利用职位发布数据进行研究

Data ScienceAkshat Mittal(i)+12125267823akshat.mittal@barclays.comBCI, USVincenzo Potatii)+44 (0)20 3134 2097vincenzo.pota@barclays.comBarclays, UKRenate Marold(v)+1 212 526 2484renate.marold@barclays.comBCI, US Thedatagobacktolate2oo7,coveringalmost7mnactivejobpostingsatitspeakrightaftertheCOvID pandemic trough, and averaging 5.5mn since 2024 (Figure 1). LinkUp covers worldwidejob postings, which enables us to extend our previously Us-focused analysis to other markets,like Europe.FIGURE 1. Active job openings show the expected pattern of job availability, including the drop during the CovID pandemic and subsequent22233忆aActive job postings calculated as the total number of active postings on last day of each month across all companies and geographies.postings begins in 2012, with data available for c.200 countries in recent years (Figure 2). Theglobal distribution of the data is heavily skewed towards the US, with the next biggestcontributors being the UK, Canada, India and the Netherlands (Figure 3).9FIGURE3....butisheavilydominatedbyUSjobseLinkUp Job Posting Countc2502001501005002019202120232025USAOtherGBRCANINDDEULinkUp CountryGBR represents The UK; DEU represents The NetherlandsSource: LinkUp, Barclays ResearchForUKjobpostings,therewasaninitialbumpincumulativejobpostingsin2o14(Figure4),whichis whenthevendorstarted scrapingcompany websitesforUK-based job postings.Inthisnote,theUk jobpostingsdataweanalyzefromLinkUpstartin2o20.Figure5showsthepoint-in-time monthly coverage ofjob postings for the share of companies in the S&P 500 and thecoverage of the STOxx 600 has been steady since around 2020.2 Closer to source: pros and consGetting job postings directly from company websites should result in higher-quality data thanscraping job postings from aggregators, for a couple of reasons.: Scraping jobs from a company's own website should obviate the need of extra parsing tomatch the job opening to the company. This should result in greater precision of mappingbetween companies and theirjob postings.: The data generating process should ensure uniqueness of job postings. This contrasts withsame posting is not repeated across multiple sources. In the past, we have found that suchde-duplication is done imperfectly onthe vendor's end.In general, we expect these data to be of higher quality for companies that exhaustivelymaintain their career page than for those that post job openings on aggregator websites, or acombination of the two. Care must be taken when doing systematic studies includingcompanies that may preferentially, or even exclusively, hire through job aggregators or otherthird parties.Scraping job postings primarily from company websites also comes with its own shortcomings.With this, the set of job postings we can track is limited by the company websites the vendorchooses to scrape. This makes it much more likely to miss job postings by smaller, lesser knowncompanies, as well as postings forjobs like nannies and handyman that show up in our otherjob posting data set, Lightcast.Potential biasesAll alternative data come with their own sets of biases. To understand the potential biases inthese data, we need to understand how the data are generated. A job posting is included in thedata set if the vendor scrapes the company website where the job is posted. This implies thatthe volume of job postings can change because the vendor added new websites to scrapes, asopposed to a change in the volume of postings for existing companies.Keepingthepanelof companiesfixed overtimeduringanalysis canmitigatethis selectionbiasto some extent, but the appropriateness of such a solution depends on the analysis itself. This isalso demonstrated later, when the comparison of LinkUp data with official job openings Use Case l: Deletions can indicate company healthand ease of hiringFor each job posting, LinkUp includes the deletion date, which shows when the job posting wastaken down. This enables us to track the job postings for a company in three different ways:Active job postings: the number of job postings that are active on a day.New job postings: the number of previously unseen job postings encountered on a day: Deleted job postings: the number of job postings deleted on a day.Knowing when job postings are taken down holds orthogonal value to what is known fromwhen new job are posted, which opens up new opportunities for research.: Ajob posting deletion date allows us to calculate the length of time a job opening staysunfilled, which we can then track across companies and job types. A lengthy unfilled jobopening may indicate hiring difficulties, high thresholds to hire, or low hiring intent(companies might continue to scour the labor force), providing additional insight into the job,the company and the general workforce.: The rate of job posting deletions relative to total employee count of a company may provide asignal for the financial health of a company. For example, a company with a high rate of jobposting deletions b