Content uploaded by Seamus Mcguinness
Author content
All content in this area was uploaded by Seamus Mcguinness
Content may be subject to copyright.
For Peer Review
Assessing the Impact of Skill Shortages on the Productivity Performance
of High-Tech Firms in Northern Ireland
Journal: Applied Economics
Manuscript ID: APE-06-0419
Journal Selection: Applied Economics
JEL Code:
J24 - Human Capital|Skills|Occupational Choice|Labor Productivity
<J2 - Time Allocation, Work Behavior, and Employment
Determination/Creation < J - Labor and Demographic Economics,
J30 - General < J3 - Wages, Compensation, and Labor Costs < J -
Labor and Demographic Economics, J31 - Wage Level, Structure;
Differentials by Skill, Occupation, etc. < J3 - Wages, Compensation,
and Labor Costs < J - Labor and Demographic Economics
Keywords: Skill Shortages , Productivity, High-tech Industries
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
peer-00582157, version 1 - 1 Apr 2011
Author manuscript, published in "Applied Economics 41, 06 (2009) 727-737"
DOI : 10.1080/00036840601007450
For Peer Review
Assessing the Impact of Skill Shortages on
the Productivity Performance of High-Tech
Firms in Northern Ireland
Page 1 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
Abstract
This paper utilises data from three separate skill related surveys of firms in the
Northern Ireland IT, Electronic Engineering and Mechanical Engineering industries in
order to assess the extent to which the performance of high-tech firms are being
constrained as a result of hard-to-fill and / or unfilled vacancies. Whist it was found
that the determinants of skill shortage varied somewhat depending upon the
definitional approach adopted, a high degree of correlation was found. With regards
to the impacts of skill shortages on firm level performance, it was found that both
hard-to-fill and unfilled vacancies had reduced output per worker levels by between
65 and 75 per cent in affected firms, however, these impacts were only detectable
after controlling for selection effects. The evidence suggests that standard OLS
procedures can generate highly misleading results in studies of this nature.
Key Words: Skill Shortages, Productivity, High-tech Industries
JEL Codes: J24, J30, J31
Page 2 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
1
Introduction
Skills related issues are now centre stage in UK economic policy, having been
identified as one of the five drivers of productivity performance. A critical aspect of
government policy has been the attempt to alleviate skill shortages within key sectors
of the economy which have led to the establishment of Sector Skills Councils which
are charged with the responsibility of reducing skills gaps and shortages within their
respective sectors. Thus, policy is underpinned by the assumption that skills
shortages within the economy have serious negative impacts on productivity
performance. There are certainly a number of reasons why we might expect skill
shortages to adversely impact productivity levels, for instance, firms may be forced to
lower their recruitment standards and fill positions with less productive workers, and /
or workers in the affected occupations may exploit their bargaining power to
disproportionately enhance their employment conditions; either way, we would expect
the level of output per worker to fall. In the event of skill shortages becoming a wide
spread phenomenon, it has been argued that the economy will move towards a low
skills, bad jobs, lower wages equilibrium trap, characterised by persistent low
productivity levels, under investment in training and few skilled jobs (Snower, 1996;
Redding 1996). However, UK evidence supporting such views is limited and
somewhat mixed, a factor which is at least partially related to a lack of available
datasets linking skill shortages and firm level performance. Forth & Mason (2004)
using the 1998 Technical Graduates Employers Survey (TGES) combined with Dun
&Bradstreet data found no clear differences in the sales per employee levels of
companies that experienced recruitment difficulties compared to those that did not.
Similarly, McGuinness & Bonner (2002) failed to find a link between unfilled
vacancies and productivity levels within a sample of Northern Ireland (NI) IT firms,
Page 3 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
2
whilst McGuinness & Doyle (2005) came to the same conclusion in relation to the NI
construction industry. Nevertheless, there does exist some evidence to support the
view that skill shortages incur real costs on the firm. Haskel and Martin (1993a)
report that the increased skill shortages during the mid-1980s reduced productivity
growth by around 0.7 per cent per annum whilst Nickell and Nicolatsis (1997) found
that skilled labour shortages at industry level were significantly and negatively related
to a number of productivity corollaries such as investment in fixed capital and R&D.
This paper utilises a dataset constructed using information collected during three
separate skill related surveys of firms in the NI IT, Electronic Engineering and
Mechanical Engineering industries in order to assess the extent to which the
performance of high-tech firms are being constrained as a result of skill shortages.
Data and Methods
The data was collected by the Priority Skills Unit (PSU) of the Economic Research
Institute of Northern Ireland (ERINI) who work under the direction of the NI Skills
Task Force, the body tasked with identifying sectors of importance to the NI economy
that are potentially being constrained as a result of labour market shortages. The data
consists of a merged dataset comprising information collected in three separate
surveys of the electronic engineering sector (surveyed 2000), the mechanical
engineering sector (surveyed 2001) and the IT sector (surveyed 2002). In each of the
studies, the inter-departmental business register (IDBR) provided the sampling frame.
Each sample was skewed towards the larger firms in order to ensure the largest degree
of employment coverage; however care was taken to ensure that sufficient numbers of
Page 4 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
3
small and medium sized firms were also included to ensure the representativeness of
each sample. Given that the surveys were collected at slightly different points in time
and related to different sectors, the information contained within each dataset was not
fully compatible, however, whilst some data was lost during the merger process, we
were able to retain a lot of information on both skill composition and company
performance
1.At the end of the data reconciliation process we were left with information on 242
high-tech firms employing 39,328 workers. In terms of the employment breakdown
11 per cent of the sample was employed in the IT sector, 40 per cent in the electronic
engineering sector and 49 per cent in the mechanical engineering sector. The
achieved sample is broken down by firm size in Table 1 and the dominance of large
firms becomes wholly apparent with the largest 20 per cent of firms accounting for
over 80 per cent of employment. However, it is important to note that the distribution
of employment in Table 1 is consistent with the observed facts for each of the sectors
and, as such, the samples reflect the respective industry structures.
As stated, the IDBR provided the populations from which the respective samples were
drawn with each sample selected on the basis of the Standard Industrial Classification
1992 (SIC92). The electronics sector was principally defined as those firms operating
within divisions 30, 32 and 33. From our final sample of 125 firms we were left with
valid responses for 63 firms which gave us a survey response rate of 50.4 per cent, of
whom 43 were located in SIC groups 30, 32 and 332. On the basis of our 43
respondent IDBR firms, we had a population coverage rate of 99.5 per cent suggesting
that official sources had seriously underestimated the size of the NI sector. The IT
sector was defined as consisting of the computer services industry (division 72 of the
Page 5 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
4
SIC 92). A sample was drawn of 121 firms from which a total of 77 firms employing
2,888 agreed to participate in our study giving a survey response rate of 76 per cent
(measured in terms of employment) and a population coverage rate of 56 per cent3.
Finally, the mechanical engineering industry was defined in terms of the
amalgamation of five two-digit sectors (SIC 28, SIC 29, SIC 31, SIC 34, and SIC 35)
where demand for mechanical engineering skills tends to be most heavily
concentrated. Our final sample consisted of 160 firms employing 20,581 persons
from which a total of 98 firms employing 16,537 agreed to participate in the study
giving a survey response rate of 80 per cent and a target population coverage rate of
77 per cent.
Turning to the key variable within the dataset, it is important to note that the notion of
askills shortage is subject to substantial ambiguity with different surveys using
different methodologies, terminologies and phraseologies. The official definition
from the Department for Education and Skills (DfES) defines skill shortage vacancies
as a “situation where there is a genuine shortage in the accessible external labour
market of the type of skill being sought, and which leads to a difficulty in
recruitment”. This rather ambiguous definition reflects the fact that there is no clear
cut approach to measuring skill shortages and generally, studies have tended to use
data on either unfilled vacancies or some measure of hard-to-fill vacancies. However,
it is far from certain that both measurement approaches will yield similar results given
that the evidence would tend to suggest that they have different determining factors.
This study has the advantage of containing information on both unfilled and hard-to-
fill vacancy data and as such both measures are tested against productivity as
measured by the log of output per worker.
Page 6 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
5
<Insert Table 1 Here >
The Incidence of Skill Shortages
Our first measure of skill shortage relates to hard-to-fill vacancies. Firms were asked
to assess the degree of difficulty associated with recruiting staff within a set of
occupational categories that could be standardized across the various sectors. It
should be noted that this, along with the unfilled vacancy rate, is the standard measure
of skill shortage used within the UK Skills Monitoring Surveys. As we would expect,
the level of recruitment difficulty increases the more experienced and qualified are the
staff concerned (Table 2). Just under 70 per cent of firms employing graduates with
over two years experience, project leaders and senior managers described the
recruitment of such staff as being difficult or very difficult. This is in contrast to the
labour market for new graduates (those with no previous experience), which appears
to be relatively well supplied with only 24 per cent of firms employing such staff
perceiving the labour market to be tight. For all other occupational categories,
recruitment appears much less problematic with over 50 per cent of employers
describing the process as easy or very easy. In terms of the sectoral breakdown the
pattern is generally similar however, some differences are observed. Whilst over half
of electronics firms believed the recruitment of new graduates to be problematic, less
than 15 per cent of IT and mechanical engineering respondent firms felt that
recruitment at this level was difficult (Table 3). The recruitment of craft and
technician level staff was much less of a problem for firms operating in the IT sector,
with 16 per cent and 10 per cent of firms respectively describing this process as
Page 7 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
6
difficult or very difficult compared to over half of firms in electronic and mechanical
engineering sectors who described these labour markets as tight.
Whilst hard-to-fill vacancies is a useful measure of labour market difficulty it can also
be problematic as such an approach cannot provide us with a concrete empirical
measure of shortage nor is it possible to control for sources of subjective bias arising
from differences in how respondents define a term such as “difficult”. Consequently,
information was also collected on the number of unfilled vacancies occurring in the
twelve months preceding our surveys, thus enabling us to generate estimates of the
actual rate of shortfall occurring within each of the various occupational groupings.
Of the 242 respondent firms, 60 firms reported a total of 420 unfilled vacancies in the
twelve months prior to the surveys (Table 4), with 76 per cent of these vacancies
occurring at below graduate level. In terms of raw numbers, recruitment problems
were most acute at the operator and craft levels. When respondents were asked to
attribute these unfilled vacancies to a range of problems, it was obvious that a lack of
skilled and qualified workers, as opposed to an inability to pay were the key drivers of
unfilled posts. Disaggregating the results by sector, some differences are observable
with a lack of technical ability as the most important factor in explaining unfilled
vacancies in both the electronics and IT sectors, whilst a shortage of qualified
applicants was cited as the most significant problem for firms in the mechanical
engineering sector4.Thus it is clear from this descriptive analysis that the variable
used here is capturing the effects of skill shortage and if it transpires that the hard-to-
fill and unfilled vacancies variables are highly correlated, this will also provide
support for the view that hard-to-fill vacancy information is a good proxy for skill
shortages5.
Page 8 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
7
We then proceeded to calculate actual vacancy rates6and these were derived for each
occupational aggregate by standardising the number of unfilled vacancies in each
category by the number of persons employed in that category. The highest rate of
shortage, at 4.5 per cent, was recorded for experienced graduates (more than two
years experience) suggesting that the recruitment of staff to this level has been
particularly difficult, a finding broadly consistent with the more subjective measure.
To a lesser extent, problems were also experienced in the recruitment of technician
level staff (HNC/D) and new graduates with these categories recording an unfilled
vacancy rate of 3.2 and 2.8 per cent respectively. Table 5 shows the comparative
vacancy rates for each sector. The rate of shortfall for technician level staff is similar
in both the electronic and mechanical engineering sectors, however the similarity ends
here. Vacancy rates for all of the occupational groupings are much higher in the
electronics sector with difficulties particularly acute for graduates with more than two
years experience. Electronics vacancy rates are also very high for new graduates and
craft level staff at 12.1 and 10.4 per cent respectively.
<Insert Table 2 Here >
<Insert Table 3 Here >
<Insert Table 4 Here >
<Insert Table 5 Here >
<Insert Table 6 Here >
Whilst both measures of skill shortage yield generally similar patterns, it is unclear to
what extent these key measurement variables are correlated and subject to the same
Page 9 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
8
driving factors. To investigate this matter further we modeled the likelihood of hard-
to-fill and unfilled vacancies using standard econometric analysis. A composite
measure of hard-to-fill vacancies was derived by attributing the level of assessed
recruitment difficulty for each non assembler grouping a value between 1 and 47and
then calculating the average for each firm. If the firm reported an average recruitment
difficulty level of greater than three then they were deemed to have been affected by
hard-to-fill vacancies and the relevant variable was assigned the value one (and zero
otherwise). In relation to the unfilled vacancy measure, the dependent variable takes
the value 1 if the firm experienced an unfilled vacancy in the previous 12 months, and
zero otherwise8. As both variables are binary in nature and likely to be highly
correlated we estimate a bivariate probit model, the use of binary variables is also
useful as they will facilitate the estimation of selection controls within the
productivity regressions. In terms of explanatory variables, we use a number of firm
level characteristics which include employment structure, location, ownership, salary,
age, size, R&D and sector.
Table 7 gives the results from the Bivariate probit and the first thing to note is that the
model yields a positive and significant correlation coefficient indicating, not
surprisingly, that that firms affected by shortage are likely to report both unfilled and
hard-to-fill vacancies. In relation to the explanatory variables both measures of
shortage share a number of determining factors. The coefficient on the wage variable
and the female graduate share of total employment are both positive and highly
significant. The wage impact is in line with the findings of Dickerson (2003), who
reported that higher local relative wages are associated with both higher vacancy
incidence and higher vacancy propensity. In contrast to this, Green et al (1998) find
Page 10 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
9
that in higher wage establishments there is a reduced probability of having skill
shortages or hard-to-fill vacancies, whilst Haskel & Martin (2001) observed a
significant negative effect on the wage variable on hiring difficulties9.In relation to
female employment structure, there is no obvious explanation; however, it may be the
case that as females are more likely to exit the labour market due to family reasons,
this may in turn lead to higher levels of recruitment activity and consequently skill
shortages. Our results also demonstrate that firms employing a lower proportion of
new and inexperienced graduates will be more likely to report unfilled vacancies.
This implies that firm level preferences and Human Resource Management (HRM)
practices are a significant driver of skill shortages, specifically, those firms setting
higher entry requirements and thus recruiting a relatively low proportion of new and
inexperienced graduates are the firms most likely to have vacancies that remain
unfilled. Therefore, firms, by restricting their available pool of labour, through low
levels of new graduate utilization, are leaving themselves more open to unfilled
vacancies; McGuinness & Bonner (2002) also report this result. Finally, in keeping
with our more descriptive analysis, skill shortages were found to be more common
within the electronics sector.
Anumber of variables are unique to one or other of the measures of shortage,
however, this was particularly the case within the unfilled vacancy model. The
location variable was significant indicating that firms located in central Belfast, the
main urban conurbation in Northern Ireland, are less likely to suffer from unfilled
vacancies. This is perhaps unsurprising given that the regions universities are both
located in the Belfast area thus providing increased access to a source of highly
skilled and educated labour. Firms reporting unfilled vacancies were also more likely
Page 11 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
10
to view new labour market entrants from the regions educational and training
institutions as being deficient in either, technical, interpersonal or business awareness
skills. This negative perception is likely to influence the recruitment activities of
firms and specifically, may result in firms restricting their available labour pool by
holding out for more experienced staff which in turn increases the likelihood of an
unfilled vacancy occurring. Again this result is consistent with the findings of
McGuinness & Bonner (2002). Therefore, to conclude, whilst there are many
similarities in the determinants of both measures of skill shortage, some differences
are also present. On the grounds that both hard-to-fill and unfilled shortages may
impact firms’ behavior in different ways, it is also possible that both variables will
generate distinct impacts with respect to productivity and therefore, whilst potentially
correlated, they cannot be thought of as interchangeable.
<Insert Table 7 Here >
Firms reporting unfilled vacancies were then asked to outline the main difficulties
arising as a result of these recruitment shortfalls. The impact on company
performance was believed to be very important with the majority of firms reporting
that their development was being constrained in one or more ways (Table 8). Over
half of affected firms asserted that their organizations ability to meet deadlines was
severely impeded with a further 47 per cent of firm’s experiencing lower productivity
as a direct result of unfilled vacancies, thus providing preliminary evidence of a link
between skill shortages and output. Reduced credibility and higher running costs
Page 12 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
11
were also reported to be a direct result of skill shortages. Generally speaking, the
results for the individual sectors broadly reflected those reported in Table 8.
<Insert Table 8 Here >
Finally, in terms of the policies introduced by firms when attempting to alleviate skill
shortages (Table 9), the majority of affected firms did appear to have undertaken
some form of longer-term strategy. Sixty-three per cent of firms were prepared to
increase salary, with a further 58 per cent choosing to train and up-skill their existing
staff to fill these positions. Forty-seven per cent of firms were also willing to recruit
staff from other backgrounds and train with a third of firms stating that they would be
willing to adjust internal structures and practices. Therefore, in addition to the
obvious response of increasing wages, additional training and flexible HRM practices
appear to have a very important role in combating skill shortages within the high tech
sectors. In relation to the sub-sectoral analysis, the up-skilling of existing staff was
the most common reaction within the IT and electronics sectors whilst the offering of
higher wages was the dominant strategy within mechanical engineering10.
<Insert Table 9 Here >
Productivity Impacts
Whilst the descriptive analysis indicates that just under 50 per cent of firms incurring
unfilled vacancies believe that their productivity has been adversely affected, it is not
always the case that such opinions are validated when tested within the context of
econometric models (McGuinness & Bonner (2002)). Consequently, OLS models
Page 13 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
12
were estimated to assess the productivity / skills shortage relationship more formally
both in the context of unfilled and hard-to-fill vacancies11.On the grounds that both
variables were correlated and potentially highly endogenous with respect to each
other, separate models were estimated on efficiency grounds. As before a
parsimonious approach was adopted. Under both models it was found that
productivity was lower the higher the R&D intensity of the firm, however, output per
worker was higher if the firm was R&D intensive and large (Tables 10 and 11). Thus
the interaction term indicates higher returns to scale in relation to R&D investment
whereas the negative effect may indicate that many small firms within the sectors, all
of which were experiencing rapid growth at the time of survey, were still very much
in the investment stage of the R&D cycle. Externally owned firms were found to have
higher productivity within both models. Over and above the R&D and foreign
ownership effects, the models yielded slightly different results with the model
containing hard-to-fill vacancies somewhat better specified than that including
unfilled vacancies. Within the hard-to-fill vacancies specification firms less than two
years old had higher productivity whilst a higher proportion of female graduates
tended to lower output; within the unfilled vacancy model there was some evidence to
suggest that firms employing a higher proportion of new and inexperienced graduates
were less productive. The finding relating to new and inexperienced graduates is
consistent with the findings of McGuinness & Bonner (2002) and would perhaps
explain why firms are apparently willing to exclude this type of labour from their
recruitment pool despite the fact that such a strategy will lead to unfilled and hard to
fill vacancies (Table 6). The finding in relation to female graduates is slightly more
contentious as it tends to suggest substantially lower rates of human capital
accumulation amongst female professionals due, for example, to more demanding
Page 14 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
13
family commitment or periodic absences from the labour market. However, most
significantly, despite the fact that almost half of firms (47 per cent) with unfilled
vacancies reported that their productivity levels were being adversely affected, no
statistical relationship was established. Conversely, we found that firms with hard-
to-fill vacancies had higher productivity12.This result runs contrary to ex ante
reasoning and previous studies in this area and, given the early descriptive evidence of
Table 9, may lead us to believe that the HRM policies and organizational restructuring
introduced to combat hard-to-fill vacancies have had the additional impact of
improving productivity performance.
However, we also know from the bivariate probit that the probability of a firm
experiencing skill shortages is not randomly distributed and, if this proves to be the
case with respect to productivity, the results observed in Tables 10 and 11 may be
biased. Given the lack of any obvious instrument for skill shortages, treatment
models were estimated following the framework developed by Heckman (1979). The
adopted methodology gets around the problem of selection bias by estimating a two
stage model; stage 1 involves a probit model describing the key characteristics of
firms experiencing skill shortages whereas stage 2 involves the productivity OLS
augmented with a control term, drawn from stage 1, that explicitly accounts for any
differences in the characteristics of firms experiencing skill shortages. Assuming the
error terms from both models are drawn from a bivariate normal distribution we can
derive the following:
1111
(/ , 1) .
ii i
ii
EP x m x
==
+
(1)
Page 15 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
14
Where
i
P
measures firm level productivity,
i
m
is a dummy variable indicating firm
level unfilled / hard-to-fill vacancies,
1
x
is a vector of explanatory variables,
is the
correlation coefficient between the stage 1 and stage 2 error terms,
1
is the standard
deviation of the probit model and
i
is the Inverse Mills Ratio (IMR) given by:13
22 2
22 2
(/)
(/)
i
i
i
x
x
=(2)
Where
and
are density and distribution functions of the standard normal
distribution and
2
is the standard deviation of the OLS. In order to ensure that the
models were properly identified (see Himler, 2001) we used the evidence from the
bivariate probit and the parsed productivity regressions to select at least one variable
that, whilst determining the likelihood of a shortage occurring, had no impact on
productivity. Consequently, the electronics variable was omitted from the unfilled
vacancy model whilst the electronics and perceived shortages variables were omitted
from the hard-to-fill treatment models14.
The results of both treatment models are reported in Tables 12 and 13 with the results
generally consistent with the bivariate probits and OLS models reported earlier with
the critical exception of the skill shortage impacts within the selection consistent
productivity models. The results reveal that firms experiencing skill shortages, of
whatever kind, have ex ante productivity levels approximately 50 per cent higher than
the average and that these skill shortages substantially reduce productivity levels by
65 per cent in the case of hard-to-fill vacancies and 75 per cent in the case of unfilled
vacancies. The analysis thus provides weight to the assertion that skill shortages incur
Page 16 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
15
real costs on firms within key economic sectors and demonstrate that a failure to
account for non-randomness in the incidence of skill difficulties can produce highly
misleading results.
Summary and Conclusions
This paper sought to examine the extent and nature of both hard-to fill and unfilled
vacancies amongst a group of high-tech firms and the extent to which such skill
shortages impacted on productivity performance. Not surprisingly, problems of skill
shortage were more acute for experienced graduates across each of the technical
disciplines with firms attributing unfilled vacancies to a combination of supply side
shortages and the existence of skill gaps amongst applicants. Consistent with the
finding of previous research the results indicate that the determinants of skill shortage
vary depending upon the definitional approach adopted, nevertheless, a high level of
correlation was found. In relation to the impacts of skill shortages on firm level
performance, it was found that ex ante high productivity firms were much more likely
to experience skill shortages and that such labour market constraints had the effect of
reducing productivity levels by between 65 and 75 per cent. The results suggest that
skill shortages can almost completely eradicate the productivity advantage of the best
performing high-tech firms. However, on the grounds that the highest rates of
shortage were found for experienced professionals it is not clear that the solution to
the problem is be found by simply expanding higher education supply, this is
particularly the case given that many firms appear to actively desist from employing
new graduates and are of the view that they are not being adequately equipped with
the requisite skills at university. In support of the skills gap hypothesis there was
some evidence of a productivity cost associated with employing large numbers of new
Page 17 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
16
graduates. Finally, from a methodological perspective, the results highlight the
importance of controlling for selection effects when estimating the impact of labour
market constraints on firm level performance.
<Insert Table 10 Here >
<Insert Table 11 Here >
<Insert Table 12 Here >
<Insert Table 13 Here >
1All financial information within the study is expressed in 2002 prices.
220 of our respondent firms were added by key informants and were classified within these SIC groups
by IDBR.
3In addition to the IDBR firms, a further 4 firms employing 200 persons were added to the sample by
key informants giving us a total of 81 IT firms for whom we have full information.
4Results available from the authors.
5Unfortunately a question was not asked on the reasons behind unfilled vacancies.
6The unfilled vacancy rate is calculated as (unfilled vacancies / employment + unfilled vacancies)*100
7Where 1 = Very Easy, 2 = Quite Easy, 3 = Difficult and 4 = Very Difficult.
8This variable was not restricted to professional level vacancies as only a few firms’ reported unfilled
vacancies at assembler level only; however, when the sample is restricted all subsequent findings hold.
9The authors point out that this finding was not detected in their previous studies, for example, Haskel
&Martin (1993b).
10 Results available from the authors.
11 Due to incomplete information the sample size was reduced to 227. This consists of 162 firms who
provided turnover information and a further 65 whose financial data was obtained from databases held
within ERINI. All information is given at 2002 current prices.
12 Given that unfilled vacancies and firms perceived hard-to-fill vacancies may be correlated and thus,
in addition to being technically inefficient, may have been canceling each other out, we re-estimate the
regression excluding one variable alternatively and we find our results are still robust (results available
from authors).
13 It should be noted that equation five was estimated using a maximum likelihood procedure which
solves the model simultaneously as opposed to adopting a two-step process.
14 The models were estimated in Stata using maximum likelihood.
Page 18 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
17
References
Department for Education and Skills, (2004), National Employers Skills Survey 2003:
Key Findings
Dickerson, A. (2003) The Distribution and Determinants of Job Vacancies: Evidence
from the 2001 Employers Skill Survey, Institute for Employment Research, University
of Warwick. Paper Presented at the Royal Economic Society Annual Conference,
2003.
Forth, J. & Mason G. (2004), The Impact of High-Level Skill Shortages on Firm-
Level Performance: Evidence from the UK Technical Graduate Employers Survey,
National Institute of Economic and Social Research, Discussion Paper No. 235
Green, F., et al. (1998), The Meaning and Determinants of Skills Shortages, Oxford
Bulletin of Economics and Statistics, Vol. 60, No. 2, pp. 165-187. Blackwell
Publishers Ltd.
Haskel, J. & Martin, C. (1993a), Do Skill Shortages Reduce Productivity? Theory and
Evidence from the United Kingdom, The Economic Journal, Vol.103 (417), pp. 386-
394.
Haskel, J. & Martin, C. (1993b), The Causes of Skill Shortages in Britain, Oxford
Economic Papers, New Series, Vol.45, No.4, pp.573-588.
Page 19 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
18
Haskel, J. & Martin, C. (2001), Technology, Wages and Skill Shortages: Evidence
from UK Micro Data, Oxford Economic Papers, Vol.53, No.4, pp.642-58
Himler, J.H. (2001), A Comparison of Alternative Specifications of the College
Attendance Equation with an Extension to Two Stage Selectivity-Correction Models,
Economics of Education Review, Vol.20, No.3, pp.263-78.
McGuinness & Bonner, (2002), Employer Characteristics and Practices as Drivers of
Unfilled IT Vacancies, The Service Industries Journal, Vol.22, No. 4, pp. 137-152.
McGuinness, S. & Bennett, J. (2005), Examining the link between Skill Shortages,
Training Composition and Productivity Levels in the Construction Industry: Evidence
from Northern Ireland. The International Journal of Human Resource Management,
Vol. 17, No.2, pp. 265-279
Nickell, S. & Nicolatsis, D. (1997), Human Capital, Investment and Innovation: What
are the Connections?, Centre for Economic Performance Discussion Paper No.20,
London School of Economics.
Redding, S. (1996), The Low-Skill, Low-Quality Trap: Strategic Complementarities
Between Human Capital and R&D, Economic Journal (RES Conference Volume),
Vol. 106 (435).
Page 20 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
19
Snower, D. (1996), Transferable Training and Poaching Externalities (in Boot, A. &
Snower, D. (1996), Acquiring Skills. Market Failures, their Symptoms and Policy
Responses, Cambridge University Press).
Page 21 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
20
Table 1: Size Distribution of All Firms
Firm Size N % N Employment % Employment
1–20 89 37 895 3
21 – 50 64 26 2,139 5
51 – 150 48 20 4,234 11
151 – 500 19 8 4,915 12
501 – 1500 19 8 15,310 39
1501 + 3 1 11,835 30
Total 242 100 39,328 100
Source: ERINI, 2005
Page 22 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
21
Table 2: Ease / Difficulty of Recruitment – All Firms (%)
VEasy Quite Easy Difficult V. Difficult N
Operators / Assemblers 22 32 29 17 144
Non-Grad Tech Support – Level 3 26 32 29 13 120
Non-Grad Tech Support – HNC/D 24 31 29 17 108
Graduates with no experience 36 39 12 12 105
Graduates with < 2 yrs experience 29 28 26 18 98
Graduates with 2 + yrs experience 6 26 32 35 62
Project Leaders 13 19 34 35 107
Senior Managers 14 17 26 42 99
Source: ERINI, 2005
Page 23 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
22
Table 3: Firms Describing Recruitment as Difficult / Very Difficult by Sector
(%)
Electronics
IT
Mechanical
Engineering
Operators / Assemblers 42 - 49
Non-Grad Tech Support – Level 3 57 16 52
Non-Grad Tech Support – Level 4 51 10 59
Grads no experience 57 12 14
Grads < 2 yrs experience 78 24 48
Grads 2 + yrs experience 80 - 59
Project Leaders 82 58 66
Senior Managers 72 62 71
Page 24 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
23
Table 4: Number and Rate of Unfilled Vacancies in last 12 months
–All Firms (%)
No. Firms
No.
Vacancies
Total
Employment
Vacancy
Rate
Operators / Assemblers 22 168 18523 0.9
Non-Grad Technical Support – Level 3 24 109 6219 1.7
Non-Grad Technical Support – HNC / D 17 41 1244 3.2
Graduates with no experience 9 16 547 2.8
Graduates with < 2 yrs experience 7 19 2085 0.9
Graduates with 2 + yrs experience 16 44 932 4.5
Project Leaders 13 20 1352 1.5
Senior Managers 3 3 840 0.4
Total 60 420 31742 1.3
Source: ERINI, 2005
Page 25 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
24
Table 5: Comparative Vacancy Rates
Electronics IT Mechanical
Operators / Assemblers 1.3 - 0.5
Non-Grad Technical Support – Level 3 10.4 0 1.0
Non-Grad Technical Support – HNC / D 4 0 3.6
Graduates with no experience 12.1 0.6 0
Graduates with < 2 yrs experience 7.8 0 0
Graduates with 2 + yrs experience 14.3 - 0.4
Project Leaders 5.8 0.5 0.2
Senior Managers 0 0.4 0.5
Total 1.7 0.2 0.7
Source: ERINI, 2005
Page 26 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
25
Table 6: Reasons for Unfilled Vacancies - All Firms (n= 60)1
%
Shortage of
Qualified
Applicants
Lack of
Technical
Ability
Shortage of
Sufficient
Experience
Cannot Pay
Enough
Competition
from Other
Employers
Very Important 60 67 53 16 59
Important 20 21 33 30 21
Not Important 20 13 13 54 21
Source: ERINI, 2005
1The percentages in each of these tables below vary as not all firms were able to rate the level of
importance for each reason given.
Page 27 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
26
Table 7: Bivariate probit – Hard-to-Fill and Unfilled Vacancies
Had-to-fill Unfilled
Constant -1.949 (0.427)*** -1.781 (0.429)***
Externally Owned (dummy) -0.531 (0.305)* -0.118 (0.300)
Located in Belfast (dummy) -0.000 (0.28) -0.579 (0.307)**
Technical staff share of total employment -0.017 (0.581) 0.303 (0.572)
New and inexperienced graduate share of employment -4.704 (2.830)* -8.462 (4.056)**
Graduate project leader share of total employment 0.677 (1.271) 1.157 (1.378)
Lost workers to competitor companies (dummy) -0.046 (0.286) 0.143 (0.293)
Perceived deficiencies in of new entrants (dummy) 0.034 (0.229) 0.477 (0.224)**
Mean technical salary (logged) 0.091 (0.029)*** 0.067 (0.028)**
Proportion of Firms activity concentrated in R&D -0.001 (0.006) 0.005 (0.006)
Firm is less than 2 years old (dummy) -0.706 (0.506) -1.185 (0.558)**
Female graduate share of total employment 1.634 (0.618)*** 1.800 (0.751)**
Firm Size (logged) 0.072 (0.091) 0.064 (0.088)
Electronic Engineering Firm (dummy) 1.210 (0.294)*** 1.057 (0.282)***
IT Firm (dummy) -0.067 (0.375) -0.710 (0.419)*
No. of Observations 242
Wald Chi2(28) 68.89***
P0.620***
Notes: The figures are regression coefficients with the standard errors in parentheses.
The significance of each coefficient is also noted, *** denotes significance at the 99% level; ** at the
95% level; and * at the 90% level.
Page 28 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
27
Table 8: Difficulties Arising from Unfilled Vacancies (n=58)
%
Lower
Productivity
Loss of
Orders
Lower
Quality
Product
Higher
Running
Costs
Inability to
Develop
New
Products
Failure to
Meet
Deadlines
Reduced
Credibility
Very Important 47 34 28 43 31 57 43
Important 33 33 26 19 22 24 28
Not Important 21 33 47 38 47 19 29
Source: ERINI, 2005
Page 29 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
28
Table 9: Actions Taken Regarding Unfilled Vacancies - All Firms (%)
Total (n = 60)
Upskill Existing Staff 58
Increase Salary 63
Recruit from Other Backgrounds and Train 47
Recruit Less Qualified Persons 40
Change Internal Structure and Practices 33
Other 12
Source: ERINI, 2005
Page 30 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
29
Table 10: Results from Productivity Level Regressions – Hard-to-Fill
Unparsed Parsed
Constant 10.986 (0.241)*** 10.925 (0.074)***
Hard-to-fill Vacancies 0.318 (0.144)** 0.277 (0.128)**
Externally Owned (dummy) 0.428 (0.140)*** 0.402 (0.109)***
Located in Belfast (dummy) -0.041 (0.125)
Technical staff share of total employment -0.056 (0.269)
New and inexperienced grad share of employment -0.639 (0.615)
Graduate project leader share of total employment -0.481 (0.531)
Lost workers to competitor companies (dummy) 0.037 (0.129)
Perceived deficiencies in new entrants (dummy) -0.097 (0.116)
Participates in on-the-job training (dummy) -0.046 (0.171)
Participates in off-the-job training (dummy) -0.029 (0.122)
Proportion of Firms activity concentrated in R&D -0.012 (0.006)** -0.015 (0.005)***
Firm is less than 2 years old (dummy) 0.318 (0.198) 0.315 (0.185)*
Female graduate share of total employment -0.522 (0.329) -0.586 (0.280)**
Firm Size (logged) 0.021 (0.048)
Firm size * Prop R&D 0.002 (0.001)* 0.003 (0.001)**
Electronic Engineering Firm (dummy) -0.164 (0.146)
IT Firm (dummy) 0.052 (0.184)
No. of Observations 227 227
F-Stat 2.42*** 6.13***
R20.164 0.143
Notes: The figures are regression coefficients with the standard errors in parentheses.
The significance of each coefficient is also noted, *** denotes significance at the 99% level; ** at the
95% level; and * at the 90% level.
Page 31 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
30
Table 11: Results from Productivity Level Regressions - Unfilled
Unparsed Parsed
Constant 11.015 (0.243)*** 10.978 (0.070)***
Firm has Unfilled Vacancies (dummy) 0.082 (0.136)
Externally Owned (dummy) 0.382 (0.140)*** 0.402 (0.109)***
Located in Belfast (dummy) -0.033 (0.127)
Technical staff share of total employment -0.038 (0.273)
New and inexperienced grad share of employment -0.725 (0.622) -0.960 (0.508)*
Graduate project leader share of total employment -0.448 (0.537)
Lost workers to competitor companies (dummy) 0.040 (0.130)
Perceived deficiencies in new entrants (dummy) -0.090 (0.119)
Participates in on-the-job training (dummy) -0.077 (0.172)
Participates in off-the-job training (dummy) -0.012 (0.123)
Proportion of Firms activity concentrated in R&D -0.012 (0.006)** -0.012 (0.005)**
Firm is less than 2 years old (dummy) 0.322 (0.200)
Female graduate share of total employment -0.381 (0.327)
Firm Size (logged) 0.024 (0.048)
Electronic Engineering Firm (dummy) -0.099 (0.147)
IT Firm (dummy) 0.017 (0.186)
Firm size * Prop R&D 0.002 (0.001)* -0.002 (0.001)*
No. of Observations 227 227
F-Stat 2.11*** 7.39***
R20.146 0.1175
Notes: The figures are regression coefficients with the standard errors in parentheses.
The significance of each coefficient is also noted, *** denotes significance at the 99% level; ** at the
95% level; and * at the 90% level.
Page 32 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
31
Table 12: Results from Treatment Model for Hard-to-Fill Vacancies
Results from Probit:
Constant -1.408 (0.412)***
Externally Owned (dummy) -0.762 (0.309)**
Located in Belfast (dummy) 0.103 (0.277)
Technical staff share of total employment 0.668 (0.524)
Lost workers to competitor companies (dummy) 0.154 (0.261)
New and inexperienced graduate share of total employment -5.326 (3.439)
Graduate project leader share of total employment -0.376 (1.207)
Perceived deficiencies in the quality of new entrants (dummy) 0.306 (0.223)
Proportion of Firms activity concentrated in R&D -0.012 (0.015)
Firm is less than 2 years old (dummy) -0.607 (0.502)
Female graduate share of total employment 2.579 (0.797)***
Firm size * prop R&D 0.004 (0.004)
Firm Size (logged) 0.041 (0.092)
Electronic Engineering Firm (dummy) 0.981 (0.264)***
IT Firm (dummy) -0.691 (0.389)*
Results from Selection Model:
Constant 11.048 (0.243)***
Firm has Hard-to-Fill Vacancies (dummy) -0.641 (0.240)***
Externally Owned (dummy) 0.320 (0.140)**
Located in Belfast (dummy) -0.046 (0.132)
Technical staff share of total employment 0.050 (0.283)
Lost workers to competitor companies (dummy) 0.034 (0.136)
New and inexperienced graduate share of total employment -1.013 (0.655)
Graduate project leader share of total employment -0.355 (0.560)
Perceived deficiencies in the quality of new entrants (dummy) -0.025 (0.123)
Female graduate share of total employment -0.001 (0363)
Participates in on-the-job training (dummy) -0.043 (0.167)
Participates in off-the-job training (dummy) 0.008 (0.115)
Proportion of Firms activity concentrated in R&D -0.012 (0.006)*
Firm is less than 2 years old (dummy) 0.327 (0.209)
Firm Size (logged) 0.036 (0.050)
Firm size * Prop R&D 0.002 (0.002)*
IT Firm (dummy) -0.113 (0.189)
Q0.563 (0.125)
Wald Chi2(16) 39.14***
No. of Observations 227
Notes: The figures are regression coefficients with the standard errors in parentheses.
The significance of each coefficient is also noted, *** denotes significance at the 99%
level; ** at the 95% level; and * at the 90% level.
Page 33 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
For Peer Review
32
Table 13: Results from Treatment Model for Unfilled Vacancies
Results from Probit:
Constant -1.319 (0.433)***
Externally Owned (dummy) -0.147 (0.285)
Located in Belfast (dummy) -0.487 (0.300)*
Technical staff share of total employment 0.873 (0.521)*
Lost workers to competitor companies (dummy) 0.258 (0.270)
New and inexperienced graduate share of total employment -6.968 (3.936)*
Graduate project leader share of total employment -0.098 (1.373)
Perceived deficiencies in the quality of new entrants (dummy) 0.556 (0.203)***
Proportion of Firms activity concentrated in R&D -0.005 (0.016)
Firm is less than 2 years old (dummy) -1.012 (0.543)*
Female graduate share of total employment 2.148 (0.789)***
Firm Size (logged) 0.032 (0.092)
Firm size * prop R&D 0.003 (0.003)
IT Firm (dummy) -0.956 (0.402)**
Electronic Engineering Firm (dummy) 0.871 (0.248)***
Results from Selection Model:
Constant 11.117 (0.246)***
Firm has unfilled vacancies (dummy) -0.753 (0.272)***
Externally Owned (dummy) 0.380 (0.139)***
Located in Belfast (dummy) -0.128 (0.135)
Lost workers to competitor companies (dummy) 0.036 (0.135)
Technical staff share of total employment 0.125 (0.284)
Graduate project leader share of total employment -1.089 (0.652)*
Participates in on-the-job training (dummy) -0.066 (0.169)
Participates in off-the-job training (dummy) -0.010 (0.115)
Proportion of Firms activity concentrated in R&D -0.011 (0.006)*
Female graduate share of total employment -0.021 (0.359)
Firm is less than 2 years old (dummy) 0.295 (0.208)
Firm Size (logged) 0.042 (0.050)
Firm size * prop R&D 0.003 (0.002)*
IT Firm (dummy) -0.201 (0.197)
Q0.529 (0.156)
Wald Chi2(15) 39.39***
No. of Observations 227
Notes: The figures are regression coefficients with the standard errors in parentheses.
The significance of each coefficient is also noted, *** denotes significance at the 99%
level; ** at the 95% level; and * at the 90% level.
Page 34 of 34
Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK
Submitted Manuscript
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
peer-00582157, version 1 - 1 Apr 2011
