2.6 The E ff ect of Learning on Job Search E ff ort Decisions and Occupational
2.6.2 The E ff ect of Learning on Transition Outcomes
Due to search frictions, job search effort is only a part of the equation determining occu- pational transition outcomes of workers. How learning affects the latter therefore needs to be examined separately. To this end, I first simulate labour market transitions for a large number of workers whose true acquisition probabilities are drawn from the popula- tion distributions ofθ1andθ2, and predict their occupational affiliation distribution after 10
years of (potential) experience for this simulated sample. Then, I compare the result with a counterfactual occupational affiliation distribution for the same period simulated under a scenario where learning is shut down. Specifically, I consider a scenario where workers do not update their initial belief distributions ofθ1 andθ2 despite receiving new information,
thus holding on to their initial beliefs.
Table 2.14 presents the simulated distributions of employed workers in the three occu- pations under the two scenarios with and without learning, respectively. Comparing these two distributions shows that learning increases the likelihood of working in occupation 2, with this difference offset by a comparable decrease in the proportion of employed workers
Table 2.14: Predited Occupational Choice Distributions in the 41st Period Occupation 1 Occupation 2 Occupation 3 With learning 0.277 0.624 0.099 Without learning 0.306 0.598 0.096 Difference −0.029 0.026 0.003
in occupation 1. However, the differences between the two distributions are small even if a 10 year period of labour market experience is considered.
This small difference is partly due to the composition of workers with different skill acquisition probabilities in the simulated sample. With workers of different skill acquisition probabilities affected by the learning process differently, their changes can offset each other to some degree. Indeed, it is more illumining to analyze the effect at a more disaggregated level.
Therefore, I simulate occupational affiliation distributions for samples of workers with the same skill acquisition probabilities, and present the results in Figure 2.10. The three surface graphs on the left-hand side of the figure show that the fractions of workers working in given occupations in the 41st quarter since their labour market entry under the counter- factual no-learning scenario. All the graphs are flat, indicating that without learning, the occupational affiliation distribution vary little with skill acquisition probabilities.
The three graphs on the right-hand side of Figure 2.10 present the simulated occupa- tional affiliation distribution in the 41st quarter with learning. Following the way learn- ing influences workers’ job search effort decisions presented in Figure 2.9, the proportion working in occupation 1 increases with θ1, and this change is largely offset by a fall in
the proportion working in occupation 2. There is a weaker pattern between occupational choice andθ2.
Figure 2.10: Predicted Occupational Affiliations in the 41st Quarter No Learning With Learning
occupation 1. However, such effort faces high search frictions associated with this occupa- tion, and as a result, their occupational affiliation outcomes do not match their job search effort. Around two fifth of them are employed in occupation 1, with a majority of them are in occupation 2 instead. Because of search frictions, their occupational affiliation out- comes do not reflect the extent of the changes in job search effort allocations. In contrast, the search frictions are less of a concern for workers with lowθ1, because they direct their
job search effort away from occupation 1 toward occupation 2 as learning progresses. They indeed attain occupational affiliation where more than three fifths of them are employed in occupation 2. The search frictions associated with occupation 2 are substantially lower than those associated with occupation 1, and job search effort to find work in occupation 2 is more likely to be successful in generating transitions to this occupation.
Not only does the low likelihood of finding a job in occupation 1 substantially affect occupational transitions of workers with high skill 1 acquisition probabilities, it also influ- ences learning in a different way. Specifically, it slows down workers’ learning about θ1
because the likelihood of receiving a signal about θ1 is the highest in occupation 1. This
effect can be seen by simulating labour market transitions where the arrival of signal is made more frequent. For example, I consider an environment where signals for both θ1
andθ2arrive every period when workers are employed in any occupation so that the diffi-
culty to find work in occupation 1 does not hinder the learning process. Under this faster learning process, I simulate the occupational affiliation distributions in the 41st quarter, and compare them with the counterfactual distribution without learning. These distribu- tions are presented in Figure 2.11. Comparing the distributions presented in Figure 2.11 with those in Figure 2.10 shows that the faster learning process makes larger differences in the workers’ labour market outcomes. With the faster learning process, workers whoseθ1
are at either end of the skill 1 acquisition probability distribution attain more transitions to the occupations targeted by their job search effort. With more signals, the learning process
Figure 2.11: Predicted Occupational Affiliations in the 41st Quarter with the Faster Learn- ing Process
accelerates changes in workers’ job search effort decisions, which in turn generates larger differences in occupational transitions outcomes.