Toy Microsimulation of Labour Mismatch

Usage | Examples | Codebook | References

This python module implements a simple microsimulation of labour mismatch based on a static interpretation of the two-sided matching model developed by Zinn (2012). The purpose of the simulation is to produce matching outcomes of a labour market populated by a set of workers and a set of jobs with pre-defined characteristics and compute the output of multiple mismatch measures, which can then be analysed to gain a deeper understanding of their construct. For the description of the economic model, available mismatch measures and a demonstration, please see the slides.

Available mismatch measures:

• Realised Matches (RM)
• Indirect Self-Assessment (ISA)
• Pellizzari and Fichen (2017) (PF)

Please note that this microsimulation is based on a stylized (toy) model and should be taken as more of a methodological experiment and illustration than an inference tool.

Any suggestions regarding either the code or economic model are most welcome!

To use the module

1. Create a new directory where you want the simulation output files to be generated and make it your working directory
2. Clone the repository to that directory
3. Make sure pandas package is installed
4. Import the module:

import toysimlm

5. Type help(toysimlm) to view the module's description or help(toysimlm.simulate) for the simulation parameters
6. Run the simulation with default parameters:

stats, workers, jobs, log = toysimlm.simulate()

7. This produces 4 CSV files:

• stats — main output file containing the simulation statistics for each repetition, including the shares of under, well and over-matched workers according to 3 different measures
• workers and jobs — additional files containing the corresponding characteristics in the last repetition that illustrate the pre-defined distributions
• log — additional file containing matching outcomes for each iteration of the matching procedure for the last repetition illustrating the algorithm

8. Analyse the simulation output using a statistical software of your choice

Other examples

• Run the simulation for 1000 reps instead of 100; with probability 0.25 increase workers' education by 1.8 (two standard deviations):

stats, workers, jobs, log = toysimlm.simulate(reps = 1000,
                                              spec = '_n100x1000_edu_inc',
                                              w_treatment = True,
                                              w_treatment_prob = 0.25
                                              w_edu_treatment_mag = 1.8)

• Set the numbers of workers and firms to 200 each instead of 100; with probability 0.5 decrease jobs' skill requirement by 92 (two standard deviations):

stats, workers, jobs, log = toysimlm.simulate(n = 200,
                                              spec = '_n200x100_skillreq_dec',
                                              j_treatment = True,
                                              j_treatment_prob = 0.5
                                              j_skill_treatment_mag = -92)

• Run the simulation while saving the data after each iteration and every time a worker is taken through the matching procedure (as opposed to upon the completion). If either n and reps are large, this may increase the completion time of the simulation considerably but could be useful for troubleshooting.

stats, workers, jobs, log = toysimlm.simulate(sim_rep_updates = True,
                                              worker_updates = True)

Codebook

Stats

• iterat - number of iterations
• uw_uj - unmatched workers matching with unmatched job
• uw_mj - unmatched workers matching with matched job
• mw_uj - matched workers matching with unmatched job
• mw_mj - matched workers matching with matched job
• match_fail - failed matches
• w_rejects - rejections by the workers
• j_rejects - rejections by the firms
• mututal_reject - mutual rejections
• av_aspir_w1 - average aspiration level for workers before the simulation
• av_aspir_w2 - average aspiration level for workers after the simulation
• av_aspir_j1 - average aspiration level for firms before the simulation
• av_aspir_j2 - average aspiration level for firms after the simulation
• av_compat_1 - average compatibility level for workers after the first iteration (zero in the first iteration)
• av_compat_fin - average compatibility level for workers after the simulation
• w_edu - average workers' education
• w_skill - average workers' skill
• occupat_0 - number of workers in occupation 0
• occupat_1 - number of workers in occupation 1
• occupat_2 - number of workers in occupation 2
• well_matched - well-matched indicator
• rm_under - share of under-matched workers according to RM
• rm_well - share of well-matched workers according to RM
• rm_over - share of over-matched workers according to RM
• isa_under - share of under-matched workers according to ISA
• isa_well - share of well-matched workers according to ISA
• isa_over - share of over-matched workers according to ISA
• pf_under - share of under-matched workers according to PF
• pf_well - share of well-matched workers according to PF
• pf_over - share of over-matched workers according to PF
• j_edu - average firms' education requirement
• j_skill - average firms' skill requirement
• j_wage - average wage
• w_treated - share of treated workers
• j_treated - share of treated firms

Workers

• edu - Education level
• skill - Skill level
• compat - Compatibility level if matched and 0 if unmatched
• match - Firm's index if matched and 9999 if unmatched
• r_edu - Required education level if matched and 0 if unmatched
• r_skill - Required skill level if matched and 0 if unmatched
• aspir - Current aspiration level
• aspir_init - Initial aspiration level
• occupat - Skill-based occupation group classification
• treatment - Treatment indicator (1: treated, 0: not treated)
• isa - Skill mismatch according to ISA (-1: under, 0: well, 1: over)
• mode_edu - Occupation-specific mode of education level
• sd_edu - Occupation-specific standard deviation of education level
• rm - Educational mismatch according to RM (-1: under, 0: well, 1: over)
• well_matched - Well-matched indicator
• pf_min - PF lower bound
• pf_max - Pf upper bound
• pf - Skill mismatch according to PF (-1: under, 0: well, 1: over)

Jobs

• edu - Education requirement
• skill - Skill requirement
• wage - Wage
• compat - Compatibility level if matched and 0 if unmatched
• match - Worker's index if matched and 9999 if unmatched
• aspir - Current aspiration level
• aspir_init - Initial aspiration level
• treatment - Treatment indicator (1: treated, 0: not treated)

References

• Pellizzari, M. and Fichen, A. (2017). A new measure of skill mismatch: theory and evidence from PIAAC. IZA Journal of Labor Economics, 6(1):1–30.
• Zinn, S. (2012). A mate-matching algorithm for continuous-time microsimulation models. International journal of microsimulation, 5(1):31-51.