| This study deals with three research questions that pertain to the effects of human resource management (HRM) on firm performance: (1) Does a 'best' HRM system exist which outperforms other types of HRM systems in improving firm performance? (2) Which HRM systems or individual HRM practices matter to firm performance? and (3) Does a certain HRM practice work equally well regardless of the HRM system in which it is implemented?;First, a HRM system has been hypothesized to become more effective for the improvement of firm performance as it employee involvement and commitment strengthen (Osterman, 1994; Pfeffer, 1994; Huselid, 1995). Second, it has been hypothesized that systems of HRM practices, as opposed to individual HRM practices, matter to firm performance (Ichniowski, Shaw, & Prennushi, 1997). Third, according to the complementarity theory (Milgrom & Roberts, 1995), certain HRM practices are hypothesized to have larger effects in HRM systems which are composed of a set of complementary HRM practices, but smaller effects in HRM systems where supporting HRM practices have not been adopted.;The above hypotheses were tested using a two-year-interval longitudinal data set. The first hypothesis was not supported by either the cross-sectional analyses or by two-year-interval longitudinal analyses. A high-involvement HRM system was found to be the least effective in improving firm performance, as represented by net sales per employee, whereas an open flexibility HRM system was found to be the most effective. Indirect support was found for the second hypothesis. Very few HRM practices were found to be significant in explaining variation of firm performance, whereas the HRM system was found to significantly contribute to explaining variation in firm performance. Support was found for the third hypothesis. Some high-involvement HRM practices (i.e., organization-based pay, work teams, QC, TQM, and employee participation) were the most effective when they were implemented within a high-involvement HRM system. The results and implications of these findings are discussed. |
