October 4, 2012
October 4, 2012
By Josh Lehner
Oregon Office of Economic Analysis Blog
One of the indicators used in our office’s Oregon Index of Leading Indicators is Help Wanted Ads. The theoretical basis for this is straight forward: if a business wants to hire another employee, the firm places an ad to notify prospective employees, therefore one would expect the number of help wanted ads to increase before actual employment increases. In practice this is generally the case however there is quite a bit of hiring that takes place in the economy without any official job announcement (i.e. “It’s not what you know but who you know.”), while some postings are general recruitments that are designed to gauge the pool of applicants, not necessarily to fill a position. Hopefully these situations effectively cancel each other out and the number of help wanted ads are representative of the underlying labor demand in the economy.
Unfortunately there are structural and technical issues that make using help wanted ads problematic. The primary structural issue is the decline of the newspaper industry and the rise of online job postings. Prior to the widespread use of online job posting sites, using the number of advertisements in local newspapers was a great way to gauge future employment growth. However today the number of jobs posted in the paper is a faint shadow of its former self, and are about 96 percent below their levels of 15-20 years ago. That is not a typo, at least not for the series we use. The technical issues pertain to data availability – the help wanted online data begins in mid-2005, however the structural changes started long before then – and methodology – how to combine the newspaper ads with the online ads.
Fortunately there are ways to alleviate these concerns and merge the two series together. Up until now our office had been detrending the newspaper ads and using the movements around the trend as a leading indicator. This method works ok but was always a stop-gap measure until an better solution could be obtained. Well, now we have that solution. Thanks to a methodology created by former Federal Reserve economist Regis Barnichon and first brought to our attention by another Fed economist, David Andolfatto, a number of months ago. The methodology breaks down into three components. The first is pre-internet, or prior to the rise of internet use. The second is the period when the use of the internet is increasing however the online help wanted data is not available. Here you can use a technique to estimate the structural decline of newspapers and therefore, by extension, the online demand. The third component covers the period when both measures are available and can be blended together into one series. For all the nuts and bolts and Greek letter equations, please see the first link above. The final results can be seen below.
Note that the composite series should be used as an index and not taken literally as the number of help wanted ads in the local economy.
Finally, how does the new series perform as a leading indicator? While the series is a bit noisy, particularly in the pre-online ads period, it does pretty well, as shown in the following graph.
Like our overall OILI itself, the new help wanted series gives a false signal following the Asian financial crisis in the late 1990s and also did not lead the downward swing of the double dip in 2003, however the remainder of the series is very good. For the Great Recession, help wanted turned negative on a six month basis ten months prior to employment doing so but help wanted turned severely negative seven months prior. On the upswing, the new series turned positive seven months before employment followed suit. In recent years OILI overall, the help wanted series and Oregon employment have all fluctuated in fits and starts, however the underlying trends have been positive.
no comments yet