From the TUC

Misleading claims of 4% earnings growth

12 Dec 2014, by in Labour market

It is of limited relevance to the population as a whole that some people are lucky enough to be enjoying stable employment with rises in earnings of four per cent.

The figure on earnings growth for employees in full-time work for over a year issued by the ONS with their recent ASHE (annual survey of hours and earnings) data is not only used inappropriately, but is also more difficult to interpret than commonly thought. For example, in the Parliamentary debate on food banks earlier in the week, Iain Duncan Smith proclaimed that if you had been in work for over a year ‘your salary rate is about on average 4 per cent’. In their Autumn Statement document, the Treasury claimed:  “those in continuous employment for the past year saw wage growth of 4.1%”, showing the below Chart 1.5 in support. 


The purpose of the ONS analysis was to illustrate composition effects that the TUC have previously emphasised:

Looking only at jobs in which the employee had been in post for at least one year (thereby removing the influence of the changes in the composition of the labour market), growth rates are consistently higher. For the public sector, the annual increase in weekly earnings becomes 3.7% (compared with 1.0%) and for the private sector it becomes 4.1% (compared with 0.7%). (pages 19 and 20)

A chart goes on to look at the ‘Annual percent change in median full-time gross weekly earnings for all employees and those in continuous employment’, not unlike the HMT one (though HMT have added CPI inflation, presumably to make us feel even better).

But the use of the median in this sort of analysis of growth rates can be rather misleading; they are not so obviously relevant as indicators of central tendency. The  critical point is that 4.1 per cent is the rise in median pay, not the median rise in pay. 

Yet the tendency is to think that the median growth is the middle value, so that half the population are seeing earnings rises above four per cent and the other half below four percent. Though common sense should perhaps tell us that there is something wrong with the idea that so many people are enjoying pay rises on this scale.

The ASHE survey is more normally used to assess various distributional aspects around earnings, rather than changes over time. All the 4.1 per cent figure says is that at a specific point in the income distribution, this is what the increase is (for whatever specific subset of the population). It tells us nothing about increases at other points of the income distribution. In the extreme it could be that all other deciles have seen a zero per cent growth, implying a more representative ‘average’ of closer to 0.4 per cent.  [= (9 x 0 + 1 x 4)/10]

While other growth rates derived with the ASHE dataset are potentially prone to the same issue, published information for most measures includes distributional detail. Unfortunately because the continuous employment measure is relatively new, the distributional information are not available. 

To illustrate, the chart below shows the distribution of earnings growth for the ASHE sample as a whole, at each consecutive 10th percentile across the income distribution:


So here, for the median, growth of 0.6 per cent is one of the highest growth rates, and is about double the average obtained by summing the figures at each percentile point (and greatly larger than the ONS’s calculated average figure of -0.05 per cent). The middle value of these growth rates is also around 0.25 (at the seventieth percentile). All such growth figures therefore need careful handling and interpretation. 

It is hard to avoid the conclusion that the 4.1 figure is being used as representative of wages growth for the economy as a whole. Notwithstanding important compositional factors, this is wrong and very misleading.  

Even the UK Statistics Authority gave the figures great prominence at a recent set-piece event, ‘Better Statistics, Better Decisions An expert seminar’ (Tuesday 25 November 2014). In his keynote address, Andrew Dilnot seemingly offered them as important and informative. Given the importance of the earnings debate, they and the ONS should think more on how these figures are calculated and presented.