The relative importance of human resource management practices for innovation

ABSTRACT

Human resource management (HRM) practices are generally expected to stimulate a firm's innovation performance. However, which of these practices really pay off? Based on a unique dataset that includes detailed information for both a firm's innovation activities and a broad set of HRM practices, we find that primarily new workplace organization practices seem to enhance a firm's innovation activities. Flexible practices of working time management and incentive payment schemes show only small effects on both innovation propensity and innovation success. Further training does only affect innovation success, but not innovation propensity. Overall, we find a stronger linkage between HRM practices and innovation propensity than with innovation success. Further, we find that innovation propensity increases, first, with the number of combinations of HRM practices adopted by a firm but not with the number of combinations of HRM practices from different groups of HRM practices adopted by a firm.

KEYWORDS:

• Human resource management
• workplace organization
• innovation performance
• cumulative effects

JEL CLASSIFICATION:

• O31

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. The Swiss Innovation Survey has been conducted every third year since 1990 by the KOF Swiss Economic Institute, ETH Zurich on behalf of the Swiss State Secretariat for Economic Affairs (Seco).

2. However, we do not cover organizational changes such as vertical integration, mergers, outsourcing, and offshoring within multinational firms (see Bloom, Sadun, and Van Reenen 2010 on this point).

3. Nevertheless, the empirical evidence for part-time work and fixed-term contract work is mixed and depends on the overall conditions of the labor market and its institutional framework (see, e.g., Hutchinson and Brewster 1994; Michie and Sheenan 2005).

4. The industries are defined according to the NACE classification. We excluded the value of the observation itself in order to ensure the exogeneity of these variables. If the number of observations of a specific three-digit industry is lower than five, we used the average score at the NACE two-digit level.

5. We also estimated a bivariate Heckman selection model with maximum likelihood. Also in this case we did not find any evidence that selection is an issue.

6. We alternatively estimated the propensity model separately for product and process innovation. However, results differ only marginally between the two types of innovation (see Table A5 in the appendix).

7. A better method would probably be to look at interaction effects of different practices to test whether complementarities exist between different practices. We also tried this approach, but as the variables for the single practices and the respective interacted variables are highly correlated, we did not get any meaningful results. As noted by Laursen and Foss (2014), given the large number of individual practices considered in empirical literature, the “bundles” or “systems” approach is dominant, even it only confers an indirect test of complementarity.

8. As the variables are scaled quite differently, we use the median of the underlying distributions of the original variables as cut-off value. The only exception is the variable Δ_hierarchical_levels, which is coded as 0/1.

9. This variable is constructed as follows: for every HRM practice with above-median use intensity each firm receives the value of 1. The sum of these values for every practice with above-median intensity use gives the number of practices applied by a firm.