Extracting More Value from Job Vacancy Information (Methodology Part 2)
| Autor | Jeisson Cárdenas Rubio |
| Páginas | 118-149 |
118 C
A WEB -BASE D APPRO ACH TO M E A SU R E S KILL MISM ATCHE S AND S KILL S PROFI LES FO R A DEV ELOPI NG COUN TRY
6. Extracting More Value from
Job Vacancy Information
(Methodology Part 2)
119
c
6.1. Introduction
The previous c hapter has shown that infor mation from job portals m ight pro-
vide detailed labour demand in formation such as educational requ irements
and experience, among other vacanc y characteristics. However, what makes
job portals a pot ential and remark able source of data is that t hey might provide
detailed information in real-time about the skills and occ upations demanded
by companies. As d iscussed in Chapter 2, the dynamic between the skills or
occupations offered by individuals and the sk ills or occupation s demanded
by employers is a relevant factor that has str ong implications on outcomes
for productivit y, wages, job satisfaction, turnover rates, unemployment, etc.
(OECD 2016a; Acemoglu and Autor 2011). Indeed, the m ismatch between the
supply and demand for skills mig ht explain a considerable share of unem-
ployment and infor mality rate s in Colombia (see Chapters 2 and 3). Despite
the relevance of t his topic, detailed informa tion (from official sources suc h as
ONS) for the analy sis of the labour demand for sk ills is relatively sca rce due to
methodological is sues and the high cost of collecti ng detailed labour demand
information (Chapter 4). Thus, the key ta sk of this chapter is to describe the
techniques t hat can be utilised to e xtract informat ion on skills and occu pations.
As mentioned in Ch apter 5, information from job por tals is not categorised
with stati stical analysis in mind. For instance, non-categor ised information
related to ski lls and occupations (for the Colombian case) ca n be found in job
descriptions a nd job titles, respective ly. Consequently, this chapter expla ins the
steps requir ed to organise and categorise sk ills and occupational i nformation
from the vacanc y database. Section 6.2 of t his chapter develops a methodolog y
to identify sk ill patterns i n job vacancy description s based on internationa l skill
descriptors, such as the ESCO. However, there might be some country-spe-
cific skil ls that are not listed in the ESCO d ictionary, or its internationa l skills
descriptors m ight not be updated according to the mos t current labour dema nd
C c
120 C
A WEB -BASE D APPRO ACH TO M E A SU R E S KILL MISM ATCHE S AND S KILL S PROFI LES FO R A DEV ELOPI NG COUN TRY
requireme nts. Therefore, Section 6. 3 proposes a methodology to automat ically
identify country-speci fic or new skills from information from job portals.
The classification of job titles into occupations is a cr itical stage for
vacancy ana lysis. Correctly coding the job t itle variable requires different
and advanced dat a mining techniques. Therefore, Section 6.4 describes and
applies techniques such as manua l classificat ion, software classifiers, a nd
machine lear ning to organise job titles i nto occupational groups. Th is section
also proposes a met hod that uses un structu red information from job titles
and skil l requirements (variable s created in the previous steps) to ident ify the
occupational g roups of hard-coding vacancies. W ith this last procedure, the
vacancy database is completely organised.
Once the vacancy database is organ ised and categorised into occupa-
tional group s, educational require ments, etc., this helps to identi fy duplication
problems at this stage. A job vacancy advertisement might be repeated as an
employer might adver tise the same vacancy m any times on the same job por tal
or between dif ferent job portals. T hus, Section 6.5 deals w ith duplication issues.
With the vaca ncy data variables orga nised and categorised and t he dupli-
cation problems mi nimised as much as possi ble, an imputation process can b e
conducted for certa in variables. As shown in Chapter 5, vacancy data mig ht
contain a considerable number of missing values in the var iables of interest
(e.g. educational requirements and wages of fered). This m issing information
might create bia ses in the later analys is of labour demand requ irements. Thus,
Section 6.6 outlines how missing va lues were imputed for the “educational
requirement” and “wage offered” variables by usi ng predictors such as occu-
pation, city, and experience requir ements, among others. Finally, Section 6.7
presents consolid ated, organised, categorised, cleaned, and imputed data for
the analysis of the Colombian labour demand using job por tal sources. Next,
Figure 6.1 provides a su mmary of the above descr ibed steps that were imple-
mented to organise Colombian vacancy information.
Para continuar leyendo
Solicita tu prueba