Labour markets in Canada and around the world are evolving rapidly with the digital economy. Traditional data are adapting gradually but are not yet able to provide timely information on this evolution and the changing nature of jobs. The increasing use of the internet for job searches provides an opportunity both for workers to access information about job vacancies and for researchers to learn what types of jobs people look for.
In this interactive tool, we present the results of a collaboration between Microsoft Research and the Bank of Canada using Bing online search data to characterize job searches in Canada.
Microsoft Research has been exploring the possibility of using Bing search data to help researchers understand developments and emerging trends in the United States labour market. To this end, researchers at Microsoft classified almost 250 million online job search queries from 2015–16 into 14 employment sectors—architecture/engineering, arts, business, construction, education, finance, food, health care, leisure/hospitality, manufacturing, retail, science, technology, and transportation—and a generic residual job type . Details on the methodology can be found in Chancellor and Counts (2018). The job search data across the United States can be explored in an interactive tool developed by Microsoft.
In a collaboration, Microsoft Research and the Bank of Canada used the same methodology to look at job searches in Canada. Though the approach to the Canadian case is broadly the same, small adjustments in the job title dictionary take into account unique characteristics of the Canadian labour market. The interactive tool presents the job search data for 2017. Queries are aggregated by geographical area where the search was originated (293 census divisions), and both the proportion of jobs sought by type and actual employment distribution (from Census 2016) are plotted for every census division in Canada.
A few characteristics of the data are worth highlighting. First, this tool focuses solely on searches for jobs that belong to one of the predefined employment sectors. These represent 47 per cent of online job searches in Canada conducted using Bing. The other 53 per cent are considered generic job searches and are not displayed in the dashboard. Second, job searches are based on the location of the person doing the search and not on the location of the job. For example, a person in Ottawa looking for a job in Toronto constitutes a job search in Ottawa. All job search queries conducted in a language other than English were excluded from the results since the original methodology developed for the United States did not include other languages. In Canada, the exclusion of French queries may result in some employment sectors being underrepresented, specifically in the province of Quebec, where French is the primary language of communication . Future analytical work will include French queries. Lastly, job search queries for the oil, mining and gas sector have not been explicitly classified, despite the importance of this sector in Canada, because the original methodology did not separate out this sector .
This interactive tool also allows us to correlate job search data with different demographic and socio-economic traits (from Census 2016 unless otherwise stated), including education, population, income, age and employment . The data presented are for the 293 census divisions. All demographic and socio-economic traits are displayed in percentile rank .
To understand patterns of online job searches, we performed a series of bivariate regressions of the proportion of jobs sought by type at the census division level on each of the demographic and socio-economic characteristics . For this exercise, we exclude Quebec due to the lack of representative information from the classification. We highlight some of the results from this simple exercise in the next section.
The first observation that sticks out from the data are that 20 per cent of online job searches in Canada using Bing is for occupations in health care; this contrasts with 16 per cent employment in the sector. There seem to be more searches for jobs in health care in areas with a larger population of dependent-age individuals, both young and elderly (a decrease of 10 percentage points in the prime-age population in a region is associated with an increase of 2 percentage points in job searches in health care).
Jobs in education and business are also relatively popular among online job seekers; each corresponds to 11 per cent of job-related online queries. Jobs in business, such as human resource specialist, publicist or project manager, appear highly sought after in areas where the average education level of the population is higher (the rate of job searches in business is higher by 2 percentage points in areas where the proportion of adults with at least a bachelor’s degree is 10 percentage points higher). Jobs in the education sector, while not apparently related to the education level of the population, seem relatively more popular in areas with a higher proportion of the prime-age population (an increase in 10 percentage points in the prime-age population correlates with a 1.5-percentage-point increase in the number of job searches in the education sector). Finance is another relatively popular sector for online job searches, comprising 7 per cent of searches. It is positively associated with income and education levels in the area.
Job searches related to science—such as mathematician, chemist and archeologist—and technology—such as programmer, data scientist and systems engineer—represent 3.5 per cent and 4 per cent of total searches, respectively. Searches for these jobs, especially technology-related jobs, take place more often in areas with higher average levels of education and areas with higher average levels of income. Jobs in construction (6 per cent of all online job searches) and the arts (4 per cent of all online job searches) seem more attractive to job seekers in areas with lower average levels of income and areas with lower average levels of education.
Even though the distribution of job queries seems to match the analogous distribution of employment, there are some differences worth mentioning. Jobs in the retail and food sectors tend to be less represented in searches than in employment. In geographical areas characterized by lower average levels of income and education, jobs in the retail sector seem to be less represented in searches than in employment. In areas with higher population density and a higher average level of education, jobs in the food sector appear less sought after compared with the actual number of jobs.
If you have comments, questions, or ideas about these data, contact us at email@example.com.
Chancellor, S. and S. Counts. 2018. “Measuring Employment Demand Using Internet Search Data.” In CHI ’18. Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, April 21–26. Montréal, QC: Association for Computing Machinery.
The authors are grateful to Bob Fay and Marc-André Gosselin for helpful comments and suggestions. We also thank Carole Hubbard and Alison Arnot for valuable editorial suggestions.
Bank of Canada staff analytical notes are short articles that focus on topical issues relevant to the current economic and financial context, produced independently from the Bank’s Governing Council. This work may support or challenge prevailing policy orthodoxy. Therefore, the views expressed in this note are solely those of the authors and may differ from the official Bank of Canada views. No responsibility for them should be attributed to the Bank.
 Job search queries are identified as queries containing keywords such as “employment,” “job(s),” “positions” and “career(s).” The researchers created a dictionary with a collection of related words for each employment sector.
 In the rest of Canada, excluding Quebec, English queries comprise more than 95 per cent of all job search queries. In Quebec, English queries make up only 4 per cent of the job search queries.
 Job searches in the oil, mining and gas sector are likely included in other classifications such as construction and science, which are overrepresented in Alberta compared with the national average.
 Employment status and age are added to the Canadian analysis, although they are not included in the US analysis.
 The education characteristic is based on the percentage of the population aged 15 years and over with a bachelor’s degree or higher level of education (Catalogue number 98-400-X2016242). Population data are from the census divisions’ population aged 15 years and over (Catalogue number 98-400-X2016365). The income measure is derived from the census program’s average employment income (Catalogue number 98-400-X2016120) translated into real terms to account for different purchasing power across the country. The measure of purchasing power comes from the Survey of Household Spending, which provides an estimate of the annual household total current consumption for all 10 provinces (CANSIM Table 11-10-0222-01) and the capital of each of the three territories in Canada (CANSIM Table 11-10-0233-01). The relative consumption value is calculated by dividing each of the provincial and territorial consumption values by the national average for the reference year 2015—the latest data point for the territories. Average employment income is then deflated by the relative price for each area to create a measure of real income by census division. The age trait relates to the share of the population in the prime age (those aged 25 to 54). The share is computed by dividing the population aged 25 to 54 years by the population aged 15 years and over. The employment attribute is based on the census divisions’ employment rate, provided directly from the census program (Catalogue number 98-400-X2016365).
 The socio-economic and demographic variables used in this exercise are (i) proportion of population with a bachelor’s degree or higher level of education, (ii) population in natural logarithm, (iii) income in natural logarithm, (iv) proportion of prime-age population, and (iv) employment rate.
% w/ a least a BA
15 years or older
Average real income
% in prime age (25-54)
|% Jobs Searched||% Jobs Worked|