation	
of	a	new	or	significantly	 improved	production	or	distribution	method"	((OECD,	2005),	p.58).	
Marketing	innovation	refers	to	"implementing	a	new	marketing	method	with	significant	changes	
in	product	design	or	packaging,	product	positioning,	promotion	or	pricing"	((OECD,	2005),	p.59).	
Finally,	organisational	innovation	concerns	the	"implementation	of	a	new	organisational	method	
in	 the	 company's	 business	 practices,	 in	 the	 organisation	 of	 its	 workplace	 or	 in	 its	 external	
relations"	((OECD,	2005),	p.61).	
At	 the	 firm	 level,	 several	 studies	 show	 a	 positive	 relationship	 between	 productivity	 and	

different	 types	 of	 innovation	 (Moreno	 and	 Surinach,	 2014).	 Based	 on	 the	 former	 OECD	
classification,	(Hall,	2011)	concluded	that	there	are	significant	impacts	of	product	innovation	on	
productivity,	but,	on	the	other	hand,	the	impact	of	process	innovations	is	more	ambiguous	and	
may	even	be	negative.	(Tavassoli	and	Karlsson,	2015)	analysed	the	persistence	of	the	innovation	
behaviour	by	firms	and,	similar	to	(Hall,	2011),	concluded	that	the	degree	of	persistence	is	not	
the	same	for	all	types	of	innovations,	being	greater	for	product	innovations:	product,	process,	
and	organizational	innovations	have	a	"true"	dependency,	that	is,	the	decision	to	innovate	in	the	
next	 period	 is	 influenced	 by	 the	 success	 of	 the	 innovation	 of	 the	 previous	 period,	 while	
marketing	innovations	have	a	"false"	dependency	because	the	persistence	effect	does	not	remain	
in	marketing	innovations.	(Mohnen	and	Hall,	2013)	updated	the	previous	research	analysing	the	
existing	 evidence	 on	 the	 effects	 of	 technological	 and	 non‐technological	 innovations	 on	 the	
productivity	of	 firms	and	the	existence	of	potential	complementarities	between	the	different	



4
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

types	of	innovation.	The	authors	concluded	that	all	types	of	innovations	contribute	to	improved	
productivity	performance.	(Moreno	and	Surinach,	2014)	also	sought	to	empirically	analyse	the	
relationship	between	the	business	sector	innovation	and	productivity	growth	in	25	EU	Member	
States,	 as	 well	 as	 in	 Iceland,	 Norway	 and	 Turkey	 for	 the	 period	 1998‐2000	 and	 2002‐2004,	
concluding	that	innovation	has	a	positive	impact	on	productivity	growth,	but	this	impact	is	more	
important	in	the	case	of	process	innovations	than	output	innovations,	which	can	be	justified	on	
the	basis	that	by	introducing	a	new	production	process	firms	become	more	efficient,	reducing	
costs	and,	thus,	increasing	productivity.	
To	sum	up,	it	is	consensual	in	the	literature	that	innovation	boosts	economic	growth	and,	from	

a	 business	 sector	 perspective,	 the	 introduction	 of	 innovations	 by	 firms,	 whether	 a	 product,	
process,	 organisational	 or	 marketing	 innovations,	 leads	 to	 an	 increase	 in	 the	 respective	
productivity,	which	should	also	lead	to	faster	economic	growth.	The	present	studies	contribute	
to	the	existing	literature	by	carrying	out	a	comprehensive	analysis	of	the	relationship	between	
firms’	innovation	activity	and	economic	growth	taking	into	account	a	wide	array	of	indicators	
(technological	 and	 non‐technological)	 to	 measure	 business	 sector	 innovation	 (inputs	 and	
outputs).		

BUSINESS	SECTOR	INNOVATION	IN	THE	EU28:	A	COMPARATIVE	ANALYSIS	

In	this	section	we	differentiate	the	28	EU	Member	States	in	terms	of	the	innovation	activity	
carried	 out	 by	 the	 respective	 business	 sector	 (firms),	 grouping	 countries	 into	 different	 sets	
according	to	their	average	performance	in	terms	of	inputs	and	outputs,	i.e.	differences	between	
innovation	efforts/investments	and	the	results	of	 innovation	activity	 for	 the	countries	under	
analysis,	relative	to	the	EU28	average.	
The	business	sector	innovation	indicators	related	to	inputs	considered	in	this	study	are	R&D	

expenditures,	innovation	expenditures	in	the	business	sector,	percentage	of	firms	that	promote	
the	training	of	their	employees,	total	staff	in	R&D	activities	and	researchers	in	R&D	activities.	
Outputs	 are	 represented	 by	 the	 following	 indicators:	 number	 of	 patent	 applications	 by	 the	
business	sector	to	the	European	Patent	Office	(EPO),	number	of	patents	granted	by	the	USPTO	to	
the	business	sector,	number	of	registrations	of	trademark	applications,	percentage	of	Small	and	
Medium	Enterprises	(SMEs)	that	introduce	product	or	process	innovations,	percentage	of	SMEs	
that	 introduce	 marketing	 or	 organizational	 innovations,	 percentage	 of	 SMEs	 that	 innovate	
internally,	 and	 percentage	 of	 innovative	 SMEs	 that	 cooperate	 with	 others.	 Table	 A.1	 in	 the	
appendix	 summarises	 the	 indicators	 used,	 as	 well	 containing	 a	 detailed	 description	 and	
classification	of	each	indicator	according	to	the	European	Innovation	Scoreboard	(EIS)	2017.		
The	countries	under	analysis	were	grouped	based	on	their	average	performance	in	terms	of	

inputs	and	outputs	of	business	sector	innovation	relative	to	the	EU28	average	for	the	period	
2008‐2015.	We	started	by	computing,	 for	each	country,	 the	average	 indicator	 for	 the	period	
2008‐2015.	 Next,	 we	 computed	 the	 ratios	 of	 each	 of	 these	 indicators	 relative	 to	 the	 EU28	
average	 (EU28=100).	 Finally,	 in	order	 to	 analyse	 the	 overall	performance	 in	 terms	 of	 either	
inputs	or	outputs,	we	computed	the	average	of	the	ratios	for	each	of	these	two	categories,	based	
on	the	relative	performance	of	each	indicator.	Four	different	groups	were	identified.		
The	first	group	of	countries	(group	1)	 is	made	up	of	countries	with	average	performances	

above	that	of	the	EU28	in	terms	of	both	inputs	and	outputs	of	business	sector	innovation.	This	
group	comprises	ten	countries:	Austria,	Belgium,	Denmark,	Finland,	France,	Germany,	Ireland,	
Luxembourg,	the	Netherlands	and	Sweden.	In	the	EIS	classification	for	2015,	these	countries	are	
classified	 as	 innovation	 leaders	 (Sweden,	 Netherlands,	 Luxembourg,	 Finland,	 Denmark,	 and	
Germany)	and	strong	innovators	(Ireland,	France,	Belgium,	and	Austria).		
Figure	1	contains	information	on	the	relative	performance	of	these	member	states	in	terms	of	

inputs	 and	 outputs	 for	 the	 period	 2008	 to	 2015	 relative	 to	 the	 EU28	 average.	 According	 to	
Figure	1,	all	these	member	states	have	higher	performances	in	inputs	and	outputs	compared	to	



	 		
Cláudia	Caseiro,	Marta	Simões	 5	

the	 EU28	 average,	 with	 Finland,	 Sweden,	 Denmark,	 Austria,	 Germany	 and	 Luxembourg	
recording	the	highest	ratios.	However,	this	aggregate	information	masks	different	performances	
in	 terms	 of	 the	 various	 indicators	 used	 to	 measure	 inputs	 and	 outputs.	 Only	 Germany	 and	
Sweden	perform	above	the	EU28	 in	all	 indicators.	The	remaining	member	states	have	 lower	
performances	than	the	EU	in	some	indicators.	For	instance,	Austria,	Belgium,	Denmark,	Finland,	
France,	Ireland,	Luxembourg	and	the	Netherlands	show	levels	of	expenditure	on	innovation	by	
the	 business	 sector	 (excluding	 R&D)	 below	 the	 EU28	 average.	 On	 the	 output	 side,	 Austria,	
Denmark	 and	 Luxembourg	 have	 performances	 above	 the	 EU28	 average	 in	 all	 indicators,	
especially	for	registration	of	trademark	applications	in	Luxembourg,	an	outlier.	Belgium,	Finland	
and	France	record	 lower	 than	average	trademark	registrations	applications,	while	 in	 Ireland	
EPO	performance	is	lower	than	average,	as	well	as	the	trademark	registrations	applications1.	

	

	

Figure	1.	Group	1:	inputs	and	outputs	of	business	sector	innovation	relative	to	the	EU28	
average,	2008‐15	

Source:	Own	elaboration.	
	

																																																													
1	The	data	for	these	more	detailed	indicators	are	available	from	the	authors.	



6
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

	

Figure	2.	Group	2:	inputs	and	outputs	of	business	sector	innovation	relative	to	the	EU28	
average,	2008‐15	

Source:	Own	elaboration.	
	
The	second	group	of	countries	(group	2)	includes	member	states	which,	on	average,	perform	

worse	than	the	EU28	in	terms	of	both	inputs	and	outputs	of	business	sector	innovation.	This	
group	 includes	 fifteen	countries:	Bulgaria,	Croatia,	 Czech	 Republic,	 Estonia,	 Greece,	 Hungary,	
Italy,	Latvia,	Lithuania,	Malta,	Poland,	Portugal,	Romania,	Slovakia	and	Spain.	In	the	EIS	typology	
for	 2015,	 these	 member	 states	 are	 classified	 as	 modest	 innovators	 (Romania	 and	 Bulgaria),	
moderate	innovators	(Croatia,	Czech	Republic,	Greece,	Hungary,	Italy,	Lithuania,	Latvia,	Malta,	
Portugal,	Poland,	Slovakia	and	Spain)	and,	in	the	case	of	Estonia,	a	strong	innovator.	Figure	2	
contains	information	on	the	inputs	and	outputs	of	this	group	2	for	the	period	2008‐2015	relative	
to	the	EU28	average.	According	to	Figure	2,	overall	these	countries	have	a	performance	in	terms	
of	inputs	and	outputs	of	business	sector	innovation	below	the	EU28	average,	although	Romania,	
Bulgaria	and	Latvia	stand	out	with	levels	of	inputs	and	outputs	which	are	more	distant	from	the	
EU28	average	than	the	other	member	states	in	this	group.	
On	the	other	hand,	 the	Czech	Republic,	Estonia	and	Malta	record	 levels	close	 to	 the	EU28	

average.	Within	this	group,	there	are	also	substantial	differences	in	the	performance	of	certain	
indicators.	For	instance,	the	Czech	Republic	performs	above	the	EU28	average	in	terms	of	R&D	
activities	personnel;	Croatia,	Slovakia,	Estonia,	Greece,	Latvia,	Lithuania,	Malta,	Poland	and	the	
Czech	Republic	record	higher	than	average	expenditure	on	innovation	in	the	business	sector;	
and	in	Croatia,	Slovakia,	Malta,	Portugal	and	the	Czech	Republic	the	percentage	of	companies	
promoting	 employee	 training	 is	 also	 above	 average.	 For	 output	 indicators:	 trademark	
applications	are	above	the	EU28	average	in	Malta;	the	same	applies	to	product	and/or	process	



	 		
Cláudia	Caseiro,	Marta	Simões	 7	

innovations	and	the	percentage	of	SMEs	innovating	internally	in	Estonia,	Greece,	Italy,	Portugal	
and	 the	 Czech	Republic	and	 in	 marketing	 and/or	organizational	 innovations	 in	Greece,	 Italy,	
Portugal	and	the	Czech	Republic2.	

	

	

Figure	3.	Group	3:	inputs	and	outputs	of	business	sector	innovation	relative	to	the	EU28	
average,	2008‐15	

Source:	Own	elaboration.	
	

	

Figure	4.	Group	4:	inputs	and	outputs	of	business	sector	innovation	relative	to	the	EU28	
average,	2008‐15	

Source:	Own	elaboration.	
	
The	third	group	(group	3)	is	characterised	by	higher	than	average	inputs	and	at	the	same	time	

lower	than	average	outputs.	This	group	consists	only	of	Slovenia,	classified	by	the	EIS	in	2015	as	
a	strong	innovator	and	its	innovative	activity	is	very	close	to	the	EU	average.	According	to	Figure	
3,	the	performance	is	higher	than	average	for	inputs	(in	three	out	of	the	five	indicators	and	very	
close	 for	 the	 others),	 but	 below	 average	 for	 outputs.	 Slovenia	 performs	 well	 in	 most	 input	
indicators	but	records	patent	levels	and	registrations	of	brand	applications	well	below	the	EU28	
average.	
Finally,	the	fourth	group	(group	4)	includes	countries	with	inputs	below	the	EU28	average	

and	outputs	higher	than	the	EU28	average.	Two	countries,	Cyprus	and	the	United	Kingdom,	are	
included	in	this	group.	According	to	the	EIS	in	2015,	Cyprus	is	classified	as	a	moderate	innovator	
and	the	UK	as	a	strong	innovator.	In	figure	4,	it	is	possible	to	see	that	in	Cyprus	the	performance	
for	 inputs	 is	 lower	 than	average,	while	outputs	are	slightly	above	 the	EU28	average.	Cyprus	
records	 high	 innovation	 expenditures	 and	 percentage	 of	 firms	 that	 promote	 the	 training	 of	
officials	and	above	the	EU28	average,	but	low	levels	of	patents	(applications	and	concessions).	In	
the	United	Kingdom	inputs	are	close	to	the	EU28	average,	although	slightly	below,	but	the	UK	
records	 a	 higher	 than	 average	 performance	 in	 three	 out	 of	 the	 five	 indicators.	 Outputs	 are	
below‐average	due	essentially	to	trademark	applications,	the	introduction	of	the	product	and/or	
process	 innovations,	 and	 the	 percentage	 of	 SMEs	 innovating	 internally	 (three	 out	 of	 seven	
indicators).		
In	the	econometric	analysis,	since	Slovenia	has	a	large	number	of	output	indicators	(four	in	

seven)	which	are	close	to	the	EU28	average,	this	country	will	be	included	in	group	1.	The	United	

																																																													
2	The	data	for	these	more	detailed	indicators	are	available	from	the	authors.	



8
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

Kingdom	 will	 also	 be	 considered	 in	 group	 1	 as	 inputs	 are	 close	 to	 the	 EU28	 average,	 while	
Cyprus	will	be	included	in	group	2	since	it	performs	poorly	in	inputs.	These	changes	also	make	
the	groups	more	in	line	with	the	EIS	classification.	

EMPIRICAL	MODELS,	METHODOLOGY	AND	RESULTS	

In	this	section,	we	present	the	two	empirical	models	estimated,	the	methodology	used	and	the	
main	results	obtained	for	the	total	sample	as	well	as	for	the	two	groups	defined	in	the	previous	
section.	

Empirical	Models	

In	order	to	study	the	relationship	between	business	sector	innovation	and	economic	growth,	
we	consider	the	global	sample	of	28	EU	Member	States	as	well	as	the	two	country	groups,	1	and	
2,	defined	 in	 the	previous	section.	Two	periods	were	considered	 in	 the	analysis	due	 to	data	
availability	for	some	indicators	of	innovation	activities:	1990‐2015	(model	1)	and	2008‐2015	
(model	2).	
The	 first	 empirical	 model	 (model	 1)	 is	 similar	 to	 that	 of	 (Moreno	 and	 Surinach,	 2014),	

corresponding	to	an	ad	hoc	growth	regression,	with	variables	measured	at	5‐year	intervals,	and	
is	given	by	equation	(1):	
	

, 	 	 	 , 	 	 	 	 	 , 	 	 	 _
	 	 	 	 1 	 	 1	 	 2	 	 	 	 	 	 	 	 			(1)	

	
where:	δ0	–	Constant;	ui,t	–	error	term.;	i	–	countries	(i	=	1,	2,	…,	N).	t	–	5‐years	sub	periods	(t	=	5);	
t‐1	–	initial	year	of	each	sub	period.	
	
The	dependent	variable	in	the	first	model,	equation	(1),	is	the	annual	growth	rate	of	real	GDP	

per	capita	in	PPPs	(Purchasing	Power	Parity)	(ΔlnGDPrpc)	for	5‐years	sub‐periods.	The	data	for	
this	variable	are	from	the	Penn	World	Table	9.0	(Feenstra	et	al.,	2015)	and	the	World	Bank3.	The	
main	 explanatory	 variable	 of	 the	 model	 is	 INOV1	 (proxy	 for	 the	 business	 sector	 innovation	
activity),	and	the	proxies	used,	alternatively,	are	six	indicators	of	business	innovation	(see	table	
A.1	 in	the	Appendix)	that	constitute	measures	of	both	 inputs	and	outputs,	selected	based	on	
their	importance	in	explaining	the	relationship	between	business	sector	innovation	and	growth,	
as	reported	in	previous	literature	((Hasan	and	Tucci,	2010);	(Hong,	Oxley	and	McCann,	2012);	
(Wang,	 2013)).	 We	 expect	 a	 positive	 sign	 for	 all	 the	 estimated	 coefficients,	 in	 line	 with	 the	
endogenous	growth	predictions	of	(Romer,	1990)	and	(Jones,	1995)	and	the	results	of	studies	
such	as	(Freimane	and	Bāliņa,	2016),	(Hasan	and	Tucci,	2010)	and	(Wang,	2013).	
The	 control	 variables	 included	 were	 selected	 from	 the	 theoretical	 and	 empirical	 growth	

literature	(e.g.	(Romer,	1986);	(Barro	and	Sala‐i‐Martin,	2004);	(Hasan	and	Tucci,	2010)).	Initial	
real	GDP	per	capita	(lnGDPrpc)	is	expected	to	deliver	a	negative	estimated	coefficient	based	on	
the	convergence	predictions	of	(Solow,	1956)	and	technology	diffusion	growth	models	(Weil,	
2013)	 with	 poorer	 countries	 recording	 higher	 real	 GDP	 per	 capita	 growth	 rates.	 A	 positive	
relationship	 is	 expected	 between	 investment	 in	 physical	 capital	 (GCF)	 and	 economic	 growth	
based	on	(Solow,	1956)	and	also	between	investment	in	human	capital	(HC)	and	growth	as	this	
allows	for	higher	production	of	new	ideas/projects	and	the	accumulation	of	knowledge,	leading	
to	higher	productivity	and	(Romer,	1990)).	Government	consumption	(Gov_cons)	is	expected	to	
present	a	negative	sign,	assuming	that	these	are	non‐productive	expenditures:	(Mitchel,	2005)	

																																																													
3	The	data	up	to	2014	is	from	PWT	9.0	(Feenstra,	Inklaar	and	Timmer,	2015),	the	values	for	2015	were	
calculated	based	on	per	capita	real	GDP	growth	from	the	World	Bank.	



	 		
Cláudia	Caseiro,	Marta	Simões	 9	

argues	that	each	monetary	unit	that	the	government	spends	means	less	money	in	the	productive	
sector	 of	 the	 economy,	 which	 reduces	 the	 rate	 of	 output	 growth.	 The	 expected	 relationship	
between	globalisation	(Glob)	and	growth	is	ambiguous.	According	to	(Samimi	and	Jenatabadi,	
2014),	globalisation	allows	countries	to	allocate	resources	more	efficiently,	as	well	as	benefit	
from	economies	of	scale	and	cost	reductions,	which	leads	to	higher	growth.	On	the	other	hand,	
globalisation	 has	 negative	 effects	 on	 growth	 in	 countries	 with	 weaker	 institutions	 and	 with	
political	 instability	and	also	in	countries	specializing	in	activities	that	are	not	effective	in	the	
process	of	globalization.	
The	second	empirical	model	(model	2)	is	given	by	equation	(2):	

	
, 	 	 	 , 	 	 	 	 	 , 	 	 	 _

	 	 	 	 1 	 	 1	 	 	 2	 	 	 	 	 	 	 	 			(2)	
	

where:	δ0	–	Constant;	ui,t	–	error	term.;	i	–	countries	(i	=	1,	2,	…,	N).	t	–	years	(2008‐2015).	
	
In	model	2	the	dependent	and	control	variables	are	the	same	as	those	in	model	1,	although	the	

proxy	for	human	capital	now	corresponds	to	the	percentage	of	the	population	aged	15	to	64	
years	old	with	upper	secondary,	non‐tertiary	post‐secondary	and	higher	education	(tertiary)	
due	 data	 availability.	 The	 frequency	 of	 the	 data	 used	 is	 yearly	 since	 the	 innovation	 proxies	
related	to	SMEs	(INOV2)	are	only	available	for	a	limited	number	of	years,	from	2008	until	2015.	
According	 to	 the	 studies	 reviewed,	 a	 positive	 relationship	 is	 expected	 between	 the	 various	
business	sector	innovation	indicators	and	output	growth	((OECD,	2009);	(Hall,	2011);	(Mohnen	
and	 Hall,	 2013),	 (Moreno	 and	 Surinach,	 2014)).	 However,	 different	 types	 of	 innovations	 are	
expected	 to	 have	 different	 quantitative	 impacts.	 The	 indicators	 of	 the	 different	 types	 of	
innovation	 used	 aggregate	 product	 and	 process	 innovations	 into	 a	 single	 indicator	 and	
marketing	and	organisational	innovations	in	another.	We	thus	expect	that	the	former	will	have	a	
more	significant	impact	than	the	latter,	but	may	not	contribute	to	the	explanation	of	economic	
growth	due	to	the	inclusion	of	process	innovations,	in	light	of	the	conclusions	of	(Hall,	2011)	and	
(Moreno	 and	 Surinach,	 2014).	 Marketing	 and	 organisational	 innovations	 might	 even	 have	 a	
negative	impact	on	growth,	according	to	(Tavassoli	and	Karlsson,	2015).	
In	both	models	1	and	2	the	variables	D1	and	D2	correspond	to	time	dummies,	 included	to	

consider	the	potential	impact	of	the	financial	crisis	on	real	GDP	per	capita	growth.	In	model	1,	D1	
assumes	 the	 value	 1	 for	 the	 five‐year	 period	 2005‐2010	 and	 0	 otherwise;	 and	 variable	 D2	
assumes	the	value	1	for	the	last	five	years	(2010‐2015)	and	0	otherwise.	In	model	2,	D1	assumes	
the	value	1	for	2009,	2010,	2011	and	2012,	0	otherwise;	and	variable	D2	assumes	the	value	1	for	
2013,	2014	and	2015	and	0	otherwise.	

Methodology	and	results	

The	empirical	models	were	estimated	using	the	econometric	package	Gretl	version	2017d.	As	
a	preliminary	analysis	to	determine	the	appropriate	estimation	method	in	a	static	panel	context,	
we	applied	the	usual	diagnostic	tests	to	choose	between	pooled	OLS,	 fixed	effects	or	random	
effects:	 the	F‐test,	 the	Breusch‐Pagan	 test	 (Breusch	and	Pagan,	1980)	and	 the	Hausman	test	
(Hausman,	1978).	The	F‐test	allows	one	to	choose	between	the	pooled	OLS	and	the	fixed‐effects	
methods:	the	null	hypothesis	(H0)	considers	homogeneity	in	the	constant,	so	the	pooled	OLS	
method	is	the	most	adequate,	against	the	alternative	hypothesis	(H1)	that	there	is	heterogeneity	
in	the	constant,	so	the	most	appropriate	method,	in	this	case,	will	be	fixed	effects.	The	Breusch‐
Pagan	test,	 in	turn,	allows	to	decide	between	the	pooled	OLS	method	and	the	random	effects	
method:	 H0	 considers	 the	 homogeneity	 in	 the	 constant,	 so	 the	 most	 appropriate	 method	 is	
pooled	 OLS;	 H1	 admits	 heterogeneity	 between	 countries	 (not	 constant	 over	 time),	 so	 the	
random	 effects	 method	 is	 the	 most	 appropriate.	 Finally,	 the	 Hausman	 test	 allows	 to	 decide	



10
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

between	 fixed	 effects	 and	 random	 effects:	 H0	 considers	 heterogeneity	 across	 countries	 (not	
constant	over	time),	so	the	random	effects	method	is	the	most	appropriate;	H1	also	considers	
heterogeneity	across	countries,	but	constant	over	time,	so	that	the	fixed	effects	method	is	the	
most	appropriate.	We	performed	these	tests	for	the	baseline	specifications	given	by	equations	
(1)	and	(2)	and	the	alternative	innovation	indicators	considering	the	whole	sample	of	28	EU	
member	states.	According	to	the	results	for	F‐test,	the	fixed‐effects	method	is	preferable	to	the	
pooled	 OLS	 method.	 According	 to	 the	 Breusch‐Pagan	 test,	 the	 random	 effects	 is	 the	 most	
appropriate.	Finally,	from	the	results	of	the	Hausman	test,	we	concluded	that	the	fixed‐effects	
method	is	preferable	to	the	random‐effects	method.	Overall	these	diagnostic	tests	thus	indicate	
that	the	most	appropriate	estimation	method	for	models	1	and	2	considering	the	whole	sample	
is	the	fixed	effects	method.	We	also	carried	out	the	diagnostic	tests	for	the	sub‐samples,	groups	1	
and	2,	and	both	empirical	models.	For	model	1,	group	1,	the	most	appropriate	method	is	pooled	
OLS	 for	all	 innovation	 indicators,	except	 for	Trademark_APP	(fixed	effects);	when	group	2	 is	
considered,	 in	model	1	the	most	appropriate	methods	are:	random	effects	for	RDEpc;	pooled	
OLS	for	RD_Staff,	RES_RD	and	Patent_APP;	and	fixed	effects	for	Patent_G	and	Trademark_APP.	
For	model	2,	the	fixed	effects	method	is	the	most	appropriate	estimation	method	for	groups	1	
and	24.	

Main	results	of	the	estimation	of	model	1	

Table	1	presents	the	results	of	the	estimation	of	model	1	for	the	whole	sample	using	fixed	
effects.	In	each	column,	a	different	indicator	of	business	sector	innovation	is	considered.	Table	
A.2	in	the	Appendix	contains	the	results	of	the	estimation	for	group	1	and	table	A.3	contains	the	
results	for	group	2.	
Regarding	the	control	variables,	 the	results	obtained	for	the	whole	sample	 for	 lnGDPrpci,t‐1	

confirm	the	predictions,	negative	and	statistically	significant	coefficients	in	all	columns,	and	thus	
the	convergence	predictions	of	(Solow,	1956)	and	technology	diffusion	models.	These	results	
also	apply	to	groups	1	and	2.	The	estimated	coefficients	for	the	investment	rate	also	confirm	
predictions	when	the	EU28	or	group	2	are	considered,	with	a	positive	and	statistically	significant	
sign	(except	column	V	for	EU28	and	VI	for	group	1).	The	estimated	coefficients	for	human	capital	
for	 the	 EU28	 are	only	 statistically	 significant	 in	 columns	 II	 and	 V.	 In	 group	 1	 the	 estimated	
coefficients	confirm	the	predictions,	positive	and	statistically	significant,	except	when	trademark	
applications	are	used.	For	group	2,	we	obtain	a	negative	(although	not	statistically	significant)	
relationship	between	human	capital	and	economic	growth	in	columns	I	and	VI.	In	the	remaining	
cases,	we	get	positive	coefficients	but	not	statistically	significant.	As	 for	public	consumption,	
some	of	the	results	are	not	in	line	with	the	predictions:	for	the	EU28	the	estimated	coefficient	is	
negative,	 but	 not	 statistically	 significant	 in	 columns	 I,	 II,	 III	 and	 V,	 it	 is	 positive	 but	 not	
statistically	significant	in	column	IV	and	negative	and	significant	in	column	VI.	When	the	sample	
is	restricted	to	group	1,	the	results	show	that	an	increase	in	public	consumption	has	a	negative	
and	statistically	significant	impact	on	the	growth	rate	of	real	GDP	per	capita.	When	the	sample	is	
restricted	to	group	2,	the	results	indicate	that	the	relationship	between	public	expenditure	and	
growth	is	negative	for	columns	I,	 II,	 III,	 IV	and	VI	but	not	statistically	significant.	The	results	
found	for	the	EU28	show	that	the	globalisation	index	has	a	positive	and	statistically	significant	
impact	on	growth,	supporting	the	predictions	of	(Samimi	and	Jenatabadi,	2014).	These	results	
continue	to	apply	for	groups	1	and	2,	with	a	few	exceptions.	Overall,	these	results	indicate	that	
our	growth	regression	is	correctly	specified	since	the	majority	of	the	results	are	in	 line	with	
evidence	from	previous	empirical	growth	studies.	
Regarding	the	time	dummy	D1	(2005‐2010),	for	the	EU28	and	the	subsamples	the	majority	of	

results	show	a	negative	(and	generally	statistically	significant	 impact)	on	growth.	As	 for	 the	
results	 with	 the	 time	 dummy	 D2	 (2010‐1015),	 for	 the	 EU28	 and	 group	 1,	 in	 general,	 the	

																																																													
4	These	results	are	available	from	the	authors.	



	 		
Cláudia	Caseiro,	Marta	Simões	 11	

estimated	coefficient	is	not	statistically	significant,	while	for	group	2	the	coefficient	is	positive	
and	usually	statistically	significant.	These	results	indicate	that	the	sovereign	debt	crisis	period	
did	not	affect	all	the	EU	countries	in	the	same	way.	
Turning	now	to	the	results	for	our	main	explanatory	variables,	when	inputs	of	business	sector	

innovation	 are	 used	 (R&D	 expenditure,	 total	 R&D	 staff	 and	 R&D	 researchers),	 the	 results	
obtained	 with	 the	 EU28	 (Table	 1,	 columns	 I,	 II	 and	 III)	 indicate	 that	 these	 inputs	 do	 not	
contribute	to	growth	over	the	period	1990‐2015,	since	the	respective	estimated	coefficients	are	
not	statistically	significant.	These	results	remain	for	group	1,	as	can	be	seen	from	Table	A.2,	
columns	I,	II	and	III.	The	results	obtained	for	group	2	(Table	A.3)	are	different:	an	increase	in	
total	 R&D	 personnel	 and	 R&D	 researchers	 has	 a	 negative	 and	 statistically	 significant	 effect,	
contrary	to	predictions.	The	lack	of	statistical	significance	for	group	1	could	be	because	these	
inputs,	in	reality,	do	not	result	in	successful	innovations,	with	economic	value,	and	thus	do	not	
lead	to	productivity	increases	and	faster	growth.	It	could	also	be	the	case	that	there	is	a	time	lag	
between	R&D	activities	and	innovations	becoming	effective	(Wang,	2013).	As	for	group	25,	these	
countries	might	be	devoting	R&D	expenditures	to	the	imitation	of	innovations	from	developed	
countries	that	are	not	adequate	to	their	reality	and	thus	do	not	permit	faster	growth.	According	
to	(Galor,	2005),	the	growth	process	is	characterized	by	different	stages	in	terms	of	the	relative	
importance	of	physical,	human	capital	and	knowledge	accumulation,	so	countries	 in	group	2	
might	be	deviating	scarce	resources	from	more	important	activities	as	far	as	their	stage	in	the	
growth	process	is	concerned,	which	results	in	less	growth.	
When	outputs	of	business	sector	innovation	are	introduced	as	the	main	explanatory	variable	

(patent	applications,	patents	granted	and	trademark	applications	registrations),	the	results	for	
the	 EU28	 (Table	 1,	 columns	 IV,	 V	 and	 VI)	 show	 that	 there	 is	 a	 positive	 impact	 on	 growth,	
however,	 only	 trademark	 applications	 produce	 statistically	 significant	 effects6.	 For	 instance,	
patents	do	not	contribute	to	the	explanation	of	growth,	contrary	to	the	conclusions	of	authors	
like	(Hasan	and	Tucci,	2010).	This	result	can	be	due	to	a	number	of	reasons,	including	that	many	
innovations	do	not	result	in	patens,	in	order	to	effectively	remain	secret	and	not	be	copied,	so	
that	technical	details	are	not	revealed;	and	also	because	some	firms	use	patents	to	prevent	other	
firms	from	placing	an	invention	in	the	market	(Hasan	and	Tucci,	2010);	and	finally,	because	the	
costs	of	patenting	are	high,	becoming	unfeasible	 for	small	 firms	(Wang,	2013).	These	results	
remain	for	group	1,	according	to	table	A.2,	columns	IV,	V	and	VI.	When	the	sample	is	restricted	to	
group	 2,	 the	 results	 obtained,	 as	 shown	 in	 Table	 A.3,	 columns	 IV,	 V	 and	 VI,	 are	 in	 general	
statistically	significant	(except	patents	granted).	In	the	case	of	patent	applications,	the	coefficient	
obtained	is	in	line	with	the	findings	of	(Hasan	and	Tucci,	2010):	an	increase	in	the	number	of	
patent	applications	produces	positive	and	significant	effects	on	the	real	GDP	per	capita	growth	
rate.	
On	the	other	hand,	contrary	to	predictions,	registrations	of	trademark	applications	lead	to	a	

decrease	in	growth.	Trademark	application	records	are	not	associated	with	technology	and	are	
relevant	 to	 the	 services	 sector,	 according	 to	 the	 EIS	 (2017).	 In	 countries	 such	 as	 Bulgaria,	
Croatia,	Hungary,	Poland	and	Romania	(transition	countries)	 this	 indicator	does	not	become	
relevant	and	may	indicate	that	trademark	registrations	do	not	lead	to	productivity	increases,	as	
evidenced	 by	 the	results	of	 the	estimation	of	 the	model	 leaving	 out	 the	outliers	Cyprus	and	
Malta.	

																																																													
5	 Consisting	 of	 some	 of	 countries	 considered	 by	 the	 IMF	 as	 developing	 countries:	 Bulgaria,	 Croatia,	
Hungary,	Poland	and	Romania.	
6	Cyprus,	Malta	and	Luxembourg	represent	outliers	in	the	Trademark_APP	indicator	and	were	therefore	
withdrawn	from	the	sample	to	check	the	 impact	on	 the	results:	 trademark	applications	are	no	 longer	
statistically	 significant	 in	 any	 of	 the	 samples	 under	 analysis,	 which	 may	 indicate	 that	 if	 the	 legal	
framework	does	not	protect	intellectual	property,	this	indicator	is	not	relevant	for	growth.	These	results	
are	available	from	the	authors.	



12
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

Looking	at	the	AIC,	BIC	and	HQ	information	criteria	for	the	EU28	models	in	Table	1,	we	can	
conclude	 that	 the	 model	 that	 best	 explains	 growth	 is	 the	 model	 that	 considers	 patent	
applications	since	the	criteria	present	the	lowest	value.	The	same	applies	for	groups	1	and	2	
(tables	A.2	and	A.3,	respectively)7.	
	
Table	1.	Results	of	the	estimation	of	model	1	(fixed	effects),	28	EU	countries,	1990‐2015	

Explanatory	
variables	

Dependent	variable	‐	a	Growth	rate	of	real	GDP	per	capita	

I		 II		 III		 IV		 V		 VI		

Constant		 0.7386	***		 0.8976	***		 0.8894	***		 0.948	***		 1.1756	***		 1.1877	***		

lnGDPrpci,t‐1		 −0.0932	
***		

−0.1167	
***		

−0.1169	
***		

−0.1249	***		 −0.1456	
***		

−0.1297	
***		

GCFi,t		 0.1891	***		 0.216	***		 0.2292	***		 0.2677	***		 0.0475		 0.2278	***		

HCi,t‐1		 0.0042		 0.00787	*		 0.00755		 0.0060		 0.00885	*		 0.000794		

Gov_consi,t		 −0.0744		 −0.0552		 −0.0434		 0.0091		 −0.0191		 −0.3460	
***		

Globi,t		 0.0019	***		 0.0023	***		 0.0024	***		 0.0025	***		 0.0027	***		 0.0018	***		

RDEpci,t		 −1.0516e‐5		 ‐		 ‐		 ‐		 ‐		 ‐		

RD_Staffi,t‐1		 ‐		 −3.1771e‐6		 ‐		 ‐		 ‐		 ‐		

RES_RDi,t‐1		 ‐		 ‐		 −3.4140e‐8		 ‐		 ‐		 ‐		

Patent_APPi,t		 ‐		 ‐		 ‐		 0.000121		 ‐		 ‐		

Patent_Gi,t		 ‐		 ‐		 ‐		 ‐		 0.0001073		 ‐		

Trademark_APPi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 4.6620e‐5	
***		

D1		 −0.00995			 −0.0111		 −0.0120		 −0.0146	*		 0.000197		 −0.00953		

D2		 0.0143	*		 0.0148		 0.0135		 0.0152	*		 ‐		 0.0231	**		

LSDV	R2		 0.6097		 0.6361		 0.6369		 0.6649		 0.7718		 0.6644		

R2	dentro		 0.4591		 0.5561		 0.5429		 0.6189		 0.7437		 0.5465		

AIC	(Akaike)		 −635.3512		 −623.6562		 −605.0257		 −648.7218		 −554.6305		 −536.4121		

BIC	(Schwarz)		 −531.8441		 −519.6036		 −502.0724		 −542.8226		 −459.4830		 −439.1948		

Hannan‐Quinn		 −593.2916		 −581.3732		 −563.1937		 −605.6875		 −516.0262		 −496.9802		

Notes:	Column	I	‐	R&D	expenditures;	column	II	‐	total	staff	in	R&D	activities;	column	III	‐	researchers	in	R&D	
activities;	column	IV	‐	patent	applications;	column	V	‐	granted	patents;	column	VI	‐	registration	of	trademark	
applications.	***,	**,	*	indicate	statistical	significance	at	the	1,	%	and	10%	levels,	respectively.		

Source:	Own	elaboration.	
	
Taking	the	results	as	a	whole,	we	can	conclude	these	differ	according	to	the	group	of	countries	

and	time	period	under	analysis.	Between	1990	and	2015,	in	the	EU28	and	group	1,	inputs	do	not	
contribute	to	the	explanation	of	aggregate	output	behaviour,	while	only	R&D	personnel	and	R&D	

																																																													
7	We	also	estimated	model	1	with	data	for	the	period	2008‐2015.	The	results	obtained	with	the	inputs	
indicators	show	that	R&D	expenditures	have	a	relevant	role	in	explaining	an	increase	in	the	growth	rate	of	
real	GDP	per	capita,	as	predicted.	The	results	with	the	outputs	indicators	are	different	depending	on	the	
sample	under	analysis.	For	the	EU28,	registrations	of	trademark	applications	are	no	longer	statistically	
significant,	with	patents	granted	explaining	economic	growth.	The	coefficients	obtained	with	outputs	in	
group	1	show	that	patent	applications	and	trademark	application	have	significant	negative	and	positive	
impacts,	respectively.	The	results	with	outputs	for	group	2	show	that	patent	applications	in	this	period	
fails	to	explain	the	behavior	of	output.	These	results	are	available	from	the	authors.	



	 		
Cláudia	Caseiro,	Marta	Simões	 13	

researchers	play	a	relevant	role	in	explaining	growth	(but	with	a	negative	sign)	in	group	2.	On	
the	other	hand,	restricting	the	period	to	2008‐2015,	R&D	expenditures	present	a	positive	and	
statistically	 significant	 role.	 As	 for	 outputs,	 between	 1990	 and	 2015	 trademark	 applications	
contribute	to	the	explanation	of	growth	due	to	the	outliers	Cyprus,	Luxembourg	and	Malta	(and	
also	patent	applications	in	group	2).	Between	2008	and	2015,	the	results	relative	to	the	longer	
period	change:	for	the	EU28,	patents	granted	have	positive	and	relevant	impacts	on	growth	and	
brands	 lose	 their	 significance;	 for	 group	1,	 patent	applications	have	 negative	 and	significant	
growth	impacts;	and	for	group	2	patent	applications	lose	statistical	significance.	

Main	results	of	the	estimation	of	model	2	

Table	 2	 presents	 the	 results	 of	 the	 estimation	 of	 model	 2	 for	 the	 EU28	 with	 fixed	 effects	
considering	each	indicator	of	business	sector	innovation	alternatively.	Tables	A.4	and	A.5	in	the	
Appendix	contain	the	results	for	groups	1	and	2,	respectively.	
Concerning	the	control	variables,	according	to	Table	2	for	the	EU28	initial	real	GDP	per	capita	

presents	 a	 negative	 and	 statistically	 significant	 coefficient	 in	 all	 models	 (at	 the	 1%	 level	 of	
significance),	 as	 expected,	 and	 the	 same	 applies	 for	 groups	 1	 and	 2	 (Tables	 A.4	 and	 A.5,	
respectively).	The	results	obtained	for	the	 investment	rate	are	also	as	expected:	we	obtain	a	
positive	 relationship	 and	 statistically	 significant	 for	 the	 EU28	 and	 group	 2.	 The	 relationship	
between	human	capital	and	the	real	per	capita	GDP	growth	rate	for	the	EU28	and	both	groups	1	
and	2	is	positive	and	statistically	significant.	The	results	obtained	for	public	consumption	are	
negative	and	statistically	significant	for	the	EU28	and	group	2.	However,	when	the	sample	is	
restricted	to	group	1,	the	results	are	not	statistically	significant	(some	exceptions	apply).	Finally,	
as	far	as	globalisation	is	concerned,	the	results	are	positive	and	statistically	significant	in	the	
EU28	and	group	2	as	expected,	but	negative	and	not	statistically	significant	in	group	1.	
The	time	dummy	D1	(2009‐2012)	does	not	influence	economic	growth	in	the	EU28	sample,	so	

that	the	2007‐08	financial	crisis	shows	no	negative	growth	effects.	However,	the	latter	occurs	in	
group	 1,	 while	 in	 group	 2	 (Table	 A.5)	 the	 estimated	 coefficient	 is	 positive	 and	 statistically	
significant.	The	results	for	the	time	dummy	D2	(2013‐15)	for	the	EU28	(Table	2)	indicate	that	
this	was	a	post‐crisis	recovery	period,	with	a	positive	and	in	most	cases,	statistically	significant	
coefficient	for	D2.	The	coefficients	obtained	for	group	1	(Table	A.4)	contrast	with	the	previous	
ones	since	negative	coefficients	were	obtained	although	without	statistical	significance.	
Turning	now	to	the	results	of	the	estimation	of	model	2	with	inputs	indicators	(innovation	

expenditure	and	training	of	employees),	the	results	obtained	for	the	EU28	(Table	2,	columns	III	
and	IV)	show	that	only	an	increase	in	the	percentage	of	firms	that	promote	employee	training	
contributes	positively	to	growth.	The	results	obtained	for	group	1	(table	A.4,	columns	III	and	IV)	
remain	the	same.	When	the	sample	is	restricted	to	group	2,	the	results	indicate	that	no	input	
indicator	contributes	to	the	explanation	of	the	growth	rate	of	real	GDP	per	capita,	as	can	be	seen	
in	table	A.5,	columns	III	and	IV.	The	lack	of	statistical	significance	of	innovation	expenditure	in	
group	1	can	be	justified	by	the	relatively	small	investment	in	innovation	in	the	form	of	dedicated	
machinery	and	equipment	(within	this	group	only	Germany	and	Sweden	have	a	ratio	 in	 this	
indicator	 that	 is	higher	 than	 the	 EU	average),	 which	 is	not	enough	 to	 spark	 innovation	 that	
promotes	 productivity	 improvements.	 For	 group	 2,	 the	 lack	 of	 significance	 of	 innovation	
expenditure	can	be	justified	by	the	lack	of	absorptive	capacity.	This	group	includes	transition	
countries	with	a	low	level	of	technological	and	scientific	development,	and	thus,	investments	in	
innovation	 (equipment	 and	 machinery)	 do	 not	 translate	 into	 successful	 innovations.	
Additionally,	the	lack	of	statistical	significance	for	the	percentage	of	firms	that	promote	training	
to	develop/upgrade	the	ICT	skills	of	employees	for	group	2	can	be	justified	by	the	fact	that	this	
type	of	 innovation	input	 is	more	relevant	 in	more	advanced	economies,	according	to	the	EIS	
Methodology	Report	2017,	so	it	does	not	translate	into	productivity	gains	in	this	second	group.	



14
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

As	for	the	results	with	indicators	of	outputs	of	business	sector	innovation	(product/	process	
innovations,	 marketing/	 organizational	 innovations,	 SMEs	 that	 innovate	 internally	 and	 SMEs	
that	cooperate	with	others),	 the	results	reveal	 that	 for	the	EU28	all	 these	 indicators	present	
negative	 coefficients,	 but	 only	 statistically	 significant	 for	 marketing	 and/or	 organizational	
innovations.	 The	 results	 for	group	 1	 are	 not	statistically	 significant.	 For	 group	2	 (Table	 A.5,	
columns	V,	VI,	VII	and	VIII),	the	results	show	that	outputs	have	statistically	significant	negative	
effects	on	growth	(except	the	percentage	of	SMEs	that	cooperate	with	others,	without	statistical	
significance).	The	lack	of	statistical	significance	of	the	outputs	indicators	for	group	1,	according	
to	the	EIS	Methodology	Report	of	2017	lies	on	the	fact	that	adoption	of	product	and	process	
innovations	 is	especially	relevant	 to	 the	manufacturing	 industry	and	since	 the	present	study	
studies	 business	 sector	 innovation	 (which	 encompasses	 all	 business	 sectors,	 not	 only	
manufacturing),	it	is	not	enough	to	explain	growth	in	this	group	of	countries,	especially	because	
this	 is	 not	 the	 most	 important	 sector	 of	 activity	 in	 advanced	 countries.	 Another	 possible	
explanation,	this	time	for	the	lack	of	significance	of	marketing/organisational	innovations	is	that	
the	adoption	of	new	organisational	methods	requires	a	period	of	adaptation	that	often	does	not	
translate	into	success,	because	in	many	firms	there	is	"resistance	to	change".	Third,	the	lack	of	
significance	 of	 cooperation	 activities	 related	 to	 innovation	 among	 group	 1	 firms	 could	 be	
explained	by	the	fact	that	this	type	of	activity	has	not	enough	relevance	in	the	overall	activities	of	
the	 firms.	On	the	other	hand,	a	possible	 justification	for	the	 lack	of	relevance	of	cooperation	
activities	 between	 group	2	 firms	 may	 be	 due	 to	 the	 fact	 that	many	 cooperation	 agreements	
require	 complex	 innovation	 levels,	 which	 in	 the	 case	 of	 countries	 such	 as	 Bulgaria,	 Croatia,	
Hungary,	 Poland	 and	 Romania	 may	 not	 be	 successful	 as	 these	 countries	 lack	 human	
capital/more	 skilled	 workers.	 The	 negative	 impact	 of	 output	 indicators	 (product/	 process	
innovations,	 marketing/	 organizational	 innovations	 and	 SMEs	 that	 innovate	 internally)	 for	
group	2	may	indicate	that	the	resources	devoted	to	innovation	activities,	such	as	skilled	workers,	
may	be	necessary	for	other	more	fundamental	activities	in	terms	of	the	growth	process	of	these	
countries,	resulting	in	slower	growth.	
We	 also	 considered	 as	 explanatory	 variables	 exports	 of	 medium	 and	 high‐tech	 products,	

exports	 of	 knowledge‐intensive	 services	 and	 sales	 of	 new‐to‐firm	 and	 new‐to‐market	
innovations.	The	results	for	the	EU28	countries	and	both	country	groups	are	not	statistically	
significant	(Tables	2,	A.4	and	A.5,	columns	I,	II	and	IX,	respectively).	One	possible	justification	for	
this	lack	of	significance	is	that	exports	of	these	types	of	products	and	services	are	just	a	small	
share	of	total	exports,	and	the	same	applies	to	sales	of	innovations	as	a	share	of	total	sales.	
From	the	AIC,	BIC	and	HQ	information	criteria	for	the	different	models	using	the	EU28	(table	

4),	we	conclude	that	the	best	model	corresponds	to	the	one	considering	the	percentage	of	firms	
that	promote	employees	training	as	the	innovation	variable.	The	same	is	true	for	group	1	(Table	
A.4).	 For	 group	 2,	 the	 best	 model	 is	 the	 one	 using	 the	 percentage	 of	 SMEs	 that	 innovate	
internally	(Table	A.5).	
Taking	the	results	as	a	whole,	again,	the	role	of	business	sector	innovation	for	growth	differs	

according	to	the	group	of	countries	and	the	innovation	indicator	considered.	In	most	cases	for	
the	EU28	and	group	1,	the	results	with	inputs	show	that	only	employee	training	has	a	relevant	
(and	positive)	role;	while	for	group	2	inputs	do	not	have	a	growth	effect.	In	terms	of	outputs,	for	
the	 EU28	 only	 marketing/organisational	 innovations	 have	 a	 significant	 (although	 negative)	
effect;	 for	 group	 1	 outputs	 do	 not	 play	 a	 relevant	 role	 and	 for	 group	 2	 product/	 process	
innovations,	 marketing/organisational	 innovations	 and	 internally	 innovating	companies	 have	
negative	and	significant	impacts	on	economic	growth.	
	
	 	



	 		
Cláudia	Caseiro,	Marta	Simões	 15	

Table	2.	Results	of	the	estimation	of	model	2	(fixed	effects),	28	EU	countries,	2008‐2015	

Explanatory	
variables	

Dependent	variable	‐	a	Growth	rate	of	real	GDP	per	capita	
I		 II		 III		 IV		 V		 VI		 VII		 VIII		 IX		

Constant	 2.64***		 2.66***		 2.62***	 2.71***		 2.69***		 2.76***		 2.70***		 2.66***		 2.61***		
lnGDPrpci.t‐1		 −0.3653***		 −0.3736	

***		
−0.369	
***		

−0.3839	
***		

−0.3737	
***		

−0.3868	
***		

−0.3753	
***		

−0.3714	***		 −0.3711	
***		

GCFi.t		 0.6476	***		 0.6504	
***		

0.6708	
***		

0.615	***		 0.6615	
***		

0.7029	
***		

0.6674	
***		

0.6524	***		 0.6389	
***		

HCi.t‐1		 0.0094	***		 0.0097	
***	

0.0095	
***		

0.0096	
***		

0.0097	
***		

0.0102	
***		

0.0099	
***		

0.0097	***		 0.001	***		

Gov_consi.t		 −1.1774	
***		

−1.1703	
***		

−1.081***		 −1.218***		 −1.124***		 −1.149***		 −1.1***		 −1.16***		 −1.19***		

Globi.t		 0.007***		 0.007***		 0.007***		 0.007***		 0.007***		 0.007***		 0.007***		 0.007***		 0.007	***		
Export_HTi.t		 −0.000518		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
Export_Servi.t		 ‐		 0.000674		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
INOV_Ei.t		 ‐		 ‐		 −0.0121		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
TRAIN_EMPi.t		 ‐		 ‐		 ‐		 0.0032	**		 ‐		 ‐		 ‐		 ‐		 ‐		
ProdProc_INOVi.t		 ‐		 ‐		 ‐		 ‐		 −0.000554		 ‐		 ‐		 ‐		 ‐		
MarkOrg_INOVi.t		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.0011	*		 ‐		 ‐		 ‐		
INO_SME.t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.0012	 ‐		 ‐		
COOP_INOVi.t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.0002	 ‐		
Sales_INOVi.t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 0.0008	
D1		 0.0106		 0.0107		 0.0085		 0.0102		 0.0105		 0.0095		 0.0095		 0.0107		 0.0103		
D2		 0.0279	*		 0.0282	*		 0.026		 0.0267	*		 0.0259		 0.0251		 0.024		 0.0277	*		 0.0284	*		
LSDV	R2		 0.6122		 0.612		 0.6155		 0.6234		 0.6127		 0.6195		 0.616		 0.6118		 0.6133		
R2	within		 0.5437		 0.5434		 0.5475		 0.5568		 0.5442		 0.5522		 0.548		 0.5431		 0.5449		
AIC	(Akaike)		 −765.103	 −764.99	 −766.73	 −770.81	 −765.33	 −768.79	 −766.99	 −764.87	 −765.64	
BIC	(Schwarz)		 −647.09	 −646.98		 −648.72	 −652.80	 −647.32	 −650.78	 −648.98	 −646.85	 −647.62	
Hannan‐Quinn	
(HQ)	

−717.33	 −717.2	 −718.95	 −723.03	 −717.56	 −721.02	 −719.21	 −717.09	 −717.86	

Notes:	Column	 I	 ‐	exports	of	medium	and	high‐tech	products;	column	 II	 ‐	exports	of	knowledge‐intensive	
services;	 column	 III	 ‐	 innovation	 expenditures;	 column	 IV	 ‐	 training	 of	 employees;	 column	 V	 ‐	 product/	
process	 innovations;	 column	 VI	 ‐	 marketing/	 organizational	 innovations;	 column	 VII	 ‐	 %	 of	 SMEs	 that	
innovate	internally;	column	VIII	–%	of	SMEs	cooperating	with	others;	column	IX	‐	sales	of	innovations.	***,	**,	
*	indicates	that	the	coefficients	are	significant	at	1,	5	and	10%	levels,	respectively.		

Source:	Own	elaboration.	

CONCLUSION	

This	study	investigates	the	relationship	between	business	sector	innovation	and	real	GDP	per	
capita	growth	in	the	28	European	Union	member	states	using	static	panel	data	methods	for	the	
estimation	of	two	growth	regressions,	assessing	how	inputs	and	outputs	of	 firms'	 innovation	
activities	play	a	role	in	economic	growth	and	distinguishing	between	two	groups	of	countries	
identified	 according	 to	 their	 innovation	 performance	 (above	 or	 below)	 relative	 to	 the	 EU28	
average	in	terms	of	selected	business	sector	innovation	indicators.	
The	results	obtained	indicate	that	the	role	of	business	sector	innovation	in	economic	growth	

varies	according	to	the	sample	of	countries	and	the	period	under	analysis,	as	well	as	according	to	
the	 proxy	 used	 to	 measure	 business	 sector	 innovation	 activities.	 Overall,	 the	 innovation	
indicators	that	play	a	relevant	role	in	explaining	economic	growth	for	group	1,	the	group	with	
the	highest	innovation	performance	in	the	EU,	have	a	positive	impact	on	the	growth	rate	of	real	



16
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

GDP	 per	 capita,	 except	 patent	 applications	 (2008	 ‐2015).	 On	 the	 other	 hand,	 group	 2,	
characterized	by	below‐average	performances	 in	business	sector	 innovation,	mainly	presents	
indicators	that	have	negative	growth	impacts	(R&D	personnel,	R&D	researchers	and	trademark	
registrations,	product/	process	innovations,	marketing/	organizational	innovations,	and	SMEs	
that	innovate	internally	between),	with	the	exception	of	patent	applications	for	the	1990‐2015	
period	 and	 R	 &	 D	 expenditure	 between	 2008	 and	 2015.	 One	 possible	 justification	 for	 this	
divergence	 lies	 in	 absorptive	 capacity,	 which	 is	 associated	 with	 the	 level	 of	 development	 of	
countries,	as	the	success	of	the	innovations	in	terms	of	increasing	productivity	and	generating	
faster	 growth	 depends	 on	 different	 aspects,	 namely	 human	 capital	 availability,	 accumulated	
knowledge,	adequate	infrastructure,	technological	and	financial	support/development.	Group	1	
includes	the	more	developed	member	states	(higher	real	GDP	per	capita)	thus	has	in	principle	
higher	 absorptive	 capacity,	 and	 so	 innovation	 in	 this	 group	 has	 more	 positive	 impacts,	 and	
higher	business	sector	innovation	indeed	translates	into	faster	growth.	
On	the	other	hand,	group	2	includes	less	developed/transition	member,	so	that	innovations	do	

not	translate	into	productivity	increases	because	they	lack	the	necessary	absorptive	capacity.	
The	 negative	 growth	 impacts	 obtained	 could	 also	 be	 the	 result	 of	 competition	 for	 scarce	
resources	between	different	types	of	activities,	with	different	weights	in	the	process	of	growth	of	
these	 countries,	 with	 business	 sector	 innovation	 activities	 using	 resources	 that	 would	 have	
higher	growth	returns	in	other	sectors	of	activity.	The	findings	also	allowed	us	to	conclude	that	
not	 all	 inputs	 and	 outputs	 of	 business	 innovation	 are	 relevant	 to	 the	 explanation	 of	 the	
behaviour	of	output	in	the	28	EU	member	states,	which	again	could	be	explained	by	the	lack	of	
certain	 types	 of	 absorptive	 capacity	 so	 that	 not	 all	 R&D	 activities	 carried	 out	 translate	 into	
successful	innovations.	
In	 terms	 of	 policy	 implications,	 public	 policies	 can	 promote	 innovation	 through	 different	

instruments,	e.g.	either	directly	by	providing	funds,	through	human	capital	formation,	enforcing	
patent	protection	laws,	through	tax	benefits	or	indirectly	through	regulatory	policies	((Griffith,	
2000);	 (OECD,	 2007)).	 Also,	 government	 policies	 can	 also	 provide	 incentives	 for	 firms	 to	
innovate.	In	light	of	the	results	obtained,	the	policy	design	should	be	different	depending	on	the	
target	group	of	countries.	First,	the	results	for	the	most	recent	period	support	policies	that	allow	
for	an	increase	in	R&D	expenditures	in	the	business	sector.	For	group	1,	the	results	suggest	that	
policies	that	promote	the	acquisition	of	machinery	and	equipment	for	R&D	activities	are	key	to	
growth.	For	group	2,	the	results	suggest	that	similar	to	(Pradhan,	Arvin,	Hall	and	Mahendhiran,	
2016),	 policies	 must	 ensure	 an	 efficient	 allocation	 of	 resources,	 avoiding	 that	 innovation	
competes	with	other	key	sectors	for	resources.	Besides,	investment	in	innovation	infrastructure	
is	still	necessary	for	business	sector	innovation	to	translate	into	productivity	gains	and	higher	
growth.	
Future	research	should	focus	on	a	deeper	understanding	of	the	differences	in	results	across	

samples	and	periods,	as	well	as	between	inputs	and	outputs	of	business	sector	innovation.	

REFERENCES	

Barro,	R.	J.	&	Sala‐i‐Martin	X.	(2004).	Economic	Growth.	Cambridge,	Mass:	MIT	Press.	
Barro,	R.	&	Jong‐Wha	L.	(2013).	A	New	Data	Set	of	Educational	Attainment	in	the	World,	1950–
2010.	Journal	of	Development	Economics,	Vol.	104,	pp.	184‐198.	

Breusch,	T.	&	Pagan,	A.	 (1980).	The	Lagrange	Multiplier	Test	and	Its	Applications	to	Model	
Specification	in	Econometrics.	Review	of	Economic	Studies,	Vol.	47,	pp.	239–253.	

Feenstra,	R.C,	 Inklaar,	R.	&	Timmer,	M.P.	 (2015).	The	Next	Generation	of	 the	Penn	World	
Table.	American	Economic	Review,	Vol.	105,	pp.	3150‐3182.	

Freimane,	R.	and	Bāliņa	S.	 (2016).	Research	and	Development	Expenditures	and	Economic	
Growth	in	the	Eu:	A	Panel	Data	Analysis.	Economics	and	Business,	Vol.	29,	pp.	5‐11.	



	 		
Cláudia	Caseiro,	Marta	Simões	 17	

Galor,	O.	(2005).	From	Stagnation	to	Growth:	Unified	Growth	Theory,"	P.	Aghion	and	S.	Durlauf,	
Handbook	of	Economic	Growth.	Elsevier:	171‐293		

Goel,	R.K.	&	Ram,	R.	(1994).	Research	and	Development	Expenditures	and	Economic	Growth:	A	
Cross‐Country	Study.	Economic	Development	and	Cultural	Change,	Vol.	42,	pp.	403–411.	

Griffith,	 R.	 (2000).	 How	 Important	 Is	 Business	 R&D	 for	 Economic	 Growth	 and	 Should	 the	
Government	Subsidise	It?	The	Institute	for	Fiscal	Studies	Briefing	Note	No.,	12.	

Hall,	B.	H.	(2011).	Innovation	and	Productivity.	NBER	Working	Paper	No.,	17178.	
Hasan,	I.	and	Tucci,	C.	L.	 (2010).	The	Innovation‐Economic	Growth	Nexus:	Global	Evidence.	
Research	Policy,	Vol.	39,	pp.	1264‐1276.	

Hausman,	J.	(1978).		Specification	Tests	in	Econometrics.	Econometrica,		Vol.	46,	pp.	1251–1271.	
Hong,	S,	Oxley	L.	&	McCann,	P.	(2012).	A	Survey	of	the	Innovation	Surveys.	Journal	of	Economic	
Surveys,		Vol.	26,	pp.	420‐444.	

Jones,	C.	(2005).	"Growth	and	Ideas,"	P.	Aghion	and	S.	Durlauf,	Handbook	of	Economic	Growth.	
Elsevier:	1063‐111.	

Jones,	 C.	 I.	 (1995).	 R&D‐Based	 Models	 of	 Economic	 Growth.	 Journal	 of	 	 Political	 Economy,	
103(41):	759‐84.	

Lucas,	R.	(1988).	"On	the	Mechanics	of	Economic	Development."	Journal	of	Monetary	Economics,	
22(1):	3‐42.	

Maradana,	R.	P.,	Rudra,	P.	P.,	Saurav,	D.,	Kunal,	G.,		Manju,	J.	&	Debaleena,	C.	(2017).	"Does	
Innovation	 Promote	 Economic	 Growth?	 Evidence	 from	 European	 Countries."	 Journal	 of	
Innovation	and	Entrepreneurship,	6(1):	2‐23.	

Mitchel,	D.	J.	(2005).	"The	Impact	of	Government	Spending	on	Economic	Growth."	The	Heritage	
Foundation	Working	Paper	No.,	1831.	

Mohnen,	P.	&	Hall,	B.	H.	(2013).	"Innovation	and	Productivity:	An	Update."	Eurasian	Business	
Review,	3(1):	47‐65.	

Moreno,	R.	and	Surinach,	J.	(2014).	"Innovation	Adoption	and	Productivity	Growth:	Evidence	
for	Europe."	Research	Institute	of	Applied	Economics	Working	Paper	No.,	2014/13.	

OECD.	2007.	Innovation	and	Growth:	Rationale	for	an	Innovation	Strategy.	Paris:	OECD.	
____.	2009.	Innovation	in	Firms:	A	Microeconomic	Perspective.	Paris:	OECD.	
____.	2005.	Manual	De	Oslo:	Diretrizes	Para	a	Coleta	E	Interpretação	De	Dados	Sobre	Inovação	
Tecnológica.	Paris:	OECD.	

Pece,	A.	M.,	Simona,	O.	E.	O.,	Salisteanu,	F.	 (2015).	 "Innovation	and	Economic	Growth:	An	
Empirical	Analysis	for	Cee	Countries."	Procedia	Economics	and	Finance,	26:	461‐67.	

Pradhan,	R.	P.,	Mak	B.	A.,	Hall,	J.H.	&	Nair,	M.	(2016).	"Innovation,	Financial	Development	and	
Economic	Growth	in	Eurozone	Countries."	Applied	Economics	Letters,	23(16):	1141‐44.	

Romer,	P.	(1990).	"Endogenous	Technological	Change."	Journal	of	Political	Economy,	98(5):	S71‐
102.	

____.	1986.	"Increasing	Returns	and	Long‐Run	Growth."	Journal	of	Political	Economy,	94(5):	1002‐
37.	
Samimi,	 P.,	 &	 Jenatabadi,	 H.S.	 (2014).	 "Globalization	 and	 Economic	 Growth:	 Empirical	
Evidence	on	the	Role	of	Complementarities."	PLoS	ONE,	9(4):	e87824.	

Schmookler,	J.	(1966).	Invention	and	Economic	Growth.	Cambridge,	MA.:	Harvard	University	
Press.	
Solow,	R.	M.	(1956).	"A	Contribution	to	the	Theory	of	Economic	Growth."	Quarterly	Journal	of	
Economics,	70(1):	65‐94.	

Tavassoli,	S.	&	Karlsson,	C.	(2015).	"Persistence	of	Various	Types	of	Innovation	Analyzed	and	
Explained."	Research	Policy,	44:	1887‐901.	

Ulku,	H.	(2004).	"R&D,	Innovation,	and	Economic	Growth:	An	Empirical	Analysis."	International	
Monetary	Fund	Working	Papers	No.,	WP/04/185.	

Wang,	C.	 (2013).	"The	Long‐Run	Effect	of	 Innovation	on	Economic	Growth."	Working	Papers	
School	of	Economics,	UNSW,	Sydney	No.,	2052.	

Weil,	D.	N.	(2013).	Economic	Growth.	Boston:	Pearson/Addison	Wesley.	



18
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

	
	

APPENDIX	

Table	A.1.	Variables	in	the	empirical	models:	description	and	sources	

	 Variables	 Model		 Description	 Source		
Dependent	
Variable		 lnGDPrpci,t		 1	e	2		

Average	real	GDP	growth	rate	per	capita	at	constant	2011	
prices	(USD)	in	PPPs	 PWT		

Control	
Variables		

lnGDPrpci,t‐1		 1	e	2		 Initial	value	of	real	GDP	per	capita	at	constant	2011	prices	
(USD)	in	PPPs	

PWT		

GCFi,t		 1	e	2		 Gross	Capital	Formation	as	%	of	GDP		 PWT		
HCi,t‐1		 1	e	2		 Average	years	of	total	schooling	of	the	population	aged	15	

and	over		
(Barro	
and	Lee,	
2013)	

Gov_consi,t		 1	e	2		 Public	consumption	as	%	of	GDP		 PWT		
Globi,t		 1	e	2		 Globalization	index	 KOF		

Investments/	
Inputs		

RDEpci,t		 1		 R&D	expenditures	in	the	business	sector	per	capita	(euros)	–	
source	Eurostat	

Eurostat		

INOV_Ei,t		 2		 Expenditure	on	innovation	(excluding	R&D)	as	%	of	
turnover–	source	EIS	

EIS		

TRAIN_EMPi,t		 2		
Firms	that	promote	training	to	develop	or	upgrade	the	ICT	
skills	of	employees,	%	of	total	firms–	source	EIS	 EIS		

RD_Staffi,t‐1		 1		 Total	staff	in	R&D	activities	in	the	business	sector	per	
million	inhabitants–	source	PORDATA	

PORDATA		

RES_RDi,t‐1		 1		 Researchers	in	R&D	activities	in	the	business	sector	per	
million	inhabitants–	source	PORDATA	

PORDATA		

Innovation	
Activities/		
Outputs		

Patent_APPi,t		 1		 Number	of	patent	applications	to	the	EPO	by	the	business	
sector	per	million	inhabitants–	source	Eurostat	

Eurostat		

Patent_Gi,t		 1		 Patents	Granted	by	the	USPTO	to	the	business	sector	per	
million	inhabitants–	source	Eurostat	

Eurostat		

Trademark_APPi,t		 1		 Number	of	European	Union	trademark	applications	(EUTM)	
per	million	inhabitants–	source	Eurostat	

Eurostat		

ProdProc_INOVi,t		 2		 SMEs	that	introduce	product	or	process	innovations	as	%	of	
the	total	number	of	SMEs–	source	EIS	

EIS		

MarkOrg_INOVi,t		 2		 SMEs	that	introduce	marketing	or	organisational	
innovations	as	%	of	the	total	number	of	SMEs–	source	EIS	

EIS		

INO_SMEi,t		 2		 SMEs	that	innovate	internally	as	%	of	the	total	number	of	
SMEs–	source	EIS	

EIS		

COOP_INOVi,t		 2		 Innovative	SMEs	that	cooperate	with	others	as	%	of	total	
SMEs–	source	EIS	

EIS		

Impacts		

Export_HTi,t		 2		 Exports	of	Medium	and	High‐Tech	products	as	%	of	total	
exports	of	products		

EIS		

Export_servi,t		 2		 Exports	of	knowledge‐intensive	services	as	%	of	total	
exports	of	services	

EIS		

Sales_INOVi,t		 2		 Sales	of	new‐to‐market	and	new‐to‐firm	innovations	as	%	of	
turnover	

EIS		

Temporal	
Dummies		

D1		 1	e	2		
Temporal	Dummy	1	(Model	1:	1	for	the	period	2005‐2010,	0	
otherwise;	Model	2:	1	for	2009,	2010,	2011	and	2012,	0	
otherwise)	

‐		

D2		 1	e	2		 Temporal	Dummy	2	(Model	1:	1	for	the	period	2010‐2015,	0	
otherwise;	Model	2:	1	for	2013,	2014	and	2015,	0	otherwise)	 ‐		

Source:	 Own	elaboration	 based	 on	 data	 from	 the	 Penn	 World	 Table	 9.0	 (PWT),	Barro‐Lee	 dataset,	 KOF,	
Eurostat,	European	Innovation	Scoreboard	(EIS)	and	PORDATA.	
	



	 		
Cláudia	Caseiro,	Marta	Simões	 19	

Table	A.2.	Results	of	the	estimation	of	model	1,	EU12	(group	1),	1990‐2015	

Explanatory	
variables	

Dependent	variable	‐	a	Growth	rate	of	real	GDP	per	capita	

I		 II		 III		 IV		 V		 VI		
Constant	 0.1878		 0.3344		 0.2280		 0.2510		 0.5379***		 1.5891	***		

lnGDPrpci,t‐1		 −0.0280		 −0.0429	*		 −0.0318		 −0.0330	*		 −0.0558	***		 −0.1572	***		

GCFi,t		 0.2180	***		 0.1989	**		 0.2118***		 0.2170	***		 0.0284		 −0.1266		

HCi,t‐1		 0.0071	**		 0.0076	**		 0.0073	*		 0.0074	**		 0.0055	**		 0.0062		

Gov_consi,t		 −0.3376	**		 −0.3975	***		 −0.3541	**		 −0.3982	***		 −0.4278	***		 −0.9902	***		

Globi,t		 0.000759		 0.000856		 0.000730		 0.0006584		 0.000858	**		 0.0023	*		

RDEpci,t		 −7.2578e‐6		 ‐		 ‐		 ‐		 ‐		 ‐		

RD_Staffi,t‐1		 ‐		 1.6809e‐6		 ‐		 ‐		 ‐		 ‐		

RES_RDi,t‐1		 ‐		 ‐		 −1.2094e‐7		 ‐		 ‐		 ‐		

Patent_APPi,t		 ‐		 ‐		 ‐		 2.3588e‐5		 ‐		 ‐		

Patent_Gi,t		 ‐		 ‐		 ‐		 ‐		 7.2195e‐5		 ‐		

Trademark_APPi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 9.4847e‐5	***		

D1		 −0.0238	**		 −0.0231	**		 −0.0238	**		 −0.0251***		 −0.00999		 −0.0044		

D2		 0.0103		 0.0122		 0.0104		 0.0115		 ‐		 0.0406	***		

R2		 0.3685		 0.3703		 0.3620		 0.3851		 0.5481		 ‐		

Adjusted	R2		 0.2654		 0.2675		 0.2534		 0.2887		 0.4690		 ‐		

LSDV	R2		 ‐		 ‐		 ‐		 ‐		 ‐		 0.8439		

R2	within		 ‐		 ‐		 ‐		 ‐		 ‐		 0.8131		

AIC	(Akaike)		 −268.3492		 −268.5145		 −257.0800		 −278.4762		 −259.9794		 −255.5067		

BIC	(Schwarz)		 −249.8052		 −249.9705		 −238.8518		 −259.6271		 −245.0097		 −218.0827		

Hannan‐Quinn		 −261.1260		 −261.2912		 −250.0129		 −271.1033		 −254.3223		 −241.3641		

Notes:	Column	I	‐	pooled	OLS	for	R&D	expenses;	column	II	‐	pooled	OLS	for	the	total	staff	in	R&D	activities;	
column	III	‐	pooled	OLS	for	researchers	in	R&D	activities;	IV	column	‐	pooled	OLS	for	patent	applications;	
column	 V	 ‐	 pooled	 OLS	 for	 granted	 patents;	 column	 VI	 ‐	 fixed	 effects	 for	 the	 registration	 of	 trademark	
applications.	***,	**,	*	indicate	statistical	significance	at	the	1,	5	and	10%	levels,	respectively.	
Source:	Own	elaboration.	
	
	 	



20
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

	

Table	A.3.	Results	of	the	estimation	of	model	1,	EU16	(group	2),	1990‐2015	

Explanatory	variables	
Dependent	variable	‐	a	Growth	rate	of	real	GDP	per	capita	

I		 II		 III		 IV		 V		 VI		
Constant	 0.7373	***		 0.766124	***		 0.7719	***		 0.8930	***		 1.2841	***		 1.0914	***		

lnGDPrpci,t‐1		 −0.0907	***		 −0.0948377	***		 −0.0953***		 −0.1114	***		 −0.1640	***		 −0.1247	***		

GCFi,t		 0.2005	***		 0.179851	**		 0.1692	**		 0.2341	***		 0.1575	*		 0.3654	***		

HCi,t‐1		 −0.0013		 0.000828091		 0.000564		 0.000264		 0.0072		 −0.0066		

Gov_consi,t		 −0.0507		 −0.103619		 −0.0979		 −0.0774		 0.1050		 −0.0822		

Globi,t		 0.0021	***		 0.00226445	***		 0.0023	***		 0.0024	***		 0.0027***		 0.0021	***		

RDEpci,t		 1.0100e‐5		 ‐		 ‐		 ‐		 ‐		 ‐		

RD_Staffi,t‐1		 ‐		 −8.6045e‐6	*		 ‐		 ‐		 ‐		 ‐		

RES_RDi,t‐1		 ‐		 ‐		 −1.7627e‐5	*		 ‐		 ‐		 ‐		

Patent_APPi,t		 ‐		 ‐		 ‐		 0.000463	*		 ‐		 ‐		

Patent_Gi,t		 ‐		 ‐		 ‐		 ‐		 0.000974		 ‐		

Trademark_APPi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.000138	***		

D1		 −0.0078		 −0.0076		 −0.0053		 −0.0116		 0.000567		 −0.0033		

D2		 0.0136	*		 0.0151	*		 0.0176	**		 0.0158	*		 ‐		 0.0440	***		

R2		 ‐		 0.5639		 0.5829		 0.6404		 ‐		 ‐		

Adjusted	R2		 ‐		 0.5111		 0.5308		 0.5999		 ‐		 ‐		

LSDV	R2		 ‐		 ‐		 ‐		 ‐		 0.8421		 0.7133		

R2	within		 ‐		 ‐		 ‐		 ‐		 0.8256		 0.5628		

AIC	(Akaike)		 −355.5393		 −348.6601		 −344.4415		 −370.7207		 −311.6848		 −313.4456		

BIC	(Schwarz)		 −334.9252		 −327.8027		 −323.8274		 −349.2825		 −262.0305		 −262.3944		

Hannan‐Quinn		 −347.3242		 −340.3319		 −336.2264		 −362.1255		 −292.1235		 −293.4016		

Notes:	Column	I	 ‐	random	effects	 for	R&D	expenditure;	column	II	 ‐	pooled	OLS	 for	the	total	staff	 in	R&D	
activities;	 column	 III	 ‐	 pooled	 OLS	 for	 researchers	 in	 R&D	 activities;	 IV	 column	 ‐	 pooled	 OLS	 for	 patent	
applications;	column	V	 ‐	fixed	effects	for	granted	patents;	column	VI	 ‐	 fixed	effects	for	the	registration	of	
trademark	applications.	***,	**,	*	indicate	statistical	significance	at	the	1,	5	and	10%	levels,	respectively.		
Source:	Own	elaboration.	
	
	 	



	 		
Cláudia	Caseiro,	Marta	Simões	 21	

	

Table	A.4.	Results	of	the	estimation	of	model	2	with	fixed	effects,	EU12	(group	1),	2008‐15	

Explanatory	
variables	

Dependent	variable	‐	a	Growth	rate	of	real	GDP	per	capita	

I		 II		 III		 IV		 V		 VI		 VII		 VIII		 IX		
Constant	 6.3944***		 6.1876	***		 6.2550	***		 6.6236	***		 5.9766	***		 6.3595	***		 6.1626	***		 6.2453	***		 6.4380	***		
lnGDPrpci,t‐1		 −0.6641	

***		
−0.6421	
***		

−0.6468	
***		

−0.6782	
***		

−0.6270	
***		

−0.6540	
***		

−0.6371	
***		

−0.6451	
***		

−0.6698	
***		

GCFi,t		 0.1869		 0.1768		 0.1598		 0.1538		 0.1521		 0.1953		 0.1735		 0.1632		 0.1516		
HCi,t‐1		 0.0134	***		 0.0130	***		 0.0133	***		 0.0121	***		 0.0132	***		 0.0129	***		 0.0130***		 0.0131***		 0.0135***		
Gov_consi,t		 −0.7687		 −0.8443		 −0.8478		 −0.8567	*		 −0.8805	*		 −0.8466		 −0.8339		 −0.8067		 −0.7162		
Globi,t		 −0.0027		 −0.0024		 −0.00245		 −0.0027		 −0.0021		 −0.0023		 −0.0025		 −0.0025		 −0.0023		
Export_HTi,t		 0.000739		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
Export_Servi,t		 ‐		 0.000533		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
INOV_Ei,t		 ‐		 ‐		 0.0068		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
TRAIN_EMPi,t		 ‐		 ‐		 ‐		 0.0029	***		 ‐		 ‐		 ‐		 ‐		 ‐		
ProdProc_INOVi,t		 ‐		 ‐		 ‐		 ‐		 0.0013		 ‐		 ‐		 ‐		 ‐		
MarkOrg_INOVi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.000365		 ‐		 ‐		 ‐		
INO_SMEi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 0.000327		 ‐		 ‐		
COOP_INOVi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 0.000559		 ‐		
Sales_INOVi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 0.0016		
D1		 −0.0306	*		 −0.0291	*		 −0.0294	*		 −0.0262	*		 −0.0292	*		 −0.0287	*		 −0.0294	*		 −0.0309	*		 −0.0318	*		
D2		 −0.0074		 −0.0060		 −0.0066		 −0.0051		 −0.0034		 −0.0050		 −0.0062		 −0.0074		 −0.0091		
LSDV	R2		 0.6465		 0.6457		 0.6461		 0.6848		 0.6517		 0.6465		 0.6460		 0.6466		 0.6490		
R2	within	 0.6284		 0.6275		 0.6279		 0.6686		 0.6338		 0.6284		 0.6279		 0.6285		 0.6310		
AIC	(Akaike)		 −335.5293		−335.3313		−335.4253		−345.1485		−336.7694		−335.5249		−335.4088		−335.5508		−336.1278		
BIC	(Schwarz)		 −286.9129		−286.7150		−286.8090		−296.5322		−288.1531		−286.9085		−286.7925		−286.9345		−287.5115		
Hannan‐Quinn		 −315.9859		−315.7879		−315.8820		−325.6051		−317.2261		−315.9815		−315.8654		−316.0074		−316.5844		

Notes:	Column	 I	 ‐	exports	of	medium	and	high‐tech	products;	column	 II	 ‐	exports	of	knowledge‐intensive	
services;	 column	 III	 ‐	 innovation	 expenditures;	 column	 IV	 ‐	 training	 of	 employees;	 column	 V	 ‐	 product/	
process	 innovations;	 column	 VI	 ‐	 marketing/	 organizational	 innovations;	 column	 VII	 –%	 of	 SMEs	 that	
innovate	internally;	column	VIII	‐	%	of	SMEs	cooperating	with	others;	column	IX	‐	sales	of	innovations.	***,	**,	
*	indicate	statistical	significance	at	the	1,	5	and	10%	levels,	respectively.		
Source:	Own	elaboration.	
	
	 	



22
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	1‐22)
	 	

	

Table	A.5.	Results	of	the	estimation	of	model	2	with	fixed	effects,	EU16	(group	2),	2008‐15	

Explanatory	
variables	

Dependent	variable	‐	a	Growth	rate	of	real	GDP	per	capita	 	

I		 II		 III		 IV		 V		 VI		 VII		 VIII		 IX		

Constant	 2.5022	***		 2.4888	***		 2.4548***		 2.5312	***		 2.6043***		 2.4820	***		 2.2812***		 2.5214	***		 2.4941	***		
lnGDPrpci,t‐1		 −0.3670	

***		
−0.3681	
***		

−0.3675	
***		

−0.3746	
***		

−0.3829**
*		

−0.3840	
***		

−0.3538	
***		

−0.3716	
***		

−0.3704	
***		

GCFi,t		 0.7714***		 0.7674	***		 0.7924	***		 0.7502	***		 0.8471	***		 0.8735***		 0.8341***		 0.7757	***		 0.7662	***		
HCi,t‐1		 0.0088	***		 0.0089	***		 0.0089	***		 0.0087	***		 0.0094	***		 0.0109	***		 0.0103	***		 0.0088	***		 0.0090	***		
Gov_consi,t		 −1.2168	

***		
−1.2079	
***		

−1.1076	
***		

−1.2289	
***		

−0.9851	
***		

−1.1182	
***		

−0.8771	
**		

−1.2014	
***		

−1.2349	
***		

Globi,t		 0.0086***		 0.0087	***		 0.0088***		 0.0088	***		 0.0085	***		 0.0092	***		 0.0080	***		 0.0087	***		 0.0087	***		
Export_HTi,t		 −0.00029

6		
‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		

Export_Servi,t		 ‐		 −0.00011
8		

‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		

INOV_Ei,t		 ‐		 ‐		 −0.0108		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		
TRAIN_EMPi,t		 ‐		 ‐		 ‐		 0.0017		 ‐		 ‐		 ‐		 ‐		 ‐		
ProdProc_INOVi,
t		

‐		 ‐		 ‐		 ‐		 −0.0024	**		 ‐		 ‐		 ‐		 ‐		

MarkOrg_INOVi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.0019	
**		

‐		 ‐		 ‐		

INO_SMEi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.0030**		 ‐		 ‐		
COOP_INOVi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 −0.00040

6		
‐		

Sales_INOVi,t		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 ‐		 0.000465		
D1		 0.0297**		 0.0295	**		 0.0272	*		 0.0292	**		 0.0316	**		 0.0256*		 0.0233		 0.0301	**		 0.0297	**		
D2		 0.0415	*		 0.0408	*		 0.0393	*		 0.0417	*		 0.0351		 0.0312		 0.0247		 0.0416	*		 0.0424*		
LSDV	R2		 0.6666		 0.6665		 0.6704		 0.6696		 0.6839		 0.6845		 0.6891		 0.6665		 0.6670		
R2	within	 0.5830		 0.5828		 0.5877		 0.5867		 0.6046		 0.6054		 0.6112		 0.5829		 0.5835		
AIC	(Akaike)		 −440.920

8		
−440.866

5		
−442.196

1		
−441.925

8		
−446.8741		 −447.105

2		
−448.748

6		
−440.899

3		
−441.063

7		
BIC	(Schwarz)		 −375.676

8		
−375.622

5		
−376.952

1		
−376.681

9		
−381.6301		 −381.861

3		
−383.504

6		
−375.655

3		
−375.819

8		
Hannan‐Quinn		 −414.449

2		
−414.395

0		
−415.724

5		
−415.454

3		
−420.4026		 −420.633

7		
−422.277

0		
−414.427

7		
−414.592

2		
Notes:	Column	 I	 ‐	exports	of	medium	and	high‐tech	products;	column	 II	 ‐	exports	of	knowledge‐intensive	
services;	 column	 III	 ‐	 innovation	 expenditures;	 column	 IV	 ‐	 training	 of	 employees;	 column	 V	 ‐	 product/	
process	 innovations;	 column	 VI	 ‐	 marketing/	 organizational	 innovations;	 column	 VII	 –%	 of	 SMEs	 that	
innovate	internally;	column	VIII	‐	%	of	SMEs	cooperating	with	others;	column	IX	‐	sales	of	innovations.	***,	**,	
*	indicate	statistical	significance	at	the	1,	5	and	10%	levels,	respectively.		

Source:	Own	elaboration.	
	
	
	

Article	history:	 Received:		November	8,	2018	
Accepted:		May	30,	2019	

	
	
	



	

DOI: 10.28934/ea.19.52.12.pp23-35	
	
	
ORIGINAL	SCIENTIFIC	PAPER	
	
	

Comparative	Analysis	of	Insurance	Premiums	in	Serbia	and	
Bosnia	and	Herzegovina	‐	Multiple	Linear	Regression	
Analysis	Model	

Tijana	Kaličanin1*			|			Sandra	Kamenković1			|			Ivana	Simeunović1	
1	University	Union,	Belgrade	Banking	Academy	‐	Faculty	of	Banking,	Insurance	and	Finance,	Belgrade,	Serbia	
	
	
ABSTRACT	
In	the	last	few	years,	the	global	insurance	market	has	shown	a	trend	of	concentration	growth,	which	
was	conditioned	by	the	processes	of	mergers	and	acquisitions	in	insurance.	The	aim	of	this	paper	is	
to	 make	 a	 comparative	 analysis	 of	 insurance	 premiums	 in	 Serbia	 and	 Bosnia	 and	 Herzegovina.	
Dynamic	analysis	of	market	concentration	indicators	calculated	on	the	basis	of	absolute	amounts	of	
premiums	indicates	that	the	insurance	market	in	Bosnia	and	Herzegovina	is	characterized	by	low	
concentrated	supply,	i.e.	there	is	greater	equality	of	market	share	in	relation	to	high	inequality	and	
high	 concentration	 among	market	 participants	 in	 the	 insurance	 sector	 of	 the	 Republic	 of	 Serbia.	
Having	 applied	 the	 multi‐linear	 regression	 model	 in	 order	 to	 analyze	 the	 impact	 of	 selected	
macroeconomic	indicators	on	the	amount	of	insurance	premiums	in	the	period	2000‐2017,	it	can	be	
concluded	that	the	greatest	impact	on	the	amount	of	the	premium	in	Bosnia	and	Herzegovina	had	
Gross	 Domestic	 Income	 and	 Wage	 and	 Salaries	 Workers.	 In	 the	 Republic	 of	 Serbia,	 the	 greatest	
influence	on	the	amount	of	premium	in	the	observed	period	had	the	Average	Net	Salary,	Households	
and	Final	Consumption	Expenditure	and	Gross	Domestic	Income.	

	
Key	words:	insurance	premiums,	insurance	sector,	multiple	linear	regression	analysis	model	
	
JEL	Classification:	C30,	G22,	L19	

	

INTRODUCTION	

In	the	last	two	decades,	countries	of	Central	and	East	Europe	have	experienced	tremendous	
changes	in	the	political,	cultural,	social	and	economic	environment.	Central	and	East	European	
financial	 system	 has	 been	 rapidly	 developing	 during	 the	 last	 couple	 of	 years,	 contemporary	
regulations	 have	 been	 introduced	 and	 new	 financial	 institutions	 have	 been	 established	
contributing	 to	 the	 maintenance	 of	 macroeconomic	 stability	 in	 the	 region.	 Nonetheless,	 the	
macroeconomic	sector	in	the	region,	underdeveloped	even	before	the	recession	took	place,	has	
been	a	highly	risky	place	 for	 investments	 and	 unstable	 in	comparison	 with	 Western	Europe	
(Kaličanin	&	Hanić	2016a).	The	insurance	market	in	the	Western	Balkans	is	characterized	by	
significant	changes	caused	by	different	economic	growth	pace.	Countries	preparing	to	become	
members	of	the	European	Union	are	carrying	out	faster	reforms,	and	there	is	also	a	significant	
inflow	of	foreign	capital	due	to	a	reduction	of	financial	and	political	risks	(Novović‐Burić	et	al.	
2017).		
In	this	paper,	the	insurance	sectors	were	analyzed	in	the	Republic	of	Serbia	and	Bosnia	and	

Herzegovina.	 In	 addition	 to	 the	 transition	 process	 in	 both	 countries,	 these	 countries	 have	
																																																													
*	Corresponding	author,	e‐mail:	tijana.kalicanin@bba.edu.rs	



24
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	23‐35)
	 	

undergone	similar	political	and	economic	changes	in	the	past	decade.	Both	insurance	markets	
have	passed	through	the	process	of	 integration	and	internationalisation.	 In	addition	to	many	
similarities	 characteristic	 for	 these	 two	 markets,	 there	 are	 substantial	 differences	 that	 are	
reflected	in	the	number	of	residents,	the	number	of	insurance	companies,	the	market	structure,	
the	participation	of	the	market	leader,	and	the	number	of	companies	with	foreign	capital.	
The	 market	 of	 a	 country	 is	 as	 developed	 as	 its	 competition	 is	 able	 to	 function	 on	 it.	

Competition	 has	 to	 be	 constantly	 stimulated	 and	 protected	 by	 mechanisms	 in	 line	 with	 the	
European	 integration	 processes	 and	 policy	 focused	 on	 market	 economy	 development.	
Competition	as	such	has	been	a	particularly	sensitive	 issue	 in	 transitional	 countries	such	 as	
Serbia	and	countries	in	the	region.	Changing	the	number	of	insurance	companies	on	the	market	
influenced	significantly	the	formation	of	a	group	of	leaders	in	the	insurance	market	as	well	as	
strengthening	the	position	and	increasing	the	individual	market	share	of	the	leader	(Kaličanin	&	
Lazić,	2018).	
	Initial	 structural	 changes	 raised	 an	 issue	 and	 brought	 about	 the	 need	 for	 a	 higher	

competitiveness	in	Serbia.	Every	country	aspiring	to	join	the	EU	and	integration	processes	ought	
to	develop	legal	norms	and	apply	the	EU	regulations	(Kaličanin	&	Hanić	2016a).	In	recent	years,	
the	insurance	sector	has	become	a	significant	factor	in	the	development	of	the	overall	financial	
and	 economic	 system.	 In	 the	 first	 part	 of	 the	 paper,	 insurance	 markets	 and	 level	 of	
competitiveness	were	analyzed.	Competitiveness	in	the	market	has	led	to	changes	in	the	balance	
sheet	 structure	 of	 the	 entire	 financial	 sector	 as	 well	 as	 the	 position	 of	 individual	 insurance	
companies	in	previous	years.	The	insurance	sector	is	extremely	important	for	the	economy	of	a	
country,	not	only	from	the	point	of	view	of	security	and	protection	from	different	types	of	risks	
but	also	from	the	point	of	view	of	the	overall	economic	development	and	improvements	in	the	
functioning	of	the	financial	market.	In	the	second	part	of	the	paper,	the	focus	is	on	the	analysis	of	
insurance	premiums	and	the	impact	of	selected	macroeconomic	indicators	on	premiums	using	
the	multi‐linear	regression	model.	

LITERATURE	REVIEW	

There	 are	 many	 analyses	 which	 deal	 with	 insurance	 premiums	 and	 economic	 growth.	
Outreville	(1990),	Zhi	(1998),	Beck	and	Webb	(2003),	Webb	et	al.	(2002)	have	shown	a	very	
strong	interaction	between	insurance	premiums	and	GDP	despite	different	periods	and	country	
patterns.	Analyses	mainly	suggest	that	higher	GDP	growth	rates	have	an	impact	on	economic	
activity	growth,	which	leads	to	assumptions	about	a	positive	correlation	between	GDP	growth	
rate	and	demand	for	insurance.	
Haiss	and	Sumegi	presented	very	extensive	research	in	2008,	which	led	to	the	conclusion	that	

there	is	a	correlation	between	insurance	and	GDP	growth	in	EU‐15	countries	with	developed	
financial	 markets	 as	 well	 as	 short‐term	 linkages	 between	 GDP	 and	 non‐life	 premiums	 on	 a	
sample	 of	 CEE	 countries.	 Serbia	 and	 Bosnia	 and	 Herzegovina	 are	 selected	 for	 this	 research	
because	very	few	authors	have	analyzed	this	region	from	the	insurance	aspect.	
Novovic‐Buric	 et	 al.	 (2017)	 explored	 the	 influence	 of	 certain	 factors	 on	 the	 purchase	 of	

insurance	products	through	a	panel	analysis.	Western	Balkan	countries	were	analyzed	in	the	
period	from	2005	to	2015,	and	the	results	show	that	most	of	the	economic	factors	affect	total	life	
premiums.	 The	 demand	 for	 life	 insurance	 has	 a	 significant	 and	 positive	 impact	 on	 GDP	 and	
wages,	while	the	influence	of	unemployment	and	interest	rates	is	negative.	
Dragos	(2014)	used	the	fixed	and	random	effects	model	in	the	analysis,	which	entailed	17	

countries	 in	Asia	and	Central	Eastern	Europe.	The	aim	was	to	compare	emerging	markets	 in	
Europe	 characterised	 by	 market	 economies	 and	 emerging	 markets	 in	 Asia,	 which	 are	
predominantly	planned	economies	in	terms	of	the	impact	of	economic	performance	on	life	and	
non‐life	insurance.	It	has	been	noted	that	the	differences	certainly	exist.	The	results	concerning	
the	CEE,	taking	into	consideration	the	countries	which	are	analyzed	in	this	research	have	shown	



	 		
Tijana	Kaličanin,	Sandra	Kamenković,	Ivana	Simeunović	 25	

that	income	and	education	have	a	positive	impact	on	the	insurance	demand,	while	urbanisation	
has	shown	a	significant	impact	on	non‐life	insurance.	On	the	other	hand,	 income	distribution	
negatively	influences	the	demand	for	insurance.	
Kjosevski	 (2012)	 analysed	 the	 determinants	 of	 life	 insurance	 demand	 in	 Central	 and	

Southeastern	Europe	by	using	fixed‐effect	panel	models	in	the	period	1998‐2010.	The	results	
show	that	in	terms	of	life	insurance,	the	most	significant	predictors	are	the	following:	high	GDP	
per	capita,	inflation,	health	expenditure,	level	of	education	and	the	rule	of	law.	
Mitra	 (2017)	 analysed	 the	 impact	 of	 economic,	 demographic	 and	 cultural	 factors	 on	 life	

insurance	 consumption	 in	 28	 EU	 countries.	 The	 focus	 was	 on	 the	 emerging	 East	 European	
economies,	given	that	in	the	analysed	period	2009‐2014,	there	were	significant	reforms	of	the	
insurance	sector	 in	 these	countries.	 The	results	 show	that	 the	higher	GDP	 is	an	 indicator	of	
higher	 wages	 and	 higher	 levels	 of	 economic	 activity,	 the	 more	 positive	 impact	 on	 insurance	
demand.	
Ward	 and	 Zurbruegg	 (2002)	 analysed	 37	 countries	 in	 the	 period	 1987	 ‐	 1998	 with	 the	

intention	 to	 point	 to	 the	 links	 between	 insurance	 premiums	 and	 various	 legal	 and	 political	
factors,	as	well	as	economic	and	social	factors.	The	analysis	has	shown	that	the	consumption	of	
life	insurance	products	is	under	stronger	influence	of	GDP	in	Asia	than	in	OECD	countries,	which	
is	an	expected	result	given	that	in	the	OECD	countries	there	is	a	considerably	higher	average	
income	level	and	that	‘S	curves’	suggest	that	at	higher	levels	of	income	the	demand	for	insurance	
is	less	susceptible	to	the	revenue	growth.	
Bianchi	 et	 al.	 (2011)	 analysed	 the	 insurance	 market	 in	 Central,	 Eastern	 and	 Southeastern	

Europe	and	both	countries	which	were	analyzed	in	this	research	are	contained	in	the	sample.	
They	used	panel	regression	(a	cross	section	with	fixed	effects)	to	examine	the	impact	of	GDP	
growth	 on	 insurance	 premium	 growth.	 The	 results	 suggest	 that	 economic	 development	 and	
catching‐up	processes	mainly	condition	premium	growth,	and	that	in	unstable	periods,	it	shows	
increased	volatility.	

COMPARATIVE	ANALYSIS	OF	THE	INSURANCE	MARKET	

Figure	1	shows	the	total	number	of	insurance	companies	operating	in	the	insurance	market	in	
Serbia	 as	 well	 as	 in	 Bosnia	 and	 Herzegovina	 in	 the	 last	 ten	 years.	 The	 largest	 number	 of	
insurance	 companies	 in	 Serbia	 was	 present	 in	 the	 period	 2011‐2013	 when	 there	 were	 28	
companies,	followed	by	the	decline	in	the	number	of	insurance	companies,	ranging	from	25	in	
2014	to	21	at	the	end	of	2017.		
	

	
Figure	1.	Number	of	insurance	companies,	2007‐2017	

	
With	the	number	of	 insurance	companies	 in	the	market	 in	Bosnia	and	Herzegovina,	 fewer	

oscillations	can	be	noted	in	the	observed	period,	ranging	from	24	to	27.	The	largest	number	of	

0

10

20

30

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

20
24

26 26
28 28 28

25 24 23
21

25 26 26 26 25 25 25 24 24
27 27

Number	of	insurance	companies	‐	Serbia Number	of	insurance	companies	‐	BiH



26
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	23‐35)
	 	

insurance	companies	at	the	B&H	market	operated	in	the	last	two	years	of	the	observed	period,	
i.e.	2016‐2017	with	a	total	of	27	companies.	
In	addition	to	the	banking	sector	with	the	largest	share	of	the	balance	sheet	total	in	the	total	

financial	sector	‐	about	90%,	the	balance	sheets	of	the	leasing	companies,	pension	funds	and	
insurance	companies	are	included.	Figure	2	shows	a	fall	in	the	share	of	the	insurance	sector's	
balance	sheet	total	in	the	total	financial	sector,	which	is	notable	in	both	markets	in	the	period	
2005‐2008.	After	2008,	the	share	of	the	insurance	sector	balance	sheet	total	in	Serbia	rose	from	
4.2%	in	2008	to	6.3%	in	2017.	The	same	trend	is	present	in	the	insurance	market	of	Bosnia	and	
Herzegovina,	where	the	growth	of	the	balance	sheet	total	in	the	total	financial	sector	increased	
in	the	same	period	for	the	same	percentage	‐	2.1%,	from	3.45%	to	5.57%	in	2017.	
	

	

Figure	2.	Contribution	of	the	balance	sheet	total	of	the	insurance	sector	to	the	overall	financial	
sector,	2005‐2017	

	
Figures	 3	 and	 4	 show	 the	 ratio	 of	 the	 total	 number	 of	 insurance	 companies	 and	 foreign	

companies	in	the	period	2007‐2017.	It	can	be	concluded	that	the	number	of	foreign	insurance	
companies	in	Serbia	has	changed	proportionally	with	the	total	number	of	companies	present	in	
the	market.	The	number	of	foreign	companies	in	the	insurance	market	of	B&H	did	not	change	
significantly	in	the	observed	period.	There	were	10	foreign	insurance	companies	in	the	period	
2008‐2014,	 after	 which	 this	 number	 increased	 to	 11	 and	 12	 in	 2015,	 2016	 and	 2017,	
respectively.	
	

	

Figure	3.	Relationship	between	the	total	number	of	insurance	companies	and	foreign	
companies,	2007‐2017,	Serbia	

4,42 3,75 3,46 3,45 3,72 3,86
4,43 4,77 4,86

5,19 5,38 5,57 5,57

5,3
4,3 4,1 4,3 4,2 4,2

4,4 4,5
4,8

5,2
5,8 6,1

6,3

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

BiH Serbia

20
24 26 26

28 28 28
25 24 23 21

13
17 19 19

21 21 21 19 18 17 15

0

5

10

15

20

25

30

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Total Foreign



	 		
Tijana	Kaličanin,	Sandra	Kamenković,	Ivana	Simeunović	 27	

	

	

Figure	4.	Relationship	between	the	total	number	of	insurance	companies	and	foreign	
companies,	2007‐2017,	B&H	

	
Given	that	the	subject	matter	of	the	analysis	is	the	amount	of	 income	earned	by	insurance	

companies,	 the	companies’	market	shares	are	calculated	on	 the	basis	of	 the	amount	of	 total	
premiums	at	the	end	of	the	year	taken	from	the	balance	sheet	of	insurance	companies.	Figure	5	
shows	a	comparative	analysis	of	the	concentration	of	the	largest	companies	in	B&H	and	Serbia	
according	to	the	criterion	of	the	total	balance	sheet.	Dunav	insurance	company	was	the	market	
leader	in	Serbia	according	to	the	mentioned	criterion	and	has	absorbed	about	one‐quarter	of	the	
entire	market	in	the	observed	period.	By	the	end	of	2015,	Sarajevo	Insurance	was	the	market	
leader	in	B&H	according	to	the	criterion	of	the	total	premium.	For	the	next	two	years	of	the	
observed	 period,	 Uniqa	 insurance	 achieved	 the	 largest	 amount	 of	 total	 premiums.	 The	
concentration	of	leader	ratio	in	B&H	is	much	lower	than	in	Serbia,	so	the	leader	in	this	market	
has	a	share	of	about	one‐tenth	of	the	entire	market.	
	

	

Figure	5.	CR1	according	to	the	criterion	of	the	total	premium,	2007‐2017	
	

The	 number	 of	 companies	 involved	 in	 the	 calculation	 of	 this	 indicator	 is	 determined	 by	
government	 agencies	 that	 are	 monitoring	 the	 degree	 of	 bid	 concentration	 in	 that	 particular	
country,	provided	that	this	coefficient	is	used	as	an	official	indicator	(Martin,	2002).	CR	1,	CR3	
and	CR5	are	most	frequently	accrued	in	the	reports	of	the	responsible	insurance	institutions,	
according	to	the	criterion	of	total	premiums,	total	non‐life	 insurance	premiums	and	total	 life	

25 26 26 26 25 25 25 24 24
27 27

9 10 10 10 10 10 10 10 11 11
12

0

5

10

15

20

25

30

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Total Foreign

013 012 012 012 013 013 012 011 010 009 008

031

025
028 026 027

029
027

025 027 026 027

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

BiH SRB



28
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	23‐35)
	 	

assurance	premiums.	Figure	6	shows	the	concentration	ratio	of	the	top	five	insurance	companies	
with	the	highest	total	premium.	Although	there	is	a	mild	tendency	of	market	penetration	at	both	
markets,	at	the	Serbian	insurance	market,	the	first	five	insurance	companies	have	a	much	larger	
share	than	it	is	the	case	in	Bosnia	and	Herzegovina.	(as	can	be	seen	in	Lorenz	curves	8	and	9).	
CR5	had	fallen	from	84.16%	in	2007	to	77.19%	in	2017.	The	mentioned	concentration	ratio	in	
B&H	dropped	from	45.97%	to	39.21%,	from	2007	to	2017.	
	

	

Figure	6.	CR5	according	to	the	criterion	of	the	total	premium,	2007‐2017	
	
Figure	7	shows	the	dynamic	analysis	of	the	Herfindahl‐Hirschman	index	in	the	period	2007‐

2017	 for	both	markets.	The	significance	of	 the	 index	 is	 reflected	 in	 the	 fact	 that	although	 it	
respects	the	individual	market	share	of	all	companies	in	the	branch	it	particularly	responds	to	
the	 presence	 of	 companies	 with	 large	 market	 participations,	 which	 significantly	 increase	 its	
value	 (Lipczynski	 &	 Wilson,	 2001)	 Given	 the	 above	 and	 taking	 into	 account	 the	 high	 CR5	
concentration	 ratio	 on	 the	 insurance	 market	 of	 the	 Republic	 of	 Serbia,	 high	 values	 of	 the	
Herfindahl‐Hirschman	index	are	not	surprising.	
	

	

Figure	7.	Herfindahl‐Hirschman	index	according	to	the	criterion	of	the	total	premium,	2007‐017	
	
With	the	increase	in	the	number	of	insurance	companies	in	Serbia,	this	index	has	declined,	so	

the	 highly	 concentrated	 bid	 with	 the	 index	 of	 2050	 in	 2007	 reached	 the	 level	 of	 a	 medium	
concentrated	bid	(it	is	considered	thin	the	market	is	medium	concentrated	if	the	HHI	index	value	

046 045 045 046 046 045 046 045 042 041 039

084 082 079 077 077 077 076 077 076 075 077

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

BiH SRB

2050
1820

1654
1520 1551 1595 1495 1496 1558 1495 1543

655 623 622 643 636 640 636 610 585 555 534

0

500

1000

1500

2000

2500

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Serbia BiH



	 		
Tijana	Kaličanin,	Sandra	Kamenković,	Ivana	Simeunović	 29	

ranges	 from	 1000	 to	 1800).	 It	 can	 be	 concluded	 that	 from	 2008	 until	 today,	 the	 insurance	
market	 in	 Serbia	 is	 medium	 concentrated.	 Herfindahl‐Hirschman	 index	 according	 to	 the	
criterion	of	 the	 total	premium	differs	significantly	 in	 the	market	of	Bosnia	and	Herzegovina,	
where	the	permanent	low‐concentration	bid	is	present	in	the	observed	period.	Although	the	HHI	
index	fell	from	655	to	534,	from	2007	to	2017,	it	can	be	concluded	that	the	insurance	market	in	
B&H	is	inconsistent,	i.e.	that	there	is	greater	equality	of	market	share	compared	with	the	high	
inequality	among	market	participants	in	the	insurance	sector	of	the	Republic	of	Serbia.	
Figure	8	shows	Lorenz	curves	based	on	the	amount	of	total	premiums	for	the	initial	and	the	

last	 year	 of	 the	 observed	 period,	 i.e.	 2007	 and	 2017.	 First,	 insurance	 market	 leaders	 have	
reduced	their	market	share,	i.e.	the	top	20%	of	insurance	companies	in	2007	had	a	cumulative	
80%	of	the	total	market	share	measured	by	the	total	premium,	while	the	same	percentage	of	
leaders	 in	 2017	 had	 less	 than	 60%.	 With	 the	 exception	 of	 market	 leaders	 (the	 first	 four	
companies)	of	18	companies	in	2007,	14	companies	shared	18.85%	of	the	total	market,	while	in	
the	year	2017	some	15	companies	shared	18.69%,	which	suggests	that	the	number	of	companies	
increased	 in	 the	market	with	smaller	market	share,	primarily	due	 to	 the	appearance	of	new	
companies	 in	 the	 market.	 This	 finding	 contributes	 to	 the	 value	 of	 the	 Herfindahl‐Hirschman	
index	as	well	as	the	value	of	CR5.	
	

	 	

Figure	8.	Lorenz	curve	and	Herfindahl‐Hirschman	index	according	to	the	criterion	of	the	total	
premium,	2007‐2017,	Serbia	

	
Constructing	Lorenz's	curve	in	terms	of	the	degree	of	concentration	of	supply	in	the	relevant	

market,	 the	 distribution	 of	 the	 entire	 market	 between	 business	 entities	 can	 be	 clearly	 seen.	
(Kaličanin	&	Hanić	2016b).	Lorenz	curves	for	the	insurance	market	of	Bosnia	and	Herzegovina	
constructed	based	on	the	premiums	shown	in	Figure	9	differ	considerably	from	those	presented	
in	 Figure	 8.	 The	 distribution	 of	 the	 market	 share	 of	 insurance	 companies	 in	 the	 stipulated	
market	 is	 characterised	 by	 greater	 equality,	 which	 can	 be	 inferred	 from	 the	 Lorenz	 curve	
distance	 from	 the	 equal’s	 curves	 (in	 case	 of	 equal	 distribution	 of	 market	 share	 among	 all	
participants	 in	the	market).	Fewer	companies	have	been	able	to	 increase	their	market	share	
over	the	observed	period,	while	market	leaders	reduced	their	cumulative	market	share.	
	

0

20

40

60

80

100

0 20 40 60 80 100

2007

equals

lorenz

0

20

40

60

80

100

0 20 40 60 80 100

2017

equals lorenz



30
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	23‐35)
	 	

	 	

Figure	9.	Lorenz	curve	and	Herfindahl‐Hirschman	index	according	to	the	criterion	of	the	total	
premium,	2007‐2017,	B&H	

	
In	2017	in	Bosnia	and	Herzegovina,	there	were	27	companies,	i.e.	6	companies	more	than	in	

the	Republic	of	Serbia.	In	addition	to	the	difference	in	the	number	of	companies,	there	is	greater	
inequality	in	the	distribution	of	market	shares	in	the	Serbian	insurance	market	as	well	as	greater	
concentration	on	the	supply	side	compared	to	the	low	concentration	of	the	insurance	market	of	
Bosnia	and	Herzegovina.	

METHODOLOGY	

As	emphasised	before,	the	aim	of	this	paper	is	to	analyse	the	insurance	market	in	Serbia	and	
Bosnia	and	Herzegovina	with	reference	to	certain	macroeconomic	indicators	that	play	a	key	role	
in	 the	 development	 of	 insurance.	 In	 this	 respect,	 we	 investigated	 the	 mutual	 influence	 of	
individual	variables	on	the	movement	of	total	premium	income	earned	in	these	countries.	In	this	
analysis,	a	time	series	is	used	for	the	period	2000‐2017,	i.e.	for	a	period	of	18	years.	For	the	
mentioned	period,	the	following	independent	variables	were	analysed:		

 	–	Average	net	salary	(current,	RSD/KM)	
 	–	Unemployment,	total	(%	of	tthe	otal	labor	force)	
 	–	Gross	domestic	income	(constant	LCU)	
 	‐	GDP	per	capita	(current	US$)	
 	‐	Wage	and	salaried	workers,	total	(%	of	total	employment)	
 	‐	Households	and	NPISHs	Final	consumption	expenditure	(current	US$)	

while	the	dependent	variable	 	was	–	Total	premium	(current	RSD/KM).		
	
For	 this	 research,	 multiple	 regression	 analysis	 was	 carried	 out	 using	 the	 SPSS	 statistical	

software.	
The	 choice	 of	 independent	 variables	 is	 based	 on	 empirical	 fundaments	 that	 relate	 to	 the	

studied	 variable,	 as	 well	 as	 on	 the	 information	 contained	 in	 the	 relevant	 literature.	 Six	
independent	variables	 were	 included	 in	 the	 analysis	 to	 identify	 those	 that	could	explain	 the	
major	part	of	the	variability	of	the	studied	(dependent)	variables.	A	model	involving	variables	
whose	calculated	regression	coefficient	is	significant	at	5%	level	is	selected.	

EMPIRICAL	RESEARCH	AND	DISCUSSION	

Usually,	the	first	part	of	the	study	contains	basic	indicators	of	descriptive	statistics.	Since	all	
variables	 included	in	the	analysis	were	measured	on	the	ratio	scale,	 the	values	of	arithmetic	
mean	and	standard	deviations	are	presented	as	indicators	of	a	descriptive	statistic	(Table	1	and	
Table	2).	
	

0

20

40

60

80

100

0 20 40 60 80 100

2007

equals
lorenz

0

20

40

60

80

100

0 20 40 60 80 100

2017

equals
lorenz



	 		
Tijana	Kaličanin,	Sandra	Kamenković,	Ivana	Simeunović	 31	

Table	1:	Descriptive	statistics	for	different	variables	–	Serbia	
	
	 Mean	 Std.	Deviation	 N	
Total	premium	 49372718.5000	 24831359.35237	 18	
Average	net	salary	 345729.3333	 183811.84680	 18	
Unemployment,	total	(%	of	total	labor	
force)	

17.6239	 3.61340	 18	

Gross	domestic	income	(constant	LCU)	 2855342777777.7780	 458570337497.97485	 18	

GDP	per	capita	(current	US$)	 4614.9873	 1798.25314	 18	

Wage	and	salaried	workers,	total	(%	of	
total	employment)	

69.4582	 1.98410	 18	

Households	and	NPISHs	Final	
consumption	expenditure	(current	US$)	

25301236558.1667	 9670960437.31504	 18	

Source:	Authors	using	SPSS	
	
Table	2:	Descriptive	statistics	for	different	variables	–	BiH	

	 Mean	 Std.	Deviation	 N	
Total	premium	 430206288.0000	 141434498.99548	 18	
Average	net	salary	 1035.7778	 262.30260	 18	
Unemployment,	total	(%	of	total	labor	
force)	

27.3079	 2.00943	 18	

Gross	domestic	income	(constant	LCU)	 24073220032.1667	 3036558289.97812	 18	
GDP	per	capita	(current	US$)	 3848.6667	 1366.86275	 18	
Wage	and	salaried	workers,	total	(%	of	
total	employment)	

66.5221	 6.48730	 18	

Households	and	NPISHs	Final	
consumption	expenditure	(current	US$)	

12811176132.8333	 2684992943.38614	 18	

Source:	Authors	using	SPSS	
	
Below	is	an	output	that	refers	to	the	multiple	regression	model	which	was	implemented	on	

the	data	obtained	for	both	countries	involved	in	the	analysis.	
	

Table	3:	Model	summary	along	with	the	values	of	R	and	R	square	–	Serbia	

Model	 R	
R	

Square	

Adjusted	
R	

Square	

Std.	Error	of	
the	Estimate	

Change	Statistics	
R	Square	
Change	

F	
Change	

df1	 df2	
Sig.	F	
Change	

1	 .962a	 .925	 .920	 7023512.91992	 .925	 196.491	 1	 16	 .000	

2	 .981b	 .962	 .957	 5152189.94256	 .037	 14.733	 1	 15	 .002	
3	 .992c	 .983	 .980	 3548389.97989	 .021	 17.624	 1	 14	 .001	

4	 .996d	 .993	 .990	 2444365.32992	 .009	 16.502	 1	 13	 .001	

a.	Predictors:	(Constant),	Average	net	salary	
b.	 Predictors:	 (Constant),	 Average	 net	 salary,	 Households	 and	 NPISHs	 Final	 consumption	 expenditure	
(current	US$)	

c.	 Predictors:	 (Constant),	 Average	 net	 salary,	 Households	 and	 NPISHs	 Final	 consumption	 expenditure	
(current	US$),	Gross	domestic	income	(constant	LCU)	

d.	 Predictors:	 (Constant),	 Average	 net	 salary,	 Households	 and	 NPISHs	 Final	 consumption	 expenditure	
(current	US$),	Gross	domestic	income	(constant	LCU),	Unemployment,	total	(%	of	total	labor	force)	

Source:	Authors	using	SPSS	
	



32
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	23‐35)
	 	

In	the	case	of	variables	describing	the	insurance	market	in	the	Republic	of	Serbia,	four	models	
were	generated,	where	the	last	model	was	selected	having	the	highest	value	of	the	coefficient		
.	The	 	value	in	this	case	shows	that	99.3%	variations	in	total	premium	can	be	explained	by	

selected	four	variables	and	this	model	can	be	considered	appropriate	to	develop	the	regression	
equation.	The	 independent	variables	selected	by	 the	above	mentioned	model	are	as	 follows:	
Average	 net	 salary,	 Households	 and	 NPISHs	 Final	 consumption	 expenditure,	 Gross	 domestic	
income	and	Unemployment,	total	(Table	3).	
The	following	table	shows	the	unstandardized	and	standardised	regression	coefficients	for	all	

models.	In	the	last	model,	t‐values	for	all	the	two	regression	coefficients	are	significant	as	their	
significance	 values	 (p‐values)	 are	 less	 than	 0.05.	 Therefore,	 it	 can	 be	 concluded	 that	 the	
previously	 selected	 4	 independent	 variables	 significantly	 explain	 the	 variations	 in	 the	 total	
premium.	

 
Table	4:	Regression	coefficients	of	selected	variables	in	different	models	along	with	their	 ‐
values	and	partial	ccorrelations	–	Serbia	

Model	
Unstandardized	Coefficients	

Standardised	
Coefficients	

t	 Sig.	
Correlations	

B	 Std.	Error	 Beta	
Zero‐
order	

Partial	 Part	

1	 (Constant)	 4460520.406 3606405.932	 1.237 .234	 	 	
Average	net	salary	 129.906 9.267 .962 14.018 .000 .962 .962 .962	

2	 (Constant)	 14150355.208 3656718.281	 3.870 .002	 	 	
Average	net	salary	 178.058 14.268 1.318 12.479 .000 .962 .955 .628	
Households	and	
NPISHs	Final	
consumption	
expenditure	

‐.001 .000 ‐.405 ‐3.838 .002 .753 ‐.704 ‐.193	

3	 (Constant)	 ‐66019515.273 19262250.240	 ‐3.427 .004	 	 	
Average	net	salary	 102.779 20.448 .761 5.026 .000 .962 .802 .174	
Households	and	
NPISHs	Final	
consumption	
expenditure		

‐.002 .000 ‐.640 ‐6.977 .000 .753 ‐.881 ‐.242	

Gross	domestic	
income	

4.254E‐5 .000 .786 4.198 .001 .934 .747 .145	

4	 (Constant)	 ‐49098413.472 13907538.352	 ‐3.530 .004	 	 	
Average	net	salary	 107.565 14.135 .796 7.610 .000 .962 .904 .182	
Households	and	
NPISHs	Final	
consumption	
expenditure	

‐.001 .000 ‐.550 ‐8.200 .000 .753 ‐.915 ‐.196	

Gross	domestic	
income	

3.880E‐5 .000 .717 5.512 .000 .934 .837 .132	

Unemployment,	
total	(%	of	total	
labor	force)	

‐783350.327 192833.041 ‐.114 ‐4.062 .001 .250 ‐.748 ‐.097	

a.	Dependent	Variable:	Total	premium	

Source:	Authors	using	SPSS	
 
Using	the	values	of	the	unstandardized	regression	coefficients	of	the	last	model	presented	in	

Table	4,	the	following	regression	model	can	be	shown:	
	

	
49098413.472 107.565	 	 	

0.001 	 	 	 	 	
	0.0000388	 	 	 	 783350.327	 	 		



	 		
Tijana	Kaličanin,	Sandra	Kamenković,	Ivana	Simeunović	 33	

Concerning	 the	 same	 analysis	 carried	 out	 for	 the	 data	 from	 Bosnia	 and	 Herzegovina,	 two	
models	were	generated,	where	the	second	one	was	selected	for	which	the	value	of	0.983	was	
assigned	for	the	determination	coefficient	(Table	5).	In	this	way,	a	multiple	hierarchy	model	was	
formulated	which	combines	two	independent	variables	as	follows:	Gross	domestic	income	and	
Wage	and	salaried	workers,	total.	Since	the	F‐value	for	this	model	is	highly	significant,	the	model	
is	reliable.	Also,	the	regression	coefficients	in	the	stipulated	model	are	statistically	significant,	
and	it	is	considered	that	the	selected	variables	have	a	great	predictive	significance	in	estimating	
the	value	of	the	Total	premium.	

 
Table	5:	Model	summary	along	with	the	values	of	R	and	R	square	‐	BIH	

Model	 R	
R	

Square	
Adjusted	R	
Square	

Std.	Error	of	the	
Estimate	

Change	Statistics	

R	Square	
Change	

F	
Change	

df1	df2	
Sig.	F	
Change	

1	 .975a .951	 .948 32128444.21790 .951 313.443 1 16 .000	

2	 .992b .983	 .981 19491967.96979 .032 28.470 1 15 .000	

a.	Predictors:	(Constant),	Gross	domestic	income	(constant	LCU)	
b.	Predictors:	(Constant),	Gross	domestic	income	(constant	LCU),	Wage	and	salaried	workers,	total	(%	of
total	employment)	
Source:	Authors	using	SPSS	
 
The	last	table	contains	the	values	of	the	regression	coefficients	 for	the	previously	selected	

model.	
	
Table	6:	Regression	coefficients	of	selected	variables	in	different	models	along	with	their	 ‐
values	and	partial	correlations	‐	BIH	

Model	

Unstandardized		
Coefficients	

Standardised	
Coefficients	

t	 Sig.	
Correlations	

B	 Std.	Error	 Beta	
Zero‐
order	

Partial	 Part	

1	 (Constant)	 ‐663490820.738	 62238137.572	 	 ‐10.661	 .000	 	 	 	
Gross	domestic	
income	

.045	 .003	 .975	 17.704	 .000	 .975	 .975	 .975	

2	 (Constant)	 ‐827546449.311	 48694106.929	 	 ‐16.995	 .000	 	 	 	
Gross	domestic	
income		

.035	 .002	 .759	 14.452	 .000	 .975	 .966	 .483	

Wage	and	
salaried	
workers,	total	
(%	of	total	
employment)	

6110923.803	 1145287.155	 .280	 5.336	 .000	 .866	 .809	 .178	

a.	Dependent	Variable:	Total	premium	

Source:	Authors	using	SPSS	
	
Regression	equation,	which	can	explain	the	variability	of	the	observed	Total	premium	variable	

analysing	trends	in	selected	independent	variables	goes	as	follows:				
	

	
827546449.311 0.035	Gross	domestic	income

6110923.803	Wage	and	salaried	workers	



34
	 	 	

Economic	Analysis	(2019,	Vol.	52,	No.	1,	23‐35)
	 	

CONCLUSION	

After	2008,	the	share	of	the	insurance	sector	balance	sheet	total	 in	Serbia	and	Bosnia	and	
Herzegovina	 increased	 in	 the	 same	 period	 for	 the	 same	 percentage	 –	 2.1%.	 The	 number	 of	
insurance	companies	with	 foreign	equity	 in	Serbia	has	changed	proportionally	with	the	total	
number	 of	 companies	 present	 in	 the	 market	 whereas	 the	 number	 of	 the	 foreign	 insurance	
companies	in	Bosnia	and	Herzegovina	remained	almost	the	same	although	the	total	number	of	
the	 insurance	 companies	 increased.	 	 The	 concentration	 of	 leader	 ratio	 (CR1)	 in	 Bosnia	 and	
Herzegovina	is	much	lower	than	in	Serbia,	so	the	leader	in	this	market	has	a	share	of	about	10%	
of	the	entire	market	while	in	Serbia	it	is	about	27%.	Although	there	is	a	moderate	tendency	of	
market	penetration	at	both	markets,	at	the	Serbian	insurance	market,	the	first	 five	insurance	
companies	have	a	much	larger	share	than	it	 is	the	case	in	Bosnia	and	Herzegovina.	 It	can	be	
concluded	 that	 in	 the	 last	 decade	 the	 insurance	 market	 in	 Serbia	 is	 medium	 concentrated	
according	to	Herfindahl‐Hirschman	index	calculated	by	total	premium	and	it	differs	significantly	
in	the	market	of	Bosnia	and	Herzegovina,	where	permanent	low‐concentration	bid	is	present	in	
the	 observed	 period	 ‐	 there	 is	 greater	 equality	 of	 market	 share	 compared	 with	 the	 high	
inequality	among	market	participants	in	the	insurance	sector	of	the	Republic	of	Serbia.	
In	this	paper,	the	multi‐linear	regression	model	was	applied	in	order	to	analyse	the	impact	of	

selected	macroeconomic	indicators	on	the	amount	of	insurance	premiums	in	the	period	2000‐
2017.	We	can	conclude	that	the	greatest	impact	on	the	amount	of	the	premium	in	Bosnia	and	
Herzegovina	had	Gross	Domestic	Income	and	Wage	and	Salaries	Workers.	 In	the	Republic	of	
Serbia,	the	greatest	influence	on	the	amount	of	premium	in	the	observed	period	had	Average	Net	
Salary,	Households	and	Final	Consumption	Expenditure	and	Gross	Domestic	Income.	Therefore,	
there	is	a	positive	correlation	between	household	income	(including	salary	and	other	income)	
and	premiums.	Also,	in	the	periods	when	household	expenditure	was	higher,	the	amount	of	total	
insurance	premium	was	also	higher.	
Both	of	the	selected	economies	have	undergone	changes	in	the	financial	sector,	particularly	in	

terms	of	 recent	regulatory	reforms.	This	paper	provides	 information	on	 the	determinants	of	
insurance	 demand	 in	 the	 Republic	 of	 Serbia	 and	 Bosnia	 and	 Herzegovina.	 This	 research	 is	
limited	 to	a	macro	 level	analysis	of	 the	 insurance	demand.	Further,	detailed	analysis	can	be	
performed	on	individual	life	insurance	products,	which	may	result	in	more	reliable	findings.	The	
period	2000‐2017	was	analysed,	so	 future	studies	can	 further	segregate	 the	CEE	region	 into	
developed	and	developing	economies	and	make	a	detailed	time	series	analysis	 incorporating	
both	pre‐crisis	and	post‐crisis	period.		

ACKNOWLEDGEMENTS	

This	paper	is	a	result	of	research	projects	under	the	code	179015	(Challenges	and	Prospects	
of	 Structural	 Changes	 in	 Serbia:	 Strategic	 Directions	 for	 Economic	 Development	 and	
harmonisation	with	EU	requirements)	 financed	by	the	Ministry	of	Science	and	Technological	
Development	of	the	Republic	of	Serbia.	

REFERENCES	

Beck,	 T.	 &	 Webb,	 I.	 (2003).	 “Economic,	 demographic	 and	 institutional	 determinants	 of	 life	
insurance	 consumption”,	 The	 World	 Bank	 Economic	 Review,	 17(1):	 51–88.,	 doi:	
10.1093/wber/lhg011.	

Bianchi,	T.,	Ebner,	G.,	Korherr,	R.	&	Ubl,	E.	(2011).	“The	Austrian	Insurance	Industry	in	CESEE:	
Risks	and	Opportunities	from	a	Financial	Stability	Point	of	View.”	Financial	Stability	Report	22	
[Internet],	 pp.	 88–106.	 Available	 at:	
https://www.researchgate.net/profile/Eva_Ubl/publication/227462654_The_Austrian_Insur



	 		
Tijana	Kaličanin,	Sandra	Kamenković,	Ivana	Simeunović	 35	

ance_Industry_in_CESEE_Risks_and_Opportunities_from_a_Financial_Stability_Point_of_View/l
inks/0deec53b4fe428fd37000000.pdf	(Accessed:	28	March	2019).	

Burić	Novović,	M.,	Cerović	Smolović,	J.,	Lipovina	Božović,	M.	&	Lalević	Filipović,	A.	(2017).	
"Impact	 of	 economic	 factors	 on	 life	 insurance	 development	 in	 Western	 Balkan	
Countries."	Zbornik	 radova	 Ekonomskog	 fakulteta	 u	 Rijeci	35,	 br.	 2:	 331‐352.	
https://doi.org/10.18045/zbefri.2017.2.331.	

Dragos,	S.	(2014).	“Life	and	non‐life	insurance	demand:	the	different	effects	of	influence	factors	
in	 emerging	 countries	 from	 Europe	 and	 Asia”.	 Economic	 Research	 Ekonomska	 Istraživanja,	
27(1):	169–180,	doi:	10.1080/1331677x.2014.952112.	

Haiss,	P.	&	Sümegi,	K.	(2008).	“The	relationship	between	insurance	and	economic	growth	in	
Europe:	a	theoretical	and	empirical	analysis“	Empirica,	35(4):	405–431,	doi:	10.1007/s10663‐
008‐9075‐2.	

Kaličanin,	T.	&	Hanić,	A.	(2016a).	"Comparative	Analysis	of	Levels	of	Banking	Sector	Markets	
Concentration	in	CEE	Region."	Economic	Analysis,	49(1‐2)	59‐72.	

Kaličanin,	T.	&	Hanić,	A.	 (2016b).	"Analiza	koncentracije	u	finansijskom	sektoru."	 In:	Pravci	
strukturnih	promena	u	procesu	pristupanja	Evropskoj	uniji,	ed.	Minović,	Jelena,	Duško	Bodroža,	
Ivan	Stošić,	and	Božo	Drašković,	308‐324.	Belgrade:	Institute	of	economic	sciences.	

Kaličanin,	T.	&	Lazić,	M.	(2018).	"Evaluating	the	Level	of	Market	Concentration	in	Insurance	
Sector:	the	case	of	Serbia."	In:	Western	Balkans	Economies	in	EU	Integration,	past,	present	and	
future,	ed.	Richet,	Xavier,	Dejan	Erić,	Srđan	Redžepagić,	Ivan	Stošić	and	Duško	Bodroža,	202‐
221.	Nice:	CEMAFI	International	Association.	

Kjosevski,	J.	(2012).	“The	determinants	of	life	insurance	demand	in	central	and	Southeastern	
Europe”.	 International	 Journal	 of	 Economics	 and	 Finance,	 4(3):	 pp.	 237–247,	 doi:	
10.5539/ijef.v4n3p237.	

Lipczynski,	J.	&	John,	W.	(2001).	Industrial	Organization:	An	Analysis	of	Competitive	Markets.	
Prentice‐Hall,	London.	

Martin,	S.	(2002).	Advanced	Industrial	Economics.	Blackwell	Publishers	Ltd,	Oxford.	
Mitra,	A.	(2017).	“Influencers	of	Life	Insurance	Investments:	Empirical	Evidence	from	Europe”.	
Australasian	 Accounting,	 Business	 and	 Finance	 Journal,	 11(3):	 87–102.	
https://doi.org/10.14453/aabfj.v11i3.7.	

Outreville,	 F.	 (1990).	 "The	 Economic	 Significance	 of	 Insurance	 Markets	 in	 Developing	
Countries."	The	Journal	of	Risk	and	Insurance	57,	no.	3:	487‐98.		doi:10.2307/252844.	

Ward,	 D.	 &	 Zurbruegg,	 R.	 (2002).	 “Law,	 Politics	 and	 Life	 Insurance	 Consumption	 in	 Asia”,	
Geneva	Papers	on	Risk	and	Insurance,	27(3):	395–412.	doi:10.1111/1468‐0440.00181.	

Webb,	I.,	Martin,	G.	&	Skipper,	H.	(2002).	“The	Effect	of	Banking	and	Insurance	on	the	Growth	
of	Capital	and	Output”.	Centre	for	Risk	Management	and	Insurance,	Working	Paper,	No.	02‐1.	
Robinson	College	of	Business,	Georgia	State	University,	Atlanta.	

Zhi,	Z.		(1998).	Die	Nachfrage	nach	Lebensversicherungen:	Eine	empirische	Analyse	für	China,	
Mannheimer	 Manuskripte	 zu	 Risikotheorie,	 Portfolio	 Management	 und	
Versicherungswirtschaft	112,	Universität	Mannheim.	

www.nbs.rs	
www.azobih.gov.ba	
www.data.worldbank.org	
	
	

Article	history:	 Received:		April	1,	2019	
Accepted:	May	29,	2019