Hollowing out – deeper than it sounds


It’s rare that technical economic concepts make their way out of the halls of academia and into the mainstream. Yet the idea that labour markets in advanced economies are increasingly ‘polarising’ between ‘lovely’ and ‘lousy’ jobs – with middle-of-the-road jobs being ‘hollowed out’ and wage inequality rising as a result – is one that has popular appeal. This crossover is helped by the powerful use of metaphor – especially the argument that the ‘robots are taking our jobs’ – but it also has the benefit of intuition: the less dramatic description of new technology helping higher-skilled workers become more productive while simultaneously replacing mid-skilled jobs and proving largely irrelevant for lower-skilled ones just ‘feels’ right.

And it should. After all, study after study has found evidence of this pattern. But the truth is of course more complicated than the pop-language implies. And it is subject to change over time. Our new analysis considers whether patterns of polarisation have become more or less marked in the UK in the period since the financial crisis of 2008-09. In presenting some preliminary, descriptive findings ahead of a full report later this year, we note some potential changes in the nature of the ‘hourglass’ – some of which might have implications not just for the theory but for economic policy. In this blog we describe three complexities that lie beneath the concept.

First, the basic patterns of polarisation are consistently evident over the past two decades, but the hollowing-out-of-the-middle metaphor may be misleading. Ranking detailed occupation classes by their initial wage (as a proxy for skill levels), we find consistently strong growth in employment share in the top 30 per cent since 1993, and an ongoing relative decline in the 50 per cent of occupations below that. But what’s going on in the bottom fifth of occupations is murkier. In the mid-1990s their employment share grew, but since then it has been mostly falling or flat.

Indeed, the summary in the figure below (using a higher level of occupational descriptions for simplicity), shows that over the whole period since 1993 the only low-skilled occupation experiencing relative growth is ‘caring personal services’. So rather than the ‘U-shape’ that hollowing-out-of-the-middle would imply, polarisation since the mid-1990s more closely resembles a Nike ‘Swoosh’, with growth in top jobs always exceeding growth at the bottom. Our new, up-to-date analysis suggests that low-skilled jobs may have gained share a little during and since the crisis (particularly if we include the self-employed), but growth at the top remains much stronger.


Second, technological change does seem to be an important driver of employment shifts, but it is certainly not the only factor at play. The analysis we’ve published today replicates and updates previous modelling by Martin Goos and colleagues, in order to show that the ‘routineness’ of occupations – how easily they can be automated – is a strong predictor of whether their share of UK employment has grown or shrunk between 1993 and 2014. The changes shown in the figure above attest to this – the mid-skilled jobs that have experienced the greatest relative declines in labour share involve manual and office tasks that can be most easily substituted by technology.

However, it’s not all about robots taking our jobs, as even major technology-driven polarisation theorists have recently made clear. A deeper review of the literature highlights that ‘supply-side’ factors such as rising educational attainment, immigration and welfare reform have also contributed to employment shifts. In addition, more time-and-place-specific changes in demand for services look to have played a role. For example, the strong growth in caring occupations, seen in the figure above, will reflect the UK’s changing demography. And the post-crisis collapse in demand in the construction industry, which has led to a sharp decline in the share of employment in (mid-skilled) construction occupations in recent years, reflects cyclical patterns quite separate from the longer-term upward march of technology. In sum, the causes of job polarisation are multi-faceted.

Third and finally, it’s not clear that job polarisation is a key driver of wage inequality. Although mid-skilled (initially middle-paying) jobs have experienced falling labour shares relative to the top and bottom, the shape of the occupational pay distribution hasn’t shifted all that much. This reflects the changing wage structure of occupations over time: other jobs can move into the middle as initially mid-skilled jobs decline, and completely new jobs can be created. The implication is that job polarisation provides only partial insights into overall wage trends, and that increasing wage inequality relates in a large part to divergence between things like age groups, regions or sectors, rather than occupational shifts.

This final complexity within the headline polarisation narrative does not mean that the trends we observe are not important: they have implications for career progression in industries in which traditional, mid-skilled jobs have declined; and they offer important insights for skills policy. As well as further exploring the complex ways in which the shape of the UK jobs market has changed over time, it is to these real-world implications that our further work on polarisation will turn.