The sample of online job postings found on Indeed in many of the world’s largest economies is broadly representative of the overall universe of job postings as measured by various governmental statistical agencies, according to a recent benchmarking analysis. The analysis found that the universe of Indeed job postings largely aligns with governmental measures of job openings and new hires across several dimensions, including in the aggregate and across sectors, as well as for wage data.
No single private or governmental source can be expected to fully capture the entire universe of jobs, job openings, and new hires by themselves. The survey-based methods used by many governmental agencies will lose some visibility into the full market based solely on known challenges in collecting adequately sized and representative survey samples. And by definition, online platforms – even those, like Indeed, that extensively aggregate (and deduplicate) postings from multiple sources in addition to sourcing proprietary postings – can only observe that portion (albeit a large one) of open jobs that are advertised digitally.
But understanding how well these two main sources of labor market data overlap, and where there are differences, is crucial to giving researchers confidence that any insights gained or trends identified using Indeed data are a true signal of real forces acting in the market. As more job market activity moves online, the analysis gives confidence that the strong correlation between public and private data observed today is likely to continue in the future. A version of this benchmarking analysis was included as part of a recent paper by Hiring Lab economists and the Central Bank of Ireland that used wage data in Indeed job postings to construct what has subsequently become the Indeed Wage Tracker.
The authors found that aggregate online job postings correlate well with overall trends in job vacancies from official sources, and line up with expected economic patterns. For example, the sharp decline in job openings experienced in many countries during the initial stages of the pandemic was observed in both the online data and official sources, along with the strong rebound in openings once economies began reopening. While one online job posting on Indeed does not necessarily equal one job opening as measured by an official source, both approximate the demand for new workers and trend closely together across most of the nations analyzed. These patterns should inspire confidence that Indeed’s online job posting dataset captures overall trends in job openings.
Composition
The authors also compared the composition of Indeed online job postings and published job vacancy statistics. Here, differences in how data are classified by governments versus by Indeed do present a challenge in directly comparing the series. Indeed’s online postings are classified by job title and occupational category, while many government vacancy data series are available only by sector. In many occupations, including in the construction, transport, and retail sectors, it can be relatively easier to map a given job title (i.e. “driver”) to a given sector (i.e. “transportation”). But for other occupations or titles (i.e. “manager”) an obvious sectoral match is less clear. Where there was no obvious direct occupation/title-sector match, the authors assigned occupations to sectors based on occupation-sector share published in various governmental labor force surveys.
The first takeaway is that Indeed job postings capture all major sectors and occupations in each country’s economy, rather than pertaining to only low- or high-skilled jobs (as was the case with some online job posting datasets in the past). While not a perfect match, the sectoral composition of Indeed job postings and job vacancies from national statistics is very similar. In all countries, the Indeed data is over- or under-represented in certain sectors, and there are some patterns, but they do not apply to all countries equally.
For example, when the analysis was conducted (in 2023, using data from 2018-2022), construction was the only sector that appeared to be over-represented in online job postings in all eight countries. Accommodation and food was consistently under-represented in all countries, perhaps due to a high share of informal/difficult-to-observe hiring methods in that sector, or because it is a sector where each online posting relates to more than one vacancy to a greater extent than in other sectors. Transport and storage, wholesale/retail, administrative, and professional services tended to be over-represented in online job postings in most (but not all) countries, although to a different extent in each one. In contrast, manufacturing, finance, and healthcare tended to be under-represented.
Overall, this exercise suggests that Indeed’s online dataset captures a slightly different sectoral mix of jobs than survey-based vacancy estimates, though some of those differences could be driven by the imperfect ability to map postings to sectors. For instance, if a high share of vacancies in the healthcare sector is for administrative roles, then we will underestimate the number of postings in the healthcare sector and overestimate the number of postings in the administrative services sector. The impact is difficult to test since vacancies in official surveys are not broken down by occupation. Even so, despite some apparent differences in the sectoral mixes of Indeed postings and official vacancy data, the quarterly sector shares from the two are strongly positively correlated, suggesting postings and vacancies evolve similarly in most sectors over time.
New Hires
The authors also examined the intersection of Indeed job postings and official statistics with data on new hires, with a focus on occupations rather than sectors, given better availability of data. Overall, this analysis found the occupational mix of the online job posting data is very similar to that of new hires. For most countries, including the UK, Ireland, France, Germany, and the Netherlands, the comparison is very good, with correlation coefficients of the quarterly occupation shares in the two datasets ranging from 0.72 to 0.86. For Italy and Spain, the correlation is weaker (but still positive) at 0.65 and 0.53, respectively. In both Italy and Spain, a relatively high concentration of online job postings in a small number of occupations – such as ’Retail & Sales’ and ’Software Development’ – drives this result.
Wages
Finally, the authors also analyzed the level and composition of wage data published in online Indeed job postings with other publicly available wage data. The results suggest that wage data from Indeed job postings effectively captures the percentiles of the new hire wage distribution. If there is a part of the Indeed wage distribution that is comparably lower than in the official data, it is the highest-paid workers, but only in certain countries. In Ireland, France, the Netherlands, and the UK, the top decile of wages from the population was above the top decile in the Indeed data at the time of the analysis, for two possible reasons. First, higher-paid jobs may simply be less likely to include advertised wages in the job posting. Second, advertised pay in some higher-paying jobs may be more akin to a floor or starting point for potential negotiations.
For each country analyzed, the authors calculated the absolute difference between the official, governmental survey estimate of new hires’ earnings at each percentile, and the estimate from the Indeed data, and took an average across percentiles. Relative to the population estimate from the official surveys, the Indeed data tends to be closest in Italy, France, and Germany (where the statistic ranged from 0.035 to 0.055 at the time of analysis). For the US, UK, and Ireland (where it ranged from 0.070 to 0.092) it is slightly less aligned. The authors also compared average wages by sector from the online job posting data with average monthly wages by sector from national statistical agencies and found that average wages by sector extracted from online job postings match well with average wage levels by sector from official statistics. According to both datasets, workers in the IT and/or Professional sectors tend to have the highest average wages in most countries, while average wages tend to be lowest in the Accommodation and Food, Other Services, Transport and Storage, and Administration sectors.
Conclusion
The overall conclusion from these validation exercises is that Indeed’s online job posting data captures all major sectors and occupations in the labor market. It aligns well with job vacancies and new hires across both sectors and occupations, while wages extracted from online Indeed job postings are fairly representative of the wages of new hires.
