Exposed: The collapse of green space provision in England and Wales

For 15 years the UK has stumbled from crisis to crisis, from recessions to the pandemic, through numerous political upheavals and scandals. Amongst the noise, little attention has been given to the health of the public parks and green spaces which breathe life into our neighbourhoods. As it became apparent that austerity budget cuts were savaging local government’s ability to maintain the public realm and driving privatisation of public space, some raised their voices. Parliamentary committees chuntered, but little changed. Between 2013 and 2021, the proportion of parks in ​‘good condition’, as reported by local authority park managers themselves, slipped from 60% to just over 40%.

But the decline in Britain’s ​‘green estate’ goes far further than just cuts to maintenance budgets. New analysis by NEF reveals a much broader decline in green space provision, linked to how the UK planning system is designed and how it builds communities.

We combined data on public green space provision from the Office for National Statistics (ONS), data on the average age of local housing stock from Datadaptive and new national survey data on public perceptions of local green space from Natural England. Doing so allowed us to compare experiences of green space in communities across England and Wales based on the generation of housing which dominates in the area, and hence the generation of planning law which governed its design.

We find stark differences in green space provision across development periods. For example, in neighbourhoods where most of the housing was built between 1930 and 1939, the median size of a neighbourhood’s nearest park was around 61,500 m² (around 8.5 football pitches). The equivalent figure for developments dominated by post-2000 housing is 36,200 m² (5.0 football pitches) – a 40% decline. The total amount of green space found within 1km of a development declines steadily the younger the housing stock (see figure 1). For example, around 13% of the space found within early 20th-century developments is typically devoted to green space, compared to just 9% within post-2000 developments – a decline of around a third.

Figure 1: Neighbourhoods dominated by the most recent generation of housebuilding (2009 – 2021) have up to 40% less green space provision than neighbourhoods dominated by late 19th- and early 20th-century housing

Figure 2: The more 21st-century housing stock a neighbourhood has, the less of the local area is dedicated to green space

If you are living in a neighbourhood built after the millennium, your local park is more likely to be smaller and you are likely to have to travel further to reach it (Figure 2). The shift in provision of green space extends into the private realm too. This trend is broadly the same in both England and Wales, Scottish data was not available. Our analysis of the People and Nature Survey suggests individuals living in post-2000 neighbourhoods are almost twice as likely to report having no access to a garden than those in developments dominated by housing from 1930 – 1999. If they do have access, they are 30% more likely to report that they don’t like their garden.

These gaps, driven by factors such as the collapse of council house building and the freedom given to developers to cut public amenities like green space in order to make higher profits, has a societal impact. Not only are communities living in post-2000 dominated developments 30% more likely to report that they do not ​“feel part of nature”, but they visit green space less too. Our statistical analysis suggests the decline in new green space provision after 2000 can now be associated with at least 9m fewer trips to green space a year (see methods below). Individuals living in post-2000 developments were around 5% less likely to visit green space once a week after other key variables (deprivation, age and dog ownership) were controlled for. The provision of green space in our newest neighbourhoods is substandard, and the impact can be seen in individual behaviours.

Local public green space is not just window dressing: it is a fundamental building block for a good quality of life. Its benefits, and role as an equaliser of wellbeing and opportunity have also been the subject of a recent surge in scientific investigation. New research demonstrating the role public green space plays in better educational outcomes, physical health and multiple dimensions of mental and social wellbeing, resilience, and recovery has edged green space provision into the scientific and public health mainstream. But there is no sign, as yet, of an entry into the national political mainstream. This needs to change.

NEF are supporting the Nature For Everyone petition, which is calling for a legal right to nature. The Nature for Everyone coalition, comprised of more than 80 nature, planning, health and equality organisations, is calling on the government to make equal access to a high-quality natural environment a key measure of success for the levelling-up agenda, establish a new legal duty for developers and public bodies to provide equal access to nature-rich green and blue spaces, and extending the Levelling Up Fund to green infrastructure projects.

Methods

Statistics on public perceptions of the availability of parks and people’s connection with nature, as well as data on visits to green space are derived from the Natural England People and Nature Survey (PANS) waves 2 – 10 spanning April-December 2020 (n=18,746). This sample is pandemic-impacted.

Headline figures on the provision of green space in neighbourhoods dominated by different age classes of housing are derived by combining ONS and Valuation Office Agency (VOA) (aggregated by Datadaptive) datasets for England and Wales using LSOA codes. The VOA data describes the median housing age in each Lower Layer Super Output Area (LSOA) by age class categories, as shown in the figure above. ONS green space provision data describes various parameters of green space provision within 1km of a postcode, averaged to each LSOA. The combined dataset can then be subjected to simple descriptive analysis to generate the outputs described above.

In order to estimate the number of lost visits to green space resulting from the declining level of green infrastructure provision in neighbourhoods dominated by more recent developments we link Datadaptive’s housing age class data with the PANS raw data, waves 2 – 10, again using LSOA codes. Once each survey respondent’s neighbourhood development age is identified, we conduct regression analysis to look at the correlation between neighbourhood age and individuals’ frequency of visits to green space, as captured in PANS.

As is standard in this sort of analysis (for example see this study in the British Medical Journal) we utilise a logistic regression model to look at the probability an individual who lives in each housing age class will visit their local green spaces at least once per week. A logistic regression performs well when analysing datasets with a categorical dependent variable and multiple categorical independent variables and has been established as the academic standard for studies in this field (see here for example). In this case, in addition to the housing development age, independent variables used as study control included: the index of multiple deprivation score of the area, the individual’s age, and whether they own a dog – all factors known to heavily influence rates of green space use.

Multiple model designs were tested in order to identify the model with the best fit. McFadden’s pseudo-R2 was utilised to test for model fit. Our final model obtained a McFaddens R2 of 0.038, an adequate result for a model of this nature. Age, deprivation, and dog ownership were all highly statistically significant. All 20th-century housing stock age classes were associated with higher probability of visiting green space when compared with 21st-century classes. Different housing stock age groups were associated with different levels of statistical significance, with P‑values ranging from 0.73 (1973 – 1982) to <0.01 (1993 – 1999). The full regression output can be downloaded here.