Hamilton (2007) therefore recommend longitudinal research designs as the only designs th a t offer the appropriate insights into the growth change process. The
4.3 Data Generation, Collection and Analysis
4.3.1 Data sources
A proprietary dataset was initially developed in M icrosoft Excel and then
transferred to SPSS20 for quantitative analysis. The dataset contains perform ance
variables constructed from eight years financial inform ation for all firms in the
cohort. Both profitability and share value inform ation was gathered for th e tw o
years preceding state investm ent to establish a base line or pre-investm ent
perform ance measure. The year o f the state investm ent was treated as year zero
(the tre a tm e n t intervention year). This was necessary to create a break betw een
th e 'before and after' perform ance measures and so develop an 'interrupted t im e -
series logic' (Yin, 2009). Five years post investm ent data was also collected from the
a n n u a l a c c o u n ts ( t h e p ost - t e s t m e a s u r e ) - i.e. T h e v a lu e o f t h e s h a r e h o l d e r fu n d s
on t h e b a la n c e s h e e t at y e a r e n d a n d also t h e a f t e r t a x p r o f it f o r t h e y e a r w a s
e x t r a c t e d f r o m t h e p r o f it a n d loss a c c o u n ts . T h e d a t a s e t also c o n t a in s t h e s a lie n t
g e o - d e m o g r a p h i c v a ria b le s f o r all 5 1 fi r m s in t h e s tu d y (See: C h a p t e r 5 f o r d e ta ils ).
I n f o r m a t i o n on ea ch o f t h e p r o p o s e d e x p l a n a t o r y v a ria b le s w a s g a t h e r e d f r o m
v a rio u s so u rce s such as t h e F A M E d a t a b a s e , C o m p a n ie s R e g is tra tio n O ffic e (CRO),
V is i o n n e t , w o r l d w i d e w e b , E n t e rp ris e Ir e la n d A n n u a l R e p o rt s ( 1 9 9 8 - 2 0 1 1 ) a n d t h e
i n d iv id u a l f i r m w e b s i t e s . O v e r a ll t h e p e r io d u n d e r in v e s tig a tio n w a s 1 9 9 7 - 2 0 1 0
w h e n t h e 'b e f o r e a n d a f t e r ' m e a s u r e f o r e a c h f i r m is in c lu d e d . All fir m s in t h e
d a t a s e t w e r e cli en ts o f E n t e rp ris e Ir e la n d a n d h ad r e c e iv e d a t least € 6 3 5 , 0 0 0
( I R £ 5 0 0 , 0 0 0 ) o f p ub lic v e n t u r e c a p ita l i n v e s t m e n t in o n e o f t h e y e a rs 1 9 9 9 - 2 0 0 5
( E n t e r p r i s e Ir e la n d a n n u a l re p o rts : 1 9 9 9 - 2 0 0 5 ) .
Table 4.2 - Firm sector breakdown
In d u s tr ia l S e c to r N o . o f % o f % o f V a lu e
F irm s t o t a l in v e s t m e n t
€'000
C o n su m er prod ucts - F u rn itu re /c e ra m ic /c a rp e t
3 6 9
4591
Food and n a tu ra l resources - A g rip ro d u c ts /c o n s u m e r 7 15 20 1 0 089
fo o d s /n a tu r a l resources
C lea n te ch , m ed ic al devices an d in dustrial products
25
“ “ “S o ftw a re , ICT and in te rn a tio n a lly tra d e d services 29 54 45 22 6 5 2
... ..._ ' ‘ V
T o tal 51 100% 100% 50376
(Source: Enterprise Ireland, Fame database, Visionet, CRO, Firm websites)
4 .3 .2 Data analysis techniques
B in o m ial Logistic regression
The logistic function (the dependent variable) is particularly useful as it can take as
input any value from negative infinity to positive infinity whilst outputting values
betw een zero and one (Garson,2012). This is th e most appropriate model here -
once th e desired outcom e is an estim ation o f w h e th e r shareholder value creation in
preferable to shareholder value destruction (Arnold, 2007). The study is also
interested in w h e th e r th e states venture capital will be repaid and so this will also
be of interest as an altern ate dependent variable. The outcom e categories can then
be expressed as a dichotomous variable - was value created over the five years post
state investm ent period (1) or was value destroyed (0). W hilst return on invested
capital could in some circumstances be m odelled as a discrete or continuous
variable, it needs to be m odelled in binary fashion in this case as some o f the firms
in the study had returned negative profit figures year on year which fu rth er
exacerbated th e shareholder value decrease over tim e. This m eant th a t it was not
possible to obtain meaningful Return on Invested Capital (ROIC) figures and thus a
dichotomous variable is the only suitable choice in capturing the value
creation/decrease construct in the cohort of firms in the study.
The outcom e variable in binary Logistic regression differs from OLS regression in
th a t is expressed in probabilistic rather than numeric term s and its outcomes needs
to be interpreted differently (Garson, 2012). Since the probability of an event must
lie betw een zero and one, it is impractical to model probabilities w ith linear
regression techniques because the linear regression model allows the dependent
variable to take values greater than one or less than zero (Collett, 2003).
Cross-case analysis
Case analysis is one o f the most popular research designs in th e social sciences (Yin,
2009) and th e international business and m anagem ent fields (Piekkari & W elch,
2011). W hilst case study design has traditionally been associated w ith qualitative
research it has much w ider application and can incorporate both qualitative and
quantitative elem ents w ithin an overall design.
Chapter six and seven in this study are cross-case analyses. Q uantitative and
qualitative data was collected on ten firms from th e cohort o f firms in the study.
The resultant analyses w ere then w ritten -u p as descriptive case studies using
Storey's (1994) and Smallbone & W yer's (2006) fram ew ork. (See: Chapter 2:
Literature review ). The purpose was to identify possible determ inants of and
influences on the growth trajectories and growth experience of each firm . Data
from these cases was then utilised in th e cross-case analyses in Chapter six and
seven and also as input to the contribution analysis in chapter eight. Since th ere
w ere only fifty one firms in the study, it was appropriate to use a case study
approach. The case study analysis also provides corroborative m aterial fo r the
quantitative findings in Chapter five. The prim ary data used in the case studies was
collected through sem i-structured depth interviews w ith current or ex-CEO's o f the
case firms - the key inform ant's (Marshall, 1996; Fletcher & Plakoyiannaki,
2011).This data was supplem ented w ith archival inform ation, inform ation from the
firm 's literatu re and digital assets and reported inform ation in the media. These
provide th e m ultiple sources o f evidence suggested by Yin (2009; Chap. 4).
Contribution analysis
Chapter eight com pletes th e empirical analysis in th e overall study by conducting a
contribution analysis (M ayne, 2001). This is a theory based impact evaluation
m ethodology (TBIE) (W h ite, 2010). It is a structured iterative analytic technique
which looks at th e 'theory o f change' proposed by th e policy instrum ent under
analysis. It take the evidence assembled in chapters five, six and seven and conducts
a meta-analysis to answer - as definitively as possible - th e research questions
posed and objectives set at the outset of th e study. Blaney and McKenzie (2007)
make a distinction in TBIE betw een those approaches which are 'realist evaluations'
(Pawson & Tiley, 1997) and those approaches th a t develop an explicit program m e
th eo ry o f change - (Chen, 1990; Weiss, 2000; M ayne, 2001). The approach which
has gained in popularity since it was first proposed is M ayne's 'Contribution
Analysis' which developed from his w ork on results m onitoring systems. It was
developed whilst he was considering w hat could be said about causality o f an
intervention w hen only m onitoring (weak) data was available (M ayne, 2012). W h at
distinguishes Contribution Analysis from oth er theory-based approaches in
evaluation is its m ore systematic approach to arriving at creditable causal claims.
M ayne (2012) notes:
From an evaluation perspective, the issue was what could be done to make credible causal claims in the absence of experimental approaches. Many evaluations seemed either to be silent on causality or, perhaps worse, made causal claims based solely on the views of interviewees (p.271).
The objectives articulated by M ayne agree in principal w ith both Storey's (2000) and
th e OECD's (2008) approach on impact assessment. How ever M ayne is more
118 |P a g e
pragm atic in recognizing th e lim itation s on data availability and the difficulties in
creating creditable counterfactuals and thus estim ating 'additionality' - particularly
w hen myriad influencing and determ ining factors are considered. The aim o f the
analysis is to reduce uncertainty about th e 'contribution' th a t the intervention is
making to observed results through an increased understanding of w hy results did
or did not occur and the roles played by the intervention and oth er influencing
factors. In sum th e analysis eith er confirms the postulated 'theory o f change' or
suggests revisions to the theory w here results prove otherwise. M ayne (2012) notes
th a t a 'contribution analysis' will rarely provide definitive proof. Causality is
provided in probalistic term s. The six stage process is an iterative process which
builds a chain o f evidence and argum ent to get to a conclusive situation w here
'plausible association' does or does not exist (Hendricks, 1996). CA's six stage
process which can be tailored for specific policies or programmes in differing fields
(Delahais & T o u le m o n d e , 2012; Wimbush e t a i , 2012; Lemire e to i., 2012). The
generic six stage process is as follows: