TWIG’s JobsTO relies on job posting information scraped from a variety of internet sites. It is then processed and “massaged” to produce information about the current and past state of job openings in Toronto. The result is almost up-to-the-minute information on the conditions in Toronto’s labour market. It also allows for comparison over time and the possibility of identifying trends. Job posting data is the only easily attainable way to get current labour market information on job openings.
However, this approach has it limitations. The observations are not statistically sampled resulting in less reliability and potential bias in the current estimates and historical trends. Job posts are repeated because multiple websites are analyzed, and a single job may be posted on multiple sites. Or, an employer finds the cost and/or number of steps required too onerous and forgoes posting the job. Online job posting tends to favour professional, retail, and service jobs while skilled trades, union, and manufacturing openings are not all posted. It can be difficult to precisely categorize a job posting to an occupation code and industry, and, sometimes, even the location is not clear as the home office is listed but the actual work location is not or is “various”. A job posting may not actually represent a job opening – a company will post the job in anticipation of needing to fill it but may not have an immediate need.
Despite these limitations, online job posting data is still widely used because of its availability and timeliness. It’s often a case of “hold your nose and carry on”. A more nuanced approach is to use online job posting data in conjunction with more statistically validated labour market data such as the Labour Force Survey, Census, or Job Vacancy and Wage Survey (JVWS).
Recently, the Labour Market Information Council released a “demonstration report” on Benchmarking Online Job Postings to the Job Vacancy and Wage Survey to Improve Vacancy Estimates. The report applies machine learning and statistical benchmarking to improve the accuracy of job vacancy estimates using online job postings. This report is geared toward a technical audience and is not for the econometrically faint of heart. However, by comparing online posting data with the JVWS, it develops some alternative estimation processes that markedly improve the reliability of online posting data while still allowing for the timeliness provided by the online data.
