Discussions

Hi everyone,

I’m Shaokun, a researcher analyzing a large job-posting dataset that includes O*NET occupation codes, posting dates, and location attributes. My research focuses on urban- and spatially related occupations, and one pattern stands out very strongly.

What I observe

1. Job postings classified under O*NET major group 49 (Installation, Maintenance, and Repair) show an extreme surge concentrated in a very short time window.
2. The peak occurs around 2024-08-24, with daily counts exceeding 130k postings, far above surrounding days.
3. When aggregated weekly and compared against all other occupations, the same peak appears across the entire dataset, not just O*NET 49.
4. Spatially, the spike is visible across many states/regions simultaneously, rather than being localized.

What makes this puzzling

1. O*NET 49 jobs are typically tied to urban infrastructure, utilities, facilities maintenance, and construction-adjacent roles.
2. There were no obvious nationwide infrastructure policy announcements or shocks exactly matching this date.

My question

For those familiar with large-scale job posting datasets or labor market data pipelines:

1. How are posting dates typically generated in these datasets? Are they true employer posting dates, crawler ingestion dates, or batch-normalized timestamps?
2. Could this kind of spike be caused by: A large job board (e.g., Indeed, LinkedIn, government portals) being bulk-ingested?

Any insight into how these job datasets are constructed — especially how post_date is defined — would be extremely helpful.

Thanks in advance!