Can online job post data be used to predict labour market movements in South Africa?

71point4 > Blog archive > 2020 > July > 22 > Can online job post data be used to predict labour market movements in South Africa?
71point4 > Blog archive > 2020 > July > 22 > Can online job post data be used to predict labour market movements in South Africa?

Can online job post data be used to predict labour market movements in South Africa?

Posted by: Jessica Robey
Category: Covid-19, Data analytics, Employment

In this blog, Abri de Beer, Frances Whitehead and Hanjo Odendaal , look at the potential benefits of using online job ads to track labour market changes.

The exponential increase in global internet usage over the past few decades has fueled a growing reliance on the internet for our daily needs, including applying for jobs. This has significantly affected the nature of the employer-employee matching process, addressing the asymmetry of information which has costly outcomes for both parties. The introduction of online job advertisements has reduced geographical barriers and response times while also allowing for detailed job descriptions, a high volume of applications and consequently, more efficient matches [1].

Considering the increasing prominence of this rich dataset in recent years, 71point4 has started tracking advertisements posted on a well-known online job-listing website. This tracking process started in December 2019 and the data is updated daily. To date, approximately 90,000 job postings have been collected.

In other economies, this data has been shown to be highly correlated with total hiring trends, and hiring trends, in turn, are closely linked to rates of unemployment.

Recent data on online job postings in South Africa illustrates the significant impact of lockdown. After the lockdown was announced at the end of March, there was a monthly drop of 44% in the total number of ads posted online. There has been no sign of recovery since the lockdown was eased.

A provincial analysis of job advertisement data shows that Limpopo and Mpumalanga experienced a decline of up to 70% in the number of advertisements posted between March and April. In Gauteng and the Western Cape, where employment rates are highest, the number of online job advertisements posted dropped by 55% and 62% respectively.

These numbers closely resemble trends in the UK published by the Bank of England. Their researchers saw that daily job ad posting on one of the country’s biggest job searching platforms fell by more than 50% since the start of the pandemic. This data is part of a study aimed at combining macroeconomic and epidemiological models to understand the relationship between the virus and the economy [2].

One of the major benefits of online job advertisements is that they provide data that is far timelier than labour force surveys, allowing for the early detection of job vacancy and employment trends. The real-time nature of this alternative labour market data allows for a weekly comparative analysis and provides access to granular data with a high number of data fields embedded in a single job advertisement [1].

A study by Templin and Hirch (2013) compared data from different job posting platforms in the United States (US) with their Quarterly Workforce Indicators (QWI). The results of the study indicated that data from online advertisements can be used as a reliable leading indicator for hiring trends. Although the quarterly online job advertisement volume and total hires data were found to be poorly correlated, this could be accounted for by lagging the total hires data by one quarter [3]. The image below shows that online job advertisements can also serve as a reliable proxy for job openings in the economy if collected correctly. The Job Openings and Labor Turnover Survey (JOLTS) data referred to in the figure is data that is collected on a monthly basis by the US Bureau of Labor Statistics from selected establishments – stores, offices, factories, and other employers. Data collection is done by the way of telephonic interview and an online portal [4].

The image below showcases a study conducted by the Economic Policy Institute on job openings and unemployment. It is clear that job openings and unemployment in the US are inversely related (i.e. as the number of job openings increase, overall unemployment decreases) [5]. In a struggling economy, the number of available job openings will decrease, thereby reducing the labour absorption rate and fueling unemployment.

Although this alternative data source has some merit, online job advertisement data has its limitations. The data is prone to error based on the data extraction processes and algorithms used by job websites [6]. In addition, online job advertisement data may not necessarily paint a complete picture of labour demand; not all ads are posted online, and those that are posted have been found to display a bias toward relatively high-skilled white-collar jobs [1]. These limitations make this type of data more suitable for understanding the formal jobs in South Africa as opposed to the labour market as a whole.

Despite the limitations of online job advertisement data, more traditional labour force survey data does not offer the same level of detail nor timeliness. The two data sources should therefore be seen as complementary. Where government survey data lags, online job advertisement data is timely; where online job advertisement data may not cover the market, government survey data is designed to be nationally representative.

Unfortunately, there is no equivalent to the United States’ Job Openings and Labor Turnover Survey (JOLTS) data in South Africa. With sufficient time however, there is certainly a case to be made for the use of online job advertisement data in conjunction with QLFS data in order to track unemployment and labour market trends in a more timely manner. Recent labour market shocks caused by the COVID-19 pandemic strengthen the case for this approach. Online job advertisement data could potentially reflect the impact of COVID-19 almost immediately, while the QLFS data will only reveal how the pandemic has impacted the labour market in the next quarterly release.

Authors: Abri de Beer, Frances Whitehead & Hanjo Odendaal

References

[1] Carnevale et al. 2014. Understanding Online Job Ads Data. https://cew.georgetown.edu/wp-content/uploads/2014/11/OCLM.Tech_.Web_.pdf (Accessed 15 July 2020).

[2] Gibney, E., 2020. This physicist-turned-economist is modelling the pandemic’s financial fallout. Nature 2020 581:7807.

[3] Templin & Hirsch. 2013. Do Online Job Ads Predict Hiring? https://www.gc.cuny.edu/CUNY_GC/media/365-Images/Uploads%20for%20LMIS/Reports%20and%20Briefs/NYCLMIS-RESEARCH-BRIEF-Do-Online-Ads-Predict-Hiring.pdf (Accessed 15 July 2020).

[4] U.S. BUREAU OF LABOR STATISTICS, 2015. Job Openings and Labor Turnover Survey – Data Collection [WWW Document]. URL https://www.bls.gov/jlt/jltcoll.htm (accessed 7.22.20).

[5] Gould, E., 2020. The U.S. economy remains in an enormous jobs deficit [WWW Document]. Economic Policy Institute. URL https://www.epi.org/indicators/jolts/ (accessed 7.22.20).

[6] Kurekova et al. 2013. Online job vacancy data as a source for micro-level analysis of employers’ preferences. A methodological enquiry.

Like our blogs? Subscribe to receive email updates when new blogs are published.

Author: Jessica Robey

Leave a Reply