دو مرحله در جستجوی کار شواهد از تایوان

This study decomposes on-the-job search into two-step procedures. The first step is concerned with the dissatisfaction with the current job, and is very sensitive to the wage gap and other observed factors. The second step is concerned with initiating an on-the-job search, and is insensitive to the wage gap and most other observed factors. It is likely that the one-step approach overstates these observed factors, while it understates the unobserved factors that give rise to the subjective probability of finding a better job in the behavior of on-the-job search.

Research on on-the-job search has been conducted over a period of more than three decades. Empirical studies have usually embarked on determining who are the on-the-job searchers. These studies are usually concerned with whether a worker is searching for a new job as an on-the-job searcher. Regression techniques are then used to analyze which factors are important. This would appear to constitute an appropriate approach to understanding the problem of on-the-job search. However, on-the-job search involves a two-step procedure. It seems that most empirical studies have inappropriately reduced the two-step procedure into a one-step procedure. The first step of the two-step procedure consists of dissatisfaction with the current job, and the second step the actual activity involved in searching for a job. A worker must first be dissatisfied with his/her current job, or else believe that he/she deserves a better job, and if he/she considers that searching for a job is an advantageous choice, he/she will then become an on-the-job searcher. Neglecting either of these two steps might lead to an incomplete conclusion. The first step of this two-step procedure, the dissatisfaction with the current job, is usually a response to the gap between the current wage and the average wage of all workers’ with similar traits. The gap can be interpreted as the predicted gross gain from changing jobs. The second step in the on-the-job search procedure, that is, initiating a search for a job, is likely to be more dependent on one's subjective probability of finding a better job, and hence more insensitive to the predicted gross gain, than the first step in the on-the-job search. While a person may not be satisfied with his/her current job, searching for a new job does not mean that he/she will definitely find a better job. The subjective probability of finding a better job varies across individuals. In other words, individuals are heterogeneous in terms of their subjective probability of finding a better job. This heterogeneity might be due to the degrees of optimism and of motivation, or, in sum, due to some unobserved characteristics. Consequently, the sensitivity of the predicted gross gain of changing jobs in the second step of on-the-job search is much lower than that in the first step of on-the-job search. Furthermore, the implementation of the search for a job is not costless. In the case of the first step, the cost of searching for a job and changing jobs is not a concern since the job search has not yet been implemented. By contrast, the cost of searching for a job and changing jobs will become a concern if an individual starts with an on-the-job search. Studies that reduce the two-step on-the-job search procedure to a one-step procedure might not be able to disentangle these two-step procedures, and not only will they not be able to explore the insights of the behavioral characteristics of on-the-job search, but might also arrive at a misleading conclusion. For instance, if the argument in the last paragraph is true, that is, the first step is sensitive but the second step is insensitive to the predicted gross gain from changing jobs, then using only the first-step as a substitute for the behavioral characteristics of on-the-job searchers will probably overestimate the significance of the predicted gross gain. It seems that ignoring the first-step response and directly using the second-step search behavior will be a better methodology. However, this is not a proper substitute either, and it will be discussed in the next section. The remainder of this study is organized as follows. Section 2 points out the invalidation of the one-step model. Section 3 introduces the data used in the present study. Section 4 establishes two econometric models: the bivariate censored probit model and the ordered probit model. Section 5 discusses the results of these two models, and also applies the data set to three probit models that comprise. In Section 6, the results of all these models demonstrate the incomplete conclusions regarding the behavior of job searchers that arise through the use of a traditional probit or logit model in terms of either the first-step response, or the second-step search procedure. Section 7 concludes.

As a result of this study, we see that data for which there is only a one-part question regarding either the desire to change jobs or the action taken in actually searching for a job are likely to be insufficient in terms of explaining the behavior of on-the-job search. Either of the parts will misrepresent the level of dissatisfaction with the current job as compared with the behavior of job searchers. To ensure that the conclusions from the on-the-job research are correct, it is necessary to have a two-part questionnaire on hand. If what we are really concerned with is on-the-job search behavior, not on-the-job wanting to change jobs behavior, then the traditional approach has, generally speaking, overestimated the link between on-the-job search behavior and selected factors, thereby leading to a doubtful relationship. The motivation in relation to a desire to change jobs is more sensitive than implementing searching for a new job in terms of the return and cost side factors. The return side factors are conversely more insensitive to the on-the-job search behavior. These results imply that the subjective probability of finding a better job and search cost do not influence the decision to want to change jobs, but are likely to influence the on-the-job search. This is because a low probability of finding a better job lowers the sensitivity of the return variables to the on-the-job search behavior, making it easier for workers to give up their intentions to change jobs. The sensitivity of the cost side factors to the on-the-job search behavior is also lower, compared to the motivation to change jobs, but since the costs faced in the on-the-job search are much more certain than the corresponding returns, the cost side factors still significantly affect job search behavior. The insignificant results in relation to the behavior of the on-the-job searchers could arise from some unobserved heterogeneity among full-time workers. The behavior of workers with the same observed characteristics but with different wage rates might be a reflection of some other unobserved characteristics, such as optimism, sociability, creativity, motivation, integrity, a team-work attitude and so on. These unobserved characteristics might be those factors that distinguish between workers who want to change jobs and workers who are searching for a better job. However, since observed factors still explain workers who want to change jobs very well, those unobserved factors mentioned above are not likely to be important factors categorizing workers who want to change jobs and workers who do not. Unfortunately, using only one of the two steps leads to a conclusion that overstates those observed factors, while it probably understates those unobserved factors in terms of the behavior of on-the-job search.