UC Berkeley

We first investigate properties of employee replacement costs, using a panel survey of California businesses in 2003 and 2008. We establish that replacement costs are substantial relative to annual wages and that they are associated negatively with the use of seniority in promotion. We also find some evidence, albeit not under all specifications, that replacement costs are positively associated with establishment size, which is consistent with monopsony. Bivariate scatterplots, pooled regressions and panel-based estimates suggest a positive (although not robust) relationship between replacement costs and the wage. This result constitutes an anomaly for hiring and separation models, such as Manning (2003), in which the negative wage elasticity of replacement costs is a key assumption. We also examine particular occupational groups, and find that although there is not necessarily a clear incongruity for blue collar workers, there is one for professional and managerial workers. To resolve the anomaly, we propose several possible solutions including for one the role of heterogeneity, and on the other hand expanding Manning's model to incorporate the effects of wages on productivity.

We also study the relationship of wages with High Performance Workplace Organization (HPWO) practices, a broadly-defined set of changes in establishment policies, often involving flexible policies, that have been adopted by many firms in the U.S. in the past few decades. HPWO practices include, for example, employee meetings to discuss workplace problems, self-managed teams, and job rotation. There is no empirical consensus in the literature on the relationship between wages and these practices, and the data required for its study is somewhat uncommon. We use the same panel survey of California businesses to investigate the issue. We find some evidence that adoption of these practices is correlated with increased entry wages for blue collar workers, and correlated with increased entry and highest tier wages for professionals and managers. There is also evidence that HPWO incidence is associated with increased wages for workers at manufacturing establishments, establishments with few professional or managerial workers, and workplaces with a union presence. Regarding individual practices, we find support for a positive correlation between higher wages and increased worker participation in self-managed teams, especially among blue collar workers and at establishments with few professional or managerial workers. But there is a negative relationship between team adoption and wages for professionals and managers. Finally, the adoption of formal quality management programs is associated with higher wages for professionals and managers. We conclude that the relationship between HPWO intensity and wages is complex and heterogeneous, varying for different groups of workers and different practices, and in fact some of these correlations may lead to cancelling effects when one looks only at aggregate relationships. Thus any analysis of high performance practices should be examined with a focus on fine levels of interaction.

We also study the interplay among tenure, firm size and firm age. The positive association between firm size and the average tenure of employees at a firm is well-known. We present evidence, at the establishment level, that a large portion, though likely not all, of this association can be explained by the following observation: there is a positive association between an establishment's size and its age so that when combined with the mechanical relationship between establishment age and average employee tenure, this fact leads to a size-tenure correlation. We investigate this issue by making use of the panel survey of California establishments; it is very useful for studying the topic because it is unusual in having data on all three of these quantities.

We also study the relationship of certain employer policies with voluntary quit rates, using the panel survey of California businesses. We particularly focus on two policies, namely (i) how much employers give preference to current employees when hiring for a position above entry level, and (ii) how much employers take into account seniority when promoting to a position they have already decided to fill internally. This question has been studied before by Fairris (2004), but the data required is scarce; moreover, with this survey of California establishments, we can answer new related questions. We find some evidence from pooled regressions that (i) is associated positively with quit rates. Balanced panel regressions do not confirm these results; however, the standard errors are quite large due to small sample sizes. We find no clear evidence of associations between (ii) and the overall voluntary quit rate. Moreover, the data set allows us to investigate questions that have not yet been studied, by breaking down quits into finer categories. We find, albeit with non-robust results, that (i) is positively associated in particular with quits to accept another job. We find as well that (i) and (ii) are both negatively associated with employee turnover rates due to unsatisfactory performance.

We also examine a controlled randomized health care experiment known as the RAND Health Insurance Experiment, which tested the effect of free insurance on use of medical services and on health outcomes, relative to control groups which paid a larger share of health costs out-of-pocket. We focus on using the experiment to test econometric techniques, by comparing experimental estimates of the benefit of free insurance on certain health outcomes to estimates one might obtain using nonexperimental econometric techniques if an analogous ``natural experiment'' were all that was available, following Lalonde (1986). Like Lalonde, our results cast doubt on the use of these nonexperimental methods. We then present a closer inspection of both the setup of the experiment and some findings of previous researchers. This work yields a slight caveat in the form of one piece of evidence that the result we have used in the test may not be robust. We also consider other health outcomes, and identify one beneficial impact of free health insurance that does not appear to have been noticed previously by RAND researchers, as well as finding several apparently unnoticed beneficial effects that vary by income category. Like the RAND researchers, we do not find very many beneficial impacts of the free plans, but when interpreting the experimental results, one should bear in mind the low maximums on total out-of-pocket expenditures under the control group insurance plans.