Using High-Frequency Data to Monitor Real-Time Labor Market Updates
The average time-lag is around 4 months across the Arab world between the publication of a labor force survey (LFS) and the measured time period, compared to a 3.5 week lag in countries such as Japan and the U.S. In Saudi Arabia, the LFS for Q4, 2022 still not been published as of March 29, 2023, nearly 3 months since the measurement period. This time-lag is particularly problematic for policymakers because 1) it inhibits their ability to respond in a timely manner to rapid changes in private sector employment, and 2) it reduces the value from such statistics. Specifically, one can imagine that there is little to no value for a policymaker to learn about the unemployment rate for Q4, 2022 while we are starting Q2, 2023.
As a result, I argue, as I have in the past, that the General Authority for Statistics (GASTAT), and policymakers in Saudi Arabia should utilize high-frequency data sources such as online job postings. In fact, even in countries like the U.S. where LFS data are published every four weeks, policymakers at the Federal Reserve have demonstrated that using online employment data to monitor the health of the labor market is as informative as monthly LFS statistics, albeit costing significantly less than conventional surveys. This is because the textual data from online jobs ads can provide us with information that are either 1) not captured in conventional LFS publications, or 2) not measured at the same granularity. I will provide three examples for such labor market indicators (LMIs) and an explanation for each use-case.[1]
What type of useful LMIs can policymakers gather from online job posts that the LFS does not capture?
1. Real-Time Tracking of The Nationalization Rate
In Saudi Arabia’s context, policymakers frequently monitor the nationalization or Saudization rate for private sector jobs on quarterly basis, but this metric is tracked as a lagging indicator using the LFS data. However, I argue that the nationalization rate can be also monitored as a leading indicator utilizing high-frequency data to enable decision-makers to anticipate future changes in employment. Moreover, by extracting keywords from job descriptions from sites such as Glassdoor and Indeed, I created both weekly and monthly measures of what I call, a supply-side nationalization rate. This essentially captures the monthly proportion of jobs where Saudi firms are strictly looking to hire a citizen to fulfill their nationalization quotas. I estimate that on average, 19% of monthly online job postings were mainly looking to hire Saudi workers.[2] This can be valuable for policymakers not only because it provides them with real-time information about the labor market but also because it can be then compared against actual nationalization rates. For instance, one may explore if the estimated 19% monthly nationalization rate was below or higher than the real employment data amongst citizens. If the actual nationalization rate is lower than the monthly average of 19%, then this allows policymakers to further investigate why firms could not find enough citizens. This might be because 1) the imposed nationalization rates are unfeasible to satisfy (i.e., due to an insufficient number of Saudis wanting those jobs), or 2) the wages offered by such jobs are too low.
2. Monitoring Firms’ Preferences to Employ Saudi Women
I perform the same technique above to compute the proportion of jobs where certain words indicate that firms are strictly looking to employ a Saudi female despite the job being suitable for both genders. More specifically, I exclude jobs where firms need to only hire women for segregation purposes such as a female-only elementary school.[3] On average, I estimate that 9% of online job ads are mainly looking to hire a Saudi women. Although this figure may seemingly appear negligible, but one should keep in mind that this only reflects the share of firms that are explicitly stating their preference to hire Saudi women over Saudi males or foreign labor. The actual figure might be higher than this estimate, however. Interestingly, the most common occupations where firms are strictly recruiting Saudi women were concentrated in mid to low-skill jobs such as:
· Hotel receptionists,
· Executive assistant and similar secretarial roles, and
· Waitresses and café baristas.
3. Does a preference to employ women translate into higher earnings for women?
Anecdotally, it’s not uncommon to see some Saudi men complaining on Twitter about females overtaking employment opportunities and the well-paid jobs. I showed evidence for the first concern, and now I will test whether the second concern is valid. To test this, I compute the average monthly wage offered by firms in the following graph. I categorize wages across four categories depending on their gender and nationality hiring preferences:
· Jobs that are strictly looking to hire Saudi women,
· Jobs that are strictly looking to hire women but with no nationality preferences,
· Jobs that are indifferent in terms of gender, but are strictly looking to hire citizens, and
· Jobs that are indifferent in terms of both gender and nationality.
Interestingly, the data below debunks the claim that firms are paying Saudi women higher wages than Saudi or foreign men. In fact, the average offered wage for Saudi females was the lowest at 4,500 Riyals, versus 7,725 Riyals for firms that want a Saudi worker but are indifferent with their gender. While I am cautious in citing this as evidence of a gender pay discrimination, it certainly discredits the idea that Saudi women are taking over the highest paid jobs.
In conclusion, I believe the aforementioned examples show the value of high-frequency data and the type of insights that can be generated from online job postings, which are not possible to collect using conventional surveys. Furthermore, I believe that employing high-frequency data is not a substitute but rather a complementary method to existing LFS data. Finally, in upcoming articles, I will showcase how online job ads can be used to monitor crucial labor market patterns that are otherwise difficult to measure with conventional data sources.
[1] All data here are based on online job postings for the past 6 months.
[2] This is based on all unique job postings in Saudi Arabia for the past six months.
[3] Other examples include a physical training instructor in a female-only gym or a nurse in a female-only healthcare facility.
