There has been much research over the last decade or so that has attempted to answer the question: ‘Do HR practices make a positive impact on organizational performance?’ This is summarized in Table 5.1.
Comments on the research
With one exception, the studies summarized above revealed that there was a link between certain HR practices and whatever criteria were used to indicate the level of performance. But Guest (1997: 274) commented that: ‘At present the studies report a promising association between HRM and outcomes, but we are not yet in a position to assert cause and effect’. If there is a significant link, how is the impact achieved? It is not enough to justify HRM by proving that it is a good thing. What counts is what can be done to ensure that it is a good thing. Ulrich (1997b: 304) pointed out that: ‘HR practices seem to matter; logic says it is so; survey findings confirm it. Direct relationships between investment and attention to HR practices are often fuzzy, however, and vary according to the population sampled and the measures used.’ Wood (1999: 408) analysed 15 studies of the HRM/ performance link and found that they had ‘concentrated on assessing the link between practices and performance, with an increasing disregard for the mechanisms linking them. This has meant that there has been no systematic link between HR outcomes and performance.’ The longitudinal research by Purcell et al (2003) tried to remedy that weakness.
There are a number of problems that have been identified in other studies. Paauwe and Richardson (1997: 258) commented on ‘the difficulty in establishing convincing accounts of chains of cause and effect’. This was explained by Boselie et al (2005: 75), who referred to the causal distance between an HRM input and an output such as financial performance: ‘Put simply, so many variables and events, both internal and external, affect organizations that this direct linkage strains credibility’.
Another problem is the assumption some people make that correlations indicate causality – if variable A is associated with variable B then A has caused B. It might have, but again it might not. As Wall and Wood (2005) pointed out, most of the efforts to prove causal relationships have so far produced little but persuasive associations. This is linked to the issue of ‘reversed causality’, which is the assumption, as Purcell et al (2003: 2) put it: ‘That although it is nice to believe that more HR practices leads to higher economic return, it is just as possible that it is successful firms that can afford more extensive (and expensive) HRM practices’. Although they also comment that when successful firms invest heavily in HRM they may do so to help sustain high performance.
Purcell et al (2003: 61) cast doubts on the validity of some of the attempts through research to make the connection: ‘Our study has demonstrated convincingly that research which only asks about the number and extent of HR practices can never be sufficient to understand the link between HR practices and business performance. As we have noted, it is misleading to assume that simply because HR policies are present that they will be implemented as intended.’
Another methodological problem with much of the research on the HR/ performance link is that the analysis has been done the wrong way round – identifying current HR practices and measuring performance before the practices were introduced, thus making the dubious assumption that the performance of the firm was affected by the practices that were not yet in place. As Boxall et al (2007: 6) remarked: ‘A huge proportion of the studies measuring HR practices of some kind and firm performance have found associations – but between the former and past performance, thus leaving us poorly placed to assert that causality runs from the selected HR practices to performance’. The best approach is the type of longitudinal study undertaken by Purcell et al (2003), which examined the impact of existing HR practices over succeeding years.
However, it must be accepted that HR practices do make a difference. It may be difficult to prove exactly how they do so in terms of the precise link between cause and effect but, as described below, it is possible to develop propositions that explain in broad terms what happens in the black box.