Online job vacancy attractiveness: Increasing views, reactions and conversions

• Published in: Electronic commerce research and applications Vol. 55; p. 101192
• Main Authors: Košťálová, Zuzana, Lyócsa, Štefan, Štefánik, Miroslav
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2022
• Subjects: E-recruitment; Machine learning; Morphological text analysis; Online job vacancy; Webpage attractiveness; Webpage views; E-recruitment; MAE; Webpage views; Morphological text analysis; Machine learning; Webpage attractiveness; OJV; MCS; Online job vacancy; MSE
• ISSN: 1567-4223; 1873-7846
• DOI: 10.1016/j.elerap.2022.101192

•We study the predictors of online job posting attractiveness: views, reactions and conversions.•In terms of the MSE and MAE, the linear models do not perform well.•Random forest consistently outperforms the benchmark and linear models.•Variable importance is tested within an MCS approach.•Job classification and benefits help predict attractiveness.•The morphological characteristics of the job title are also a useful predictor. E-commerce development has not bypassed the labor market, and employers seeking to hire are posting vacancies on specialized online platforms. Operators of such online platforms and employers are assessing the attractiveness of job vacancies, as they are interested in making online job vacancies (OJVs) more attractive to job seekers, i.e. to increase the number of views of the posted job ad, the number of times job seekers submit an application form and conversions (the ratio of the two). However, given that job types and job seekers are extremely heterogeneous, it is difficult for employers and specialized online platform operators to design attractive job ads. In collaboration with a leading OJV platform in Slovakia, we study whether machine-learning methods can improve the predictions of OJV attractiveness on a sample of 32482 OJVs that offer as many as 883 job features. Our study shows that, as opposed to various linear models, considerable prediction improvements can be achieved using random forest, which exploits possible nonlinear relationships between combinations of explanatory variables and measures of attractiveness. Based on this insight, we perform a statistical evaluation of key variables of importance. We find that the job classification, job benefits, and variables related to a simple morphological description of the job and job title are relevant. The results of this study can help the operators of specialized job vacancy platforms and employers improve their job ads to attract more job seekers.