Mitchell’s Musings 11-5-2017: Help-Wanted Online

03 Nov 2017 4:50 PM | Daniel Mitchell (Administrator)

Mitchell’s Musings 11-5-2017: Help-Wanted Online

Daniel J.B. Mitchell

Every month, I get an email notice from the Conference Board which contains a chart on help-wanted online advertising (HWOL) as a labor-market indicator. As can be seen below on Figure 1, something happens around 2015-2016 that appears to break the series. Up to that point, the index of online help-wanted ads tended to move up and down with employment. For example, the dip of the Great Recession is apparent on Figure 1. And from the depths of the Great Recession, the labor-market recovery appears as the line reverses its decline and moves upward. But after 2015, the number of HWOL ads heads down although employment remained (and remains) on an upward trend.


Figure 1


In fact, the Conference Board used to publish its HWOL chart with employment also shown as a second line in order to demonstrate the correlation of the two series. However, with the break in the HWOL series after 2015, the Board assumed that something had gone wrong with its methodology and began to omit employment from its chart and to publish the following note with it:

NOTE: Recently, the HWOL Data Series has experienced a declining trend in the number of online job ads that may not reflect broader trends in the U.S. labor market. Based on changes in how job postings appear online, The Conference Board is reviewing its HWOL methodology to ensure accuracy and alignment with market trends.

An interesting question is whether – because the HWOL series stopped moving up with the employment numbers – labor-market trends have really ceased to be reflected. A related question is what value the HWOL series is supposed to be adding. Note that if the HWOL series was 100% correlated with employment, you could certainly say it was reflecting “market trends.” But you could also say that collecting the series had little value to analysts since HWOL was not telling you more than you already knew from the pre-existing employment series, at least at a macro level.

The Conference Board makes available its data for the HWOL series going back to mid-2005. For many years before the HWOL series began, it had published data on newspaper help-wanted advertising. But the value of that information eroded as employers moved much of their advertising for new hires onto the internet. Of course, the internet, the technology for accessing it, and the access of potential workers to it, have all been evolving. Employers presumably are adjusting their use of the internet for help-wanted advertising over time. Some employers still use traditional newspaper ads. And there are other traditional ways of recruiting such as through word-of-mouth with the existing workforce, recruiting agencies, walk-ins, etc.


Figure 2


Apart from the number of HWOL ads per month, the Conference Board has a data series on how many of the ads each month are new, i.e., not just continuations of ads from prior periods. From those data, you can compute an implicit series on how many ads are being dropped each month. Basically given the number of existing ads in a given month, the number of ads in the next month is the result of adding the new ads to the existing pool and then subtracting those ads that are being dropped from the total. Since we know the number of existing ads each month and we know the number of new ads entering the pool, the dropped ads can easily be calculated.

It may well be that the absolute number of HWOL ads decreased after 2015 (despite the continuing increase in employment) because employers starting moving their ads to online platforms not captured by the Conference Board. Or perhaps employers began using other methods of recruitment. We don’t know what happened and apparently neither does the Conference Board. But we can look at new HWOL ads as a ratio to existing ads and at (calculated) dropped ads as a ratio to existing ads. That is, we can look at what can be learned from the world of the sources the Conference Board is tracking, even if the Board is missing some alternative online ad sources or is missing a shift to other means of recruiting.

On Figure 2 above, we show the ratios of new ads to existing ads and of dropped ads to existing ads. Because the two series are noisy, we smooth them with 12-month moving averages. Both ratios suggest that online ads probably don’t stay up online for very long. The two ratios vary between .43 (43%) and .52 (52%). So during the period shown on the chart, typical new ads each month were likely not remaining online for more than two or three months. Put another way, there seems to be a very high churn or turnover in online help-wanted advertising. We don’t know for sure why employers removed the ads that they had previously posted. But surely one reason was that they had quickly found the applicants they needed in response to the ads.

One might further suspect that during recessions, and in the early recovery period after recessions, there would be many displaced applicants looking for jobs. Thus, posting a job vacancy online would quickly produce sufficient job seekers to satisfy the recruiting employer. As the labor market tightens, however, there would be fewer potential applicants; HWOL ads might then need to stay up longer to attract a sufficient number of possible hires.

In broad terms, Figure 2 seems to support those suppositions. During the slack period of the labor market, HWOL ads were coming and going faster than they were just before the Great Recession. And now, when other measures of the labor market suggest a return to labor-market tightness and there are anecdotal reports of labor shortages, HWOL ad turnover has again slowed.

U.S. Bureau of Labor Statistics (BLS) data on the job openings (vacancy) rate show an even tighter labor market now than in the pre-Great Recessions period, as Figure 3 below illustrates. With that market development, we might expect ads to stay up even longer now than in the pre-Great Recession period, and there is some indication of that outcome on Figure 2.

In short, despite the changing practices in employer hiring and recruitment technology, the Conference Board’s HWOL data series likely is reflecting labor market trends even though it stopped rising with employment. It’s just that the information it is providing is not in the absolute number of HWOL ads, but in the implicit insight into the rate of churning of HWOL ads. It seems to be indicating that in booms ads need to remain online longer from the employer viewpoint than in busts. More generally, the labor-market information contained in the HWOL survey seems to be in the duration of ads rather than their absolute number.

It would be useful to obtain more information on ad duration such as median and average lifetimes of ads and their general duration distribution. There may be some ads with very brief lives. Others may remain online for long periods, particularly in occupations where turnover is high and there is continual replacement hiring. We get BLS data on the duration of unemployment of job seekers. Ad duration distribution might provide parallel insight into employers’ searches for new workers.


Figure 3: Job Openings (Vacancy) Rate for Non-Farm Sector


Employment Policy Research Network (A member-driven project of the Labor and Employment Relations Association)

121 Labor and Employment Relations Bldg.


121 LER Building

504 East Armory Ave.

Champaign, IL 61820


The EPRN began with generous grants from the Rockefeller, Russell Sage, and Ewing Marion Kauffman Foundations


Powered by Wild Apricot Membership Software