• Keine Ergebnisse gefunden

2.2 Empirical Motivation

2.2.3 Wages and On the Job Search

Wage Losses from Employment to Employment Transitions

One of the most important potential channels for enlarging the set of acceptable job offers to the worker is the ability to continue searching on the job. In this case, the trade-off is between earning more than unemployment benefits now and reduced search efficiency on the job. If search on the job is fairly efficient, this will substantially enlarge the range of acceptable offers13.

Studies of on the job search typically calibrate on the job search efficiency to observed employment to employment flow rates. Meanwhile, Fallick and Fleischman (2004) find in the CPS that a worker who reports to be actively searching on the job is more likely to be unemployed next month. Likewise, Fujita (2011) uses a question in the UK labor force survey that asks for the reason of on the job search.

He finds that of those reporting searching, 12% do so because they fear to lose

12We also perform a more standard Mincer wage regression that ignores individual fixed effects.

The resulting Mm-ratio at the first and fifth percentile are 2.78 and 2.03, respectively. Hornstein et al. (2007) obtain very similar results for PSID data. Without controlling for individual effects, the Mm-ratio at the first and fifth percentile are 2.73 and 2.08, respectively. Once they control for individual effects, those numbers drop to 1.9 and 1.46.

13Of course, if search efficiency on the job is as high as in unemployment, the worker will accepts any wage offer at least as high as UI benefits.

their current job and 27% because they are unsatisfied with the current job because of non-monetary reasons. This evidence seems clearly at odds with predictions of job-ladder models. Nágipal (2005) shows within a basic job-ladder model that search on the job would have to be more efficient than during unemployment in order to replicate some characteristics of flow rates. In our view, these pieces of evidence imply other mechanisms behind the magnitude of employment to employment movements than job-ladders only. They also hint at on the job search being less efficient than previously assumed, and an accurate calibration of search efficiencies should take these concerns into account.

The SIPP asks workers that terminate a job for the reason of doing so. While this appears to be the information that perfectly suits our analysis, we want to emphasis several problems with the variable. First and foremost, less than 30% of all workers making a job to job transition provide information on this variable14 15. Second, the question targets solely individuals that change employer id, but not those that switch occupations within the same firm. Third, the answers are not mutually exclusive and the interviewer provides no additional guidelines to choose among these mutually non-exclusive answers16. Nevertheless, we find the information instructive to get a sense of the magnitude of job to job transitions that do not necessarily result from the desire to move up the job-ladder. Only 55% of those responding to the question and making a job to job transition state that they quit to take another job. 19% of jobs ended because the previous job did not provide the possibility to continue17. Another 20% answer either that they had "unsatisfactory work arrangements", or

"quit for some other reason". The remaining worker made a transition because of personal or family related reasons. Taken together, the result supports the view that a large share of identified job to job transitions does not result in a move to a better paying job.

In Table 2.2 we supply additional evidence to support that claim using a moment that does not suffer from the drawbacks mentioned above. A pervasive phenomenon

14The question is not applicable for a negligible share of transitions because only the main job changed, but the worker stays with his old employer.

15See Nágipal (2008) who also uses this information.

16For example, a possible answer is that the individual "quit to take another job" and an alternative answer is that the individual had an "unsatisfactory work arrangements".

17This includes"on layoff","job was temporary and ended","discharged/fired","employer bankrupt",

"employer sold business" and"slack work or business conditions"

2.2 Empirical Motivation

Table 2.2: Wage Cuts after Job to Job Transitions

Sample Stratification Share loss Mean loss

Whole sample 0.339 -0.22

Job characteristics

Union 0.35 -0.249

+ Health insurance 0.355 -0.249

+ Education 0.355 -0.249

Year 1993 0.3301 -0.2068

1994 0.3299 -0.23

1995 0.3638 -0.2256

Sex Male 0.3367 -0.2261

Female 0.3421 -0.2122

Age 23-30 0.3483 -0.2274

31-50 0.3368 -0.2198

51-65 0.3230 -0.1981

Notes: The table shows the share of workers incurring a nominal cut in hourly wages after a job to job move-ment for the whole population and different subsamples in the 1993 SIPP. Mean loss reports the mean wage loss in log points conditional on suffering a wage cut upon movement.

in the data is that job to job transitions result in nominal wage losses. In the whole population roughly one third of all transitions result in workers earning lower hourly wages in the month after transition compared to the last month at the previous job. The months surrounding a job to job transition might be particularly prone to reporting error. Therefore, we also constructed three-month-averages of wages after and before a movement. This robustness check does not affect any of our estimates.

In terms of real wage changes, the share of wage losses increases to roughly one half18. The top panel of Table 2.2 shows that the result is robust across different stratifi-cations of the data concerning non-monetary job characteristics. First, we exclude observations where a worker moves from a non-unionized to a unionized job. The

18In principle, the worker should only care about real wages. But in the presence of some wage rigidity, the worker expects a wage loss on his current job as well and compares nominal wages.

resulting share of workers incurring a wage loss goes up by about one percentage point. Subsequently deleting transitions where a worker moves from a job that did not provide health insurance to one that does provide it, increases this share by another half percentage point. Finally, deleting observations where the former employer did not pay for educational training, but the current employer does, has no effect on the share. A likely explanation for the robustness of the result is that even though workers value these non-monetary benefits of a job, jobs that have high non-monetary benefits are usually associated with higher wage income (see for example Dey and Flinn (2008)).

The result is also robust across different stratifications of the data concerning worker characteristics. Table 2.2 splits the sample by year, sex and age, and we always find a share around 33% of workers incurring wage losses at a job to job movement. Appendix 2.B shows that the result stay robust to stratifying the data by tenure at the former job, earnings at the former job and industries and looking at real instead of nominal wage changes.

In the interpretation of this chapter, an important part of these transitions are either the result of jobs accepted within notice period of dismissal or movements for non-economic reasons (moving in with one’s spouse, moving close to ones parents, etc.). To proxy for these causes, our model includes what Jolivet et al. (2006) label forced job movements: randomly drawn on the job offers, which the worker can accept or move into unemployment.

If idiosyncratic worker productivity uncertainty is large relative to firm dispersion, parts of these wage cuts will be the result of negative shocks to general human capital, and some of them will be simply measurement error. Consequently, our baseline model includes innovations to worker productivity and all our simulations explicitly include measurement error to account for these causes. In Appendix 2.B we give consideration to an alternative explanation brought forward by Postel-Vinay and Robin (2002) and extended by Cahuc et al. (2006). They lay out a framework, in which workers will accept wage cuts upon job to job transitions, if the option value of working at the other firm is sufficiently high. Indeed, Papp (2012) shows that this framework can rationalize a large amount of wage cuts and large frictional wage dispersion. The key operating mechanism in this class of models is that workers who experienced wage losses have on average steeper observed wage growth afterwards,

2.3 The Model