Examining the link between skill shortages, training composition and productivity levels in the construction industry: evidence from Northern Ireland1

1 This paper was written as part of a research programme funded by the Department for Employment and Learning, Belfast.

Abstract

This paper utilizes data from a comprehensive survey of construction firms in Northern Ireland to investigate the extent to which industry level skill shortages arise either as a result of a mismatch between existing industry employment structure and training provision and/or a general failure among training providers to keep pace with technological change within the industry. It was found that while there was some limited evidence linking imbalances in the existing structure of craft training with skill shortages, the incidence of unfilled vacancies was much more heavily related to a failure to keep pace with the increased demand for a more multi-skilled approach to training driven by the rise in prefabricated building techniques within the industry. However, it also found that a large scale shift towards multi-skilled training might be costly as it would tend to reduce worker productivity levels.

Keywords:

• Training composition
• skill shortages
• productivity
• construction industry

Notes

1 This paper was written as part of a research programme funded by the Department for Employment and Learning, Belfast.

2 The sample drawn consisted of 45 firms with one–five employees (Emp = 174), 43 firms with six–ten employees (Emp = 315), 54 firms with 11–20 employees (Emp = 876), 238 firms with 21–50 employees (Emp = 7,280), 50 firms with 51–100 employees (Emp = 3,403) and 22 firms with 101+ employees (Emp = 4,252).

3 Not all of these workers will hold an NVQ qualification; some will hold alternative equivalents such as City and Guilds while others have been assessed by firms to hold an equivalent level of skills as a result of experience. As such, the categorization used here should be interpreted as a proxy for particular skill levels as opposed to an exact measure of educational attainment.

4 The unfilled vacancy rate is calculated as (unfilled vacancies/employment + unfilled vacancies) × 100.

5 Only partial data were available on output per worker therefore the variable was excluded on the grounds that it restricted the sample size, however, when the regressions were run on the restricted sample the results remained unchanged (available from authors).

6 A previous study of the IT sector uncovered a similar finding, however it was found that IT firms were acting rationally as a high concentration of inexperienced labour tended to reduce productivity (McGuinness and Bonner, 2002).

7 It is important to note that in order for the model to be properly identified the first stage probit must include at least one variable that is absent from the second stage OLS, also there should exist a clear rationale as to why the omitted variable is important with respect to the latent dummy but irrelevant in terms of the continuous variable (see Himler, 2001). In relation to this study the identifying variable was based on firm's perception of labour market tightness, i.e. because of their more unique, relatively non-standard, requirements prefabricating companies were much more likely to perceive the recruitment process as being more difficult, conversely there is nothing to suggest that this perception based variable will be related to output per worker. In order to test the suitability of the identifying variable, perceived labour market difficulty was tested within a standard OLS productivity regression, and found to be of no consequence. Conversely, when the variable is included in a backward stepwise probit for prefabrication it was found to be significant at 99 per cent.

8 Contact authors for details.