Metrics matter: how (not) to measure regional inequalities
Regional inequalities are a hot topic, certainly in the UK. Earlier this week, there was an illuminating Radio 4 discussion on London’s economic, social and cultural role in the UK over the decades, appropriately titled “London — villain and victim?” When people talk about regional inequalities, it is often to note the big gap between London and other places in the UK. However, such generalised perceptions diverge quite a lot from observable reality.
For example, an estimated 14% of UK’s households in the bottom decile by income live in London — slightly more than London’s 13% share of UK population. Households in London’s poorest 20% are less-well-off than those in the South East and South West, and many are just as poor as the lowest-income people in the UK’s most deprived areas (see e.g., post #14 here).
Some of the differences between perceptions and reality (at least as measured by the ONS) are no doubt due to cognitive biases: London’s wealthy population is much more visible, for example via traditional and social media, than its poorer inhabitants. Certain phrases (which I personally think are misleading), such as “the UK has the widest regional disparities in Europe”, have become received wisdom through plentiful repetition.
In fact, comparing local areas — whether by post code, ward, parliamentary constituency, local authority, or region— based on economic metrics is surprisingly complicated, and can be a source of confusion. Inevitably, there are some data limitations; averages hide important detail within even the smallest geographical areas; different data point to different conclusions; and there is debate about the best metrics to use. The latter, in turn, depends on what exactly we are aiming to measure.
Most people participating in the “levelling up” discussion would probably say that the aim is to reduce disparities in “living standards”. But how should we measure living standards? A standard metric would be to analyse output — gross domestic product (GDP) or gross value added (GVA) — per person. There are many reasons why such calculations need to be caveated (see e.g., 6 April 2021 here), but (with the right transformations) they can often be used to compare countries with broadly meaningful results.
However, the same is not really the case when comparing local areas within the same country. Why? The main reason is a material spatial disconnect between output and incomes. One area can be a hotspot for producing a lot of valuable goods and services, but if the majority of the income from this production goes to workers or capital owners outside the region, then the area’s prosperity is not boosted as much as GDP or GVA figures would suggest. In practical terms, commuting patterns are particularly important here.
This is clearly visible in the attached chart. The local authorities classified* as “London cosmopolitan” had a gross value added per (resident) person in 2018 that was several times that of other parts of the UK (light blue square in top right hand corner of the chart). However, a lot of the income related to that output accrued to people who do not live in the corresponding local authority districts; rather they commute (or telecommute!) in. Therefore, this area’s incomes on a residency basis (dark blue circle), while still high, are “only” around 75% above the UK average.
In contrast, other areas — where less is produced but where some of the workers or capital owners live — show the opposite pattern: their incomes (as measured by gross disposable household income (GDHI), equivalised to take into account household size and composition) are relatively speaking higher, whereas output (or, “production” if you will) is relatively speaking lower. This is true, for example, for the “Rural growth”, “Affluent rural”, “Prosperous semi-rural” and “City periphery” areas, the majority of which are in London’s or other major cities’ commuting zones.
[* The characterisations used here are from the ONS’s Open Geography Portal postcode register, where local authority districts have been classified into the 24 categories shown in the chart. In case you are curious as to which local authority belongs to which category, I’ve created a (#dataisbeautiful) data visualisation here, where you can hover over a local authority to see its name and classification; or click on a category in the legend to see which local authorities belong to that group.]
What about the famous “towns”, which have become a focal point in some of the discussions around levelling up (partly for political reasons, which I won’t go into here)? Well, first of all, as I’ve pointed out elsewhere (e.g., chart 5 here), on a simple “cities— towns — villages” spectrum, it is in fact the non-London cities, not towns, that have done the worst in the last 20 years. This tendency has been significantly exacerbated by the COVID-19 crisis. As I’ll discuss in a forthcoming blog, “towns” also appear to be recovering faster from the crisis.
However, with the more granular 24-category classification used in the chart here, it is easier to get a feel for what people are talking about when they refer to these deprived areas as “towns”. While they may not match a straightforward cities vs. towns split, they do fit the image conjured up by “left behind places”. As can be seen in this other (#dataisbeautiful) data visualisation, these areas have not been affluent for at least the past two decades and, while every local authority has seen increases in both output and incomes, this has not been enough to keep up — let alone catch up — with places with stronger growth dynamics.
Both GVA and GDHI per person in these types of areas was far below the UK’s average both in 1998 and in 2018. The labels used in the area characterisations give a good indication of why they have been less prosperous in recent times: a key driver has been their sector mix. “Manufacturing legacy” (e.g., Wigan, Wakefield), “Mining legacy” (e.g., Bridgend, Neath Port Talbot) and “Industrial and multi-ethnic” (e.g., Bradford, Bolton, Wolverhampton) all point to a disproportionate share of economic activities that have, on average, been on the wane for a while.
Having said all that, a final point to emphasise is the original premise of this blog: measuring living standards or prosperity based on output (i.e., GVA or GDP) is misleading (in this context). A much better metric is to look at residents’ actual incomes (i.e., gross disposable household income, or GDHI, equivalised to take into account household size). On income measures, the places that are poorer do show up, but the contrast to higher-income places is not nearly as stark as when using GVA. Of course, one could go further and adjust incomes for housing costs — see e.g., entry 10 here. This would shrink the distance between London and other places further.
Clearly, the appropriate metric, its granularity, time-frame and other dimensions ultimately depend on exactly what question we are trying to answer. I’d love to hear arguments in favour of GVA per head, or against GDHI per head, since I have a feeling there are some further nuances that I haven’t yet covered. For example, I haven’t looked at the impact of wealth on people’s living standards (or how they feel about their living standards); nor whether people attach a different value to employment income vs. benefits. No doubt the IfS Deaton review will look at these, and many other issues. This note is already very informative.
[Note: There is a separate debate on whether even income is the right metric for comparisons. After all, if people earn little but are happier with their lives than higher-income people elsewhere, is that really something we would want to intervene with (other than to maybe help the richer people be less miserable)? This is not just a hypothetical scenario, but very real in the UK: there are many places — often rural — where both output and income per head are significantly lower than in other parts of the UK but where people are much more satisfied with their lives (see e.g., entry 7 here). This too, is a theme I will return to in a future blog.]