Uddannelses- og Forskningsudvalget 2015-16
UFU Alm.del
Offentligt
1647752_0001.png
An evaluation of the Danish Innovation
Assistant Programme
En effektmåling af Videnpilotordningen
Innovation: Analysis and evaluation 12/2013
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0002.png
An evaluation of the Danish Innovation
Assistant Programme
En effektmåling af Videnpilotordningen
By:
The Centre for Economics and Business Research (CEBR)
Copenhagen Business School
Johan M Kuhn
1
Published by:
The Danish Agency for Science, Technology and Innovation
Bredgade 40
1260 København K
Tel. +45 3544 6200
ISBN: 978-87-92776-69-3
1
Thanks to Thomas Alslev Christensen and Klaus Ammitzbøll, both DASTI, Christian M. Dahl, University of Southern Denmark,
Søren Bo Nielsen, Copenhagen Business School, Anders Sørensen and the colleagues at CEBR, and seminar participants at the
Danish Research Unit for Industrial Dynamics (DRUID) at Aalborg University for fruitful discussions and highly relevant
suggestions for improvements of the analysis.
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0003.png
An evaluation of the Danish Innovation
Assistant Programme
En effektmåling af Videnpilotordningen
Copenhagen, August 2013
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
3
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0004.png
CONTENTS
SAMMENFATNING PÅ DANSK
EXECUTIVE SUMMARY
DEUTSCHSPRACHIGE ZUSAMMENFASSUNG
1. INTRODUCTION
2. THE INNOVATION ASSISTANT PROGRAMME (VIDENPILOTORDNINGEN)
3. DATA
DASTI data
The Statistics Denmark data
The Experian data
A first look at the data
4. INDIVIDUAL-LEVEL ANALYSIS
General methodological issues
Selection of controls
Empirical specification
Individual-level analysis: the regression model
Individual-level analysis: descriptive statistics
Individual-level analysis: Results
Potential employment effects
Potential earnings effects
Individual-level potential effects for different subsamples
6
9
12
14
16
17
17
17
18
18
20
20
21
22
23
24
32
32
40
47
4
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0005.png
5. COMPANY-LEVEL ANALYSIS
Empirical specification
Company-level analysis: selection of controls
Company-level analysis: the empirical model
Company-level analysis: Descriptive Statistics
Company-level analysis: Results
Potential employment effects
Potential effects on value added, net income (profits) and return on assets
Potential effects on average wage costs and labour productivity
Results for subsamples
Potential effects on the number of highly educated employees
Potential effects on the number of employees
Potential effects on value added
Potential effects on net income (profits) and return on assets
Potential effects on wages and labour productivity
6. EXTENSIONS
The survival of VP-companies
A comparison of VP-companies and companies participating in Innovation
Networks
A comparison of VP-companies and an extended sample of control companies
7. CONCLUSIONS
APPENDIX 1: ADDITIONAL TABLES OF THE COMPANY-LEVEL ANALYSIS
52
52
52
53
54
63
66
70
75
79
80
82
84
86
90
94
94
95
100
104
108
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
5
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0006.png
SAMMENFATNING PÅ DANSK
Videnpilotordningen
under
Rådet for Teknologi og Innovation
blev lanceret som en
del af
’Viden flytter ud’-tiltaget
under regeringen i 2004. Ordningen har eksisteret
siden 2005 og har som formål at øge små og mellemstore virksomheders vækst ved
at øge incitamentet til og nedbryde barrierer for ansættelsen af akademikere i disse
virksomheder.
På baggrund af den danske vækstudfordring generelt og den økonomiske afmatning
i kølvandet på finanskrisen indtager Videnpilotordningen en central rolle blandt de
politikinstrumenter, der sigter at skabe vækst og øge virksomheders kompetencer
i forhold til innovation og nytænkning. Interessen for ordningen skyldes også,
at en række tidligere analyser (f.eks. Junge og Skaksen, 2010, CEBR, 2011
2
)
har vist positive sammenhænge mellem virksomheders andel af højtuddannede
medarbejdere og deres produktivitet, og at udbygningen af ordningen kan
argumenteres for at kunne reducere den for tiden høje arbejdsløshed blandt
akademikere i Danmark.
Som led i sin løbende evalueringsstrategi har
Styrelsen for Forskning og
Innovation,
der administrerer ordningen, bedt
Centre for Economics and Business
Research (CEBR)
om at belyse, hvorvidt det kan vises, at ordningen lever op til
sin målsætning. Til dette formål har CEBR fulgt både deltagende personer og
virksomheder i et omfattende datamateriale. Denne rapport beskriver tilhørende
analyse.
Med hensyn til metodologi, analysevariation samt hvilke indikatorer, der vurderes,
er denne effektmåling af Videnpilotordningen i international sammenhæng ’best
practice’. Den kan tjene som målestok for evaluering af effekten af en specifik
indgriben i erhvervslivet, der kan udføres, hvis behandlingsgruppens etablerede
datakvalitet er ganske høj, og der findes højt detaljerede landsdækkende registre med
dataserier over tid for virksomheder og individer.
Analysen sammenligner løn- og beskæftigelsesudvikling for en stikprøve af
individer, der deltager i ordningen (videnpiloter) med andre, sammenlignelige
personer, der ikke deltager. Analysen sammenligner også vækst og
produktivitetsudviklingen i en stikprøve af virksomheder, der deltager i ordningen,
med andre (meget) sammenlignelige virksomheder, der ikke deltager.
Junge og Skaksen, 2010, Produktivitet og videregående uddannelse, CEBR, 2011, Ansættelse af Ph.D.er og
produktivitet.
2
6
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0007.png
Analysens resultater kan sammenfattes som følger:
Personer, der deltager i ordningen, øger deres beskæftigelsesrate i forbindelse med
deltagelsen i ordningen. Dette er ikke overraskende, da ansættelse er en definerende
karakteristik af selve ordningen. Efter mere end et år efter begyndelsen af
deltagelsen kan det dog ikke længere vises, at beskæftigelsesraten blandt deltagerne
er højere end i en referencegruppe af højt sammenlignelige individer – men det kan
nævnes, at analysens observationsperiode delvist ligger i en højkonjunktur med lav
arbejdsløshed blandt højtuddannede.
Personer, der deltager i ordningen, øger deres lønindkomst i forbindelse med
deltagelsen i ordningen. Lønindkomsten forbliver højere end i referencegruppen i
årene efter begyndelsen af deltagelsen, men konvergerer herefter.
Virksomheder, der deltager i ordningen, øger deres årlige vækst i antallet af
højtuddannede medarbejdere i forbindelse med deltagelsen. Det kan dog ikke
vises, at virksomheder, der deltager i ordningen, bliver ved med at ansætte flere
højtuddannede i årene efter deltagelsen i ordningen.
Virksomheder, der deltager i ordningen, er også kendetegnet ved et midlertidigt
forhøjet antal medarbejdere i årene efter deltagelsen, men det viser sig at være
svært at finde robuste sammenhænge for finansielle succesparametre som
værditilvækst, profit eller arbejdsproduktivitet. Dette skyldes ret stor variation i
nogle virksomheders udvikling i disse variable, som ikke er relateret til, hvorvidt de
deltager i ordningen.
For delstikprøver af mindre virksomheder, som ikke er kendetegnet ved større
ændringer i deres succesvariable, findes, at deltagelsen i ordningen korrelerer positivt
med stigende værditilvækst og profit. Således forøger deltagende virksomheder deres
værditilvækst i gennemsnit med op til ca. 800.000 kr. og profitten med op til ca.
400.000 kr. i årene efter deltagelsen.
Disse resultater peger i retning af eventuelle positive effekter af ordningen og er
i tråd med en tidligere analyses
3
resultater, men er behæftede med en betydelig
statistisk usikkerhed. Så selvom datamaterialet er blevet betydelig udvidet i forhold
til den tidligere analyse, er det på baggrund af de nye resultater stadig ikke muligt at
træffe sikre udsagn om, i hvilket omfang deltagelsen i videnpilotordningen forøger
værdiskabelsen eller profitten i virksomheden.
Det er ikke muligt at påvise positive sammenhænge mellem deltagelsen
i programmet og arbejdsproduktivitet, lønniveau og afkastningsgraden
(return-on-assets).
DASTI, 2010, ”Effektmåling af videnpilotordningens betydning for små og mellemstore virksomheder
Innovation: Analyse og evaluering 4/2010”
3
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
7
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0008.png
Som sammenfatning kan det siges, at eventuelle positive effekter af ordningen
kommer til udtryk i, at videnpiloter kommer hurtigere i arbejde, hvilket er forbundet
med, at de kommer på et højere lønniveau i de første år efter deltagelsen end
andre, sammenlignelige personer, der ikke deltager. Disse potentielle effekter kan
forventes at være højere i de nuværende år, som i modsætning til en stor del af
analyseperioden er kendetegnet ved en lavkonjunktur.
Resultater for virksomhedsdelen peger i retningen af, at virksomheder, som deltager
i ordningen, oplever højere vækst i værditilvækst og profit, men en betydelig
statistisk usikkerhed medfører, at disse resultater skal fortolkes med forsigtighed.
8
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0009.png
EXECUTIVE SUMMARY
The Innovation Assistant Programme under the Danish Council for Technology
and Innovation was launched as part of the
“Knowledge is moving out”-initiative
by the Danish government in 2004. The programme has existed since 2005 and
has the purpose of increasing the growth of small and medium-sized enterprises by
increasing incentives and breaking down barriers to employment of highly educated
individuals in these enterprises.
Because of Denmark’s growth problems in general and the economic downturn in
the wake of the financial crisis, the Innovation Assistant Programme plays a central
part among the policy instruments aiming at creating growth and increasing the
competences of enterprises on innovation and creative thinking. The interest in the
programme is also due to a number of previous analyses (ie. Junge og Skaksen, 2010,
CEBR, 2011
4
) that have shown positive correlations between the share of highly
educated employees in enterprises and their productivity, and that the expansion
of the programme can be argued to reduce the presently high unemployment rate
among the highly educated in Denmark.
As part of its ongoing evaluation strategy, the Danish Agency for Science,
Technology and Innovation (DASTI), which administers the programme, has asked
the Centre for Economics and Business Research (CEBR) to cast light on whether it
can be shown that the programme fulfils its objectives. For this purpose, CEBR has
followed both participating individuals and enterprises in an extensive set of data.
This report describes the corresponding analysis.
With regard to methodology, variation of the analysis and the indicators taken
into consideration, this impact analysis of the Innovation Assistant Programme is
international best practice. It may serve as a standard for intervention evaluations
that can be carried out if the established data quality of the treatment group is quite
high, and highly detailed national registers with data time series for enterprises and
individuals are available.
The analysis compares salary and employment developments for a sample of
participating individuals (innovation assistants) with other comparable individuals
not participating. The analysis also compares growth and productivity developments
for a sample of participating companies with other (highly) comparable companies
not participating.
The results of the analysis can be summarised as follows:
Individuals who participate in the programme increase their employment rate
in association with participating in the programme. This is not surprising, since
employment is a defining characteristic of the programme itself. It cannot be shown
that the employment rate among participants is higher than for a reference group of
highly comparable individuals more than a year after starting to participate.
Junge og Skaksen, 2010, Produktivitet og videregående uddannelse, CEBR, 2011, Ansættelse af Ph.D.er og
produktivitet.
4
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
9
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0010.png
However, it should be noted that the observation period of the analysis falls partly
within an economic boom period with low unemployment among the highly
educated.
Individuals who participate in the programme increase their salary income in
association with participation. Salary income remains higher than for the reference
group in the years after starting to participate, but then converges.
Companies that participate in the programme increase their yearly growth of the
number of highly educated employees in association with participation. However,
it cannot be shown that companies that participate in the programme continue to
employ more highly educated individuals in the years after participation.
Companies that participate in the programme are also characterised by a temporary
increase in the number of employees in the years after participation, but it turns
out to be difficult to find robust associations for financial success parameters such
as value added, profits or labour productivity. This is due to a quite large variation
in certain companies’ developments for these variables, which is unrelated to their
participation in the programme.
For subsamples of smaller companies that are not characterised by large changes in
their success variables, it is found that participation in the programme is positively
correlated to increasing value added and profits. Thus, participating companies on
average increase their value added by up to approx. DKK 800,000 (EUR 106,000)
and their profits by up to approx. DKK 400,000 (EUR 53,000) in the years after
participation.
These results point to possible positive effects of the programme and correspond
with the results of a previous analysis,
5
but are subject to a significant statistical
uncertainty. So even though the data material has been expanded significantly
compared to the previous analysis, it is still not possible to make any certain
claims about the extent that companies’ value added and profits are increased by
participating in the programme on the background of the new results.
It is not possible to show positive correlations between programme participation and
labour productivity, salary levels and return on assets.
In conclusion, it can be said that any positive programme effects are expressed by
innovation assistants finding employment quicker, which is associated with a higher
salary level in the first years after participating than other comparable individuals
who do not participate. These potential effects can be expected to be higher in the
present years, which unlike a large part of the analysis period are characterised by
an economic downturn.
DASTI, 2010, ”Effektmåling af videnpilotordningens betydning for små og mellemstore virksomheder
Innovation: Analyse og evaluering 4/2010”
5
10
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0011.png
For the company part of the analysis, results indicate that participating companies
experience higher growth in value added and profits, but a significant statistical
uncertainty means that these results must be interpreted with care.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
11
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0012.png
DEUTSCHSPRACHIGE ZUSAMMENFASSUNG
Die vorliegende Studie wurde vom
Centre for Economics and Business Research
(CEBR)
an der Handelshochschule Kopenhagen (CBS) für die
Styrelsen for
Forskning og Innovation (DASTI)
des Ministeriums für Forschung, Innovation und
weiterführende Bildung erstellt.
Sie betrachtet das Wissenspilotprogramm („Videnpilotordning“,
VP-Programm),
ein vom
DASTI
geführtes Innovationsprogramm. Dieses Programm existiert seit
2005 und subventioniert die Neuanstellung von Akademikern in kleinen und
mittelständischen Unternehmen mit geringem Anteil hochqualifizierter Fachkräfte
durch Gehaltszuschüsse. Ziel des Programms ist es, die Kompetenzen teilnehmender
Unternehmen zu erhöhen und deren Wettbewerbsfähigkeit zu steigern.
Die Studie folgt ca. 360 teilnehmenden Personen und ca. 320 teilnehmenden Firmen
in dänischen Registerdaten. Diese erlauben es, Aussagen über den Berufserfolg
der am Programm teilnehmenden Personen zu machen, sowie das Wachstum
teilnehmender Unternehmen zu analysieren.
Der Berufserfolg wird dabei anhand der Entwicklung des Beschäftigungsgrades
und Jahresgehaltes gemessen. Auf Unternehmensniveau betrachtet die Studie
Entwicklungen in der Anzahl hochausgebildeter Mitarbeiter, der Beschäftigung,
der Lohnkosten, sowie der finanziellen Variablen Wertschöpfung, Gewinn und
Arbeitsproduktivität.
Um den Berufserfolg der teilnehmenden Personen und das Wachstum der
Unternehmen beurteilen zu können, werden aus den umfangreichen vorliegenden
Registerdaten Kontrollgruppen von Personen oder Unternehmen ausgewählt, die
die gleichen oder sehr ähnliche äussere Merkmale aufweisen wie die Teilnehmer
im Jahr vor deren Teilnahme im VP-Programm. Die statistischen Methoden der
Studie bestehen aus Vergleichen der verschiedenen Erfolgsvariablen zwischen
den Teilnehmer- und den Kontrollgruppen. Zusätzlich dazu erlauben die Daten,
für teilnehmende Unternehmen die Entwicklungen von Erfolgsvariablen nach
Teilnahme im Programm mit den entsprechenden Entwicklungen vor der Teilnahme
zu vergleichen. Ein ähnlicher Vergleich für Unternehmen in der Kontrollgruppe
erlaubt es, auch unbeobachtbare Faktoren aus dem statistischen Modell
herauszufiltern.
Die Ergebnisse der Studie lassen sich wie folgt zusammenfassen:
Personen, die am VP-Programm teilnehmen, weisen im ersten Jahr nach
Beginn der Teilnahme am Programm eine höhere Beschäftigungsquote als
Personen der Vergleichsgruppe auf. Nach zwei und mehr Jahren haben sich die
Beschäftigungsquoten beider Gruppen jedoch weitgehend angeglichen, womit es
nicht möglich ist, einen langfristigen Beschäftigungseffekt des VP-Programms auf
individueller Ebene nachzuweisen. An dieser Stelle sei jedoch darauf hingewiesen,
dass ein grosser Teil der Beobachtungsperiode der Analyse in eine Zeit guter
Konjunktur mit allgemein geringer Akademikerarbeitslosigkeit fällt.
12
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0013.png
Personen, die am Programm teilnehmen, weisen eine bessere Gehaltsentwicklung
als Personen, die nicht teilnehmen, auf. Dieser Unterschied ist statistisch signifikant
für die ersten Jahre nach Beginn der Teilnahme.
Unternehmen, die am Programm teilnehmen, erhöhen die Beschäftigung
hochqualifizierter Mitarbeiter im Vergleich zu Unternehmen in der Kontrollgruppe,
sowie die Beschäftigung generell mit, im Durchschnitt, ca. einem zusätzlichen
Mitarbeiter in Verbindung mit der Teilnahme am Programm.
In Bezug auf die finanziellen Erfolgsvariablen lässt sich feststellen, dass es
grundsätzlich schwierig ist, potentielle Teilnahmeeffekte in den Daten zu
isolieren: erhebliche Heterogenität der Firmen in Bezug auf die Entwicklung
der Erfolgsvariablen relativ zu der Grösse der Stichprobe und der Grösse der
potentiellen Effekte führt dazu, dass die Ergebnisse der jeweiligen Analyse von der
Wahl des ökonometrischen Modells sowie der Stichprobenauswahl abhängen.
In Stichproben kleinerer teilnehmender Unternehmen mit geringer Heterogenität
in den Erfolgsvariablen und der Entwicklung dieser Variablen, sind teilnehmende
Unternehmen durch, im Durchschnitt, höheres Wachstum in der Wertschöpfung
sowie des Unternehmensgewinns gekennzeichnet. Hier liegen für teilnehmende
Unternehmen die potentiellen geschätzten Teilnehmereffekte bei bis zu ca. 800,000
Dänischer Kronen (ca. 106.000€) in Bezug auf die die jährliche Wertschöpfung
und 400,000 Kronen (53.000€) für Unternehmensgewinn in den Jahren nach
Programmteilnahme.
Diese Ergebnisse ähneln den Ergebnissen einer früheren Studie, die auf weniger
umfangreichem Datenmaterial beruht
6
, lassen sich jedoch aufgrund eines
Mangels an statistischer Signifikanz und fehlender Robustheit in Bezug auf die
Stichprobenauswahl nicht verallgemeinern.
Für die Erfolgsvariablen Rendite (return on assets), Lohnkosten (als Mass für
das Lohnniveau des Unternehmens) sowie Arbeitsproduktivität lassen sich keine
positiven potentiellen Teilnehmereffekte ermitteln. Auch in Bezug auf diese
Variablen lassen die Ergebnisse den Schluss zu, dass die Bedeutung der Anstellung
von Wissenspiloten in vielen Unternehmen von anderen Entwicklungen überlagert
wird, und dass auch das im Vergleich zu einer früheren Studie ausgeweitete
Datenmaterial noch nicht ausreicht, um gesicherte Aussagen über den Erfolg des
Programms treffen zu können.
DASTI, 2010, ”Effektmåling af videnpilotordningens betydning for små og mellemstore virksomheder
Innovation: Analyse og evaluering 4/2010”
6
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
13
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0014.png
1.
INTRODUCTION
This report presents the data, methodology, and results of an evaluation of the
Danish Innovation Assistant Programme (‘Videnpilotordningen’
- VP programme
in the following). The analysis was completed by CEBR for DASTI in 2012. It
contributes to DASTI’s strategy to continuously monitor and evaluate its innovation
support programmes, to develop and improve the designs of its initiatives, and to
improve programme evaluation techniques.
The VP programme was launched in 2005 and aims at increasing the growth and
productivity of small and medium-sized enterprises (SMEs) by increasing the
share of their employees with a higher education.
7
It is supposed to overcome any
mutual reservations between SME managers and university graduates and increase
academic knowledge in SMEs. To achieve this goal, the VP programme subsidizes
the employment of university graduates in small and medium-sized companies.
Although the programme is small-scale, especially when compared to e.g. U.S.
or European-level knowledge transfer programmes, schemes similar to the VP
programme are currently being discussed or implemented in other countries as well,
for example in a couple of local states in Germany and Austria. For this reason, the
present study might also have an interest outside Denmark. From an academic point
of view, the study furthermore contributes to our understanding of employment
subsidies for highly skilled employees and the effects of knowledge transfers to
SMEs.
The present analysis was supposed to address two questions: First, how do
individuals who participate in the programme perform with regard to their
employment and income developments? Second, how do participating companies
perform in terms of employment and productivity growth? For this purpose,
individuals and companies are followed in large-scale register data, and the success
of programme participants is compared to highly similar individuals and companies
that do not participate in the programme.
The two different questions imply that the present report is divided into two
parts. The first part addresses the question of the extent to which individuals
benefit from participating in the programme. This question has recently gained
increasing public attention in Denmark, as unemployment among especially
young university graduates is soaring in the aftermath of the recent financial crisis
and the current Danish economic slowdown. This part looks at employment and
salary developments of programme participants in association with programme
participation.
The education classifications of this study follow the International Standard Classification of Educations (ISCED).
In the following, employees with at least a post-secondary education (ISCED classifications 4,5, and 6) are re-
ferred to as ‘highly educated employees’.
7
14
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0015.png
The second part of the analysis looks at whether companies benefit from
participating in the VP programme. This company-level analysis is again based
on large-scale register data. It might be considered of primary interest, since the
purpose of the VP programme is to increase company performance, whereas any
individual employment effects are secondary.
The success parameters of interest in this part of the company-level analysis are
employment growth, the number of highly educated employees, and the growth in
value added, profits, return on assets, average wages, and labour productivity.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
15
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0016.png
2.
THE INNOVATION ASSISTANT PROGRAMME
(VIDENPILOTORDNINGEN)
An Innovation Assistant (‘videnpilot’, VP) is an academic employee with a post-
secondary or tertiary-level education. In Danish educational terminology, this
corresponds to respectively a medium-length (bachelor level) and a long higher
education (postgraduate level). The employee has to be employed in an SME to solve
one or more specific development tasks.
A VP-project is subsidised by DASTI and is supposed to contribute to the company’s
innovation, growth and productivity. The subsidy pays up to half of the VP’s salary,
with a maximum of DKK 12,500 (€1,700) a month for 6-12 months.
Privately owned small or medium-sized companies with at least 2 and at most 100
employees can apply for funding if there are at most two highly educated employees
in the company, it has existed for at least a year, and its yearly revenues surpass
DKK 1 million (€130,000).
The programme was launched in the beginning of 2005. Until 2012, approximately
500 projects have been completed.
8
For the following analysis, it is relevant to have an idea of just how VP-projects
come into life to better understand what kind of individuals and companies
participate in the programme. However, it needs to be acknowledged that there is
little if any general knowledge about how VP-company collaborations are initiated.
Anecdotal evidence suggests that it is often the VP who contacts the company
and suggests an employment relationship under the VP programme. And yet, it
might also be presumed that companies hiring new employees might exploit the
opportunity of saving wage costs in the beginning of the employment relationship.
The analysis can only consider projects for which there is information in the data after they have been started, so
the most recent projects are not part of the analysis.
8
16
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0017.png
3.
DATA
The data for the analysis is from three sources:
1. DASTI supplied information on individual VP-projects. Information includes
individual identification numbers of participating individuals, company
identifiers, and the start date of the project. These data will henceforth be
called the
DASTI data.
2. Data from companies’ financial reports from
Experian A/S,
a credit rating
agency. These data will be referred to as the
Experian data
in the following
sections.
3. Register information from
Statistics Denmark.
This is matched employer-
employee data including information on individuals (demographic
information, information on education, wage and occupation) and companies
(e.g. size, turnover). These data will be referred to as the
Statistics Denmark
data.
DASTI data
Since the start of the programme in 2005, DASTI has continuously collected
information such as individual IDs of VPs, the start-up time of VP-projects, hosting
company IDs (VP-companies in the following) and whether or not projects were
completed or aborted before schedule. Individual IDs are social security numbers
(CPR numbers) while company IDs are the numbers by which companies are
registered by the public authorities (CVR numbers).
The Statistics Denmark data
Characteristics for individuals are drawn from Statistics Denmark’s register. Data
is available up to 2010, implying that there is no information on the most recent
projects. Statistics Denmark data is typically available on an annual basis, with
census date in mid-November. It allows associating individuals with their companies
using the unique company and individual IDs.
9
Over the last decades, the data
resources of Statistics Denmark have been continuously extended, as all Danish data
with an associated individual or company ID can be merged with the existing data.
For example, the present analysis benefits from Statistics Denmark’s individual-level
information on education (degrees, focus of electives, grades) and company-level
information on turnover.
Timmermans B. The Danish Integrated Database for Labor Market Research: Towards Demystification for the
English Speaking Audience. Aalborg. 2010
9
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
17
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0018.png
The Experian data
The Experian data consists of approximately 1.7 million financial records in the
period from 2000 to 2010. The timing of the records is based on the closing dates
of the financial report periods. In case of companies filing multiple reports in a
calendar year, only one of these is selected for the analysis. The closing date of the
financial reports sets the time structure of the company-level analysis (which is
relevant to before-after comparisons). When merging information from Statistics
Denmark with the Experian data (such as information on the number of highly
skilled employees), it is the latest available information in the Statistics Denmark
registers before a given financial report’s closing date which is used in association
with the financial report in question.
10
A first look at the data
As a point of departure, there are 416 VPs in the DASTI data. Six of these cannot be
found in the registers that form the basis of the analysis, and there is no information
on the highest educational degree of 16 individuals. Since education is a control
variable of key importance for the analysis, these individuals are not included,
leaving us with 394 individuals for the individual-level analysis. For 30 of these
individuals, it has proven impossible to find highly similar controls. This implies
that the individual-level analysis is based on 364 individuals who participated in the
VP programme.
370 companies which have hosted VP projects can be found in the Experian
database the year before the start of programme participation. The remaining
companies not in the Experian data must be presumed to be unincorporated and thus
not obliged to submit financial reports to the authorities. Companies can be followed
until 2009 in the Statistics Denmark data and until 2011 in the Experian data. In
the sample of companies employed for the subsequent analysis, the companies are
observed over an average time span of 6.7 years.
The results of this report are based on DASTI’s information on the company-VP
matches. This is important to note, because the identification of hosting companies
is not always straightforward: Single companies may have several CVR numbers,
and there might be an element of randomness or selection regarding which CVR
number hosting companies use to register their VP-projects. In approximately 30
percent of the projects, the Statistics Denmark data (described in greater detail
below) suggest that the VP is employed at a company with a different CVR number
than the one stated in the DASTI data.
11
Most companies have their closing date at the end of December, which implies a short time lag between the
Statistics Denmark information (of end-November) and the financial report information. However, there are also
companies that have chosen other dates, e.g. end of March, to close books. For these companies, the information
from the Statistics Denmark registers comes with a time lag of up to one year.
10
This will of course govern robustness checks of later findings. It might be noted that some of the companies
that the Statistics Denmark data suggests are the ‘real’ hosts of the VP-projects do not fulfill the conditions for
programme participation.
11
18
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0019.png
For one of the extensions of the analysis, DASTI provided data on companies
that have participated in the so-called
Innovation Networks.
These networks are
collaborations of typically small and medium-sized companies with the purpose
of increasing knowledge transfer and innovation. The data on Innovation Networks
consist of 1923 observations belonging to 1158 companies, the discrepancy owing to
the fact that a number of companies participate in these networks more than once.
We only consider the earliest participation in any of these networks for the following
analysis.
Of the 1158 firms that participated in any of the networks, 1121 are found in the
Experian data. The discrepancy must again be assumed to be a result of non-
incorporated firms.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
19
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0020.png
4.
INDIVIDUAL-LEVEL ANALYSIS
General methodological issues
The empirical analysis addresses the basic evaluation problem: What is the causal
effect of participation in the programme on given outcome variables?
In accordance with the relevant econometrical literature, which again borrows
from the biometrics and epidemiological literature, programme participants will
subsequently also be referred to as treatments. Also, starting to participate in the
programme will also be referred to as receiving a treatment. Non-participants who
act as a control group for the statistical comparisons will be referred to as controls.
12
There are different ways of addressing the evaluation problem. One way is using a
linear regression model. This model is estimated on a sample of both participating
and non-participating individuals. The linear regression model includes a set of
conditioning variables which hold constant a set of observable characteristics and
identifies causal effects under a
conditional independence assumption,
by which
participants do – on average – not differ from non-participants in characteristics
that (a) have an impact on the outcome variables and (b) are not controlled for in the
regression model.
These characteristics, sometimes called ‘omitted variables’, prohibit interpreting
treatment-control differences in outcome variables as causal programme effects.
Instead, they offer alternative interpretations of latter results. And the above
‘identifying’ conditional independence assumption is equivalent to assuming that
there exists no other explanation for treatment-control differences in the outcome
variables than the fact that treatments have participated in the programme.
Obviously, any empirical model supposed to isolate programme effects needs to
maximise the validity of this assumption. A first step in this direction is to carefully
select a control group for the analysis by a
matching procedure.
These procedures
are explained in greater detail in the following sections. The procedures select one
(or more than one) ‘twin’ or ‘match’ for each treatment. They imply that controls are
highly similar to treatments in their observable characteristics, which also increases
the likelihood that treatments and controls are highly similar in their unobservable
characteristics.
Also, the way the dependent ‘outcome’ variable enters the model has implications
for the validity of the conditional independence assumption. For example, statistical
comparisons of individual-specific before-after developments over time or fixed
effects models will typically be preferred to cross-sectional comparisons.
And as noted earlier, a set of conditioning variables can control for any systematic
differences between the treatment and control group which might remain even if the
controls were selected in a way to make them as similar as possible to the treatments
in their observable characteristics.
The term ‘controls’ is also sometimes used for the conditioning variables in statistical models. In this report,
‘controls’ refers to subjects in a reference group and not conditioning variables.
12
20
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0021.png
Selection of controls
Obviously, the validity of any statistical comparison can be questioned if treatments
(individuals or companies) systematically differ from the controls in characteristics
that the subsequent regression model is unable to take fully into account. We
want to select individuals into the control groups that are as similar as possible
to the treatments in the most dimensions possible. The problem of finding ‘good’
matches is that there are no two absolutely identical individuals, so it should be
acknowledged that any analysis that identifies controls on the basis of a matching
procedure is nothing but a sophisticated comparison that requires additional all-else-
equal assumptions for causal interpretation.
The controls can be selected by a host of different matching procedures developed
over the last decades. Overviews of these procedures are found in
Caliendo and
Kopeinig, 2008,
and
Blundell and Costa Dias, 2009.
13
The basic idea is to find for
each participant one or more ‘twins’ that are as similar to the given participant as
possible, and to use these matches as the analysis’ control group.
The specific matching procedure depends on the nature of the data. The modeller
typically chooses between
matching on observables
and
propensity score matching,
or some combination of the two.
Matching on observables simply means that for each treatment, one or more ‘twins’
(referred to as matches in the following) are selected from the group of potential
controls that have the same observable characteristics in a number of dimensions.
For example, one could choose for each participating VP one control individual with
the same education, gender, and stays in the same geographic region. For companies,
one could select controls on the basis of industry, size, financial performance
measures, and other characteristics.
When treatments are not particularly unique and there are a lot of potential
candidates in the pool of potential controls, matching on observables might be the
preferred choice. But, matching on observables runs into a
multidimensionality
problem
when one uses too many observable characteristics as conditions in the
procedure: It becomes impossible to find controls for all participants when they are
required to be equal in too many dimensions.
Of course, one way of “solving” this problem would be to disregard a lot of
information in the data and only require equality in a few observable characteristics.
In this case one could, for each treatment, select one or more controls from the pool
of potential controls that are equal in a few dimensions (or use the entire population
of potential controls as controls and weigh them in the subsequent regressions).
Blundell, R., Costa Dias, M., 2009. Alternative Approaches to Evaluation in Empirical Microeconomics. Journal of
Human Resources 44(3., 565-640.
13
Caliendo, M., S. Kopeinig, 2008. Some Practical Guidance for the Implementation of Propensity Score Matching,
Journal of Economic Surveys (2008) Vol. 22, No. 1, pp. 31–72.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
21
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0022.png
Yet another option is to combine the benefits of the matching-on-observables-
procedure with the benefits of the
propensity-score-matching-procedure.
The
latter method has the benefit of allowing the use of vastly more information than
the matching-on-observables method: It condensates all variables which might be
considered relevant for the choice of programme participation into one single metric.
This is simply the estimated predicted probability of programme participation,
called the
propensity score.
14
This way, it is possible to find matches that are most
similar in terms of the propensity score instead of a set of observable characteristics.
The number of matches selected for each participant is set by the modeller, who
faces a trade-off between bias and efficiency: By including many matches for each
participant into the control group, the sample size is increased and the variance of
the subsequent estimators is reduced. However, increasing the number of matches
for each participant might lead to selecting subjects into the control group that are
not very similar to the treatment. This decreases the validity of the conditional
independence assumption. So there is a trade-off between the precision of the
statistical estimates and minimizing the risk of matching participants with controls
that differ in observed and unobserved characteristics.
Empirical specification
The empirical implementation is done in the following steps: (i) select a group of
controls, (ii) specify the regression model.
For the individual-level analysis, the selection of controls is from the registers of
Statistics Denmark, which contain information on the entire Danish population, and
is carried out in three steps:
First, we adjust the sample of potential controls. This is achieved by deleting
individuals with characteristics not found for any VP. For example, we drop
individuals with educations that no single VP has taken, and younger than the
youngest VP in our sample. The resulting data is referred to as the
adjusted
individual-level sample.
Second, we calculate a probability model for the likelihood of VP programme
participation for any given individual. This model provides evidence of which
individual characteristics are associated with programme participation, which
might be interesting in its own right. It is also used to calculate the propensity score
for each individual in the data and for each year, which is simply the predicted
participation probability for the given individual in the given year.
Rosenbaum, P.R., and D. Rubin (1983). The central role of the propensity score in observational studies for
causal effects. Biometrika (1983) 70(1): 41-55.
14
22
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0023.png
The conditioning variables of the propensity score matching procedure are selected
in cooperation with DASTI and include all variables potentially important for
programme participation and available in the data. The list consists of factors
such as demographic information (age, gender, marital status), information on
education, including 15 education categories, whether the individual is currently
in any education programme, the average grade of the final secondary education
examination, and focus of secondary education electives (math, language). There
are also occupational codes (17 categories including unemployment and leave), wage
income (9 categories), labour market experience (5 categories), and geographical
location of residence (9 categories).
For programme participants, these characteristics are collected for the year before
treatment, called ‘year
0’
or
t=0
in the following. This ensures that no information
affected by treatment enters the propensity score model.
Finally, we apply a single nearest-matching procedure (by employing STATA Corp.’s
psmatch2-command) on the basis of the probability model’s predicted propensity
scores (participation probabilities). In this procedure we also impose the condition
that twins are exactly equal in terms of education (approximately 2,200 different
categories in total and approximately 175 different categories for VPs), gender,
occupation (11 categories) and highly similar in age. Again, all information entering
the matching procedure is from the year prior to programme participation. We strive
for minimum bias of the later estimators and choose only one control (instead of
several controls) for each treatment.
For the following treatment-control analysis, it is necessary to define a year 0
(t=0) for controls just as has been done for treatments. This allows modelling the
dynamics of potential treatment effects in association with programme participation.
For controls, year 0 or t=0 is simply the year in which a given control is selected into
the control group. This is the year in which the given individual is most similar to its
twin in terms of observable characteristics and propensity score.
The following comparisons over time will be relative to year 0 instead of calendar
time. E.g. for treatments, t=2 is two years after the year before treatment (i.e., one
year after the start of programme treatment). For controls, t=2 is two years after
being selected into the control group.
Individual-level analysis: the regression model
The individual-level analysis is carried out using separate multivariate regressions.
We consider the following success parameters:
(a) Whether or not the individual is employed in
t=1, t=2, ..., t=5,
implemented
by indicator (dummy) variables.
(b) The increase in wage income (salary) between year 0 and year
t=1,
t=2, ..., t=5.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
23
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0024.png
The success parameters are regressed on a treatment dummy (taking the value 1 for
treatments and 0 for controls) and the following conditioning variables: age, gender,
experience, average grade of final secondary education (high school) examination,
occupation in year 0, the sum of the Statistics Denmarks unemployment index
(measuring the aggregated time an individual has been registered as being
unemployed).
Individual-level analysis: descriptive statistics
394 individuals who have participated in the VP programme can be found in the
Statistics Denmark registers. Of these, 364 can be associated with controls equal or
similar in the dimensions described in the previous section. These 364 individuals
form the basis for the subsequent analysis. TABLE 4.1 describes the adjusted
individual-level sample (the total pool of available controls), the sample of VPs, and
the samples of VPs and controls used for the subsequent analysis.
15
The variable on whether or not a person is in education at a given point in time is from Statistics Denmark’s edu-
cation registers, while the variable of having education as one’s occupation is from Statistics Denmark’s education
occupation classifications (pstill). Individuals who work while studying are classified as under education in the
education registers and as working in the occupation information.
15
24
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0025.png
TABLE 4.1: Individual-level characteristics
Adjusted
sample
excluding VPs
(N = 1.018.245)
Variable
General information
Age (years)
Female
Experience (years, since 1980)
Average grade, secondary education
(high school)
Average wage (DKK)
37.176
0.401
10.483
84.304
300574
11.358
0.490
8.169
9.173
199968
Mean
Std. dev
Treatment
group (N=394)
Analysis
sample,
Treatments
(N=364)
Mean
Std. Dev.
Analysis
sample,
Controls
(N=364)
Mean
Std. Dev.
Mean
Std. Dev.
34.226
0.419
6.104
84.265
171721
9.380
0.494
6.669
8.304
207544
34.162
0.409
6.203
84.354
178437
9.426
0.492
6.811
8.279
212787
34.110
0.409
6.332
84.511
183930
9.627
0.492
6.685
8.833
197212
Years of registered unemployment
Married
In education
Post-secondary or tertiary education
Education: arts and humanities
1.149
0.487
0.137
0.588
0.142
2.005
0.500
0.344
0.492
0.349
1.241
0.411
0.398
0.807
0.183
1.979
0.493
0.490
0.395
0.387
1.170
0.429
0.401
0.805
0.181
1.898
0.496
0.491
0.397
0.386
1.163
0.412
0.393
0.805
0.181
2.076
0.493
0.489
0.397
0.386
Education: social sciences
Education: technical sciences
Secondary education, elective
direction: no information
Secondary education, elective
direction: general
Secondary education, elective
direction: math
Secondary education, elective
direction: languages
0.273
0.253
0.606
0.193
0.125
0.041
0.445
0.434
0.489
0.395
0.331
0.198
0.274
0.355
0.330
0.231
0.226
0.157
0.447
0.479
0.471
0.422
0.419
0.365
0.288
0.346
0.332
0.231
0.223
0.157
0.454
0.476
0.472
0.422
0.417
0.364
0.288
0.346
0.363
0.187
0.245
0.151
0.454
0.476
0.481
0.390
0.430
0.359
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
25
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0026.png
Region of residence
Copenhagen
Zealand N
Zealand S
Funen, Bornholm
Jutland S
Jutland W
Jutland E
Jutland N
Region not specified
0.312
0.112
0.076
0.079
0.065
0.095
0.177
0.080
0.005
0.463
0.315
0.264
0.269
0.247
0.294
0.381
0.271
0.071
0.223
0.056
0.079
0.157
0.046
0.063
0.231
0.142
0.003
0.417
0.230
0.270
0.365
0.209
0.244
0.422
0.350
0.050
0.217
0.049
0.082
0.162
0.049
0.063
0.234
0.140
0.003
0.413
0.217
0.275
0.369
0.217
0.244
0.424
0.348
0.052
0.247
0.055
0.077
0.135
0.052
0.058
0.277
0.099
0.432
0.228
0.267
0.342
0.223
0.233
0.448
0.299
Occupation (from Statistics Denmark's variable 'pstill')
Self-employed
Manager
Employee, high level
Employee, medium level
Employee, basis level
Employee, other
Employee, no further information
0.000
0.031
0.324
0.123
0.227
0.055
0.103
0.015
0.173
0.468
0.328
0.419
0.228
0.305
0.003
0.025
0.259
0.074
0.099
0.030
0.063
0.050
0.157
0.439
0.261
0.299
0.172
0.244
0.025
0.277
0.077
0.102
0.022
0.069
0.156
0.448
0.267
0.303
0.147
0.253
0.025
0.277
0.077
0.102
0.022
0.069
0.156
0.448
0.267
0.303
0.147
0.253
26
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0027.png
Unemployed
On parental leave
On sickness pay
Non-salaried worker
Education measure
In job market training
On social benefits ("revalidering")
Unknown
Outside labour force, other
In education
0.036
0.003
0.001
0.003
0.007
0.003
0.001
0.000
0.015
0.031
0.187
0.051
0.036
0.056
0.082
0.055
0.035
0.020
0.122
0.173
0.241
0.013
0.005
0.018
0.030
0.005
0.003
0.008
0.033
0.058
0.428
0.112
0.071
0.132
0.172
0.071
0.050
0.087
0.179
0.235
0.228
0.008
0.005
0.019
0.027
0.003
0.003
0.008
0.036
0.060
0.420
0.091
0.074
0.138
0.164
0.052
0.052
0.091
0.186
0.239
0.228
0.014
0.003
0.005
0.038
0.003
0.005
0.003
0.038
0.060
0.420
0.117
0.052
0.074
0.193
0.052
0.074
0.052
0.193
0.239
Year
2005
2006
2007
2008
2009
0.436
0.080
0.082
0.077
0.122
0.496
0.272
0.274
0.266
0.327
0.234
0.142
0.152
0.140
0.152
0.424
0.350
0.360
0.347
0.360
0.225
0.137
0.148
0.143
0.162
0.418
0.345
0.356
0.350
0.369
0.225
0.137
0.148
0.143
0.162
0.418
0.345
0.356
0.350
0.369
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
27
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0028.png
We find that individuals who participate in the scheme are represented among all
occupations, age groups and income levels. There is no gender bias in programme
participation. However, many VPs are relatively young, are unemployed or recent
higher education graduates, have sparse labour market experience and low income.
A more systematic way of describing the propensity of programme participation
is to estimate a binary choice model. The results of this model (specified as a logit
model) are shown in Table 4.2 which displays a selection of coefficient estimates.
Note that this model is also the backbone of the matching procedure used to identify
one matched control for each programme participant.
A look at the estimates of the individual-level logit model reveals that they by and
large corroborate the findings of the mean comparisons of Table 4.2: Individuals
participating in the programme are often relatively young, there are regional
differences, they are not characterised by high or low secondary education grades,
and they have high unemployment rates and low salary incomes. When controlling
for these characteristics, labour market experience (as long as it is positive) does
not come out as an important explanatory factor with regard to programme
participation.
The matching procedure finds controls for 364 of the total 394 participants in the
adjusted individual sample. The remaining 30 participants remain unmatched
because no other individual in the adjusted individual sample (the total pool
of available controls) could be found who was equal to these individuals in the
dimensions of education, gender, occupation and age.
The matched sample of treatments and controls can be compared by referring to
the right hand side columns of TABLE 4.1 and 4.2. We conclude that the matching
procedure succeeded in finding a group of controls highly similar to the group of
participants. This allows us to analyse treatment-control differences in the success
factors associated with programme participation in the following section.
28
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0029.png
Tabel 4.2: Individual-level analysis. Logit estimation results. Dependent
variable: Individual participates in the VP-programme in the following year.
Selected coefficients.
Adjusted sample
N=1,018,245, LR chi2(78)
=1129.19, Pseudo R2 =
0.1618
Coeff.
General information
Female
Married
In education
-0.122
-0.004
-0.050
0.114
0.111
0.181
Sample of treatments
and controls
N=728, LR chi2(76)=
25,57, R2=0.026
Coeff.
Ste.
Ste.
-0.065
0.078
0.009
0.192
0.173
0.310
Age (in years, omitted: <25 years)
(25-29)
(30-34)
(35-39)
(40-44)
(45-49)
(50+)
0.750***
0.717***
0.611**
0.441
0.758**
0.024
0.216
0.256
0.301
0.339
0.352
0.359
0.187
0.304
0.316
0.456
0.548
0.688
0.351
0.427
0.503
0.609
0.617
0.610
Region of residence (omitted: Copenhagen)
Zealand N
Zealand S
Funen, Bornholm
Jutland S
Jutland W
Jutland E
Jutland N
Region not specified
0.325
1.446***
1.436***
1.018***
0.784***
0.775***
1.269***
-0.079
0.243
0.216
0.171
0.265
0.233
0.151
0.176
1.007
0.142
0.242
0.372
0.176
0.298
-0.063
0.516
0.406
0.344
0.273
0.402
0.386
0.237
0.302
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
29
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0030.png
Secondary education final grade average (omitted group: unknown)
(0-75)
(76-85)
(86-90)
(90+)
0.064
0.242
-0.227
0.011
0.205
0.163
0.210
0.190
0.281
0.308
0.502
0.332
0.324
0.257
0.350
0.301
Occupation (from Statistics Denmark's 'pstill' variable, omitted: pstill-category 12 ('VAT-
payer'))
Self-employed (pstill=14)
Manager
Employee, high level
Employee, medium level
Employee, basis level
Employee, other
Employee, no further
information
Unemployed
On parental leave
On sickness pay
Non-salaried worker
Undergoing education
measure
In job market training
On social benefits
("revalidering")
Unknown (pstill=57)
Outside labour force, other
In education
1.976*
0.774
0.270
0.159
0.192
0.702
0.372
1.998***
1.219**
1.681**
2.551***
1.909***
2.808***
1.292
2.645***
0.876**
0.415
1.051
0.475
0.348
0.384
0.366
0.438
0.379
0.321
0.543
0.773
0.487
0.418
0.785
1.063
0.663
0.398
0.378
-0.047
-0.090
-0.063
-0.040
-0.058
-0.016
-0.160
-0.604
0.474
0.804
-0.540
-0.265
-0.949
0.842
-0.311
-0.095
0.794
0.557
0.602
0.597
0.771
0.621
0.541
0.923
1.367
0.981
0.678
1.582
1.410
1.298
0.633
0.606
30
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0031.png
Salary (omitted: no information)
0-0.15% of sample mean
15-25% of sample mean
25-50% of sample mean
50-75% of sample mean
75-100% of sample mean
100-125% of sample mean
125-150% of sample mean
150-200% of sample mean
200%+ of sample mean
0.249
0.472**
0.275
-0.828***
-0.784***
-1.193***
-1.379***
-1.527***
-1.741***
0.183
0.219
0.201
0.287
0.275
0.287
0.302
0.313
0.408
0.087
0.265
-0.260
-0.033
0.105
-0.244
-0.115
0.027
-0.398
0.298
0.352
0.317
0.450
0.446
0.430
0.444
0.480
0.604
Notes: *, **, *** denote statistical significance at the 10, 5, and 1% level. Additional variables included in the re-
gressions, but not presented in this table are: education (15 categories), experience (five categories), high school
average grade (five categories), unemployment experience index (variable ‘sumgrad’, six categories).
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
31
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0032.png
Individual-level analysis: Results
This section presents treatment-control differences in the outcome variables wage
income (Statistics Denmark variable
slon)
and employment (Statistics Denmark
variable
pstill
with a value of less than
40).
Results are based on descriptive graphs
and estimations with conditioning variables taking any remaining treatment-control
differences into account. All conditioning variables are from t=0, i.e. they are
collected in the year before treatment or, in the case of controls, the year of selection
into the control group.
When interpreting results, it might be kept in mind that the available data suggest
that long-term employment relationships for VPs in their hosting companies are not
very common. For example, 69 VPs were hired in 2005 with the VP-company match
confirmed by the Statistics Denmark data. Of these employment relationships, 53
(77 percent) were terminated within three years. For the employment relationships
started in 2005 and 2006, 71 percent were terminated within two years.
16
Potential employment effects
In the following, employment rates of VPs are compared with the employment
rates of individuals in the control group. Employment is measured by the Statistics
Denmark variable ‘pstill’ assuming a value of less than 40.
17
Note that this
variable is conditioned on when controls were selected into the control group. As a
consequence, employment rates are exactly equal for the two groups of individuals
in year 0 (t=0).
In this project’s vintage of the Statistics Denmark data, the individual-company-match can only be followed until
the year 2008, preventing us from following individual-company relationships over longer time periods or in more
recent VP-projects.
16
17
Individuals on leave are not counted as employed.
FIGURE 4.1: Share of treatments and controls in an employment relationship
(’pstill’
<
40). By year after year 0 (on horizontal axis)
100
80
60
40
20
0
-5 -4 -3 -2 -1
0
1
2
3
4
5
Treatments
Controls
32
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0033.png
FIGURE 4.2: Share of treatments and controls, by occupation. By year after year 0
(on horizontal axis)
60
50
40
30
20
10
0
-5 -4 -3 -2 -1
0
1
2
3
4
5
Top level (TREATMENT)
Top level (CONTROL)
Other employment (TREATMENT)
Other employment (CONTROL)
Unemployed (TREATMENT)
Unemployed (CONTROL)
Other non-employment (TREATMENT)
Other non-employment (CONTROL)
A first look at the data, see FIGURE 4.1, suggests that VPs are characterised by
decreasing employment rates in the years before treatment. But in association
with treatment, the employment rate increases to almost 100 percent. This is not
surprising, since employment is a defining characteristic of the programme. This
increase is not matched by the control group’s development in year t=1. However,
employment rates of the two groups converge over time and are at the same level
two years after treatment.
FIGURE 4.2 splits up developments in occupation status by
top level employment
(pstill<33), other employment (33<pstill<39), unemployment (pstill=40),
and
other
non-employment (pstill>40).
Here, it is found that treatments and controls
are characterised by highly similar developments in these variables in the years
before year 0, suggesting that the matching procedure has been successful. The
graph further suggests that in year 0, a number of individuals in the two groups are
finishing education or have left employment in the year prior to treatment or being
selected into the control group. After treatment, a large share of treatments are
categorised as top level employees, while controls pick up and have the same shares
of individuals in this category after approximately two to three years.
Employment probabilities are more formally analysed by means of simple binary
choice logit models, with ‘the individual is employed’ at t=x, x=1,2...5 being the
dependent variable, where t=0 is the year before treatment, t=1 is the year in which
treatment takes place, etc. Estimation is by separate binary choice models for each
t=x, x=1,2...5.
Table 4.3 displays the results. The coefficient of interest is the one associated with
the treatment dummy ‘Treatment=1”.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
33
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0034.png
We find a substantial potential programme effect on employment, as a coefficient
of 2.361 implies an increase in the odds ratio of being employed by factor
(exp(2.18)=) 10.5. This large increase comes as no surprise, given that employment
is a defining characteristic of the programme, and given that we already had seen
that employment is close to 100 percent for treatments in the year after the start of
programme participation.
18
Potential programme effects for the years following programme participation are
a (exp(0.271)=) factor 1.3 increase in employment probability in year t=2, which is
not significantly different from zero, and a factor 1.9 increase in year t=3, which just
fails to be significant at the ten percent level. After more than three years after the
start of participation, the signs of the coefficients switch around zero and become
insignificant.
For the most part, the remaining variables come out as insignificant. The exception
is low-wage individuals and individuals unemployed in year 0, who have the lowest
probability of being employed in subsequent years.
We conclude that overall results indicate a presence of potential short-run
employment effects and an absence of potential long-term effects of the programme.
However, it should be noted that most of the observation period is from a boom
period with high labour demand in the Danish economy. This implies that non-
participants cannot be assumed to catch up to the same extent in current years
compared to the analysis period.
The numbers of observations of the estimations are reduced by the fact that some of the explanatory variables
completely determine the outcome variables. As a robustness check, the models for employment and salary
developments were estimated without explanatory variables. This did not change the overall results in any signifi-
cant way.
18
34
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0035.png
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
35
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0036.png
TABLE 4.3: Logit binary choice model results. Dependent variable: The
individual is employed in t=x.
Dependent variable: The
individual is employed in
t=1
Coeff.
Treatment=1
Age (years)
Female
Annual wage (DKK
1000)
(Years of unemployment
up to t=0)*1000
Year of experience since
1980
Married
Secondary education,
no information
Secondary education,
elective direction: math
Secondary education,
elective direction:
languages
Secondary education: hf
("higher preparation")
Secondary education:
average grade
2.365***
-0.0669**
-0.360
0.004***
50.480
0.000
0.300
1.785
-0.580
-0.461
Dependent variable: The
individual is employed in
t=2
Coeff.
0.271
-0.011
-0.399
0.006***
-174.5**
-0.012
0.257
0.145
-0.390
-1.380**
Ste.
0.350
0.030
0.300
0.001
74.080
0.037
0.318
2.156
0.508
0.590
Ste.
0.293
0.029
0.308
0.002
78.900
0.040
0.349
2.069
0.605
0.586
0.309
0.028
0.706
0.026
-1.492**
0.017
0.682
0.025
36
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0037.png
Dependent variable: The
individual is employed in
t=3
Coeff.
0.577
-0.031
-0.179
0.002
-147.100
Dependent variable: The
individual is employed in
t=4
Coeff.
-0.423
-0.033
0.744
-0.001
-114.500
Dependent variable: The
individual is employed in
t=5
Coeff.
0.281
-0.104
0.961
0.007
-343.500
Ste.
0.354
0.036
0.392
0.002
101.000
Ste.
0.408
0.035
0.538
0.002
96.880
Ste.
0.716
0.120
1.078
0.004
317.200
0.017
-0.304
3.091
0.837
-0.977
0.045
0.418
2.413
0.753
0.667
0.035
0.384
5.429*
0.010
-1.491*
0.047
0.438
2.910
0.782
0.837
0.229*
0.696
7.224
2.769*
1.392
0.135
0.881
6.034
1.507
1.025
0.725
1.207
-0.451
1.012
-0.051
1.343
0.042
0.029
0.0759**
0.038
0.043
0.071
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
37
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0038.png
Occupation
Top level (pstill=31, 32, omitted category)
Employee, medium level
(pstill=34)
Employee, basic level
(pstill 35)
Employee, other
(pstill=36)
Salaried employee, no
further information
(pstill=37)
Unemployed
In education
Self-employed
On leave, and other non-
employed
-1.137*
0.601
-0.415
0.655
-0.581
0.683
-0.429
0.647
0.854
-0.676
1.000
0.627
-1.268*
-0.788
0.753
0.578
-1.480***
-1.295*
0.531
0.696
-0.382
-1.040
0.559
0.802
Immigrant status: not an immigrant (omitted category)
Immigrant status: first
generation
Year: 2005
Year: 2006
Year: 2007
Year: 2008
Region: North Jutland
Constant
Number of observations
-0.702
-0.071
-0.375
-0.655
-0.722
0.302
1.967
568
596
R-squared
0.28
0.639
0.397
0.489
0.477
0.491
0.476
2.369
-0.149
1.701
486
492
0.37
0.414
2.341
-0.099
0.396
0.766
-0.447
0.616
0.398
0.513
0.395
38
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0039.png
0.732
-0.466
1.268
0.662
-0.348
-1.219
0.888
0.835
-2.046*
1.102
-0.110
0.765
0.094
1.078
0.476
1.728
-0.380
-0.392
0.638
1.233
-0.805
-1.738
0.743
1.166
-1.621
1.104
-1.106
0.745
-0.892
0.915
0.618
1.723
-0.771
-0.065
-0.295
0.614
0.461
0.456
-0.101
-0.552
0.904
0.458
-0.393
1.149
-0.202
0.109
383
386
0.38
0.585
2.576
-1.178**
-1.334
286
293
0.48
0.515
2.900
0.936
-1.156
119
129
0.57
1.017
7.844
Notes: *, **, *** denote statistical significance at the 10%, 5%, 1% significance level. All monetary values are CPI-
adjusted to base year 2009.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
39
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0040.png
Potential earnings effects
FIGURE 4.3 looks at the average salary developments (measured by the Statistics
Denmark variable ‘slon’) of treatments and controls. We find that earnings profiles
are highly similar for treatments and controls before year 0, and that VPs on average
experience increasing salaries in association with programme participation. These
increases are higher for treatments than controls. However, after two to three years
after year 0, developments converge and individuals in the control group are doing
as well as participants.
19
A look at the dynamics of the salary distributions (instead of the means) in FIGURE
4.4 suggests that this increase is driven by VPs with low salaries in year 0. VPs in
the bottom 25th percentile of the salary distribution in year 0 experience the largest
salary increases in association with programme participation, which might be
presumed to be a result of these individuals entering an employment relationship in
association with the programme. On the other hand, there are fewer VPs with very
high salaries after year 0 than is the case for controls.
The estimations behind TABLE 4.3 are based on the total sample of treatments and controls except for individu-
als who experience extreme changes in their annual salaries (e.g. increases of more than DKK400,000 between
year 0 and year 1 or more than DKK1,000,000 between year 0 and t=5). See TABLE 4.6 for results on a sample
including these individuals.
19
FIGURE 4.3: Salary developments of treatments and controls, in DKK. Means. By
years after year 0 (on horizontal axis)
400000
350000
300000
250000
200000
150000
100000
50000
0
-5 -4 -3 -2 -1
0
1
2
3
4
5
Mean (TREATMENT)
Mean (CONTROL)
40
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0041.png
FIGURE 4.4: Salary developments of treatments and controls, in DKK. Distribution
parameters. By year after year 0 (on horizontal axis)
700000
600000
500000
400000
300000
200000
100000
0
-5 -4 -3 -2 -1
0
1
2
3
4
5
Median (TREATMENT)
Median (CONTROL)
25thpercent (TREATMENT)
75thpercent (TREATMENT)
25thpercent (CONTROL)
75thpercent (CONTROL)
10thpercent (TREATMENT)
90thpercent (TREATMENT)
10thpercent (CONTROL)
90thpercent (CONTROL)
The graphs suggest positive potential programme effects on salary in the years after
treatment and an absence of long-run effects. TABLE 4.4 considers these potential
effects in a more stringent way by means of a
conditional diff-in-diff
model. The
parameters of interest are again those associated with the variable ‘Treatment=1’ that
measures the potential programme effect on income for participating individuals.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
41
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0042.png
TABLE 4.4: Linear regression results. Dependent variable: Salary (’slon’)
increase between t=0 and t=x, in DKK.
Dependent variable:
salary increase between
t= 0 and t=1
Coeff.
Treatment=1
Age
Female
Annual wage (DKK 1000)
(Years of unemployment
before t=0)*1000
Years of experience since
1980
Married
Secondary education, no
information
Secondary education,
elective direction: math
Secondary education,
elective direction:
languages
Secondary education: hf
("higher preparation")
Secondary education:
average grade
Occupation
Top level management (pstill=31, omitted category)
Employee, high level
(pstill=32)
Employee, medium level
(pstill=34)
-28418
-20162
35165
36887
56456***
-2319**
-18141*
-0.258***
0.55
303
2766
35814
-9863
-17159
Dependent variable:
salary increase between
t=0 and t=2
Coeff.
21773*
-3131***
-16614
-0.516***
-10.52**
4966***
11503
-38180
-1468
-20279
Ste.
8534
942
10183
0.04
2.63
1308
9953
53968
14459
16908
Ste.
12193
1156
12261
0.05
4.19
1510
13036
70676
20176
21190
-33721*
500
19976
638
-43647
-329
28839
855
-88562**
-120011***
37002
41621
42
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0043.png
Dependent variable:
salary increase between
t=0 and t=3
Coeff.
25310*
-5543***
-33214**
-0.541***
-9.339*
5848***
489
-106950
6464
-14886
Dependent variable:
salary increase between
t=0 and t=4
Coeff.
4721
-6457***
-41379**
-0.734***
-18.16**
7111**
-34760
36331
14275
-31307
Dependent variable:
salary increase between
t=0 and t=5
Coeff.
42096
-8382**
-43810
-0.748***
-23.13*
6018
-32646
-224476
-10233
-73838
Ste.
14264
1516
15574
0.06
5.61
2092
15552
90991
22935
29109
Ste.
18928
2207
19401
0.10
7.54
3105
21773
129246
28555
33005
Ste.
34214
3882
35452
0.19
12.00
6024
32643
224830
59345
67547
-16760
-1126
27138
1080
-8739
365
24743
1447
-6107
-3405
95472
2768
-58624
-104792*
57540
59370
-17487
-88445
58814
67119
-46003
-334935***
51369
57211
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
43
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0044.png
Employee, basic level
(pstill=35)
Employee, other
(pstill=36)
Salaried employee, no
further information
(pstill=37)
Unemployed
In education
Self-employed
On leave, and other non-
employed
-34889
-53589
-59833
38840
40962
38815
-111962***
-179786***
-140168***
41296
49799
41893
-35792
-52257
-98106**
-25577
38724
38788
43751
40412
-91996**
-157389***
-216754***
-64130
43113
47033
60192
46789
Immigrant status: not an immigrant (omitted category)
Immigrant status: first
generation
Immigrant status: second
generation
Year: 2005
Year: 2006
Year: 2007
Year: 2008
Region: North Jutland
Constant
Number of observations
R-squared
-24764
18391
28159**
24741*
41843***
-15582
-4876
160822**
596
0.28
20393
24869
12322
14035
14572
16977
14430
66487
5762
432927***
492
0.37
15799
86265
13383
106801***
1196
16645
-8490
25690
20603
13384
16282
16329
44
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0045.png
-96960
-141922*
-98519
61138
73440
63464
-93042
-62744
-58705
65682
73003
70012
-135173*
-116249*
-82539
75067
61842
95957
-62057
-77928
-247578***
-15111
62101
66937
78454
66071
-76378
-82241
-264535***
28299
68055
73711
93911
71073
-94949
-92145
-75841
-61067
73276
78694
78306
87232
34296
39432
12203
15943
28970
72979
16679
19477
21333
-34498
-23334
42702
99428
19795
-16168
-141338*
57123
73137
-14781
576595***
386
0.38
22487
115808
-2496
552475***
293
0.48
28343
148468
-2139
963247***
129
0.57
43334
266452
Notes: *, **, *** denote statistical significance at the 10%, 5%, 1% significance level. All monetary values are CPI-
adjusted to base year 2009.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
45
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0046.png
46
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0047.png
Although potential employment effects are restricted to earlier employment for VPs,
we find that potential salary effects are slightly more persistent, as coefficients come
out statistically significant (albeit only at the ten percent level) for time leads of up to
three years. TABLE 4.4 also allows calculating the total potential programme effect
as the sum of the coefficient estimates. This is approximately DKK150,000 for the
total sample of all treatments and controls, a number which might be related to the
average cost of the programme.
Individual-level potential effects for different subsamples
As an extension of the previous
analysis, the sample of VPs and associated
controls is split up by a number of project-specific and VP-specific background
characteristics. In particular, the following distinguishes between whether or not
the VP-project was completed or terminated before schedule. The sample is also
split up by the industrial sector of the companies that hire the VPs or the associated
controls, and the education and gender of the VP and the associated controls.
Findings of the estimations on the subsamples are found in TABLE 4.5 for
employment and 4.6 for salary increases. These tables are based on the same models
that were estimated earlier, but only report the relevant coefficients associated with
the treatment dummy variables.
It is found that that there is little heterogeneity in the estimated potential effects of
the programme.
20
Only completed projects are associated with larger increases in
employment. This indicates that uncompleted projects are not just aborted because
of the VP moving to another employment relationship, but becoming unemployed.
This is also reflected in the absence of any measurable potential salary effect for this
group of individuals.
It is only possible to detect statistically significant potential employment effects
in the year after treatment (t=2) for VPs with a technical sciences education. It is
possible to detect positive potential salary effects in the years after treatment only
for female VPs, VP-projects in ‘other industries’, and completed projects.
Although single coefficient estimates are in most cases not statistically significantly
different from zero, the sum of the estimates of TABLE 4.6 are still the best
guesses of any potential salary effects over the first five years after treatment. These
potential effects are largest for female VPs and VPs who are employed in service
industries, and lowest for VPs with degrees in arts and humanities or technical
sciences, and VPs with a tertiary education.
For a couple of estimations, not all coefficients could be estimated because of low variation in the data relative
to the number of observations and the number of conditioning variables.
20
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
47
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0048.png
TABLE 4.5: Linear regression results. Dependent variable: The individual is employed in t=x. By
subsamples. Results for treatment dummy variables
Dependent variable: the individual
is employed in t=1
Coeff.
All projects
N
Only completed projects
N
Only not completed projects
N
Manufacturing and construction
N
Services
N
Other industries
N
Males
N
Females
N
Tertiary-level education
N
Education in arts & humanities
N
Education in social sciences
N
Education in technical sciences
N
2.371***
328
2.635***
212
2.264***
405
2.008**
116
3.960**
70
2.948***
183
0.717
1.935
0.870
0.395
0.551
0.498
2.365***
568
3.138***
449
1.435
99
0.942
0.471
Dependent variable: the individual
is employed in t=2
Coeff.
0.271
486
0.507
377
-0.358
87
1.08
128
0.874
88
3.079*
122
0.438
182
0.072
213
0.252
387
-0.086
98
-0.438
79
1.182*
160
0.627
0.809
0.786
0.342
0.439
0.468
1.864
1.044
0.81
0.810
0.346
Ste.
0.350
Ste.
0.293
48
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0049.png
Dependent variable: the individual
is employed in t=3
Coeff.
0.577
383
1.019**
309
0.416
Dependent variable: the individual
is employed in t=4
Coeff.
-0.423
286
-0.423
246
0.460
Dependent variable: the individual
is employed in t=5
Coeff.
0.281
119
-1.603
94
1.686
Ste.
0.354
Ste.
0.408
Ste.
0.716
1.45
115
6.551
77
0.796
115
0.160
223
1.406**
131
0.280
308
0.98
-0.34
94
0.65
4.068
6.029***
36
2.109
1.150
-2.387*
67
1.390
0.502
-0.619
177
0.525
0.393
68
0.926
0.717
0.343
89
0.732
0.424
-0.451
206
0.485
-0.528
95
0.927
1.108
44
-0.641
108
1.292
0.783
-1.234*
99
0.728
Notes: *, **, *** denote statistical significance at the 10%, 5%, 1% significance level.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
49
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0050.png
TABLE 4.6: Linear regression results. Dependent variable: Salary (’slon’) increase between t=0 and
t=x, in DKK. By subsamples. Results for treatment dummy variables
Dependent variable: salary
increase between t=0 and
t=1
Coeff.
All projects
N
All projects, including outliers
N
Only completed projects
N
Only not completed projects
N
Manufacturing and construction
N
Services
N
Other industries
N
Males
N
Females
N
Tertiary-level education
N
Education in arts & humanities
N
Education in social sciences
N
Education in technical sciences
N
56456***
596
48137***
605
69542***
467
8866
129
23180
136
67198***
170
79337***
156
49065***
349
63790***
247
55809***
468
43389*
116
44390**
161
59315***
205
15017
18527
22664
9855
13465
11794
16158
17814
18829
19561
9659
9685
Dependent variable: salary
increase between t=0 and
=2
Coeff.
21773*
492
11877
501
33476**
381
-8478
111
14164
125
47811
90
69038***
149
11342
277
34185*
215
16325
391
13979
98
10731
137
5266
168
21677
30264
30402
14205
18891
17030
22550
30202
30675
29591
13573
13440
Ste.
8534
Ste.
12193
50
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0051.png
Dependent variable: salary
increase between t=0 and
t=3
Coeff.
25310*
386
24982
393
29125*
311
22334
75
44672*
126
-10601
89
26789
149
19414
232
36790
154
14992
314
-15664
72
3897
108
12788
144
26287
30019
43284
16568
22989
19759
25844
38463
26317
33461
16547
15445
Dependent variable: salary
increase between t=0 and
t=4
Coeff.
4721
293
4721
293
9146
248
24120
45
11914
97
48873
64
-7511
118
-6218
181
29105
112
-7062
233
-53186
41
4058
87
-18634
115
34162
42777
43560
22172
28294
26843
32362
39917
41021
56060
-21466
18928
Dependent variable: salary
increase between t=0 and
t=5
Coeff.
42096
129
42096
129
25853
106
123197
23
19104
40
123624
26
13542
56
60274
79
66325
50
20373
107
-30421
27
61546
40
51668
45
87173
67346
169878
39581
44998
53426
47681
135240
90804
103104
39825
34214
Aggregated dif-
ferences from
t=1 to t=5
Ste.
14264
Ste.
18928
Ste.
34214
150356
131813
167142
170039
113034
276905
181195
133877
230195
100437
-41903
124622
110403
Notes: *, **, *** denote statistical significance at the 10%, 5%, 1% significance level. All monetary values are CPI-adjusted to base year 2009.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
51
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0052.png
5.
COMPANY-LEVEL ANALYSIS
In the following, the setup and results of the company-level analysis are described.
We briefly describe the model which aims at removing as much unobserved
heterogeneity as possible from the statistical comparisons. We then take a look at
the company-level data and inspect the sample for the subsequent analysis. Finally,
we compare companies that participate (receive a
treatment)
in the programme
(‘treatments’ or ‘participants’ in the following) with highly similar companies that
act as a control group. In particular, we compare developments in:
1.
2.
3.
4.
5.
6.
the number of highly educated employees
the number of employees
value added
net income (profit) and return on assets
average wage cost
labour productivity, measured as turnover per employee
The analysis addresses the question of how VP-companies perform in terms of these
variables. This is answered by looking at the developments in these variables over
time and comparing them to developments in a control group comprised of other,
similar companies that do not participate in the VP programme.
It should be noted that the analysis of the number of (highly educated) employees,
value added and net income gives highest weight to companies experiencing the
largest changes in these variables. These are typically larger companies. For average
wage cost, return on assets and labour productivity, companies are treated equally
and, thus, higher weight is given to smaller companies.
Empirical specification
Company-level analysis: selection of controls
For the company-level analysis, the selection of controls is carried out in two steps.
First, select a pool of potential controls in the Experian data. Second, apply a
matching procedure.
Before applying the matching procedure, we go through the Experian data and
exclude observations of companies in industries without participant companies,
with ownership classifications where there are no participant companies, companies
larger than 150 employees, and companies for which a set of additional conditions is
not fulfilled.
21
The remaining sample is denoted the ‘adjusted Experian sample’.
These conditions are: equity being between DKK-20mio and 150mio., net income between DKK-20mio and
20mio., total assets between zero and DKK250 mio., short term debt between DKK15,000 and 70mio., an equity
share between -2.5 and 0.9, return on assets between -1.2 and 1, the number of employees with at least a post-
secondary education less than or equal to 25, the number of employees with a tertiary education less than or
equal to 5, and firm age less than 150 years. Imposing these conditions does not affect the number of participants
in the sample.
21
52
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0053.png
As a last step before the matching procedure, we exclude all observations
of participants that do not belong to the last financial report before starting
participation in a VP-project. We then estimate a binary choice model on the
adjusted Experian sample which is used to predict a participation probability (a
propensity score) for any given company for any given year in the reduced Experian
population.
The population is then grouped by year and industry. Within each group, a matched
twin is found for each participant company on the basis of the propensity score. This
procedure ensures equality between the participants and controls in terms of the
highly detailed industrial sector classification ‘Dansk Branchekode’ and timing.
22
This procedure implies that we identify 316 control firms for 316 participants. These
define the
analysis sample
of the study. The year in which a control company is
selected is this company’s year 0 (base year, t=0), which is the cut-off year for later
before-after comparisons. For VP-companies, year 0 is simply the last year before
participating in the programme.
23
Company-level analysis: the empirical model
We chose a model with fully specified dynamics, which is highly similar to Kaiser
and Kuhn, 2012.
24
This model is formulated as follows:
y_( i ,t)-y_(
=
i
x
+
(α
n
D (t
i
= n) + β
n
(D (treat
i
=1)
× D (t
i
= n)) ) + u
i
+ ε
i
y
i,t
–
y
i,t–1
,t-1)=x_t+∑_(n=1)^5▒▒(▒α_n D(t_( i )=n)+β_n (D(▒treat▒_(
i,t
t
)=1)×D(t_( i )=n))▒_ )+u_i+▒ε_(i,t)▒_ ▒
n=1
where
y
i,t
is the dependent variable,
i
is firm index,
t
is a time index, where
t=0
is
year 0, and
x
t
are year dummies to account for business cycle effects. The
D
are
dummy variables assuming the value of 1 if the logical conditions in their brackets
are fulfilled. This model is estimated subject to company-level fixed effects
u
i
and
has statistical errors
ε
i,t
.
The α and β are estimation coefficients, where the β measures the potential
treatment effects. Note that this model extends Kaiser and Kuhn’s analysis by
estimating post-year zero effects not just for participants but controls as well. These
are measured by the coefficient vector α, while the vector β collects the conditional
difference-in-difference estimators.
25
∑
5
The observation period is characterised by considerable business cycle movements, which implies the need to
match controls as exactly as possible with regard to the time when they are selected.
22
To be specific, the base year of participants is defined by the closing date of the last financial report before the
start of participation. This means the base year of participants is not necessarily the calender year before starting
to participate in the programme.
23
Kaiser, U., Kuhn, J.M., Long-run effects of public–private research joint ventures: The case of the Danish In-
novation Consortia support scheme. Res. Policy (2012).
24
25
Another minor extension is the clustering of statistical errors
ε
i,t
within treatment-control twin pairs.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
53
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0054.png
The fixed effects setup implies that all time-invariant factors drop out of the model,
thus making the model robust to any omitted time-constant factors which might
be correlated with the decision to participate in the programme. The set of dummy
variables generates a difference-in-difference model setup, and the coefficients
of the dummy variables in the vector β estimate the potential programme effect.
Separate dummy variables for each year after the base year allow estimating the
dynamics of the potential programme effect.
26
Company-level analysis: descriptive statistics
Out of the 434 companies that have hosted VP-projects in the DASTI data, 370
can be found in the Experian data. The remaining 64 firms that cannot be found
in these data are probably non-incorporated firms that are not obliged to publish
their financial reports by submitting them to the Danish Business Authority. Of the
firms found in the Experian data, 338 filed a report in the year prior to programme
participation. Only these firms will be considered in the subsequent analysis
comparing performance both before and after the start of participation.
When setting the sampling criteria for this analysis, we need to decide how to
treat outliers (extreme observations). This decision trades off robustness of later
results with their representativeness. In the following, we choose to describe results
for ‘typical’ VP-companies and to not consider companies in the financial sector
(reducing the sample by eleven companies) nor companies with ownership codes
that only occur very rarely in the sample of VP-companies (reducing the sample by
five companies).
27
After deleting financial sector companies and companies with atypical ownership
codes, we are left with 319 observations. Of these, 318 have started their project
before 2011 and can be followed for at least one year in the Experian data.
The controls for the latter analysis are found in the adjusted Experian sample.
In these data, there are 296,000 company-level observations in the period from
2004 onwards that are roughly similar to the participants in a few dimensions, e.g.
industrial sector and number of employees. For 316 of the 318 VP-companies, the
matching procedure succeeds in finding controls for the analysis.
Means and standard deviations of a set of characteristics of these companies
are described in the first columns of TABLE 5.1. This table also shows the
characteristics of the adjusted Experian sample – which was selected in order
to roughly resemble the group of participants, and used for the estimation of
propensity scores for the matching procedure. TABLE 5.1 allows comparing the 316
programme participants with the two Experian samples and the control group of
companies selected by the matching procedure.
Also note that taking first-differences in the outcome variables addresses any potential problems of serially
correlated unobserved characteristics.
26
For example, we drop co-operations (two occurrences), funds (one occurrence), companies with limited liability
(one occurrence), and one company with an unidentified ownership code.
27
54
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0055.png
A look at the raw figures shows that VP-companies are distributed over most
industries, with relatively large shares in trade (21 percent), consultancies (12
percent) and IT services (9 percent). These shares follow the industry distribution of
the total sample of companies in the Experian database. However, VP-companies
are underrepresented in construction and overrepresented in manufacturing, metal,
construction, advertising and cleaning.
At first sight, the VP-companies look healthy: On average, they are slightly larger
(mean 15 employees) than the average company in the Experian database (mean 11
employees) and have survived longer (15 years vs. 10 years). Many (42 percent) are
registered as exporters in the Experian database, and almost 50 to 100 percent are
owned by other companies, e.g. holding companies (compared to 34 percent for all
companies in the Experian data). Also, 11 percent own other companies (compared
to 5-6 percent of all companies).
When it comes to employee characteristics, it is found that VP companies have a
relatively large share of employees with at least a secondary education and also an
above-average share of employees with a post-secondary or tertiary-level education.
They have a relatively low share of technically trained employees.
The fact that VP-companies are not fully representative companies implies that, if
one aims at comparing these companies with other companies, one must carefully
construct a control group of similar companies for the comparison.
A first step in this process is the estimation of a binary choice model to estimate
propensity scores. This model is based on the 239,000 company observations in the
adjusted Experian sample and the 318 participants in the year before treatment.
The results of the binary choice model (formulated as a logit model) are displayed
in the left hand side columns of TABLE 5.2. Findings largely agree with what was
seen in the mean comparisons: Companies are most likely to participate if they
are not in the construction industry, are incorporated as joint stock companies,
are relatively large, have high returns on assets and a relatively low equity share, a
low average employee age, a high share of highly educated employees, and a low
share of employees with primary school as their highest level of education. The VP
programme is relatively popular in rural districts, with high propensity on the island
of Funen and both Southern and Northern Jutland.
The results of the logit model allow us to calculate predicted participation
probabilities (propensity scores). These are used to select a control group of
companies for the subsequent treatment-control analysis.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
55
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0056.png
TABLE 5.1: Means and standard deviations of key characteristics of company-level samples
Summary
of all firms,
N =296,087
Mean
Industry
Construction
Trade
IT, services
Manufacturing
Metal industries
Furniture and related
industries
Travel agencies, cleaning
services
Advertisement
Consulting, business
services
Paper&publishing
Other
Key figures
Number of employees
No number of employees
information
Number of employees=0
Number of highly educated
employees
1
Value added (DKK1,000)
No value added
information
Net income (profit,
DKK1,000)
Return on assets
11.21
0.28
0.13
0.19
4713
0.12
676
-0.41
64.13
0.45
0.34
0.31
39920
0.32
25560
42.66
0.13
0.18
0.07
0.01
0.02
0.02
0.02
0.03
0.13
0.01
0.38
0.34
0.39
0.26
0.10
0.14
0.12
0.13
0.16
0.34
0.11
0.48
Summary of
adjusted sample,
N = 238.375
Mean
Std. Dev
Summary of
treatments in
analysis sample,
N = 316
Mean
Std. Dev
Summary of
controls in
analysis sample,
N = 316
Mean
Std. Dev
Std. Dev
0.15
0.19
0.07
0.01
0.02
0.02
0.02
0.03
0.13
0.01
0.35
0.36
0.39
0.25
0.11
0.15
0.13
0.13
0.16
0.34
0.11
0.48
0.06
0.21
0.09
0.06
0.05
0.06
0.03
0.06
0.12
0.03
0.22
0.23
0.41
0.29
0.24
0.22
0.23
0.18
0.24
0.32
0.18
0.42
0.05
0.21
0.09
0.06
0.03
0.08
0.04
0.07
0.12
0.02
0.23
0.22
0.41
0.28
0.24
0.18
0.27
0.20
0.25
0.33
0.15
0.42
7.02
0.23
0.11
0.17
2903
0.08
302
0.02
12.80
0.42
0.31
0.30
5941
0.27
1654
0.23
14.75
0.03
0.01
0.22
6483
0.01
457
0.03
18.39
0.18
0.11
0.27
8425
0.11
2165
0.21
13.96
0.02
0.02
0.22
6279
0.03
567
0.04
17.46
0.15
0.14
0.29
8304
0.16
2070
0.22
Notes: 1: “highly educated” refers to post-secondary education and tertiary-level education.
56
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0057.png
Wage cost per employee
(DKK1,000)
No wage cost per employee
info.
Labour productivity
(DKK1,000)
No labour prod. Info.
Total assets (DKK1 mio.)
Equity share
Short term debt (DKK1,000)
410
0.43
3096
0.44
17.07
-1.23
7008
1540
0.49
97103
0.50
219.76
99.71
86627
400
0.37
2623
0.37
7.79
0.28
3428
660
0.48
65175
0.48
16.29
0.38
6928
395
0.09
2056
0.08
13.06
0.22
6532
217
0.29
5479
0.27
20.31
0.35
9240
377
0.09
1867
0.09
13.05
0.23
6579
163
0.28
2627
0.29
21.51
0.34
9931
Development in selected key figures (average annual increase in t=-3 to t=0)
Number of employees
Number of highly educated
employees
Value added (DKK1,000)
Net income (DKK1,000)
Wage cost per employee
(DKK1,000)
Labour productivity
(DKK1,000)
Year
2005
2006
2007
2008
2009
0.11
0.16
0.18
0.21
0.23
0.31
0.36
0.39
0.41
0.42
0.11
0.16
0.18
0.20
0.22
0.32
0.37
0.39
0.40
0.41
0.24
0.15
0.15
0.15
0.16
0.43
0.35
0.36
0.36
0.37
0.24
0.15
0.15
0.15
0.16
0.43
0.35
0.36
0.36
0.37
0.34
0.12
269
33.9
-4.2
94.0
9.25
2.64
7602
9435.8
1567.3
40919.0
0.24
0.04
154
2.1
-4.2
74.4
2.09
0.54
1233
860.6
1529.4
22814.4
0.88
0.19
448
-1.4
2.6
-114.0
3.01
0.91
1876
1412.6
161.3
2694.0
0.85
0.11
506
89.0
-17.3
-721.3
3.12
0.82
1870
995.0
239.7
11055.9
Company age and ownership information
Ownership code: joint stock
Company age
Company has mother
company
Company is mother company
Company is exporter
0.27
10.45
0.34
0.06
0.12
0.44
21.80
0.47
0.24
0.32
0.44
21.80
0.47
0.24
0.32
0.44
13.45
0.48
0.23
0.32
0.52
15.10
0.48
0.11
0.42
0.50
19.87
0.50
0.31
0.49
0.53
13.94
0.49
0.09
0.39
0.50
16.32
0.50
0.29
0.49
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
57
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0058.png
Region
Zealand N, Copenhagen
Zealand S
Funen, Bornholm
Jutland S
Jutland W
Jutland E
Jutland N
Region not specified,
overseas departments
Employee characteristics
Company: mean employee
age (years)
Company: share of
employees that is female
Company: share with a
secondary education
Company: share with a
post-secondary education
Company: share with a
tertiary education
Company: share social
sciences
Company: share arts &
humanities
Company: share technical
sciences
40.1
0.26
0.26
0.19
0.08
0.26
0.03
0.35
9.6
0.29
0.34
0.31
0.21
0.32
0.12
0.35
40.0
0.25
0.24
0.17
0.07
0.26
0.03
0.35
9.5
0.29
0.33
0.30
0.20
0.32
0.12
0.35
0.29
0.05
0.30
0.25
0.14
0.28
0.29
0.05
0.32
0.29
0.14
0.33
37.5
0.30
0.31
0.22
6.6
0.26
0.28
0.27
37.6
0.27
0.30
0.22
7.1
0.26
0.32
0.29
0.24
0.09
0.11
0.07
0.09
0.09
0.16
0.08
0.43
0.28
0.31
0.26
0.29
0.28
0.37
0.27
0.23
0.09
0.12
0.07
0.10
0.09
0.16
0.08
0.42
0.28
0.32
0.26
0.30
0.29
0.37
0.27
0.14
0.04
0.15
0.11
0.11
0.10
0.18
0.11
0.35
0.19
0.35
0.32
0.32
0.30
0.38
0.31
0.19
0.04
0.16
0.07
0.11
0.07
0.18
0.10
0.39
0.19
0.37
0.25
0.32
0.26
0.38
0.31
58
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0059.png
TABLE 5.2: Company-level analysis. Logit estimation results. Dependent
variable: The company participates in the VP-programme in the following
year
Adjusted sample
N = 238,693
Mean
Industry
Construction
Trade
IT, services
Manufacturing
Metal industries
Furniture and related industries
Travel agencies, cleaning
services
Advertisement
Consulting, business services
Paper&publishing
Other (omitted category)
-0.85***
-0.31*
-0.16
0.90***
0.16
0.55*
0.48
0.28
-0.19
0.19
0.28
0.19
0.25
0.28
0.30
0.28
0.34
0.28
0.24
0.36
Treatments and con-
trols sample
N = 632
Mean
Std. Dev
Std. Dev
0.46
0.02
0.28
0.23
0.66
-0.27
-0.31
-0.08
0.13
0.34
0.46
0.28
0.40
0.41
0.47
0.40
0.51
0.42
0.36
0.58
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
59
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0060.png
Key figures
Number of employees
Number of employees^2
No employees information
Number of employees=0
Value added (DKK 1 mio)
No value added information
Net income (DKK 1 mio)
Return on assets
Wage cost per employee
(DKK1,000)
No wage cost per employee info.
Labour productivity (DKK1,000)
No labour prod. info.
Total assets (DKK 1 mio)
Total assets (DKK1,000)^2
Equity share
Short term debt (DKK1,000)
0.04***
0.00***
-0.79
-1.18*
-0.01
-0.70
-0.04
0.64**
0.00
0.74
0.00
-0.41
0.01
0.00
-0.57***
0.00
0.01
0.00
0.74
0.69
0.02
0.63
0.04
0.32
0.00
0.48
0.00
0.50
0.01
0.00
0.17
0.00
0.00
0.00
1.37
0.40
0.02
-0.81
-0.04
0.32
0.00
0.02
0.00
-1.16
0.01
0.00
-0.35
0.00
0.02
0.00
1.22
1.04
0.03
0.95
0.07
0.53
0.00
0.93
0.00
0.78
0.02
0.00
0.33
0.00
Development in selected key figures (average annual increase in t=-3 to t=0)
Number of employees
Number of employees, missing
obs.
Number of highly educated
employees
Number of highly educated
employees, missing obs.
Value added (DKK 1 mio)
Value added, missing obs.
Net income (DKK 1 mio)
Wage cost per employee
(DKK1,000)
Wage cost per employee,
missing obs.
0.03
0.15
0.09
0.03
0.61
0.08
0.03
0.14
0.12
0.05
1.04
0.12
0.04
0.03
0.17
-0.03
0.00
0.74
0.85
0.05
0.42
0.08
0.00
0.48
0.47
-0.06
-0.02
0.00
0.00
0.02
1.26
0.09
0.76
0.11
0.00
0.93
60
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0061.png
Labour productivity (DKK 1 mio)
Labour productivity, missing
obs.
0.00
1.49***
0.01
0.45
0.01
0.56
0.02
0.71
Year
2005
2006
2007
2008
2009
0.53***
-0.27
-0.33
-0.34*
-0.23
0.19
0.21
0.21
0.21
0.20
-0.06
-0.19
-0.10
-0.22
-0.25
0.29
0.33
0.32
0.34
0.32
Company age and ownership information
Ownership code: joint stock
Company age
Company age^2
Company has mother company
Company is mother company
Company is exporter
0.30**
0.00
0.00
0.06
0.18
0.99***
0.14
0.01
0.00
0.12
0.19
0.14
-0.13
-0.01
0.00
-0.07
0.23
0.12
0.21
0.01
0.00
0.19
0.31
0.20
Region (omitted category: Copenhagen)
Zealand N
Zealand S
Funen, Bornholm
Jutland S
Jutland W
Jutland E
Jutland N
Region not specified, overseas
departments
-0.24
-0.44
0.79***
0.72**
0.41
0.23
0.34
0.67**
0.27
0.38
0.28
0.29
0.29
0.30
0.27
0.29
-0.42
-0.05
-0.06
0.64
0.14
0.44
0.06
0.26
0.40
0.56
0.43
0.46
0.44
0.46
0.40
0.45
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
61
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0062.png
Employee characteristics
Company: mean employee age
(years)
Company: share of employees
that is female
Company: share with a
secondary education
Company: share with a
post-secondary education
Company: share with a tertiary
education
Company: share social sciences
Company: share technical
sciences
-0.04***
0.00
0.16
-0.53
-0.67
-0.08
-0.73**
0.01
0.24
0.38
0.41
0.41
0.31
0.35
0.00
0.25
0.03
-0.21
0.49
0.04
-0.45
0.01
0.42
0.64
0.69
0.70
0.55
0.60
Before turning to the analysis, we need to establish an idea of just ‘how similar’ the
groups of matched treatments and controls really are. Accordingly, we will compare
the two groups of companies as follows:
First, we run a very simple test of the similarity of observable characteristics of the
two groups of companies and estimate the same logit model as earlier, but this time
on the matched treatment-control sample. The results of this exercise are displayed
in the right hand side columns of TABLE 5.2. We find that all coefficients have
decreased in absolute size and come out as insignificant, indicating an absence of
considerable differences in these variables across the two groups of companies.
Second, we look at the similarity of the two groups of companies in the matched
treatments-controls sample by simply comparing the means of observable
characteristics of the two groups, displayed in the two right hand side columns of
TABLE 5.1.
Inspection of TABLE 5.1 suggests that the matching procedure succeeded in
finding matched twin companies that highly resemble the group of treatments in
the year before treatment. Differences between the groups are typically one order of
magnitude smaller than the corresponding standard deviations, implying that none
of the differences are statistically different from zero.
So: If the VP programme significantly increases the performance variables of the
analysis, we should be able to see this by higher growth in the performance variables
after treatment than before treatment, and a greater growth increase around year 0
for treatments than for controls. This will be tested in the next section.
62
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0063.png
Company-level analysis: Results
In the following, developments in a number of performance variables for companies
that have participated in the VP programme are compared with the group of controls
selected by the matching procedure. These variables are: the number of highly
educated employees (i.e. employees with an education at a post-secondary or tertiary
level), the number of employees, value added, profits, return on assets, wage costs
per employee, and labour productivity.
TABLE 5.4 displays the results of the
conditional diff-in-diff
model with company
fixed effects.
The coefficients
‘TREAT=1 & t=1’, ‘TREAT=1 & t=2’,..., ‘TREAT=1 &
t=5’
correspond to the potential treatment effect estimates
β
n
while the coefficients
of ‘t+1’, ‘t+2’, etc. correspond to the
α
n
of the conditional diff-in-diff model
described in the previous section. The results are based on the approximately 300
programme participants and the same number of associated control companies. But
only companies that participated early in the programme can be observed after the
very first years after treatment, so results for more than a few years after year 0 are
based on a substantially reduced number of observations.
Before we look at the specific findings, it is necessary to consider how to treat
outliers. We have to do with company level data which by its very nature is highly
heterogenous, and the treatment of outliers is important to later results.
28
TABLE 5.4 is based on VP-companies and companies in the control group with
at most 50 employees that do not experience large year-to-year changes in their
numbers of employees, as well as regression-specific conditions imposed to further
reduce unobserved heterogeneity. Obviously, the results of the analysis depend on
these sampling conditions, and when interpreting later results one must be aware
that the results are only valid for companies that fulfil the conditions. In subsequent
robustness checks, these conditions are relaxed.
The results of TABLE 5.4 are summarized in the following sections.
Although there is a lot of background information in the data, we are unable to offer explanations (and, thus,
cannot control for) for a large amount of heterogeneity in the data. Clearly, we do not want to base overall results
of the analysis on single observations with extreme values - especially when it cannot be ruled out that these
values are statistical noise (e.g. due to company mergers or organisational restructuring).
28
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
63
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0064.png
TABLE 5.4: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results
Dependent variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.458***
0.318**
0.01
-0.14
-0.22
-0.03
-0.05
-0.14
-0.09
0.02
Number of employees
Value added
(DKK1,000)
2
Coeff.
219.3
374.1
165.2
124.0
-563.1
-20.5
-268.9
-10.5
243.0
468.9
Ste.
0.12
0.14
0.17
0.21
0.26
0.10
0.13
0.17
0.20
0.25
Coeff.
0.596**
0.00
0.33
-0.45
-0.69
0.00
-0.15
-0.33
-0.11
0.86
Ste.
0.30
0.34
0.40
0.60
0.65
0.24
0.32
0.41
0.56
0.59
Ste.
217.2
239.6
324.4
448.2
580.3
194.5
233.5
324.3
393.0
528.6
0.01
0.06
0.01
0.01
0.01
-0.13
-0.14
0.12
0.11
0.10
0.12
0.15
0.17
0.20
0.00
0.36
0.36
0.43
0.37
0.24
-1.477***
0.24
0.28
0.29
0.34
0.36
0.43
0.50
-298.1
308.3
156.1
348.7
277.8
-285.7
-810.8**
217.1
221.5
210.2
243.1
280.0
309.8
370.5
Constant
Number of observations:
Number of companies:
R-squared
0.12
2609
535
0.03
0.09
0.34
2727
546
0.08
0.25
240.8
2611
533
0.04
187.6
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
64
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0065.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
8.28
5.25
-21.34
16.06
-19.32
-13.08
1.67
10.89
-0.61
2.41
Labour productivity
(DKK1,000)
6
Coeff.
-27.92
57.28
-137.30
-39.23
-159.70
-36.00
88.66
108.40
24.21
97.65
Coeff.
-48.5
136.5
133.3
205.5
-103.5
24.2
-125.6
38.2
14.1
190.1
Ste.
95.2
111.4
122.1
218.1
189.2
87.2
104.2
129.0
199.6
220.3
Coeff.
-0.03
-0.04
0.00
-0.04
-0.04
-0.01
-0.03
-0.03
-0.02
-0.07
Ste.
0.02
0.03
0.03
0.04
0.06
0.02
0.03
0.03
0.04
0.05
Ste.
11.26
10.27
13.79
17.18
29.92
11.55
11.32
15.10
20.35
27.05
Ste.
93.59
107.90
91.84
134.00
215.30
87.17
91.22
97.51
117.90
203.20
-96.4
44.5
65.2
2.2
24.5
-191.8
-362.4**
90.3
86.6
84.0
95.6
113.5
130.1
159.9
-0.0426*
0.01
0.01
0.01
0.02
-0.02
-0.01
0.02
0.02
0.02
0.03
0.03
0.03
0.04
16.90***
8.99
8.00
10.54
13.41
4.35
8.81
6.23
7.21
7.97
8.97
13.39
15.66
18.74
-141.0*
-34.30
-28.89
48.53
-65.07
-180.4*
-90.63
72.66
58.79
74.78
90.79
96.66
108.80
122.50
78.7
2553
542
0.03
70.7
0.01
2669
544
0.02
0.02
-1.77
1494
346
0.01
5.88
60.65
1693
323
0.02
59.57
2. Only observations with annual change in the value added by less than DKK 10 mio.
3. Only observations with annual change in net income by less than DKK 3 mio.
4. Only observations with annual change in return on assets by less than 1, and total assets
>
DKK100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
65
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0066.png
Potential employment effects
A first question addressed in the empirical analysis is whether companies
participating in the programme do indeed increase the number of highly educated
employees (employees with an education level categorised as at least ‘post-
secondary-non-tertiary and tertiary’, ISCED 4-8) relative to companies in the
control group.
TABLE 5.5.a: Potential effects on the number of highly educated employees. Further results
Ordinary least squares
regression
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
0.445***
0.286**
-0.118
-0.241
-0.269
Firm fixed-effects model
Coeff.
0.436***
0.413**
0.097
0.092
0.214
Conditional diff-in-diff
model
Coeff.
0.458***
0.318**
0.005
-0.143
-0.221
Ste.
0.109
0.108
0.113
0.124
0.201
Ste.
0.146
0.176
0.233
0.300
0.394
Ste.
0.115
0.143
0.169
0.205
0.257
Includes firm-fixed effects
Includes year dummy
variables
Includes information from
before year zero
Includes observations of
the control group
no
no
no
yes
yes
yes
yes
yes
yes
no
no
yes
Number of observations:
Number of companies:
R2:
631
274
0.05
1354
274
0.02
2609
535
0.03
Notes: Highly educated employees are employees with a post-secondary or tertiary-level education. Only observations with annual changes in the
number of employees with a post-secondary and tertiary education
<
5. Only observations with annual changes in the number of employees of less
than 12.
*, **, *** denote statistical significance at the 10%, 5%, and 1% level.
The coefficients of a simple ordinary least squares regression, which are equivalent
to the population means and found in the leftmost columns of TABLE 5.5.a, imply
that participating companies increase their number of highly educated employees by
(0.445+0.286=) 0.7 employees in the first two years after start of participation.
66
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0067.png
The results of a company fixed effects model, which implements a before-after
comparison for programme participants, are presented in the middle columns of
TABLE 5.5.a. The similarity of this model’s results and the results of the simple
ordinary least squares regression implies that the earlier finding of an increase in the
number of highly educated employees in association with programme participation
(the results of the full-fledged model of TABLE 5.4 are replicated on the right of
TABLE 5.5.a) is not to be interpreted as a continuation of any before-participation
growth trend.
This allows the conclusion that the finding of positive potential programme effects
with regard to highly educated employees is not just the result of the developments
in (or the choice of) the group of control companies in the fully specified model
behind TABLE 5.4. This observation, and non-positive coefficient estimates of
the α
n
-coefficient associated with
‘t+1’, ‘t+2’
indicate an absence of behavioural
additivity: Companies in the control group do not experience increases in the
number of highly educated employees in the years after the selection into the control
group.
Aggregated coefficients of the fully specified model are shown graphically in
Figure 5.1.
29
Findings suggest that a participating company increases the number
of highly educated employees by 0.46 additional individuals in the year of the
treatment. The reason this number is not equal to 1.0 is that some of the projects
(and associated employment relationships) last less than one year and have already
been terminated before the census date of year 1. Also, as noted earlier, in some
cases the information on highly educated employees is registered with time lags, if
the data is from different sources (for instance, VP projects starting between the end
of November and the closing date of the company’s financial report). In these cases,
potential effects occur between
t=0
and
t=2
instead of between
t=0
and
t=1.
30
Figure 5.1 (just like the figures to follow in the next subsections) presents aggregated estimated treatment
coefficients
β
n
. These measure the average deviation of the developments of treatment companies after treatment
from the developments of the control group and the (company-specific) developments before treatment.
29
The variable ‘number of highly educated employees’ is constructed from information from Statistics Denmark.
This information can be a couple of months older than the closing date of the given company’s financial report,
which sets the time structure of the analysis. For example, VPs hired between Statistics Denmark’s closing date
at the end of November and the end of March will, in companies closing their books at the end of March, first occur
in the data in the following year.
30
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
67
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0068.png
FIGURE 5.1: Number of employees. Aggregated estimated model coefficients.
Years after treatment on horisontal axis.
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
0
1
2
3
4
5
Number of highly
educated employees
Number of employees
As with the individual-level analysis, the coefficient estimates
‘TREAT=1 &
t=1’, ‘TREAT=1 & t=2’,..., ‘TREAT=1 & t=5’
can be summed up to calculate the
total potential effect up to five years after treatment. This potential effect is an
additional
(0.46+0.32=) 0.78
individuals in the first two years and an additional
(0.46+0.32+0.00-0.14-0.22 =) 0.42
individuals in the first five years after
treatment.
31
Accordingly, a first conclusion is that VP-companies on average increase the
number of employees with a post-secondary education and above by an additional
0.8
employees in association with programme participation. However, there are
no indications that participating companies continue to increase their number of
employees in the years after programme participation: They have, on average,
lower increases (greater declines) in the number of highly educated employees
than companies in the reference group in year four and five after year zero, but this
finding is not statistically significant.
Results for employment (independent of educational level) indicate that there is an
immediate potential effect of
0.6
additional employees in the year of treatment,
which is slightly larger than the potential effect found for highly educated
employees. This indicates that VPs are often hired in association with company
growth, or that some of the VPs are categorised as having an education below
ISCED 5 or 6 in the Statistics Denmark education registers.
These numbers are high in comparison with the previous finding that long-term relationships between VPs and
their hosting companies are relatively uncommon, suggesting that VPs are replaced by other highly educated
individuals after the end of their projects.
31
68
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0069.png
As is the case for highly educated employees, there is no sign that participating
companies continue to increase the number of employees in the years after
programme participation, with negative coefficients for year 4 and 5 after treatment
resulting in an aggregate potential treatment effect over the first five years of -0.2
additional employees. Even though this number is not statistically different from
zero, it is still the best guess of any long-run treatment effect of the programme.
TABLE 5.5.b: Potential effects on the number of employees. Further results
Ordinary
least squares
regression
Firm fixed-effects
model
Conditional diff-
in-diff model
Conditional diff-
in-diff model,
dependent vari-
able: annual
employment
growth in percent
1
Coeff.
8.834***
1.517
0.812
-1.618
-5.425
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
1.164***
0.237
0.215
-0.896**
-1.168***
Ste.
0.175
0.241
0.280
0.383
0.421
Coeff.
0.642**
-0.118
0.004
-0.448
0.436
Ste.
0.278
0.411
0.520
0.651
0.789
Coeff.
0.596*
0.001
0.331
-0.446
-0.694
Ste.
0.296
0.335
0.400
0.596
0.646
Ste.
2.929
2.934
3.152
4.058
6.459
Includes firm-fixed
effects
Includes year dummy
variables
Includes information
from before year zero
Includes
observations of the
control group
no
no
no
no
yes
yes
yes
no
yes
yes
yes
yes
yes
yes
yes
yes
Number of
observations:
Number of
companies:
R2:
650
274
0.07
1399
278
0.08
2727
546
0.08
2520
525
0.07
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% signficance level.
1: Only observations with annual growth between -50 and 100 percent.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
69
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0070.png
We are again interested in whether or not the results regarding the potential
employment effects are because of higher growth in treatment companies after
participation relative to before participation, or if the results are due to control
companies having lower growth after year zero relative to before when compared
to the treatment companies. For employment developments, we find again that the
overall results do not depend on the choice of the control group, as the before-after
comparison of the fixed effects model (on the subpopulation of treatment companies)
gives estimators that are highly similar to the fully specified model.
Also, we are interested in learning how much the previous results depend on
measuring employment growth as either absolute increases or percentage-point
growth. Investigating absolute annual increases is the first choice for simple-to-
implement cost-benefit calculations, but this also implies that smaller companies
with small absolute changes in the performance parameters are given low weight in
the statistical estimations.
When considering percentage-point employment growth, we find again a statistically
highly significant positive potential employment effect in the years around treatment,
suggesting that treatment companies grow by an additional 10 percent in the first
two years after treatment. But also in this alternative model, there is no indication
that treatment companies continue to increase their number of employees in year 4
and 5 after treatment.
32
Potential effects on value added, net income (profits) and return
on assets
We now turn to the financial performance variables. The results for these variables
need to be interpreted with care, since they depend critically on the treatment of
data - first and foremost the definition and treatment of outliers, i.e. companies
experiencing large changes in the performance variables.
For the specific treatments of outliers and the given modelling choices, we find
mostly positive, albeit statistically insignificant potential treatment effects for both
value added
33
and net income (profits). Findings of TABLE 5.4 are depicted in
FIGURE 5.2 and show that participating companies gained up to an additional
DKK800,000 (EUR106,000) in annual value added and DKK400,000 (EUR53,000)
in net income. But given the lack of statistical significance, these results should be
interpreted as highly tentative.
We will also present results for percentage-point growth rates for some of the other success parameters: gross
profit, average wages, and labour productivity. There will be no such regressions for the performance measures
number of highly educated employees, net income and return on assets, because these measures often assume
the value zero or negative values – which implies that growth rates cannot be calculated.
32
This variable is from the financial statements that companies file with the public authority, where it is called
dækningsbidrag/bruttofortjeneste.
33
70
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0071.png
FIGURE 5.2: Gross profit and net income (DKK1,000) developments in small steady-
going companies. Aggregated estimated model coefficients. Years after treatment
on horisontal axis.
1000
800
600
400
200
0
-200
0
1
2
3
4
5
Value added
Net income
FIGURE 5.3: Return on assets developments (in percent) in small steady-going
companies. Aggregated estimated model coefficients. Years after treatment on
horisontal axis.
105
100
95
90
85
80
75
0
1
2
3
4
5
Return on assets
We also take a look at developments in return on assets, calculated as net income
over total assets. The reasoning is that we have already looked at company growth
variables, such as the number of employees and increases in value added, and that
return on assets is largely independent of company size (which is obviously not the
case for net income).
Cf. FIGURE 5.3, we find that companies that hire VPs on average do worse in
terms of return on assets relative to companies in the control group of highly similar
companies, but that coefficients are statistically insignificant.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
71
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0072.png
TABLES 5.5.c-d further expand on the findings for value added, net income, and
return on assets.
A look at the left hand side coefficients of TABLE 5.5.c suggests that value added
developments are on average positive for treatment companies in the first three
years after treatment, and negative more than three years after treatment. Part of the
increases in the first years after treatment can be interpreted as a continuation of
pre-treatment growth developments, as coefficients drop from DKK 421,702 to
DKK 260,056 when controlling for company fixed effects. Controlling for
developments in highly similar control companies, on the other hand, does not
change the general picture, so the selection of the control group does not appear to
be important to the overall result.
Also, for given sampling criteria, the previous (statistically insignificant) finding that
treatment companies on average have higher value added growth is confirmed by the
regression of percentage point value added growth. This regression even suggests
the presence of positive and statistically significant potential effects for year two and
four after treatment. The findings of a lack of significance for the model of absolute
value added increases and the presence of significance for the growth rate model
lends itself to the interpretation that companies with initially low value added gain
the most in association with programme participation.
Turning to net income increases, we find that there is large heterogeneity in this
variable, and as a consequence no statistically significant potential treatment effects
can be detected for any of the different models. On average, absolute net income
growth is negative for treatment companies after treatment. This can be explained
by generally adverse business developments and company-specific time trends, as
controlling with year dummies and for company-fixed effects in the regressions
reverses the sign of the point estimates, making them positive. Again, taking into
account the developments in the control group does not have any major impact on
the overall results.
With regard to return on assets, it can be noted that the estimated coefficients are
typically significantly negative in the pure before-after comparison of the company-
fixed effects model: Treatment companies experience lower increases in return-on-
assets after treatment relative to before treatment. This finding is not replicated in
the fully specified conditional diff-in-diff model, where coefficients get closer to
zero and are no longer statistically significant. This indicates that companies in the
control group also experience adverse return-on-assets developments in the years
after being chosen into the control group.
72
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0073.png
TABLE 5.5.c: Potential effects on value added (DKK1,000). Further results
Ordinary
least squares
regression
1
Firm fixed-effects
model
1
Conditional diff-in-
diff model
1
Conditional diff-in-
diff model,
dependent
variable: annual
value added growth
in percent
2
Coeff.
4.07
10.10**
5.59
12.12**
-2.81
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
421.702***
239.746*
57.844
-90.041
-884.891**
Ste.
113.743
134.653
186.946
305.445
353.632
Coeff.
260.056
181.428
237.428
396.622
-72.243
Ste.
214.888
281.134
384.165
514.582
630.232
Coeff.
219.298
374.106
165.199
124.020
-563.130
Ste.
217.152
239.569
324.385
448.158
580.313
Ste.
3.58
4.30
4.27
5.18
7.20
Includes firm-
fixed effects
Includes year
dummy variables
Includes
information from
before year zero
Includes
observations of
the control group
no
no
no
yes
yes
yes
yes
yes
yes
yes
yes
yes
no
no
yes
yes
Number of
observations:
Number of
companies:
R2:
620
272
0.03
1346
272
0.02
2611
533
0.03
2223
451
0.03
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% level.
1: Only observations with annual change in value added of less than DKK 10 mio.
2: Only observations with annual growth between -50 and 100 percent.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
73
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0074.png
TABLE 5.5.d: Potential effects on net income (DKK1,000). Further results
Ordinary least squares
regression
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
-48.481
-41.199
51.065
-36.775
-277.485
Firm fixed-effects model
Coeff.
6.965
69.773
229.337
262.090
171.135
Conditional diff-in-diff
model
Coeff.
-48.457
136.506
133.285
205.536
-103.538
Ste.
48.225
67.556
81.233
126.978
152.530
Ste.
98.949
130.681
166.055
227.975
267.157
Ste.
95.159
111.445
122.101
218.128
189.155
Includes firm-fixed effects
Includes year dummy
variables
Includes information from
before year zero
Includes observations of the
control group
no
no
no
no
yes
yes
yes
no
yes
yes
yes
yes
Number of observations:
Number of companies:
R2:
600
276
0.03
1322
276
0.02
2553
542
0.02
Notes: Only observations with annual changes in the number of employees of less than 12. Only observations with annual change in net income of
less than DKK 3 mio.
74
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0075.png
TABLE 5.5.e: Potential effects on return on assets (profits over total assets). Further results
Ordinary least squares
regression
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
-0.027*
-0.039***
-0.017
-0.036*
-0.099**
Firm fixed-effects
model
Coeff.
-0.038*
-0.066**
-0.047
-0.082*
-0.141**
Conditional diff-in-diff
model
Coeff.
-0.029
-0.036
-0.001
-0.036
-0.042
Ste.
0.013
0.015
0.021
0.021
0.046
Ste.
0.023
0.030
0.036
0.045
0.069
Ste.
0.023
0.026
0.028
0.039
0.060
Includes firm-fixed effects
Includes year dummy
variables
Includes information from
before year zero
Includes observations of
the control group
no
no
no
no
yes
yes
yes
no
yes
yes
yes
yes
Number of observations:
Number of companies:
R2:
630
277
0.04
1361
277
0.01
2669
544
0.01
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% level.
Only observations with annual change in return on assets of less than 1, and total assets
>
DKK 100,000.
Potential effects on average wage costs and labour productivity
Results for average wage costs and labour productivity (measured as turnover per
employee) are in the rightmost columns of TABLE 5.4, and illustrated in FIGURE
5.4.
34
With regard to the average wage costs per employee, it appears that any
potential treatment effects are too small relative to the variation in the data and
the number of observations. TABLE 5.5.f suggests that on average there are no
substantial changes in wage cost per employee after treatment, a finding which is
unaltered by the before-after comparisons for the subsample of treatment companies,
or when considering growth rates rather than absolute changes.
The variable ‘ wage cost per employee’ is from the balance sheet information of the KOB/Experian database,
and is characterised by a share of missing observations.
34
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
75
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0076.png
Labour productivity is measured as turnover per employee.
35
For absolute changes
in labour productivity, it is not possible to demonstrate that VP-companies have
higher productivity increases than the highly similar companies in the control
group: Negative signs for t>2 indicate that VP-companies have lower increases than
their counterparts in the reference group. However, this finding is not statistically
significant and thus highly tentative. The picture also changes when we consider
annual percentage-point growth in labour productivity rather than absolute annual
increases: In this model specification, treatment companies generally outperform
control companies in terms of labour productivity growth.
This finding – that treatment companies on average perform better than controls in
terms of percentage-point growth and not significantly better in terms of absolute
increases – implies that results are not robust with regard to model reformulation.
This should advise us against drawing too strong conclusions on the basis of the
statistical results. However, the fact that treatment companies seem to perform best
when the performance is measured in percentage-point growth rather than absolute
increases is an indication that it is in particular small companies that gain the most
from programme participation.
Turnover is from the Statistics Denmark registers instead of the Experian data. This is because (a) only compa-
nies above certain size thresholds are obliged to report this variable to the public authorities (which is why it is of-
ten missing in the Experian database) and (b) turnover is found for almost all companies in the Statistics Denmark
registers (because VAT is registered for almost all companies).
35
FIGURE 5.4: Wage and labour productivity developments (DKK1,000). Aggregated
estimated model coefficients. Years after treatment on horisontal axis.
100
0
-100
-200
-300
-400
0
1
2
3
4
5
Wage cost per employee
Turnover per employee
76
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0077.png
TABLE 5.5.f: Potential effects on wage cost (DKK1,000) per employee. Further results
Ordinary least
squares regression
1
Firm fixed-effects
model
1
Conditional diff-in-
diff model
1
Conditional diff-in-
diff model,
dependent
variable: growth
of wage cost
per employee in
percent
2
Coeff.
-3.90
-0.43
-4.95
1.36
-5.15
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
3.20
14.98*
-6.34
16.04
5.94
Ste.
5.94
7.57
9.74
12.27
13.15
Coeff.
-8.96
0.41
-17.40
7.09
-26.35
Ste.
10.86
15.59
21.95
24.88
35.12
Coeff.
8.28
5.24
-21.34
16.06
-19.32
Ste.
11.26
10.27
13.79
17.18
29.92
Ste.
2.94
2.41
3.52
4.50
6.70
Includes firm-fixed
effects
Includes year dummy
variables
Includes information
from before year
zero
Includes
observations of the
control group
no
no
no
yes
yes
yes
yes
yes
yes
yes
yes
yes
no
no
yes
yes
Number of
observations:
Number of
companies:
R2:
355
190
0.02
794
190
0.02
1494
346
0.01
1474
343
0.01
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1: Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
2: Only observations with annual growth between -50 and 100 percent.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
77
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0078.png
TABLE 5.5.g: Potential effects on labour productivity (DKK 1,000). Further results
Ordinary
least squares
regression
1
Firm fixed-
effects model
1
Conditional diff-in-
diff model
1
Conditional diff-in-
diff model,
dependent
variable: annual
labour productivity
growth in percent
2
Coeff.
2.50
6.64*
2.90
4.26
-7.57
Coeff.
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
-67.67
146.02*
-70.54
-43.03
-133.16
Ste.
53.38
78.44
69.29
103.33
83.80
Coeff.
-78.40
127.63
-45.92
-46.98
-115.44
Ste.
85.83
128.59
124.43
176.48
202.26
Coeff.
-27.92
57.28
-137.30
-39.23
-159.68
Ste.
93.59
107.95
91.84
134.02
215.28
Ste.
3.71
3.72
4.07
5.41
11.12
Includes firm-fixed
effects
Includes year dummy
variables
Includes information
from before year zero
Includes observations of
the control group
no
no
no
no
yes
yes
yes
no
yes
yes
yes
yes
yes
yes
yes
yes
Number of observations:
Number of companies:
R2:
369
171
0.02
898
171
0.02
1693
323
0.02
2186
483
0.02
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
2: Only observations with annual growth between -50 and 100 percent.
78
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0079.png
Results for subsamples
In the following, we look at whether previous findings are different for different
industries, VP- and project-specific characteristics.
This functions as a robustness check of the previous results, but it also offers an
opportunity to see under what circumstances the programme might be considered to
be most successful. In particular, the previous regression models will be applied on
the following samples:
1. All companies, with no outliers removed.
2. Only companies where the DASTI and Statistics Denmark data are in
accordance with regard to the company-VP match.
3 Only VP-projects that were not aborted before schedule.
4. Only companies without any tertiary-level educated employees in the year
prior to programme participation.
5. Only VP-projects in, respectively, manufacturing, services, and other
industries.
6. Only male VPs, only female VPs.
7. Only VPs with a tertiary education.
8. Only VPs with education degrees in, respectively, arts and humanities,
social sciences, and technical sciences subjects.
For ease of reading, the results can be found in the appendix of this report.
Aggregated regression coefficients, which measure potential treatment effects, are
for most of the subsamples illustrated graphically and discussed below.
Let us first turn our attention to the results for the sample of all companies, with
no outliers removed. For this sample, estimated standard errors are often much
larger than the absolute sizes of the coefficient estimates (TABLE A.1). Thus, for
all participant companies (including the larger ones), it is not possible to make
statements on the potential treatment effects with any degree of accuracy, with the
exception of the employment of highly educated employees.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
79
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0080.png
Potential effects on the number of highly educated employees
We find the largest potential treatment effects for companies without tertiary-level
educated employees in the year prior to treatment, and for companies hiring male
VPs, and for those hiring VPs with a technical sciences education. The lowest
potential effects are found for those hiring VPs with an education in arts and
humanities, and, especially over a time horizon beyond the very first years after
treatment, female VPs. There is only a small immediate potential effect for service
industries. However, companies in these industries increase the number of highly
educated employees in the years after treatment.
FIGURE 5.5.a: Number of highly educated employees. Estimated potential treatment
effects. Years after year zero on horisontal axis.
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
1
2
3
4
5
Companies w/o
tertiary-level educat-
ed prior to treatment
Only completed projects
Agreement on VP-
company match in
DASTI and DST data
FIGURE 5.5.b: Number of highly educated employees. Estimated potential treatment
effects. Years after year zero on horisontal axis.
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
1
2
3
4
5
Manufacturing and
construction indus-
tries
Services
Other industries
80
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0081.png
FIGURE 5.5.c: Number of highly educated employees. Estimated potential treatment
effects. Years after year zero on horisontal axis.
1.5
1.0
0.5
0.0
0
-0.5
-1.0
1
2
3
4
5
VP is male
VP is female
FIGURE 5.5.d: Number of highly educated employees. Estimated potential treatment
effects. Years after year zero on horisontal axis.
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
-2.0
0
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
81
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0082.png
Potential effects on the number of employees
For total employment, it proves to be important that the VP-project is completed and
not aborted before schedule. Again, it is companies that hire female VPs and VPs
with an education in arts and humanities that have the poorest growth performance.
For VPs with a technical sciences education, a positive potential programme effect
for highly educated employees and the absence of any detectable potential effect
for employees of all educations indicate that companies that hire these VPs would
have employed other individuals with lower educations in the counterfactual case of
non-participation.
FIGURE 5.6.a: Number of employees. Estimated potential treatment effects. Years
after year zero on horisontal axis.
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
0
1
2
3
4
5
Companies without
highly educated
employees prior to
treatment
Agreement in VP-com-
pany match in DASTI
and DST data
Completed VP projects
FIGURE 5.6.b: Number of employees. Estimated potential treatment effects. Years
after year zero on horisontal axis.
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
0
1
2
3
4
5
Manufacturing and
construction industries
Services
Other industries
82
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0083.png
FIGURE 5.6.c: Number of employees. Estimated potential treatment effects. Years
after year zero on horisontal axis.
3.0
2.0
1.0
0.0
-1.0
-2.0
-3.0
-4.0
0
1
2
3
4
5
VP is male
VP is female
FIGURE 5.6.d: Number of employees. Estimated potential treatment effects. Years
after year zero on horisontal axis.
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
-2.0
0
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
83
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0084.png
Potential effects on value added
The comparison of potential value added effects agrees to large extent with the
findings for employment: Subgroups of companies that are characterised by low
average increases in the number of (highly educated) employees in association with
programme participation are also characterised by low increases in value added.
This is notably the case for companies hiring female VPs and VPs with an education
categorised as within arts and humanities. The highest average increases are found
in the manufacturing industries and for VPs with a social sciences-related education.
With regard to value added, it is again important that the project was completed,
while there is no indication that companies without tertiary educated employees
prior to treatment gain the most in terms of value added.
FIGURE 5.7.a: Value added (DKK1,000). Estimated potential treatment effects.
Years after year zero on horisontal axis.
1500
1000
500
0
0
-500
-1000
1
2
3
4
5
Companies without
highly educated
employees prior to
treatment
Agreement on VP-com-
pany match in DASTI
and DST data
Completed VP projects
FIGURE 5.7.b: Value added (DKK1,000). Estimated potential treatment effects.
Years after year zero on horisontal axis.
2500
2000
1500
1000
500
0
-500
-1000
0
1
2
3
4
5
Manufacturing and
construction industries
Services
Other industries
84
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0085.png
FIGURE 5.7.c: Value added (DKK1,000). Estimated potential treatment effects.
Years after year zero on horisontal axis.
2000
1500
1000
500
0
-500
-1000
0
1
2
3
4
5
VP is male
VP is female
FIGURE 5.7.d: Value added (DKK1,000). Estimated potential treatment effects.
Years after year zero on horisontal axis (year zero=100).
3000
2000
1000
0
-1000
-2000
0
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
85
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0086.png
Potential effects on net income (profits) and return on assets
What was true for the developments of value added does not necessarily hold
true for net income. For example, companies hiring female VPs are on average
not characterised by less favourable net income developments. It can be noted
that companies without tertiary-level educated employees prior to treatment and
companies hiring VPs with a technical sciences education do best in terms of
return-on-assets developments.
FIGURE 5.8.a: Net income (DKK1,000). Estimated potential treatment effects. Years
after year zero on horisontal axis.
500
400
300
200
100
0
-100
0
1
2
3
4
5
Companies without
highly educated
employees prior to
treatment
Agreement on VP-
company match in
DASTI and DST data
Completed VP projects
FIGURE 5.8.b: Net income (DKK1,000). Estimated potential treatment effects. Years
after year zero on horisontal axis.
700
600
500
400
300
200
100
0
-100
-200
Manufacturing and
construction industries
Services
Other industries
0
1
2
3
4
5
86
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0087.png
FIGURE 5.8.c: Net income (DKK1,000). Estimated potential treatment effects. Years
after year zero on horisontal axis.
1000
800
600
400
200
0
-200
0
1
2
3
4
5
VP is male
VP is female
FIGURE 5.8.d: Net income (DKK1,000). Estimated potential treatment effects. Years
after year zero on horisontal axis.
1000
500
0
-500
-1000
-1500
-2000
0
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
87
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0088.png
FIGURE 5.9.a: Return on assets. Estimated potential treatment effects. Years after
year zero on horisontal axis.
0.00
0
-0.05
-0.10
-0.15
-0.20
-0.25
1
2
3
4
5
Companies without
highly educated
employees prior to
treatment
Agreement on VP-
company match in
DASTI and DST data
Completed VP projects
FIGURE 5.9.b: Return on assets. Estimated potential treatment effects. Years after
year zero on horisontal axis.
0.00
-0.05
-0.10
-0.15
-0.20
-0.25
-0.30
-0.35
0
1
2
3
4
5
Manufacturing and
construction industries
Services
Other industries
88
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0089.png
FIGURE 5.9.c: Return on assets. Estimated potential treatment effects. Years after
year zero on horisontal axis.
0.00
-0.05
-0.10
-0.15
-0.20
-0.25
0
1
2
3
4
5
VP is male
VP is female
FIGURE 5.9.d: Return on assets. Estimated potential treatment effects. Years after
year zero on horisontal axis.
0.00
-0.10
-0.20
-0.30
-0.40
-0.50
-0.60
0
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
89
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0090.png
Potential effects on wages and labour productivity
The comparison of wage costs per employee leaves us with no clear results. Instead,
erratic movements in the estimates over time suggest that these are mostly due to
statistical noise rather than underlying trends.
For labour productivity, we find that companies in other industries than
manufacturing and services, and companies that hire VPs with technical educational
degrees, do well relative to other companies. Those that hire VPs with an
educational background in arts and humanities, and those in the service industry, are
characterised by the most negative estimates.
FIGURE 5.10.a: Average wage cost per employee (DKK1,000). Estimated potential
treatment effects. Years after year zero on horisontal axis.
40
30
20
10
0
-10
-20
-30
0
1
2
3
4
5
Companies without
highly educated
employees prior to
treatment
Agreement on VP-
company match in
DASTI and DST data
Completed VP projects
FIGURE 5.10.b: Average wage cost per employee (DKK1,000). Estimated potential
treatment effects. Years after year zero on horisontal axis.
50
40
30
20
10
0
-10
-20
-30
-40
Manufacturing and
construction industries
Services
0
1
2
3
4
Other industries
90
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0091.png
FIGURE 5.10.c: Average wage cost per employee (DKK1,000). Estimated potential
treatment effects. Years after year zero on horisontal axis.
80
60
40
20
0
-20
-40
-60
0
1
2
3
4
5
VP is male
VP is female
FIGURE 5.10.d: Average wage cost per employee (DKK1,000). Estimated potential
treatment effects. Years after year zero on horisontal axis.
60
40
20
0
0
-20
-40
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
91
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0092.png
FIGURE 5.11.a: Labour productivity (DKK1,000). Estimated potential treatment
effects. Years after year zero on horisontal axis.
200
0
-200
-400
-600
-800
-1000
-1200
0
1
2
3
4
5
Companies w/o
tertiary-level educated
prior to treatment
Agreement on VP-
company match in
DASTI and DST data
Completed VP projects
FIGURE 5.11.b: Labour productivity (DKK1,000). Estimated potential treatment
effects. Years after year zero on horisontal axis.
400
200
0
-200
-400
-600
-800.
0
1
2
3
4
5
Manufacturing and
construction industries
Services
Other industries
92
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0093.png
FIGURE 5.11.c: Labour productivity (DKK1,000). Estimated potential treatment
effects. Years after year zero on horisontal axis.
200
0
-200
-400
-600
-800
-1000
VP is male
0
1
2
3
4
5
VP is female
FIGURE 5.11.d: Labour productivity (DKK1,000). Estimated potential treatment
effects. Years after year zero on horisontal axis.
500
0
0
-500
-1000
-1500
1
2
3
4
5
VP has education in
technical sciences
VP has education in
social sciences
VP has education in
arts&humanities
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
93
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0094.png
6.
EXTENSIONS
The survival of VP-companies
As a first extension of the analysis, we look at the survival/closure rate of VP-
companies in comparison with the reference group of control companies. This
is achieved by simply comparing closure rates as depicted in Figure 6.1 and an
estimation of a binary choice model which has company closure in a given year
as its dependent variable.
36
The results of this regression are displayed in TABLE
6.1 and corroborate the finding that there are no significant differences between
companies that hire VPs and other similar companies that do not participate in the
programme.
Closure is measured between year t and year t+1, where year t is the last year in which the company is found in
the Experian database. The Experian database has information on the status of companies that allow distinguish-
ing company closures from, for example, company sales or mergers.
36
FIGURE 6.1: Company closure rates, by year after year 0 (horizontal axis).
0.05
0.04
0.03
0.02
0.01
0.00
TREATMENT
CONTROL
-5 -4 -3 -2
1
0
1
2
3
4
5
94
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0095.png
TABLE 6.1: Comparison of company closure probabilities of VP-companies and companies in the
reference group. Logit binary choice regression results. Dependent variable: bankruptcy after t=x.
Dependent variables (in
first differences):
All companies in treatment and
control group
Coeff.
Ste.
0.49
0.47
0.83
0.61
0.78
Companies with less than 50
employees
Coeff.
0.67
-0.20
1.09
0.41
0.29
Ste.
0.51
0.50
0.83
0.66
0.78
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1 (omitted category)
t=2
t=3
t=4
t=5
Year dummies
2005 (omitted category)
2006
2007
2008
2009
2010
Constant
Number of observations:
Pseudo-r-squared
0.54
-0.15
1.11
0.30
0.25
0.43
-1.33
0.12
0.93
0.51
0.82
0.61
0.76
0.48
-1.17
0.01
1.05
0.55
0.83
0.67
0.78
0.13
1.32
2.24
0.98
-0.04
-4.82
1987
0.08
1.17
1.07
1.05
1.07
1.13
1.05
0.13
1.24
2.16
0.99
-0.26
-4.816
1876
0.08
1.17
1.08
1.05
1.08
1.15
1.054
A comparison of
VP-companies
and companies participating in
Innovation Networks
For one of the extensions of the analysis, DASTI provided data on companies that
have participated in the so-called
Innovation Networks.
These networks or clusters
are financially supported by DASTI and have the purpose of increasing knowledge
diffusion by providing a platform for collaborations between companies, knowledge
institutions and other cluster participants.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
95
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0096.png
These data consist of 1923 observations belonging to 1158 companies, the
discrepancy owing itself to the fact that some firms participate in these networks
more than once. In the following, these companies’
(IN-companies)
performance is
compared with the performance of the VP-companies.
First, we compare developments in some of the performance variables between
IN- and VP-companies. This comparison is highly informal since the two groups
of companies differ in observable characteristics and must be assumed to differ in
unobservable characteristics as well.
The left hand side columns of TABLE 6.2 compare VP-companies with all IN-
companies present in the Experian data that participated in the clusters after 2004.
We initially find that IN companies are on average significantly larger and have
more highly educated employees than VP-companies. Also, a larger share of the IN-
companies are in the IT industry.
To increase the comparability of the two groups of companies for the subsequent
comparisons, only companies with a net income between DKK -7 million and DKK
7 million and a maximum size of 50 employees in the year before treatment (which
is roughly the 99% percentile of the VP-companies’ distribution of this variable) are
considered.
Summary statistics of the adjusted sample used for the statistical comparison are in
the right hand side columns of TABLE 6.2. The adjustments in terms of company
size and profit have made the two groups of companies surprisingly similar in their
observable characteristics in the year before treatment, with the exception that IN-
companies are characterised by a higher share of highly educated employees.
The results of the new comparison are shown in TABLE 6.3 and are in concordance
with earlier findings based on the comparison of VP-companies with a reference
group of highly similar companies: VP-companies increase their numbers of highly
educated employees in the year of treatment and sometimes in the first years after
treatment.
However, it cannot be shown that VP-companies grow faster than IN-companies in
the number of employees. On the contrary, they appear to have lower growth, i.e.
shrink faster, than IN-companies more than three years after treatment. Additional
regressions (not shown) further indicate that this finding becomes even more
accentuated when considering percentage point employment growth rather than
absolute increases in the number of employees.
Findings also suggest that VP-companies have a lower growth in value added
and net income, but these findings are generally not statistically significant. VP-
companies have wage developments and labour productivity (turnover/employees)
developments approximately equal to the group of IN-companies in most years after
treatment and higher in single years.
96
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0097.png
In sum, earlier findings that VP-companies do not have statistically significant
higher increases in the set of financial success variables relative to the reference
group of highly similar companies are replicated in the comparison with a sample of
small companies that have participated in an Innovation Network.
TABLE 6.2: Summary statistics of companies participating in Innovation Networks (IN) vs. VP-
companies, in year t=0.
Raw data
VP-companies
N=314
Mean
Number of highly educated
employees
Number of employees
Turnover (DKK1,000)
Value added (DKK1,000)
Net income (DKK,1,000)
Return on assets
Labour productivity
(DKK1,000)
Wage cost per employee
(DKK1,000)
Industry: Construction
Industry: Trade
Industry: IT
Industry: Manufacturing
Industry: Metal
Industry: Furniture
Industry: Service
Industry: Business service
Industry: Consulting
Industry: Wood/paper
Industry: Other
2.42
17.45
43682.70
7896.05
2129.82
0.28
4898.19
218.67
0.23
0.41
0.29
0.25
0.22
0.21
0.19
0.23
0.32
0.18
0.42
Comparison sample
1
VP-companies
N=297
Std.
239.96
1043.84
3897574.00
793678.70
591004.40
0.52
2676.78
387.54
0.14
0.38
0.33
0.24
0.15
0.23
0.15
0.16
0.31
0.18
0.48
IN-companies
N=828
Mean
58.98
246.88
598268.60
130868.60
26234.95
-0.06
2063.51
477.04
0.02
0.18
0.13
0.06
0.02
0.06
0.02
0.03
0.11
0.03
0.35
IN-companies
N=479
Mean
4.34
14.12
24178.86
7304.33
213.33
-0.07
1821.75
447.00
0.01
0.20
0.15
0.05
0.02
0.05
0.02
0.04
0.13
0.04
0.29
Std.
2.42
17.45
43682.70
7896.05
2129.82
0.28
4898.19
218.67
0.23
0.41
0.29
0.25
0.22
0.21
0.19
0.23
0.32
0.18
0.42
Mean
1.72
11.22
19600.08
5234.49
357.50
0.03
2175.00
396.27
0.06
0.23
0.10
0.07
0.05
0.04
0.04
0.06
0.12
0.03
0.22
Std.
2.19
11.07
40120.67
5547.89
1311.80
0.24
5010.00
220.95
0.23
0.42
0.30
0.26
0.21
0.20
0.19
0.23
0.33
0.16
0.41
Std.
6.54
13.71
29720.16
8648.45
1871.56
0.55
1793.93
205.25
0.11
0.40
0.35
0.21
0.15
0.21
0.14
0.20
0.34
0.19
0.45
Notes: The comparison sample consists of companies with maximum 50 employees and net income between DKK 7 million and DKK 7 million in year
zero.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
97
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0098.png
TABLE 6.3: Diff-in-diff fixed effects regression results for VP- and IN-companies. Companies with up
to 50 employees in year zero.
Dependent variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2005
2006
2007
2008
2009
2010
2011
Number of highly
educated employees
1
Coeff.
0.508***
0.33
0.502*
-0.34
0.96
-0.05
0.07
-0.506**
0.41
-0.68
Number of employees
Value added
(DKK1,000)
2
Coeff.
174.4
-455.1*
-55.7
-156.5
-15.4
255.2
522.9**
328.4
855.2**
786.4
Ste.
0.17
0.22
0.28
0.35
0.87
0.13
0.18
0.25
0.32
0.84
Coeff.
0.24
-0.19
-0.53
-1.285**
-0.91
0.02
0.29
0.31
0.873*
0.57
Ste.
0.29
0.35
0.44
0.55
0.80
0.22
0.29
0.40
0.51
0.76
Ste.
214.2
255.5
315.7
386.7
491.2
165.7
218.3
290.7
373.0
487.8
0.15
0.08
0.173*
-0.07
-0.391***
0.10
0.10
0.10
0.11
0.13
0.07
0.347*
0.372**
0.01
-1.342***
-1.115***
-0.72
0.18
0.18
0.19
0.20
0.24
0.33
0.99
-211.1
50.5
-191.7
-662.9***
-1652***
-1252***
498.1
143.1
148.6
151.5
162.3
189.8
239.2
663.0
Constant
Number of observations:
R-squared
Number of companies:
0.052
3208
0.03
698
0.07
0.572***
3706
0.06
743
0.13
731.9***
4127
0.05
754
104.6
Notes: Only observations with annual changes in the number of employees by less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
98
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0099.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
-17.41
-8.09
14.72
11.02
10.40
-9.66
5.65
-21.34
19.84
-11.45
Labour productivity
(DKK1,000)
6
Coeff.
92.24
127.60
987.1***
-108.80
425.60
13.37
21.82
-1073***
281.20
-661.30
Coeff.
-433.8
-110.1
-279.2
-137.9
-580.6
687.5
413.5
486.4
352.7
844.3
Ste.
760.0
904.9
1116.0
1365.0
1730.0
581.8
767.8
1025.0
1313.0
1717.0
Coeff.
-0.0575**
0.00
0.01
-0.123**
-0.09
0.02
-0.01
0.02
0.04
0.05
Ste.
0.03
0.03
0.04
0.05
0.06
0.02
0.03
0.04
0.05
0.06
Ste.
16.00
18.91
23.58
28.84
41.93
11.81
15.32
21.11
26.66
39.18
Ste.
234.40
298.60
375.40
465.40
1855.00
171.60
239.80
326.80
416.30
1819.00
-41.2
-883.1*
-413.5
-944.5*
-980.4
-663.4
574.5
506.2
523.9
533.2
571.1
665.2
840.9
2357.0
-0.01
-0.03
-0.0569***
-0.0567***
-0.0993***
-0.04
0.04
0.02
0.02
0.02
0.02
0.02
0.03
0.08
1.93
-6.76
8.95
4.42
-1.30
12.71
28.23
9.26
9.64
9.91
10.62
12.38
16.92
49.73
-138.40
-172.50
-331.2**
-236.2*
-202.40
124.40
129.00
132.00
141.90
166.60
46.950
4224
0.0
769
368.7
0.0216*
4222
0.02
764
0.01
3.75
1917
0.02
515
6.81
116.20
2301
0.01
459
89.61
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
99
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0100.png
A comparison of VP-companies and an extended sample of control
companies
Earlier findings were characterised by a large heterogeneity in the companies’
financial success variables. This argues for a test of the robustness of earlier findings
by running the same regressions on a larger control group of companies. This
reduces the variance of the estimators but comes at the cost of lower similarity
between the group of treatments and the group of controls.
In the following, we depart from the propensity scores calculated earlier and select
five controls (instead of one) for each treatment into the control group. This time,
matching is based purely on the propensity score, without additional conditions on
industry etc.
This procedure selects 1,596 companies into the control group for the 318
participant companies. The similarity of the two groups can be assessed by
inspecting TABLE 6.4. As expected, the two groups are not as similar as the sample
of the earlier analysis, with e.g. slightly larger companies in the extended control
group. However, the conditional diff-in-diff model still allows for a meaningful
comparison between the two groups of companies, and its estimates should be less
affected by statistical noise thanks to an increase in the sample size.
100
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0101.png
TABLE 6.4: Summary statistics of VP-companies and companies in the
extended control group (5 controls per treatment). Companies with up to 50
employees.
VP-companies
N=300
Mean
Number of highly educated
employees
Number of employees
Turnover (DKK1,000)
Value added (DKK1,000)
Net income (DKK,1,000)
Return on assets
Labour productivity (DKK1,000)
Wage cost per employee
(DKK1,000)
Industry: Construction
Industry: Trade
Industry: IT
Industry: Manufacturing
Industry: Metal
Industry: Furniture
Industry: Service
Industry: Business service
Industry: Consulting
Industry: Wood/paper
Industry: Other
1.80
11.65
18437.84
5540.32
296.99
0.04
2165.72
395.80
0.06
0.22
0.10
0.07
0.05
0.06
0.03
0.06
0.12
0.03
0.21
Companies in the
extended control group
N=1,488
Std.
2.26
11.36
Mean
1.78
11.16
18103.18
5249.33
236.41
0.03
2037.00
397.12
0.04
0.22
0.09
0.05
0.04
0.05
0.04
0.07
0.13
0.03
0.23
Std.
2.62
11.86
25721.50
6234.30
1139.21
0.23
3856.43
390.80
0.20
0.41
0.29
0.21
0.20
0.22
0.19
0.26
0.34
0.18
0.42
27033.64
5739.63
981.43
0.21
4973.50
220.14
0.23
0.41
0.30
0.25
0.21
0.23
0.18
0.24
0.33
0.16
0.41
The results of the new comparisons are in TABLE 6.5. We find that increasing the
sample size only marginally reduced the standard errors of the estimates, and that
the results of this model do not alter the previous findings that VP-companies do in
general not experience statistically significant positive developments in the financial
success variables. However, for return on assets, positive (though insignificant) signs
of the relevant coefficient estimates imply that the previous finding of treatment
companies that experience lower growth in this variable in association with
treatment is not robust to changes in how the control group is selected. Also, there
are weak signs that participants experience higher growth in wage costs and value
added, and lower growth in labour productivity.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
101
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0102.png
TABLE 6.5: Diff-in-diff fixed effects regression results for VP-companies and companies in the
extended control group (5 controls per treatment). Companies with up to 50 employees in year zero.
Dependent variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2005
2006
2007
2008
2009
2010
2011
Number of highly
educated employees
1
Coeff.
0.510***
0.451***
0.06
-0.01
0.19
-0.03
-0.10
-0.179**
-0.09
-0.27
Number of employees
Value added
(DKK1,000)
2
Coeff.
235.8
33.7
-47.5
497.2**
-325.1
63.3
51.2
159.6
-11.5
164.0
Ste.
0.10
0.13
0.15
0.16
0.26
0.05
0.07
0.09
0.12
0.18
Coeff.
0.621***
0.25
0.11
-0.16
-0.975*
-0.17
-0.23
-0.29
-0.31
0.20
Ste.
0.21
0.27
0.28
0.41
0.57
0.12
0.16
0.21
0.25
0.31
Ste.
145.6
172.7
190.2
251.8
331.9
82.1
108.8
142.4
174.6
221.6
0.182***
0.173***
0.180***
0.02
0.00
0.05
0.05
0.07
0.08
0.10
0.455***
0.781***
0.645***
0.614***
-0.632**
-0.51
-2.452***
0.17
0.17
0.20
0.22
0.26
0.34
0.93
-11.5
302.9***
87.2
-343.7***
-925.4***
-633.2***
661.4
87.3
97.1
112.2
131.7
168.1
189.9
577.3
Constant
Number of observations:
R-squared
Number of companies:
-0.045
7130
0.03
1633
0.04
0.036
8088
0.07
1706
0.15
304.5***
8945
0.05
1709
77.4
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
102
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0103.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
3.64
16.17
24.43*
29.97**
28.57
6.08
0.49
2.75
5.91
11.66
Labour productivity
(DKK1,000)
6
Coeff.
-301.50
-251.70
-555.7**
-126.40
-282.50
334.80
414.90
491.6**
272.50
66.82
Coeff.
-85.9
-2.1
-100.0
29.7
-199.6
13.0
13.9
28.8
-72.2
152.1
Ste.
85.2
103.3
123.6
152.0
204.0
51.4
67.6
83.4
93.5
108.8
Coeff.
-0.02
0.06
0.06
0.00
0.01
-0.0248*
-0.08
-0.0555**
-0.0770***
-0.07
Ste.
0.03
0.07
0.04
0.05
0.06
0.02
0.06
0.02
0.03
0.04
Ste.
10.79
12.21
13.01
14.87
21.65
7.36
8.82
11.57
14.76
18.71
Ste.
248.00
304.70
272.90
205.40
493.90
248.30
349.10
222.50
188.70
430.30
114.7
129.1*
7.5
-174.4**
-221.3**
23.5
870.7***
69.8
74.9
74.8
86.0
94.5
109.4
289.8
0.0444**
0.0513**
0.03
0.01
-0.0436*
0.04
0.11
0.02
0.02
0.02
0.02
0.03
0.03
0.08
-7.87
-4.08
1.10
-16.05
-5.68
-17.47
82.72**
6.26
6.95
8.42
11.75
13.48
18.33
33.91
-189.0*
-183.60
-314.50
-365.6***
-272.3*
105.80
152.80
259.40
140.70
149.80
-26.130
7609
0.0
1494.0
63.0
-0.0326***
9094
0.01
1748.00
0.01
3.15
3106
0.02
1051.00
4.84
67.26
4500
0.00
960.00
77.79
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
103
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0104.png
7.
CONCLUSIONS
This study has taken a look at the potential effects of the Danish Innovation
Assistant Programme (‘Videnpilotordningen’, VP programme) on the individual and
company level. For this purpose, the analysis considers the employment probabilities
and salary developments of individuals participating in the programme (VPs) and
follows a number of performance variables for participating companies.
To form an understanding of the absolute potential effects of the programme, we
compare participating individuals and companies with highly similar individuals
and companies that do not participate. These comparisons indicate that:
(a) Individuals who participate in the programme have higher employment
probability than similar control individuals in the year after starting to
participate. This is no surprise, since employment is a defining element of
the programme.
(b) Individuals who participate in the programme do not have higher
employment probability than controls more than one year after starting to
participate in the programme, but earn higher wages in the first years.
Here it should be noted that the observation period falls within an
economic boom period with low unemployment. It might be assumed
that the wage and employment developments of programme participants
and non-participants do not converge at the same speed in the current
economic slow-down.
(c) Participating companies increase their numbers of highly educated
employees in association with programme participation. The analysis finds
no signs of behavioural additivity of the programme, i.e. non-
participants increasing their number of highly educated employees.
There are no indications that companies continue to increase the number of
highly educated employees in the years after programme participation.
(d) Participating companies increase the number of employees in association
with programme participation. However, in this case there are also no
indications that the companies continue to increase their employment in
the years after programme participation.
(e) It is difficult to detect statistically significant positive potential effects of
the programme on participating companies’ financial performance
variables. For subsamples of small companies that do not experience large
year-to-year changes in employment or financial measures, participant
companies on average increase their gross profit and net income
in association with programme participation. Findings are again not
statistically significant, and need to be interpreted with care.
(f) Participating companies do not experience increases in return on assets,
wage costs per employee, or labour productivity in association with
programme participation.
104
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0105.png
There are no strong findings about which particular projects are more successful
than others, but it appears that VPs with a tertiary-level education gain less from
participation than VPs with a post-secondary education, while females and VPs
finding employment in service industries gain the most.
For companies, it is important to note that results related to specific characteristics
of the VP or the hosting company are tentative, due to the presence of substantial
statistical uncertainties. This said, one can note that the largest potential effect on
the number of highly educated employees is estimated for small companies that hire
VPs with a technical sciences education as well as male VPs, and for companies that
had no tertiary-level educated employees before treatment.
Also, small companies in manufacturing do well in terms of value added and
net income (profits) developments in association with programme participation,
while participant companies that hire female VPs do relatively poorly in terms of
value added and employment, but not net income. Companies that hire VPs with
an educational background within arts and humanities are characterised by low
growth in association with programme participation, while those hiring VPs with a
technical sciences education do the best, not just in terms of increasing the number
of highly educated employees, but also with regard to net income, return on assets
and labour productivity developments.
The general finding that it is difficult to measure statistically significant potential
effects of the programme proved to be robust to comparing participant companies
with other companies that participated in a similar programme administered by
DASTI (the
Innovation Network
programme) as well as an alternative control group
consisting of several highly comparable control companies for each participant
company.
The VP programme has been analysed earlier on the basis of less extensive data.
This earlier study found potential effects of similar size to the present study.
However, it also found large unexplained year-to-year variation in the performance
variables, leading to statistically insignificant coefficient estimates.
The current analysis supports the earlier analysis’ findings. But the fact that it is still
difficult or impossible to establish statistical significance for most of the relevant
financial variables implies that we still cannot be certain that increased company
performance is a general feature of the programme.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
105
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0106.png
So while there are indications of positive potential programme effects for restricted
subsamples in our data, the general lack of statistical significance implies that any
positive effects of hiring a VP on company performance are small in the face of
the high data demands of our econometric model, a still very limited number of
observations in our data, and the large variation in the companies’ performance
measures. The latter observation also suggests that other company developments, for
example initiated by product developments, must be assumed often to be of major
importance relative to the presence of a VP in the company.
37
Fox, J.T., V. Smeets, 2011, Does Input Quality Drive Measured Differences In Firm Productivity?, International
Economic Review, vol. 52(4), 961-989.
37
106
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0107.png
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
107
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0108.png
APPENDIX 1: ADDITIONAL TABLES OF THE COMPANY-LEVEL ANALYSIS
TABLE A.1: Comparison between VP-companies and companies in the reference group. All companies
irrespective of outliers. Diff-in-diff fixed effects regression results
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.289*
0.16
-0.17
-0.380*
0.29
-0.09
-0.07
-0.23
-0.05
-0.59
Number of employees
Value added (DKK1,000)
2
Ste.
0.17
0.15
0.20
0.23
0.70
0.12
0.15
0.20
0.25
0.64
Coeff.
0.14
0.25
0.63
-0.30
-0.88
-0.41
-0.51
-0.60
-0.99
0.68
Ste.
0.74
0.52
0.55
0.86
1.41
0.45
0.63
0.80
1.09
1.35
Coeff.
48.9
73.5
66.4
-43.8
-558.7
78.8
-394.6
-103.4
-13.9
109.0
Ste.
302.4
244.8
313.4
473.8
452.6
246.0
288.3
418.1
451.8
521.9
0.08
0.02
0.02
0.17
0.14
0.00
-0.10
0.15
0.11
0.12
0.16
0.18
0.23
0.22
-0.33
0.49
0.47
1.185*
0.56
0.22
-2.647**
0.53
0.52
0.55
0.61
0.74
0.89
1.07
-382.7
219.7
361.8
446.4
251.4
-207.5
-1024.0
494.3
492.3
506.6
522.5
571.0
592.5
654.6
Constant
Number of
observations:
Number of companies:
R-squared
0.12
3046
596
0.02
0.10
0.53
2989
580
0.08
0.50
338.2
3664
611
0.03
481.1
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
108
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0109.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
44.45
-20.82
16.87
30.80
44.40*
3.72
38.96
25.52
-16.65
-25.41
Labour productivity
(DKK1,000)
6
Coeff.
-664.00
436.50
125.50
-62.61
-71.96
821.50
172.30
273.40
209.60
178.50
Coeff.
-52.7
12.3
-53.7
-476.7
185.5
35.8
57.8
69.6
299.2
10.5
Ste.
124.2
162.3
221.0
336.1
267.5
132.2
169.7
228.6
276.1
269.5
Coeff.
-0.0580*
0.05
0.00
-0.04
-2.39
0.02
-0.01
0.02
0.04
0.09
Ste.
0.03
0.04
0.04
0.06
2.38
0.04
0.06
0.08
0.06
0.17
Ste.
53.58
26.08
23.68
25.34
26.05
16.63
28.42
28.06
24.65
31.03
Ste.
680.20
318.80
315.90
310.20
593.80
814.70
481.80
588.60
614.10
833.40
-333.5
-29.1
54.4
0.3
-152.3
-468.2
-620.8*
259.2
253.6
257.4
278.0
291.4
308.2
326.0
0.19
0.16
0.14
0.12
0.147**
0.05
0.22
0.13
0.11
0.09
0.08
0.07
0.07
0.19
10.05
-1.59
-17.10
-28.85
-32.51
-39.87
9.14
12.95
13.24
19.86
28.64
39.48
42.35
25.26
-159.60
-53.96
-240.10
117.90
-142.80
-331.50
227.80
660.90
524.50
508.60
601.20
663.30
813.00
673.60
177.9
3799
626
0.02
246.3
-0.120***
3867
627
0.01
0.04
8.19
3107
588
0.01
12.37
-267.80
2856
567
0.01
477.70
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
109
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0110.png
TABLE A.2: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Only companies with
agreement on the VP-company-match in the DASTI and DST data.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.516***
0.283*
-0.06
-0.34
-0.26
0.00
0.02
-0.16
-0.07
-0.02
Number of employees
Value added (DKK1,000)
2
Ste.
0.15
0.16
0.20
0.24
0.29
0.11
0.15
0.21
0.24
0.32
Coeff.
0.57
-0.10
0.42
-0.981*
-0.96
0.14
0.16
0.08
0.64
1.549*
Ste.
0.37
0.40
0.42
0.54
0.77
0.31
0.42
0.53
0.71
0.90
Coeff.
-23.2
413.1
127.0
-0.2
-768.1
102.3
-248.9
171.8
443.6
475.8
Ste.
234.1
257.1
366.9
521.9
682.9
216.2
275.6
412.0
453.0
663.0
-0.01
-0.01
0.02
0.05
-0.02
0.01
-0.15
0.15
0.15
0.13
0.12
0.15
0.18
0.22
-0.18
-0.18
0.15
0.08
0.15
0.15
-0.14
0.26
0.26
0.31
0.33
0.40
0.47
0.61
-371.0
-371.0
401.6
98.9
343.1
272.3
-270.2
257.6
257.6
257.2
239.8
283.5
354.3
413.8
Constant
Number of
observations:
Number of companies:
R-squared
0.10
1632
289
0.05
0.10
0.34
1697
294
0.09
0.25
95.9
1627
290
0.04
203.6
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
110
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0111.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
6.42
8.99
-17.89
34.01*
-28.68
-12.53
-1.81
3.28
-26.61
-12.68
Labour productivity
(DKK1,000)
6
Coeff.
-80.61
125.30
-178.3*
-144.90
-54.19
-95.96
114.80
141.00
48.66
122.60
Coeff.
-67.7
208.0
191.4
258.7
-104.8
-29.0
-277.0**
-81.9
-177.5
-32.8
Ste.
114.2
131.1
140.0
256.9
222.2
113.3
137.1
177.0
252.0
292.2
Coeff.
-0.04
-0.0656**
0.01
-0.02
-0.08
-0.02
-0.04
-0.0709*
-0.07
-0.113*
Ste.
0.03
0.03
0.03
0.05
0.07
0.03
0.03
0.04
0.05
0.06
Ste.
13.33
12.07
15.11
19.80
34.35
13.70
12.21
16.46
22.88
31.07
Ste.
124.30
130.80
105.30
161.80
241.70
112.50
120.10
128.00
152.70
262.00
-129.6
-129.6
99.0
5.3
50.3
104.4
-71.4
107.5
107.5
101.3
94.2
117.5
147.7
185.9
-0.0597**
-0.0597**
0.02
-0.01
0.01
0.03
0.01
0.03
0.03
0.02
0.02
0.03
0.03
0.04
13.46*
13.46*
13.24
13.22
10.87
20.12
4.24
7.06
7.06
8.06
9.19
10.59
14.66
17.45
-175.6**
-175.6**
-35.98
-8.17
21.71
-77.02
-154.30
87.28
87.28
71.15
90.36
111.20
117.50
145.40
66.0
1579
291
0.02
74.9
0.02
1658
292
0.03
0.02
-3.65
978
202
0.03
5.68
63.03
1052
178
0.04
65.53
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
111
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0112.png
TABLE A.3: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Only companies with
completed VP-projects.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.555***
0.429***
0.00
-0.24
-0.25
-0.13
-0.15
-0.25
-0.24
-0.16
Number of employees
Value added (DKK1,000)
2
Ste.
0.13
0.16
0.18
0.22
0.28
0.11
0.15
0.19
0.23
0.29
Coeff.
0.708**
0.26
0.63
-0.45
-0.68
0.06
-0.16
-0.51
-0.31
0.84
Ste.
0.33
0.34
0.42
0.65
0.61
0.26
0.35
0.43
0.58
0.58
Coeff.
273.2
374.5
278.5
-85.2
-678.6
92.7
-210.3
11.4
441.7
571.3
Ste.
245.3
255.4
364.8
489.9
626.9
207.0
242.4
372.8
419.4
581.9
-0.04
0.02
0.05
0.06
0.12
-0.03
-0.01
0.14
0.12
0.11
0.14
0.18
0.20
0.24
-0.04
0.22
0.26
0.21
0.44
0.22
-1.304**
0.23
0.27
0.29
0.34
0.36
0.45
0.52
-255.3
390.3
103.4
279.2
269.8
-226.3
-749.3*
245.7
242.5
234.2
260.6
302.5
347.1
429.8
Constant
Number of
observations:
Number of companies:
R-squared
0.12
2122
431
0.04
0.10
0.38
2217
440
0.08
0.23
204.9
2120
359
0.04
207.3
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
112
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0113.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
0.71
0.47
-7.51
8.34
-26.68
3.99
21.19*
34.80**
27.78
25.04
Labour productivity
(DKK1,000)
6
Coeff.
-72.72
52.31
-163.7*
-134.50
-121.00
-68.50
73.53
88.85
-7.39
13.19
Coeff.
-58.8
228.1*
148.0
112.0
-134.4
25.5
-142.4
45.2
123.6
271.5
Ste.
107.2
122.4
124.9
236.1
195.4
97.8
116.1
146.5
216.8
240.7
Coeff.
-0.03
-0.03
0.01
-0.04
-0.04
-0.02
-0.04
-0.05
-0.03
-0.08
Ste.
0.03
0.03
0.03
0.04
0.07
0.02
0.03
0.03
0.04
0.06
Ste.
11.04
11.92
14.10
19.07
31.65
11.48
12.35
16.25
21.93
30.60
Ste.
114.10
140.10
97.86
145.80
258.60
107.60
116.40
115.30
144.10
242.40
-129.2
36.0
49.0
-15.1
-8.6
-239.4
-392.2**
101.3
95.8
92.5
106.1
126.9
149.1
184.5
-0.0530**
0.00
0.01
0.00
0.02
-0.02
-0.01
0.03
0.02
0.02
0.03
0.03
0.04
0.04
15.23**
7.18
3.37
-2.78
-7.74
-23.69
-16.50
6.89
7.80
7.01
9.59
13.91
16.81
20.86
-60.82
8.70
80.78
166.4*
3.10
-82.71
39.04
55.73
57.82
62.20
95.99
104.70
116.70
133.70
105.5
2084
438
0.03
78.6
0.01
2174
439
0.02
0.02
3.29
1175
274
0.02
6.02
-19.71
1307
386
0.03
49.68
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
113
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0114.png
TABLE A.4: Comparison between VP-companies and companies in the reference group. Companies with
up to 50 employees in year zero. Diff-in-diff fixed effects regression results. VPs in companies without
employees with a tertiary education prior to programme participation.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.690***
0.426***
0.00
-0.19
-0.27
0.02
0.03
0.08
0.07
0.10
Number of employees
Value added (DKK1,000)
2
Ste.
0.12
0.16
0.19
0.24
0.34
0.10
0.13
0.15
0.20
0.28
Coeff.
0.35
0.02
0.41
-0.13
-1.05
0.27
0.03
0.17
0.50
1.512*
Ste.
0.37
0.41
0.47
0.73
0.91
0.31
0.43
0.54
0.74
0.85
Coeff.
52.5
449.0
79.6
181.5
-478.5
132.5
-251.3
296.5
485.9
112.8
Ste.
255.3
316.4
332.1
594.6
729.5
218.4
293.7
359.3
498.9
625.3
0.07
0.09
0.07
0.05
0.00
-0.11
-0.26
0.10
0.10
0.08
0.11
0.13
0.15
0.17
-0.16
0.33
0.42
0.46
0.33
-0.08
-2.371***
0.38
0.41
0.43
0.49
0.51
0.61
0.68
-475.2*
117.2
34.4
307.5
181.1
-509.3
-1222***
286.9
270.9
263.7
321.8
351.2
379.6
453.6
Constant
Number of
observations:
Number of companies:
R-squared
0.03
1711
347
0.07
0.07
0.33
1716
348
0.12
0.36
327.3
1671
342
0.07
240.6
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
114
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0115.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
15.72
3.22
-24.15
14.08
-29.47
-15.93
0.61
19.23
-3.87
0.94
Labour productivity
(DKK1,000)
6
Coeff.
9.63
170.70
-58.23
-37.22
-74.40
-80.51
76.28
101.50
-108.10
51.62
Coeff.
58.0
155.5
95.1
114.6
-17.5
16.3
-124.6
141.5
161.9
126.8
Ste.
126.6
145.1
148.7
284.5
310.1
113.6
121.6
155.4
253.2
271.7
Coeff.
0.00
-0.02
-0.01
-0.03
0.00
0.00
-0.03
-0.02
0.00
-0.135**
Ste.
0.03
0.03
0.03
0.05
0.07
0.03
0.03
0.03
0.04
0.06
Ste.
13.69
12.19
16.50
21.63
36.69
15.31
14.47
18.95
25.53
34.50
Ste.
91.46
115.10
95.90
140.10
226.50
91.12
103.30
108.50
129.00
236.80
-124.8
-34.4
-9.4
-92.3
-86.0
-372.3**
-551.5***
97.9
106.1
88.2
113.0
127.4
149.0
191.3
0.01
0.01
0.01
0.02
-0.03
-0.01
0.00
0.02
0.02
0.02
0.03
0.03
0.03
0.04
14.56*
2.25
0.58
6.07
4.48
-4.13
7.49
7.53
8.26
10.18
12.99
16.76
20.49
23.38
-167.1*
-60.30
-73.07
13.41
-107.60
-168.70
-81.89
89.11
66.77
92.82
99.39
116.80
132.30
134.90
161.7**
1620
345
0.04
76.6
0.00
1686
347
0.03
0.02
3.07
1010
230
0.02
7.30
83.71
1168
215
0.04
71.79
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
115
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0116.png
TABLE A.5: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Companies in
manufacturing industries and contruction.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.675***
0.37
-0.25
-0.18
-0.14
-0.07
-0.22
-0.03
-0.28
0.02
Number of employees
Value added (DKK1,000)
2
Ste.
0.20
0.29
0.27
0.43
0.49
0.15
0.22
0.21
0.37
0.23
Coeff.
0.61
-0.47
0.40
-0.10
-1.61
0.39
0.29
1.09
0.95
3.974***
Ste.
0.85
0.79
0.90
1.29
1.36
0.64
0.86
1.05
1.37
1.23
Coeff.
383.2
623.6
1010.0
220.3
-2228.0
161.8
-117.0
385.7
958.8
2767.0
Ste.
506.1
610.9
760.4
1454.0
3169.0
433.3
530.1
709.5
1008.0
2760.0
-0.11
0.09
-0.05
-0.07
-0.04
-0.07
-0.22
0.27
0.25
0.21
0.22
0.27
0.27
0.29
0.45
0.91
1.43
1.40
1.06
0.14
-3.382**
0.72
0.82
0.88
0.99
0.97
1.24
1.40
-438.8
480.5
212.1
987.6
491.2
-530.1
-1705**
614.6
598.5
617.4
660.3
686.3
688.6
845.5
Constant
Number of
observations:
Number of companies:
R-squared
0.15
643
125
0.07
0.20
-0.32
667
128
0.17
0.72
100.1
640
126
0.11
527.1
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
116
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0117.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
15.16
-21.46
-25.78
38.30
-160.2*
-19.52
-13.16
7.90
-42.36
76.77
Labour productivity
(DKK1,000)
6
Coeff.
-48.80
165.80
-76.95
-5.01
-417.10
-22.15
75.43
26.12
39.84
405.70
Coeff.
0.2
67.4
378.3
159.7
-289.4
176.8
65.6
178.8
274.7
0.0
Ste.
200.4
226.3
257.3
695.7
393.7
172.5
197.9
266.5
560.0
0.0
Coeff.
-0.04
-0.04
-0.05
-0.09
-0.08
-0.03
-0.04
-0.05
-0.07
-0.14
Ste.
0.04
0.04
0.05
0.09
0.08
0.03
0.04
0.05
0.07
0.14
Ste.
15.91
14.91
22.17
29.98
82.57
15.82
16.90
23.95
27.50
67.54
Ste.
102.20
101.60
135.60
240.00
409.70
107.00
130.50
184.40
244.60
481.60
110.2
100.8
139.5
121.7
0.7
-294.3
-516.3
215.9
231.5
173.5
217.0
228.7
272.4
368.1
-0.03
-0.02
-0.02
-0.03
-0.03
-0.04
-0.05
0.03
0.02
0.02
0.03
0.03
0.04
0.05
14.69*
19.67**
4.48
29.13**
34.80*
32.23
31.22
8.25
9.03
5.95
11.75
20.13
20.59
24.77
-45.30
49.33
6.06
120.40
20.63
-74.83
-66.23
68.40
69.99
67.65
121.30
145.40
194.10
246.80
3.5
612
128
0.04
159.2
0.01
660
128
0.05
0.01
-7.64
463
98
0.07
5.32
24.76
532
98
0.05
64.36
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
117
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0118.png
TABLE A.6: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Companies in service
industries
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.27
0.24
0.30
-0.16
-0.30
0.03
0.00
-0.34
-0.01
0.00
Number of employees
Value added (DKK1,000)
2
Ste.
0.17
0.20
0.25
0.29
0.30
0.15
0.18
0.25
0.26
0.29
Coeff.
0.619**
0.11
0.55
-0.69
-0.45
-0.18
-0.35
-0.775*
-0.46
-0.07
Ste.
0.28
0.41
0.45
0.65
0.69
0.25
0.34
0.43
0.62
0.64
Coeff.
178.0
157.0
-75.1
269.1
-758.7
-95.7
-134.4
115.0
511.8
878.6*
Ste.
271.4
267.0
413.6
407.7
565.8
250.1
254.9
416.6
469.1
448.3
0.15
0.23
0.22
0.18
0.23
0.13
0.03
0.15
0.14
0.14
0.16
0.21
0.23
0.26
0.07
0.23
0.02
0.23
0.22
0.43
-0.49
0.18
0.24
0.24
0.30
0.36
0.38
0.45
-288.5
307.4
65.0
163.7
83.9
-373.1
-796.4**
243.6
225.5
206.1
243.8
291.3
365.6
399.1
Constant
Number of
observations:
Number of companies:
R-squared
-0.04
1434
300
0.02
0.12
0.433**
1492
304
0.06
0.21
340.2*
1430
293
0.04
199.4
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
118
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0119.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
1.60
23.14
-13.65
27.89
-5.02
-18.10
-2.07
-14.01
-6.00
-15.28
Labour productivity
(DKK1,000)
6
Coeff.
-99.96
-148.10
-255.60
-53.12
-15.26
-75.44
162.60
181.50
2.36
-113.40
Coeff.
-107.4
133.6
40.3
161.7
65.6
-76.4
-178.6
-22.0
56.4
271.1
Ste.
118.2
130.3
145.6
224.9
236.3
111.0
131.9
165.4
236.5
270.5
Coeff.
-0.04
-0.0702*
-0.0692*
-0.02
-0.03
-0.03
-0.02
-0.03
-0.05
-0.10
Ste.
0.03
0.04
0.04
0.05
0.10
0.03
0.04
0.04
0.05
0.08
Ste.
22.15
19.69
25.40
30.68
41.22
22.37
21.77
27.33
41.05
39.36
Ste.
184.50
221.60
172.20
255.70
466.20
169.40
177.30
179.50
220.80
356.90
-197.0*
179.0*
65.3
100.1
32.1
-105.6
-335.3*
118.3
92.0
103.4
118.0
134.5
156.8
191.1
-0.0709*
0.03
0.01
0.03
0.02
0.00
0.01
0.04
0.03
0.04
0.04
0.04
0.05
0.06
15.75
2.12
14.59
4.45
18.34
2.49
13.96
10.89
15.19
16.26
16.37
23.56
29.14
33.24
-176.80
-32.71
49.76
126.20
-16.60
-185.10
-5.86
116.90
121.00
124.00
157.40
164.20
179.40
187.10
59.4
1420
300
0.05
86.5
0.01
1444
302
0.03
0.03
1.98
634
159
0.03
12.72
13.71
722
142
0.03
107.70
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
119
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0120.png
TABLE A.7: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Companies in ’other’
industries
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.670***
0.44
-0.42
-0.11
-0.03
-0.04
0.07
0.28
0.06
0.19
Number of employees
Value added (DKK1,000)
2
Ste.
0.23
0.33
0.38
0.37
0.65
0.23
0.27
0.39
0.53
0.72
Coeff.
0.93
0.30
-0.57
-0.18
-0.92
-0.02
-0.06
-0.18
0.43
0.91
Ste.
0.74
0.64
0.94
1.57
1.58
0.60
0.71
0.92
1.21
1.25
Coeff.
253.8
541.6
-548.7
-656.4
150.0
-70.8
-467.1
-210.0
-651.9
-1383.0
Ste.
451.0
546.5
581.5
839.1
733.3
416.4
616.7
640.7
799.8
1119.0
-0.08
-0.25
-0.35
-0.28
-0.45
-0.806*
-0.45
0.26
0.20
0.23
0.30
0.37
0.46
0.55
-0.61
-0.01
-0.13
-0.44
-0.24
-0.61
-2.684***
0.47
0.53
0.51
0.63
0.82
0.96
0.97
-182.5
57.0
282.4
-85.5
324.9
-105.5
-257.5
335.3
434.0
393.3
485.6
612.0
660.1
906.3
Constant
Number of
observations:
Number of companies:
R-squared
0.374*
532
110
0.07
0.19
0.914**
568
114
0.12
0.44
233.3
541
114
0.06
298.5
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
120
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0121.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
10.97
12.11
-18.98
-31.56
40.93
2.13
18.42
41.87
49.64
8.49
Labour productivity
(DKK1,000)
6
Coeff.
91.45
238.20
-61.10
-63.79
-188.00
59.35
21.04
186.10
89.03
265.60
Coeff.
33.9
101.2
26.8
248.8
24.6
154.2
-44.4
140.6
-266.0
-259.3
Ste.
251.8
298.5
336.2
413.3
241.0
237.0
282.5
318.4
432.8
459.0
Coeff.
-0.02
-0.03
0.00
-0.09
-0.03
0.01
-0.06
-0.03
-0.02
-0.08
Ste.
0.05
0.05
0.05
0.09
0.03
0.05
0.07
0.06
0.08
0.07
Ste.
14.59
14.07
19.72
32.56
34.04
19.13
17.63
26.48
32.94
39.20
Ste.
142.50
175.10
177.20
152.20
264.70
144.00
162.50
161.50
170.50
287.80
-108.7
-354.1**
6.3
-401.5*
26.0
-307.0
-273.9
174.2
171.2
227.6
240.2
344.1
357.7
409.5
-0.03
-0.04
0.02
-0.01
0.03
-0.02
0.02
0.04
0.05
0.04
0.05
0.06
0.06
0.07
21.23*
7.11
-2.79
-6.87
-23.19
-27.97
-23.52
12.66
11.21
14.73
16.59
24.68
27.68
36.09
-201.00
-124.40
-218.50
-158.40
-279.20
-311.00
-315.10
189.30
107.80
201.00
215.50
204.20
226.30
258.40
189.3
521
114
0.06
166.5
0.00
565
114
0.03
0.03
2.66
405
89
0.04
10.41
173.80
439
83
0.03
141.40
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
121
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0122.png
TABLE A.8: Comparison between VP-companies and companies in the reference group. Companies with
up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Male VPs.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.390**
0.523***
0.18
-0.05
0.05
-0.09
-0.21
-0.31
-0.17
-0.16
Number of employees
Value added (DKK1,000)
2
Ste.
0.15
0.18
0.24
0.26
0.28
0.13
0.18
0.23
0.25
0.32
Coeff.
0.665*
0.39
0.933*
0.13
-0.19
-0.07
-0.41
-0.68
-0.32
0.32
Ste.
0.40
0.44
0.52
0.68
0.78
0.32
0.44
0.54
0.71
0.75
Coeff.
354.9
429.9
308.0
333.1
-532.8
-215.0
-391.1
-281.8
329.6
254.0
Ste.
310.2
301.0
436.8
605.4
635.7
280.9
316.6
456.2
516.9
628.9
0.02
0.16
0.06
0.07
0.24
-0.04
0.02
0.18
0.15
0.14
0.17
0.20
0.23
0.26
0.11
0.55
0.50
0.66
0.71
0.61
-0.89
0.32
0.38
0.41
0.48
0.48
0.57
0.65
-284.2
545.6*
247.3
723.6**
513.9
-48.9
-631.3
315.5
311.3
309.1
351.6
401.7
429.1
529.3
Constant
Number of
observations:
Number of companies:
R-squared
0.05
1605
331
0.03
0.13
0.08
1666
336
0.06
0.34
81.5
1585
327
0.05
269.0
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
122
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0123.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
5.67
-2.21
-27.79
33.80
-59.65
-22.36
3.95
13.10
-14.17
13.29
Labour productivity
(DKK1,000)
6
Coeff.
3.25
104.10
-211.8*
-28.19
-8.74
3.28
140.50
165.40
51.63
84.58
Coeff.
-90.9
192.3
39.8
198.4
-208.6
96.0
-108.7
188.9
263.3
552.6*
Ste.
130.5
144.9
153.4
283.4
233.9
119.4
146.8
169.9
253.6
290.8
Coeff.
-0.03
-0.04
-0.01
0.00
-0.05
-0.02
-0.06
-0.04
-0.03
-0.08
Ste.
0.03
0.04
0.04
0.05
0.08
0.03
0.04
0.04
0.05
0.06
Ste.
17.47
14.71
18.29
21.22
36.94
18.49
17.31
22.52
28.77
36.89
Ste.
122.90
139.50
117.10
181.00
276.50
114.80
134.70
136.30
162.10
251.20
-115.0
78.4
-12.9
19.6
-79.9
-295.5*
-608.3***
126.4
121.8
112.2
133.4
152.2
171.3
214.6
-0.03
0.02
0.01
0.05
0.04
-0.01
0.00
0.03
0.02
0.03
0.03
0.04
0.04
0.05
16.51**
6.10
11.56
15.95
13.88
10.00
2.84
7.78
10.08
12.43
13.89
19.70
23.55
28.28
-97.60
-44.83
-18.77
120.10
-135.60
-218.40
-135.40
66.09
71.81
78.32
105.90
125.40
139.40
154.00
113.9
1547
334
0.04
95.1
-0.01
1628
336
0.03
0.02
-1.53
898
214
0.03
8.62
46.79
1032
199
0.05
63.64
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
123
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0124.png
TABLE A.9: Comparison between VP-companies and companies in the reference group. Companies with
up to 50 employees in year zero. Diff-in-diff fixed effects regression results. Female VPs.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.580***
0.03
-0.29
-0.30
-0.80
0.07
0.16
0.11
0.01
0.31
Number of employees
Value added (DKK1,000)
2
Ste.
0.17
0.22
0.21
0.35
0.49
0.14
0.19
0.23
0.33
0.32
Coeff.
0.51
-0.54
-0.58
-1.27
-1.53
0.10
0.27
0.22
0.13
1.674*
Ste.
0.45
0.52
0.61
1.12
1.02
0.36
0.45
0.64
0.94
0.86
Coeff.
24.4
277.6
-31.2
-192.4
-642.9
248.3
-140.2
330.3
-39.9
827.2
Ste.
279.2
393.4
464.4
658.6
1242.0
240.5
344.3
440.7
601.3
1072.0
-0.03
-0.09
-0.06
-0.06
-0.361*
-0.25
-0.36
0.14
0.14
0.14
0.18
0.21
0.25
0.32
-0.25
0.04
0.12
0.06
-0.21
-0.38
-2.392***
0.36
0.37
0.38
0.43
0.57
0.65
0.77
-381.7
-72.9
7.5
-216.8
-60.3
-645.1
-1034**
233.4
249.8
210.1
270.9
331.7
396.8
434.4
Constant
Number of
observations:
Number of companies:
R-squared
0.229*
1004
204
0.06
0.12
0.747**
1061
210
0.14
0.35
498.4**
1026
206
0.05
192.7
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
124
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0125.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
13.67
18.14
-12.73
-20.79
58.27
0.98
-4.02
7.51
24.47
-35.51
Labour productivity
(DKK1,000)
6
Coeff.
-117.00
-23.48
-23.92
-63.13
-571.6*
-108.10
-14.89
-3.76
-33.94
249.60
Coeff.
15.0
51.2
279.7
241.2
176.7
-74.9
-173.5
-217.4
-418.2
-497.3*
Ste.
134.2
175.9
200.8
347.2
297.4
119.5
143.0
191.9
324.6
270.9
Coeff.
-0.03
-0.03
0.01
-0.108*
-0.03
0.01
0.01
-0.03
0.01
-0.05
Ste.
0.03
0.04
0.04
0.06
0.11
0.03
0.03
0.04
0.06
0.10
Ste.
12.56
13.81
19.98
29.18
36.94
11.76
13.02
17.31
26.46
32.44
Ste.
147.30
165.90
148.50
194.40
313.80
134.00
110.10
133.50
164.80
323.70
-63.7
11.3
183.9
-28.7
192.6
-28.4
21.2
125.3
115.4
125.5
131.5
168.6
188.8
218.8
-0.06
0.00
0.00
-0.04
-0.01
-0.04
-0.01
0.04
0.04
0.03
0.04
0.04
0.05
0.05
19.05*
14.57
3.99
3.92
11.08
-5.40
19.74
9.62
8.94
8.03
10.40
16.59
17.29
20.54
-209.30
-17.98
-48.35
-63.25
61.72
-93.17
-8.21
163.20
107.20
155.80
169.10
161.40
180.10
206.20
23.1
1006
208
0.03
102.8
0.02
1039
208
0.03
0.03
-2.26
597
132
0.04
6.70
84.41
661
124
0.03
124.00
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
125
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0126.png
TABLE A.10: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. VPs with a tertiary
education.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.533***
-0.06
-0.28
-0.15
-0.570**
-0.07
-0.05
-0.21
-0.16
0.04
Number of employees
Value added (DKK1,000)
2
Ste.
0.18
0.21
0.29
0.31
0.29
0.15
0.18
0.25
0.25
0.28
Coeff.
0.48
-0.63
0.27
-1.577**
-1.20
-0.16
-0.37
-0.62
0.50
0.55
Ste.
0.43
0.52
0.51
0.64
1.02
0.31
0.37
0.49
0.60
0.67
Coeff.
206.2
281.6
-432.1
313.6
-1277*
-211.0
-440.3
383.1
294.4
1023**
Ste.
287.9
376.1
399.5
397.5
711.0
259.5
332.4
436.5
489.5
498.8
0.26
0.10
0.26
0.23
0.30
0.02
0.14
0.16
0.17
0.18
0.20
0.23
0.27
0.30
-0.42
0.02
-0.05
-0.01
0.19
0.24
-1.182*
0.37
0.40
0.35
0.41
0.51
0.52
0.63
-87.3
328.3
179.6
198.6
313.0
-320.2
-696.4
272.5
264.6
240.2
279.3
365.7
404.2
473.5
Constant
Number of
observations:
Number of companies:
R-squared
-0.06
1177
251
0.05
0.16
0.577*
1239
257
0.08
0.34
219.8
1186
250
0.05
214.3
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
126
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0127.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
20.51
6.89
-35.96
13.84
-22.74
-20.35
-6.48
-3.69
-9.57
4.62
Labour productivity
(DKK1,000)
6
Coeff.
-79.42
7.22
-413.1**
140.30
-714.2**
-0.79
239.30
301.7*
-6.43
679.1**
Coeff.
146.6
164.1
197.4
296.2
236.7
140.3
148.6
199.4
274.7
343.0
Ste.
146.6
164.1
197.4
296.2
236.7
140.3
148.6
199.4
274.7
343.0
Coeff.
-0.04
-0.06
-0.01
-0.03
-0.167*
-0.03
-0.06
-0.05
-0.04
-0.10
Ste.
0.04
0.05
0.05
0.06
0.09
0.03
0.04
0.04
0.06
0.06
Ste.
21.29
20.56
30.15
31.59
60.93
21.33
19.92
23.96
36.18
29.02
Ste.
162.80
184.90
175.80
237.30
309.20
155.20
154.70
164.50
194.90
299.60
164.9
130.6
126.1
142.1
167.1
189.6
234.3
164.9
130.6
126.1
142.1
167.1
189.6
234.3
-0.06
0.00
0.00
0.02
0.02
0.01
0.03
0.04
0.04
0.04
0.04
0.05
0.05
0.06
10.16
-1.77
6.54
0.26
26.97
-3.31
15.84
12.16
15.68
16.66
16.46
22.13
25.36
26.69
-220.4*
-203.5*
-116.30
-90.73
-204.00
-343.4*
-216.70
127.40
116.70
132.30
151.20
147.50
178.30
182.00
111.3
1179
256
0.04
111.3
0.01
1208
256
0.03
0.03
-0.83
624
157
0.04
12.81
172.20
732
146
0.05
117.50
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
127
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0128.png
TABLE A.11: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. VPs with degrees in
art&humanities
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.40
-0.473*
-0.53
-0.32
-0.53
-0.13
-0.11
-0.04
-0.22
-0.11
Number of employees
Value added (DKK1,000)
2
Ste.
0.33
0.26
0.32
0.69
0.71
0.21
0.23
0.29
0.49
0.36
Coeff.
-0.24
-1.486*
0.33
-0.36
0.96
0.15
-0.24
-0.69
-1.830*
-0.13
Ste.
0.74
0.76
0.68
1.15
1.17
0.42
0.52
0.64
0.98
0.98
Coeff.
-585.4
90.0
-118.4
-588.1
541.7
220.7
-519.9
-241.8
619.9
316.3
Ste.
479.1
532.2
734.4
1190.0
952.4
396.2
485.4
629.0
795.6
759.0
0.16
0.22
0.580**
0.27
0.38
0.18
0.43
0.26
0.23
0.22
0.32
0.32
0.36
0.39
0.90
0.87
0.81
0.84
1.20
1.38
0.00
0.61
0.63
0.68
0.68
0.83
0.89
0.96
78.5
175.4
-59.8
-98.9
128.9
-171.3
-285.9
370.6
307.9
302.1
369.8
439.7
557.0
464.9
Constant
Number of
observations:
Number of companies:
R-squared
-0.16
366
79
0.06
0.21
-0.18
377
80
0.14
0.62
349.3
374
78
0.04
270.5
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
128
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0129.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
26.45
6.66
-5.97
-2.70
0.00
-12.28
-24.61
-15.68
4.38
-16.95
Labour productivity
(DKK1,000)
6
Coeff.
-332.90
-339.00
-293.00
-435.9*
0.00
31.47
392.70
156.80
63.65
495.30
Coeff.
-310.9
225.6
56.4
-1233.0
-250.1
-130.5
-423.7*
-304.1
909.4
341.1
Ste.
241.5
252.0
356.6
849.7
847.8
243.6
215.5
338.8
674.0
703.5
Coeff.
-0.06
-0.07
-0.08
-0.17
-0.19
-0.05
-0.07
-0.07
-0.18
-0.10
Ste.
0.06
0.07
0.08
0.17
0.19
0.05
0.07
0.07
0.18
0.10
Ste.
25.89
27.80
54.10
67.91
0.00
24.17
27.57
43.30
39.65
45.84
Ste.
304.30
432.00
323.30
255.10
0.00
376.00
409.80
361.60
232.50
327.90
18.3
47.2
99.2
143.3
102.2
145.3
76.8
176.9
134.7
147.3
166.6
221.8
260.5
294.6
-0.08
-0.08
-0.08
-0.09
-0.09
-0.10
-0.11
0.08
0.08
0.08
0.09
0.09
0.10
0.11
-3.70
-12.41
-19.93
-17.82
6.56
-16.19
29.41
26.77
32.46
21.77
27.03
33.61
42.94
49.11
-2.09
-11.39
95.89
194.20
195.40
14.77
56.61
351.50
312.40
316.00
393.80
316.70
379.90
392.80
18.2
370
79
0.06
113.1
0.08
365
79
0.08
0.08
6.67
183
44
0.07
21.30
-87.24
224
44
0.06
320.30
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
129
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0130.png
TABLE A.12: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. VPs with degrees in
social sciences.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.32
0.05
-0.12
0.07
-0.16
-0.11
-0.04
-0.67
-0.46
-0.18
Number of employees
Value added (DKK1,000)
2
Ste.
0.26
0.30
0.39
0.43
0.33
0.22
0.28
0.41
0.39
0.45
Coeff.
1.512***
-0.55
0.79
-0.38
0.31
-0.997**
-0.828*
-1.286**
-0.59
-1.65
Ste.
0.52
0.78
0.78
1.01
1.55
0.41
0.45
0.53
0.85
1.17
Coeff.
896.0**
89.0
-333.9
1194*
449.3
-805.9**
-582.5
353.1
-487.7
-22.4
Ste.
389.2
490.5
435.6
661.3
899.2
350.7
447.8
461.1
753.3
921.3
0.01
0.07
0.13
0.18
0.12
0.22
0.12
0.23
0.22
0.21
0.24
0.32
0.36
0.43
-0.45
-0.02
0.07
0.26
0.37
0.67
-0.61
0.39
0.49
0.42
0.50
0.56
0.57
0.67
-506.4
32.3
137.4
228.8
249.2
-470.2
-610.7
362.3
283.6
259.4
342.6
461.9
523.0
668.2
Constant
Number of
observations:
Number of companies:
R-squared
0.10
630
127
0.04
0.18
0.65
658
130
0.09
0.40
444.6*
621
124
0.07
239.0
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
130
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0131.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
-6.09
20.51
-48.35
35.91
33.74
-21.91
9.09
-5.14
-13.95
13.19
Labour productivity
(DKK1,000)
6
Coeff.
-2.87
-155.00
-387.40
297.70
-636.4*
-25.13
321.1**
395.6**
-126.50
581.0*
Coeff.
-31.8
-32.2
-213.2
541.7
24.8
-176.0
-169.5
369.5
-25.2
273.3
Ste.
206.1
204.4
259.8
364.3
390.6
204.1
223.3
261.0
373.0
452.1
Coeff.
-0.06
-0.07
-0.05
-0.09
-0.07
-0.02
-0.04
-0.01
-0.03
-0.09
Ste.
0.04
0.06
0.04
0.08
0.05
0.04
0.05
0.05
0.07
0.10
Ste.
27.41
30.54
32.39
29.57
21.08
26.86
25.73
31.11
43.11
38.30
Ste.
213.40
242.20
239.20
301.30
323.10
148.10
141.70
196.90
183.50
339.10
32.6
172.8
257.8*
152.4
117.9
-212.0
-364.7
160.9
148.7
145.2
168.3
226.0
237.9
317.6
0.0578*
0.0863**
0.0983**
0.07
0.05
0.11
-0.0649***
0.02
0.03
0.04
0.04
0.05
0.05
0.07
-1.98
-5.47
11.68
-3.27
22.46
-4.65
5.39
15.25
18.16
22.17
20.85
28.50
33.52
36.33
-191.30
-73.31
-90.01
-154.90
-257.90
-421.0*
-215.60
131.00
145.50
162.80
157.80
177.00
224.20
218.40
12.3
626
129
0.07
117.2
-0.0649***
638
130
0.03
0.02
0.76
368
85
0.09
15.95
172.10
416
79
0.06
125.10
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
131
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0132.png
TABLE A.13: Comparison between VP-companies and companies in the reference group. Companies
with up to 50 employees in year zero. Diff-in-diff fixed effects regression results. VPs with degrees in
technical sciences.
Dependent
variables
(in first
differences):
Treat=1 & t=1
Treat=1 & t=2
Treat=1 & t=3
Treat=1 & t=4
Treat=1 & t=5
t=1
t=2
t=3
t=4
t=5
Year dummies
2003
2004
2005
2006
2007
2008
2009
Number of highly
educated employees
1
Coeff.
0.516**
0.417*
-0.08
-0.17
-0.17
0.09
0.19
0.09
0.11
0.08
Number of employees
Value added (DKK1,000)
2
Ste.
0.21
0.24
0.30
0.34
0.39
0.17
0.23
0.25
0.31
0.40
Coeff.
-0.11
0.44
0.19
0.05
-2.006***
0.921*
0.35
0.89
0.96
3.272***
Ste.
0.58
0.53
0.72
0.93
0.70
0.51
0.72
0.88
1.06
1.00
Coeff.
-282.6
915.7*
283.7
-103.5
-1962*
745.2**
-154.7
517.3
1157*
2082**
Ste.
383.9
466.0
645.6
904.3
991.6
355.2
441.8
673.9
692.2
858.6
-0.10
0.06
-0.14
-0.20
-0.16
-0.41
-0.31
0.23
0.19
0.18
0.22
0.26
0.29
0.30
-0.01
0.31
0.47
0.32
-0.06
-0.58
-2.893***
0.45
0.52
0.57
0.73
0.79
0.90
1.05
-441.7
579.8
104.6
552.6
207.8
-625.7
-1549**
393.0
443.9
419.3
489.5
538.8
566.5
666.5
Constant
Number of
observations:
Number of companies:
R-squared
0.14
932
187
0.05
0.17
0.30
985
193
0.10
0.45
129.1
926
190
0.09
352.2
Notes: Only observations with annual changes in the number of employees of less than 12. *, **, *** denote statistical significance at the 10%, 5%,
and 1% significance level.
1. Employees with post-secondary or tertiary education. Only observations with annual changes in the number of employees with post-secondary
and tertiary education
<
5.
132
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0133.png
Net income (DKK1,000)
3
Return on assets
4
Wage per employee
(DKK1,000)
5
Coeff.
27.09
-3.78
-19.81
47.24*
-58.75
-29.64
-23.62
-3.06
-57.71*
-20.32
Labour productivity
(DKK1,000)
6
Coeff.
-55.06
230.80
-12.40
-104.30
-63.98
50.05
-8.47
-10.69
63.38
82.76
Coeff.
-168.3
337.1
260.3
134.9
-333.1
413.9***
-6.6
339.9
403.0
701.6**
Ste.
172.7
217.5
206.9
414.8
230.7
150.8
186.6
208.4
339.2
276.1
Coeff.
-0.01
-0.04
0.01
0.01
-0.10
0.00
-0.01
0.00
0.04
0.03
Ste.
0.04
0.05
0.05
0.06
0.10
0.04
0.05
0.05
0.06
0.07
Ste.
19.75
14.09
23.65
27.10
49.85
20.67
19.63
26.29
32.57
52.61
Ste.
130.90
148.80
97.78
205.20
370.80
137.80
160.20
176.70
235.40
352.10
-218.3
35.5
-169.1
-171.8
-260.2
-497.4**
-874.0***
167.4
166.6
144.1
174.6
176.9
220.2
264.0
-0.0717*
0.00
-0.05
-0.03
0.00
-0.09
-0.09
0.04
0.03
0.04
0.05
0.05
0.05
0.07
21.43***
23.56***
15.60*
40.25***
36.60
39.29
39.78
7.97
8.29
8.73
15.14
22.51
28.01
31.08
-203.60
-31.67
-155.00
111.30
-105.00
-143.90
-130.50
123.00
84.93
125.70
154.60
192.80
204.90
246.20
188.7
894
192
0.06
118.9
0.04
977
193
0.03
0.03
-11.02*
557
125
0.04
6.52
76.85
647
119
0.05
95.41
2. Only observations with annual change in the value added of less than DKK 10 mio.
3. Only observations with annual change in net income of less than DKK 3 mio.
4. Only observations with annual change in roa of less than 1, and total assets
>
DKK 100,000.
5. Only observations with number of employees
>
5. Only observations with change in average wage
<
DKK 500,000.
6. Only observations with number of employees
>
5 and change in labour productivity
<
DKK 3 mio.
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
133
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0134.png
PUBLICATIONS
Publications from the Danish Agency for Science, Technology and
Innovation in the series
Innovation: Analysis and evaluation
2013
13/2013
12/2013
11/2013
10/2013
09/2013
08/2013
07/2013
06/2013
05/2013
04/2013
03/2013
02/2013
01/2013
2012
14/2012
13/2012
12/2012
10/2012
09/2012
08/2012
07/2012
06/2012
05/2012
04/2012
03/2012
02/2012
01/2012
2011
20/2011
19/2011
18/2011
17/2011
16/2011
Analyses of Danish Innovation Programmes – a compendium of excellent
econometric impact analyses
An evaluation of the Danish Innovation Assistant Programme
The Effect of the Industrial PhD Programme on Employment and Income
Strategi for samarbejde om Danmarks klynge- og netværkindsats
De skjulte helte – eksportsucceser i dansk industris mellemklasse
An Analysis of the Level of Consistency in the Danish Innovation
Ecosystem
Key Success Factors for Support Services for Cluster Organisations
Performanceregnskab for GTS-net 2013
Kommercialisering af forskningsresultater – Statistik 2012 (Public
Research Commercialisation Survey – Denmark 2012)
Performanceregnskab for Innovationsnetværk Danmark 2013
Productivity Impacts of Business Investments in R&D in the Nordic
Countries – A microeconomic analysis
Erhvervslivets forskning, udvikling og innovation i 2013
Performanceregnskab for innovationsmiljøerne 2013
Evaluering af GTS-instituttet DFM
Evaluering af GTS-instituttet Alexandra
Evaluering af GTS-instituttet Agrotech
Let’s make a perfect cluster policy and cluster programme: Smart
recommendations for policy makers
The Perfect Cluster Programme – Nordic-German-Polish-Baltic project
The impacts of Danish and Bavarian Cluster Services – results from the
Nordic-German-Polish Cluster Excellence Benchmarking
Kommercialisering af forskningsresultater – Statistik 2011 (Public
Research Commercialisation Survey – Denmark 2011)
Performanceregnskab for GTS-net 2012
Performanceregnskab for Innovationsmiljøer 2012
Innovation Network Denmark – Performance Accounts 2012
Clusters are Individuals II: New Findings from the European Cluster
Management and Cluster Program Benchmarking
Erhvervslivets forskning, udvikling og innovation i 2012
Evaluering af innovationsmiljøerne
Access to Research and Technical Information in Denmark
Universiteternes Iværksætterbarometer 2011
Impact Study: The Innovation Network Programme
Clusters are Individuals: Nordic-German-Polish Cluster Excellence
Benchmarking
24 ways to cluster excellence – successful case stories from clusters in
Germany, Poland and the Nordic countries
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
134
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0135.png
15/2011
14/2011
13/2011
12/2011
11/2011
10/2011
09/2011
08/2011
07/2011
06/2011
05/2011
04/2011
03/2011
02/2011
01/2011
2010
12/2010
10/2010
09/2010
08/2010
07/2010
06/2010
05/2010
04/2010
Impact Study of Eureka Projects
Evaluering af GTS-instituttet Teknologisk Institut
Evaluering af GTS-instituttet DBI
Evaluering af GTS-instituttet DELTA
Kommercialisering af forskningsresultater – Statistik 2010 (Public
Research Commercialisation Survey – Denmark 2010)
Performanceregnskab for Videnskabsministeriets GTS-net 2011
Performanceregnskab for Videnskabsministeriets Innovationsmiljøer 2011
Innovation Network Denmark – Performance Accounts 2011
Erhvervslivets Outsourcing af FoU
Evaluering af GTS-instituttet FORCE Technology
Evaluering af GTS-instituttet Bioneer
Evaluering af GTS-instituttet DHI
Erhvervslivets forskning, udvikling og innovation i 2011
Økonomiske effekter af erhvervslivets forskningssamarbejde med
offentlige videninstitutioner
Analysis of Danish innovation policy – The Industrial PhD Programme
and the Innovation Consortium Scheme
Brugerundersøgelse af GTS-institutterne 2010
Universiteternes Iværksætterbaromenter 2010
Performanceregnskab for Videnskabsministeriets Innovationsmiljøer 2010
Innovationsnetværk Danmark – Performanceregnskab 2010
Performanceregnskab for Videnskabsministeriets GTS-net 2010
Kommercialisering af forskningsresultater – Statistik 2009
InnovationDanmark 2009 – resultater og evalueringsstrategi
Effektmåling af videnpilotordningens betydning for små og mellemstore
virksomheder
03/2010 An Analysis of Firm Growth Effects of the Danish Innovation
Consortium Scheme
02/2010 Erhvervslivets forskning, udvikling og innovation i Danmark 2010
01/2010 Produktivitetseffekter af erhvervslivets forskning, udvikling og
innovation
An evaluation of the Danish Innovation Assistant Programme - En effektmåling af Videnpilotordningen
135
 UFU, Alm.del - 2015-16 - Endeligt svar på spørgsmål 168: Spm. om ministeren vil fremsende relevante evalueringer af Globaliseringspuljen 2007-12, til Uddannelses- og forskningsministeren
1647752_0136.png