“Levelling up” – facts about the UK’s socio-economic landscape

Note: from 2021 onwards, "levelling up" related posts have been collated in a new repository, Tera's "levelling up" related #dataisbeautiful charts and commentary 2021.

Following on from the Brexit vote and the 2019 general election, there is now a lot of momentum behind the “levelling up” agenda: a renewed focus on tackling regional disparities. A desire to reduce place-based inequality is of course not new, and arguably, there have at times been too many attempts to fix it. This has led to fragmentation and lack of continuity (see, e.g., Chart 8 here).

Moreover, the facts about the UK’s local economies are particularly confusing, and not easily (or at least accurately) simplified into statements, such as “UK’s regional disparities are one of worst in the developed world”. The conclusion of a recent excellent article in the FT on this was that “it is complicated”. The header chart above illustrates the degree to which the various variables at the local level – ranging from demographics to employment to incomes to benefits to health and happiness – are all interconnected one way or another. So, I’m still formulating my own synthesis of what exactly is going on here.

In the meanwhile, it is comforting that there are some pretty clear facts which, as long as we don’t try to contort them into newspaper headlines or political statements, are in my view both interesting and informative. Over the course of a couple of weeks, I will be posting these as individual #dataisbeautiful charts, and this blog is the repository where I will give them a permanent home. While there is no rigid structure to the posts, they roughly cover: the past, present patterns and correlations, and the future. [Note: I’m not claming to have facts about the future, but will share here what our modelling suggests.]

1. Few local authorities have managed to improve their ranking on household incomes in the last 20 years

While rankings are an entirely relative concept, and say nothing about converging or diverging incomes, this chart does illustrate how difficult it is for any particular local authority to achieve a turnaround in their economic fortunes.

For an interactive (if slow, sorry) data visualisation where you can filter for and zoom in on your favourite local authorities see here.

2. Household incomes have grown most in urban areas, but those areas also saw the biggest declines in growth rates after the financial crisis

Given that urban areas enjoyed higher incomes than rural ones to start with, this has meant that the gap between urban and rural has widened somewhat. However, it is important to note that at the local authority level, the real issue that has been driving inequality is lack of growth in some of our large cities (see chart 5 here).

3. Life satisfaction has increased in almost all local authorities, and the difference between the top and the bottom has narrowed

Despite incomes having been essentially stagnant since 2011, almost all local authorities have reported an improvement in life satisfaction. Moreover, life satisfaction has increased most in those local authorities where residents were originally least happy. As a result, the gap between the most and least happy local authorities has narrowed from 2011 to 2018.

I have yet to figure out why this has happened. For example, despite quite a lot of data for each local authority, there’s no obvious reason why Fylde residents are so happy with their lives (for more on Fylde, see here). It’s of course plausible that happier people have moved to Fylde, or that less happy people have moved out of Fylde. Possibly, the population has aged, which tends to go with an improvement in self-reported life satisfaction (see e.g., here).

4. There is a significant gender pay gap in most local authority areas

On the International Women’s Day it is worth reflecting on what we really mean by “levelling up”. As already indicated above, it is not clear what we should be levelling up. Should this be about incomes, or life satisfaction, or something else? It is also not, in my opionion, obvious whose incomes or well-being we should be levelling up. (Richard Layard, in his latest book “Can We Be Happier?: Evidence and Ethics”, makes a compelling case for caring most about the well-being of those most deprived of it today.)

The current “levelling up” narrative is about places, and this makes some sense: many people’s fortunes are indeed heavily influenced by where they were born or where they live now. It clearly also makes political sense: geographical units, in this case parliamentary constituencies, ultimately dictate political parties’ power in government. But as we go about levelling up the regions, it is critical that we do not forget about other inequalities.

The maps below illustrate an important feature of womens’ current status in the workforce. On the left, we can see that women are close to 50% of the workforce in the vast majority of local areas in the UK. On the right, however, we can see that there is still a gender pay gap (in almost all local authorities. Out of 378 local authorities for which there is data, only 6 have a negative pay gap; and 49 have a pay gap more than 25%.

There are many reasons for the gender pay gap persisting, and we need to act on all of them. One is to do with a fundamental question: if, for example, according to Richard Layard’s book, what we as humans value most is our health (see also here), then why is healthcare workers’ pay not in line with this? I have provided some hypotheses here – but it is of course a much more complex question in reality.

5. London is not a homogenous whole, and its high-income Kensington and Chelsea also has the worst income inequality and life satisfaction in the UK

A lot of the levelling up debate uses terms like North-South divide or “London and the rest”. But zoom in on London and it’s clear that it is highly diverse. Yes, the area with the UK’s highest average houehold income is to be found in London, but the same local authority – Kensington and Chelsea – also has the highest inequality and the lowest life-satisfaction ratings in the whole country.

Yes, Londoners are relatively young and healthy. But even putting the rest of the country to one side (for a minute), there is a lot of levelling up to be done inside London alone. [Note: this is a well-konwn pattern: large cities tend to have the highest incomes but also highest inequalities. Note 2: also, after housing costs, Londoners’ household incomes plummet to the middle the UK league tables – see e.g., here.]

6. London areas where there are more women and older people also consume more sweets

Researchers at Bell Labs have put together a fascinating dataset for London, based on purchases by 1.6 million Tesco Clubcard holders over a whole year. They have complemented it with a few demographic variables, such as average age and the gender split among residents of each local ward. I have also combined it with data on household incomes by ward and a few variables by local authority.

There is a huge amount of data and some interesting findings. Of course, the sample is also not perfect, because it only covers purchases at Tesco stores. But there are a couple of interesting, even surprising findings. For example, it suggests that residents in wards with more men than women tend to consumer fattier food. Conversly, areas with more women (and, also, older populations) tend to consume more sweets.

7. There is a u-shaped relationship between incomes and life satisfaction, with people in cities earning more but being less happy on average

Many people have chosen to look at the “levelling up” ageda through the lens of productivity (see e.g., the excellent Industrial Strategy Council report here). This makes sense, because ultimately productivity contributes to an area’s competitiveness, employment and real wages – and so to people’s living standards. However, as I’ve implied above, it is more complicated than this.

When looking at more granular local areas, productivity metrics have some significant downsides (see e.g., here). Therefore, I prefer to look at household incomes. From a political perspective, this also seems more relevant, since companies do not vote. What is a bit of a conundrum, though, is the relationship between incomes and life satisfaction. People often note that the location in the UK with the highest median household incomes, Kensington and Chelsea, also scores the worst on life satisfaction.

Part of the explanation is of course that, at the level of individuals, there are many other factors that contribute to life satisfaction, such as health, employment and relationships (see e.g., the December 14 chart here). But incomes do still matter. And looking at data on occupations, there is a positive correlation, albeit one that plateaus out (see e.g., Chart 3 here).

What we find at the local authority level is a u-shaped relationship: higher income areas have higher life satisfaction, up to a point, but beyond that point, life satisfaction gets worse. Part of this could be due to the differences between cities and rural areas. Cities tend to have higher median incomes, but they also have higher levels of inequality, more unemployment, higher housing costs, more crime, and so on. Moreover, cities tend to be home to younger people, who on average report lower life-satisfaction than older people.

So when we dig deeper into the levelling up agenda, it is not quite as simple as levelling up incomes or copying the features of cities that boost earnings.

8. More prosperous places in the UK have higher quality schools based on a number of different metrics

In analysing the reasons behind low productivity (and hence, low incomes) in different areas of the UK, research has time and again pointed to the quality of human capital (see e.g., here and here). While education of younger people often falls outside the remit of industrial strategies, it is nevertheless a critical ingredient in the future prosperity of places. Not only is high quality education important for productivity; it is also associated with better health and well-being (see e.g., here).

Particularly for less-prosperous places, it can be hard to attract highly educated, highly productive workers. Labour mobility in general is much lower than people typically imagine (see e.g., here). Therefore, it is critical that the “home grown” talent is of high quality, and this is where schools come in. Sadly, at the moment, it appears that – rather than helping reduce regional disparities – the quality of schools is a factor exacerbating inter-regional inequalities.

There are no doubt many reasons for the correlation between school quality and an area’s prosperity. The causation probably goes both ways. But from a policy perspective, this seems like one of the more “fixable” problems, with a very strong evidence base to support investment in human capital. For example, it is likely to be a lot less easy to, say, encourage businesses in poorer areas to be more export oriented, more innvative, do more R&D or deploy better management practices.

9. Coastal and Northern local authorities have a higher proportion of elderly residents and a higher incidence of physical illness

Given that contracting the coronavirus seems to be particularly problematic for elderly people and those who have pre-existing health conditions, it is informative to look at where these people live. In the England, older people are more likely to live in coastal, Northern and rural local authorities. In absolute terms, large cities such as Birmingham, Leeds, Manchester, Sheffield and Glasgow have a large population of people over 65.

10. While households in London have some of the country's highest incomes, the differences shrink when housing costs are taken into account

It not a surprise that households in London have high incomes, given the education levels and occupational mix of its residents. For example, almost 40% of workers in London are in the top two occupational major groups: managers, directors and senior officials; and professional occupations. The average for the rest of the country is 30%.

However, is is also no secret that housing costs in London are very high compared to the rest of the country. So when we look at household incomes after housing costs, the differences between London and the rest of the country seem much less stark. As we level up the country, a relevant metric is probably disposable income after all essentials have been paid for. Since gas, electricity, petrol, public tranport, and telecoms costs are very similar across regions, it would seem that adjusting for housing costs is an important step.

11. Internal migration moves in and out are highest in cities with high incomes and high growth, despite high housing costs

The populations of local authorities are not static, and it is therefore interesting to look at who is moving where within the UK. The chart below has quite a few dimensions to explore, so let’s look at the key findings from left to right. [Note: this chart only looks at so called “internal migration”, i.e. movements of people between local authorities in England and Wales. It does not include international migration from other countries.]

In the pane on the left, we can see a strong relationship between the growth rate in household incomes over the last 20 years and the volume of internal migration for any particular local authority in 2018. Those areas where gross disposable household incomes grew the fastest – for example, Wandsworth, Hackney and Barnet – also attracted a lot of inward migration compared to the rest of the country. At the same time, lots of people also moved out of these places (the 2nd row), and overall, they experienced a net outflow of people to and from other local authorities. Of course, some of this would have been compensated for by international immigrants coming to stay in these areas.

The relationship between actual household incomes and and internal migration is very similar. There was a lot more “churn” in the population in areas where incomes were high, as many people moved both in and out. Again, the net effect was somewhat negative when looking at internal migration only. Interestingly, in this middle pane, as well as the others, cities lost a lot of people to more rural areas in 2018. This is indicated by the fact that the red bubbles (cities) tend to be much lower in the bottom row (net moves in) than the green ones (rural areas). I had personally not realised this picture was quite so sharp.

Finally, a very interesting finding in the last pane on the right is that these relationships become less pronounced after we consider both net household incomes (that is, after taxes and benefits) and housing costs. For example, a lot of people still move in to Islington, even though it appears less affordable after housing costs. And relatively speaking, fewer move into Kingston upon Thames, despite its relatively high incomes after housing costs.

It doesn’t make sense to draw too many conclusions from such a simple analysis, but it would seem that housing costs are not such a significant driver of people’s location decisions. Of course, this will vary materially between different segments and types of households – a topic I will return to when we look at internal migration by age group.

12. Residents in lower-income local authorities in London suffer more from physical illness and their purchases have a higher carbohydrate content

Looking further into the data on the 1.6 million Tesco Clubcard transactions over a year, it is fascinating to see if it reveals something about different diets across different wards or local authorities in London. As in Chart 6 above, I have also combined this with some other data, this time on incomes and health. [Note: this is just showing the straight data and has not been controlled for any variables that might have an impact on any of these metrics. Those other variables are bound to be plentiful.]

I’m not surprised to find that local authorities in London with lower incomes also report a higher incidence of physical illness (left hand bars). This is a well-known phenomenon.

However, I am a little surprised that residents in these poorer local authorities also consume significantly more carbohydrates. If anything, my (terribly stereotypical) hypothesis was that they’d consume more alcohol and fats. It’s plausible that the female/male ratio is also impacting these resuts (as women do on average consume more carbohydrates), as well as age of residents. It’s also plausible that carbohydrates are the cheapest source of energy in today’s supermarkets.

The point really is: at the local level everything is interconnected.

13. Internal migration between London and other local authorities (in England and Wales) tends to be to and from cities

London’s local authorities have one of the highest rates of internal migration in England and Wales: lots of people move in and out of London every year. It is interesting to see where they come from and where they move to. The answer is dominated by other cities (rather than, say, towns or more rural areas). [Note: my dataset doesn’t include detailed movements between England and Wales local authorities and Scotland and Northern Ireland.]

Quite a few of the same places have relatively large outflows of people to London, but also large inflows of people from London. Birmingham, Brighton, Leeds, Manchester, Bristol, Nottingham, Cambridge, Leicester, Coventry, and Canterbury are all in the top 15 on both lists. The vast majority of these people are between 19 and 25 of age – so most likely students moving to go to university or graduates moving from a London-based university to work somewhere else.

London continues to lose quite a few people, however, also in the 30-40 year age bracket. As can be seen in the map on the right, moves out from London are more concentrated in the South East and East of England. While in 2018, overall, more people moved out of London than into London (internally in England and Wales), on a regional basis, London was a “net gainer” of people from the North East, Yorkshire and the Humber, or Wales.

A final, important, observation on this data: the total numbers are surprisingly small. In 2018, a total of 230,000 people moved to London (from non-London local authorities) and 330,000 moved out. These figures were around 3% and 4% of London's population in 2018, respectively. Across all non-London local authorities in England and Wales, moves in and moves out were on average 5.4% and 5.1% of the population, respectively in 2018. [Note: the London figures are smaller because London consists of 33 local authority districts and the numbers above exclude moves between these areas.]

14. Within-area inequalities are significantly higher than those between areas, and London has some of the country’s poorest households

As mentioned in previous posts and here, money doesn’t guarantee happiness. There is a large, growing and evidence-based literature on what does contribute to people’s life satisfaction, and some of the most important variables – apart from absolute income – are age, sex, mental and physical health, education, relative income (and hence inequality), employment, social relationships, and not being involved in crime (see e.g., December 14th here). However, some of these variables stop being significant when other things are controlled for (see, e.g., an excellent paper here).

A particularly interesting debate concerns inequality. The “received wisdom” among economists is, first, that some inequality is inevitable. There are many reasons, but one very intuitive one is that people with different abilities and skill levels can command different wage levels. Abilities and skills in turn tend to vary quite widely between individuals. For example, in the UK in 2017, people whose highest level of qualification was GCSE (grades A-C or equivalent) made up 33% of those employed. On the other hand, 34% of those emplyed had a degree or equivalent. Median hourly wages in 2018 for occupations where most people had below-A-level qualifications (the former group) were ￡9.95, while wages for occupations with mostly graduates were on average ￡21.50.

The second part of the “received wisdom” is that inequality is bad for people’s welfare, well-being and life satisfaction. These two factors pull in different directions, implying that there could be someting like an “optimum level” of inequality. However, to my knowledge, no robust or accepted quantification of this exists. Partly this is because there is significant debate about exactly how bad inequality is for people (see e.g., McKinsey Global Institute discussion paper, ‘Tech for good’: Using technology to smooth disruption and improve well-being).

I’d also like to add to this a further question. Inequality compared to whom? Most inequality metrics focus on national level metrics, such as the GINI coefficient. Yet, the literature implies that what people really care about are social comparisons, i.e. their income relative to their peers, colleagues or neighbours. I’ve not been able to find (easily available) data on income inequality at a very granular level, but the below chart is a good start.

It shows that inequality – in this case, the difference between the 10th and 90th percentile on the household income distribution – is almost entirely driven by the top end of the distribution. In Kensington and Chelsea, the highest quintile has a household income 22 times that of the lowest quintile. In contrast, in Nottingham, this figure is only 4. The difference between the highest and lowest incomes in Kensington and Chelsea is also vastly greater than the average difference between Kensington and Chelsea and Nottingham.

Most strikingly, some of the poorest households in the whole country live in Kensington and Chelsea and other London local authorities. [For example, the 10th percentile average net household income in Kensington and Chelsea in 2016 was ￡7,750 while in Nottingham it was ￡8,960. Note that these are experimental statistics using administrative data and should not be used to draw direct comparisions about poverty or living standards.]

15. The majority of the UK’s employment is in areas with relatively high broadband download speeds

This should be helpful for homeworking during the #coronavirus outbreak.

16. Many socio-economic factors are correlated at the local authority level, but sometimes in surprising ways

While the attached chart below does not meet any kind of best practice criteria for visual ease, it does tell a fascinating, and paradoxical, story.

Many of the socioeconomic factors that come into play in the levelling up debate are highly correlated at a local level, as you’d expect. Better schools correlate with higher levels of education. Higher levels of education correlate with better incomes and better health.

But higher incomes also correlate with higher inequality. And higher inequality correlates with lower life satisfaction. Bizarrely, then, at the local authority level, there's a negative correlation between incomes and life satisfaction, and health and life satisfaction. Yet, at individual level, we know incomes (up to a point) and health (crucially) are positive drivers of wellbeing. (See Well-being in Europe: Addressing the high cost of COVID-19 on life satisfaction.)

This is, in my view, one of the most important challenges in levelling up: choosing what to level up. Should we level up incomes (knowing they don’t deliver that much happiness), or wellbeing (knowing this might be less straightforward)? My only (tentative) conclusion so far is that local areas may be the wrong unit of analysis. It’s likely and we need tailored plans not just for each locality, but people in different life circumstances.

17. COVID-19 economic impacts are likely to hurt areas that were already less prosperous

As mentioned above, the virtuous and vicious cycles at play at local level are hard to escape from. This means that reducing place-based inequality is extremely difficult. Now, with COVID-19 causing wide-spread furloughs and business closures, we need to worry about whether these might be further exacerbating regional imabalances.

Data on a granular regional level is not yet available, and won’t be for a while. While we have to be careful about extrapolating from past data, I think we can make fairly robust predictions based on factors that change slowly. As mentioned above, for example, internal migration in the UK is a relatively minor factor in the short term. Similarly, industrial and occupational structures change, but they change slowly. Third, while there will be counter-balancing factors (such as more of government support also flowing to less-prosperous regions), my view is that this is unlikely to fully mitigate any diverging impacts.

Based on these assumptions, we can make some comparisons with historical patterns and the likely impact of COVID-19.

The chart below combines two forward-looking (and hence, inevitably somewhat speculative, analyses): first, the likelihood that workers in a particular NUTS3 region are vulnerable to furloughs or job losses due to social distancing rules (X-axis). This is based on the historical (2018) occupational mix in each NUTS3 area, and our occupation-by-occupation assessment of the degree to which a job is either infeasible or unsafe to perform during the pandemic.

The second analysis is based on the sector mix in each NUTS3 area. We know what this pattern was in 2018, and we also have data on how each sector has fared till the end of June 2020. [Note: sectors have started to recover since, and in a separate post, I will be looking at the picture at the end of August. Some sectors, such as wholesale and retail, have bounced back a lot faster than, for example, hospitality. This was indeed what we expected in our earlier analysis on the shape of the recovery by sector.] Assuming that the relative sector mix of each region hasn’t changed appreciably since 2018, we can calculate the estimated drop in GVA by end of June 2020 by region.

For several poorer areas, such as Blackpool and Stoke-on-Trent, and locations reliant on tourism, both indicators suggest a worse-than-average outcome, at least in the short term. These are places where occupations are disproportionately high-proximity – i.e., require being very close to other people, or close to a large number of people. People in such occupations are more vulnerable to either catching the virus (if working, for example, as a nurse) or not being able to work (due to social distancing, lockdowns, or business closure).

The vulnerable places are also areas with more occupations that are harder to perform remotely. This also shows up in the industrial base: locations with a large proportion of professional and financial services jobs, such as London, are likely to see less overall decline in output and jobs. This is not to say that these areas are immune to problems: as mentioned in earlier posts, local inequalities tend to be the highest in high-income areas. With COVID-19 changing the shape of economic activity, within-area polarisation is also likely to be exacerbated.

Hopefully, future data sets will allow us to dig deeper into even more granular detail, for example at the local authority, parliamentary constituency, or ward level.

[Note: In the sprit of transparency, I should also say that these figures change quite a lot when one excludes health and education. This is because health and education are a larger proportion of total GVA for poorer areas; and due to the way GVA is measured for health and education, their “output” saw a big drop in GVA in the 2nd quarter of 2020. For future COVID-19 related analyses, it may make sense to look at private sector only.]