Alternative pathways to high-performance workplaces

Abstract

In this paper we outline four high workplace performance models, two of which have high-road or enabling characteristics. These are the strategic human resource management and organizational trust models, respectively. The second two models – the labour process and numerical flexibility models – motivate workers to raise productivity through the use of the stick (coercion) rather than the carrot. Based on a representative survey of Australian workers, we compare these models in terms of their capacity to explain relative workplace productivity. We find that all four models have some explanatory power. However, contrary to expectations, the low-road numerical flexibility model provides the best fit with the data. We interpret this finding by reference to recent evidence of workforce trends indicating the attraction and ability of employers to pursue this pathway toward higher productivity. We conclude with some suggestions for future research that would assist in developing this line of enquiry.

Keywords:

• High-performance work practices
• productivity
• human resource management
• flexibility
• workplace relations
• Australia

Notes

1 Owing to measurement problems, we were not able to include extent of contracting-out of work, minorities as a proportion of the workforce and extent of competition in the product market.

2 We question the construct validity of Ramsay et al.'s. (2000: 527) usage of management reports of changes in labour productivity in the last five years as a measure of work intensification. Labour productivity may have various sources aside from increased worker effort. Accordingly, we prefer to use a composite measure of worker perceptions of work intensification as indicated in Table 2.

3 In some cases, internal consistency among this larger set of items was so low that the path analyses either did not converge or had inadmissible negative error variances. This evidence casts doubt on the systemic (i.e. internally consistent) nature of HPWP.

4 We tried a variant of the model inserting ER climate as a mediator between numerical flexibility and labour productivity. This was based on the proposition that practices such as outsourcing of labour and hiring of mainly casual, part-time workers will lift worker effort through fear of possible job loss. However, this mediation model did not converge.

5 At the time of the survey (1995–6), the economy was growing at around 3 per cent a year following a recession. Structural change is indicated by reductions in tariff protection (in manufacturing from 15 per cent in 1989–90 to 8 per cent in 1995–6) and institutional change is signalled by decentralization of the industrial relations system, with workplace bargaining having been formalized in 1994 following major legislative change. By the time the survey was undertaken, workplace agreements were becoming a common alternative to arbitrated award provisions (Morehead et al., 1997: 2–7).

6 Although violations of these two assumptions are common and usually ignored, an optimal research strategy would test these assumptions, particularly multivariate normality. While univariate normality can be tested relatively easily with Q-Q plots, tests of multivariate normality are complex and none of the common statistical packages offers these plots (Sharma, 1996). However, we performed tests of univariate normality on all variables, and most did not violate the assumption of univariate normality. Even if the assumption of multivariate normality was violated, research suggests that, in practice, this assumption rarely holds (West et al., 1995).

7 To control for such potential confounding variables, we would need to (1) draw on theory and specify these confounding variables and (2) collect data on them and control for their impact statistically.