To better understand the heterogeneity of managers’ responses and the factors affecting firms’ difficulties, we construct a rich dataset linking the BMO manager’s survey to four other sources. We compiled data from firms’ balance sheets (FARE), from matched employer-employee data on social security contributions (DADS), from the labour force surveys (LFS), as well as from administrative sources to construct local geographical characteristics of the company.
Among the observable characteristics, occupations sought is the dominant explaining factor. It contributes to three-quarters of the variance explained by the model, but only to 10.5% of the total variance. Company characteristics (size, revenue, sector, etc.) contribute about 10% of the explained variance, which is 1.4% of the total variance. The geographical characteristics and economic conditions of the firm - the employment area population density, the size of the urban area, the existence of an attractive area, the local area unemployment rate - explain about 15% of the variance explained, which is 2.1% of the total variance. In particular, larger firms and firms located in an employment zone with a high population density express fewer hiring difficulties. However, the difficulties expressed by the companies increase with the wages and the level of education required.
In conclusion, our rich model leaves the lion’s share of the variance in hiring difficulties unexplained. Unobservable firm-level factors seem to play an important role in explaining the recruitment difficulties encountered. Consequently, the policy response to anticipated recruitment difficulties should be thought of in terms of targeted actions aimed at firms anticipating these difficulties.
This requires a better understanding of the reasons behind this perception, as companies of similar size, location and sector do not express the same difficulties. In addition to firm-specific issues such as HR, management quality and employer's reputation, support for hiring firms should focus on specific actions towards the hard to fill occupations for which recruitment appears to be the most problematic.
Our analysis has many limitations worth recalling: it focuses on anticipated and not experienced difficulties in hiring. The high level of aggregation of our explanatory variables is undoubtedly insufficient to capture the heterogeneity of recruitment difficulties (82 job families; 10 sectors of activity; 8 levels of density of the employment zone). Finally, this study focuses on Covid's pre-crisis situation, just before a further tightening of the labour market.