What can 40 years of data on vacancy advertising costs tell us about labour market equilibrium? – Bank Underground

Michal Stelmach, James Kensett and Philip Schnattinger

Economists frequently use the vacancies to unemployment (V/U) ratio to measure labour market tightness. Analysis of the labour market during the current inflationary period often assumes the V/U ratio is constant and compares this measure with a supposed pre-2019 equilibrium. However, the V/U ratio has trended upwards over recent decades. We explore the impact of changing vacancy posting costs on equilibrium labour market tightness through the lens of two models: an empirical error-correction model; and a simple structural ‘search and matching’ model. We find that the raw V/U ratio can be misleading for conclusions about labour market tightness. We outline an improved measure – the VU gap – which indicates that the UK labour market returned to a broadly balanced position in 2024 H2.

Chart 1: The vacancy rate and the V/U ratio have been on an upward trend over the past 40 years

Sources: Bank of England’s ‘A millennium of macroeconomic data’ and ONS.

In a standard ‘search and matching’ model firms post vacancies until the average cost of filling a vacancy equals the expected returns from employing a worker. As hiring is costly, one can think of posting a vacancy as an investment decision, whereby hiring an additional worker depends on the expected profit from filling the vacancy to the firm. Using this framework, we can think of several reasons why vacancies may be structurally higher, as suggested by Chart 1, than they have been historically:

1. Labour productivity growth. Higher productivity raises the asset value of the job, boosting job creation. Other things equal, a rise in the marginal product of labour would be associated with higher demand for labour.
2. Changes in matching efficiency. A rise in matching efficiency may entice firms to post more vacancies because the chance of filling them increases.
3. Lower cost of vacancy advertising. This allows firms to post more vacancies – or keep them live for longer – than they would otherwise do. This could result in an increase in the number of vacancies and a corresponding reduction in recruitment intensity.

While the first two explanations reflect genuine labour demand and supply, the third one could also be associated with behavioural shifts and structural changes in recruitment patterns. For example, low costs of advertising could encourage firms to be more picky in finding a suitable candidate. We find evidence to support that: the average duration of job posting has doubled in recent decades, from 3½ weeks in the 1990s to nearly seven weeks today.

Building on these foundations, we develop a new approach to estimating an equilibrium level of vacancies (V*) by utilising vacancy costs data. We draw on the quarterly Advertising Association (AA) and WARC Expenditure Report data set going back to 1982, which provides estimates of UK advertising spending on recruitment. From the composition of spending in real terms (Chart 2) we can observe the shift away from traditional advertising (pay-per-word ads in print media) and towards online-based job boards (pay-per-post) and social media platforms (with pay-per-click or free listings). The advent of online platforms coincided with a sharp fall in aggregate spending in the late-2000s, driven largely by the collapse in print media advertising following the global financial crisis (GFC).

Chart 2: Composition of real aggregate spending on recruitment advertising

Sources: AA/WARC Expenditure Report, ONS and authors’ calculations.

We use these data to construct a new index of average vacancy cost (Chart 3). It shows that on a per vacancy basis, real unit costs fell by over 80% since 2000. A key driver of that has been the lower cost of online advertisement. For example, in the early-2000s, an internet job posting on Monster.com cost less than 5% of an advertisement in the Times, despite the higher coverage via the online job posting. Since then, the divergence between online and newspaper job postings may have widened further, with online postings costing up to US$1,000 on Monster.com in 2023, versus a minimum of US$50,000 for a full-page Times advertisement. Platforms like Indeed and LinkedIn still charge for premium placements, but many allow basic job listings for free, to the same effects as firms using their internal job boards. It is possible that over time, recruitment spending has also shifted away from direct advertising and towards third-party recruiters, campus recruiting and other channels, which generate a cost but don’t show up as vacancies. This would mean that the data we use probably represent an upper bound of the impact of falling unit vacancy costs.

Chart 3: Average vacancy cost has fallen by over 80% in real terms since 2000

Sources: AA/WARC Expenditure Report, ONS and authors’ calculations.

Estimating the equilibrium level of vacancies and V*/U*

We test the relationship between the observed level of vacancies and some fundamental determinants through the lens of an error-correction model. In the baseline specification for the long-run equation, we regress the change in vacancies on the changes in real vacancy cost and hourly labour productivity to capture the key drivers of the job creation curve. This set of variables exhibits a cointegrating relationship, allowing us to extract a measure of the equilibrium level of vacancies (V*). We also control for the short-term movements in these series.

We can then combine the results with the equilibrium level of unemployment (U*). The resulting path for the V*/U* is shown alongside the observed V/U ratio in Chart 4. The equilibrium profile appears to capture the upward trend well. It has also been fairly flat since 2018, suggesting that the past six years constitute a reasonable benchmark to assess the labour market tightness within the current ‘regime’. Qualitatively, this matches priors of a prolonged period of labour market slack following the GFC, and a modest margin of tightness during 2018–19.

Chart 4: The observed V/U ratio and its estimated equilibrium level

Sources: AA/WARC Expenditure Report, ONS and authors’ calculations.

An alternative approach: the search and matching model

To provide a sensitivity check to our empirical error-correction model, we use an alternative approach based on the underlying theory from the Diamond-Mortensen-Pissarides (DMP) model, which uses the Beveridge curve and the job creation curve. The model has been calibrated to UK unemployment with vacancies normalised to fit measured vacancies.

We study the impact of a fall in vacancy costs on the vacancy-unemployment equilibrium in this model (Chart 5). A fall in unit vacancy costs leads to a higher net benefit from employing a worker, assuming productivity and other variables remain unchanged. Graphically, this leads to an anti-clockwise rotation of the job creation curve such that vacancies are higher for a given level of unemployment. The new job creation curve forms a new equilibrium with the unchanged Beveridge curve with a higher V/U ratio. As shown in Chart 5, this calibrated DMP model fits the data for the UK well.

Chart 5: The Beveridge curve and the job creation curve in a high and low cost regime

Sources: AA/WARC Expenditure Report, ONS and authors’ calculations.

Implications for our understanding of slack

Based on the results above, we think that the VU gap provides a more reasonable representation of labour market tightness than both the raw V/U ratio and the vacancy rate shown in Chart 1. Importantly, even if we think advertising costs play a notable role in the equilibrium level of vacancies – relative to, say, productivity – they don’t suggest that that equilibrium level is much above its 2019 level, because the decline in advertising costs happened largely before then.

Our empirical error-correction model’s point estimate suggests a closed or marginally negative VU gap in 2024 Q3. Alternative measures of labour demand also corroborate that picture, with the Bank of England Agents’ measure of recruitment difficulties back to levels last seen in 2017. Taken together, our results suggest that the UK labour market is now broadly in balance in an absolute sense.

Michal Stelmach and James Kensett work in the Bank’s Current Economic Conditions Division, and Philip Schnattinger works in the Bank’s Structural Economics Division.

If you want to get in touch, please email us at [email protected] or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.