Skip to main content

Inferring dwelling occupancy characteristics during lockdown and ‘staycation’ periods using smart water meter data

Date
Owen Hibbert1, Andy Newing1, Alan Smith2, Stuart Ellaway2
1University of Leeds, 2Plymouth University

The Covid-19 pandemic presented a unique opportunity to explore dwelling-level occupancy trends using data from smart water metering. We highlight the potential reuse value of these data in inferring near-real time dwelling and neighbourhood characteristics including those driven by tourism.


Project overview

As part of wider ongoing research - and building on a previous LIDA DSDP study – we explore the potential wider-reuse value of data collected by water suppliers via smart water metering. We extract dwelling-level occupancy trends for the period March 2020 – September 2021, and categorise properties according to their occupancy status during Covid-19 lockdown and staycation periods. The range of very unique dwelling occupancy patterns observed during this period enable us to establish the link between water metering data and property occupancy trends, which could have considerable benefit as a near real-time indicator of dwelling status and neighbourhood composition and change, including identification of properties with usage profiles associated with tourism. We have worked closely with the Office for National Statistics during this research as they have considerable interest in sources of non-survey data which could be used to generate or validate small-area population, housing, neighbourhood or tourism statistics.

Data and methods

We utilise high temporal-resolution water consumption data for a sample of households in Devon and Cornwall. After accounting for the impact of skewed consumption due to leakage (common in water metering data) and missing data, we work with a subset of 784 properties. Analysis draws on the magnitude of water consumed by each dwelling on an hour-by-hour basis for all days (24 hour periods) between 26th March 2020 and 22nd September 2021, coinciding with the first Covid-19 national lockdown in 2020, through to the end of summer 2021.

Whilst our interest is not to uncover household behaviour during Covid-19, the specific dwelling-level occupancy patterns observed during this period due to ‘stay at home’ guidance (lockdowns) and unusually high rates of domestic tourism (staycation), provide a unique opportunity to categorise properties according to their occupancy characteristics during these periods.

We apply the non-intrusive occupancy detection method developed by Jacob-van-Alwon and Andy Newing as part of earlier LIDA work on this project. This method identifies whether each property is inferred to be ‘occupied’ or ‘vacant’ on a given day based on the magnitude of consumption and the presence of specific water usage ‘events’. We calculated this metric on a daily basis and reported it as an ‘occupancy ratio’ (proportion of days during which a given dwelling was occupied) for five specific time periods:

  1. 26th March 2020 to 1st June 2020: First national lockdown
  2. 20th June 2020 to 22nd September 2020: Summer – lockdown restrictions eased, first ‘staycation’ summer.
  3. 5th November to 2nd December 2020: Second national lockdown
  4. 26th March 2021 to 1st June 2021: Spring – gradual easing of lockdown restrictions.
  5. 20th June 2021 to 22nd September 2021: Summer – ‘staycation’ with high rates of domestic tourism.

Based on their occupancy ratio in each of these periods, dwellings were segmented into groups that share similar occupancy characteristics across the time periods. K-means clustering was used following its success in earlier pilot work.

Key findings

Across the 18-month period, the mean occupancy rate (proportion of days during which a given property was occupied) was 93.2%.  89 properties (just over 11% of the 784 study properties) were inferred to be ‘fully occupied’ during this period (i.e. there were no nights when the property was deemed to be unoccupied) whilst four properties were empty (no evidence of occupancy) during the entire 18-month period.

There is considerable variation between properties and between time periods, illustrated in Figure 1 by comparing Spring 2020 (first national lockdown) with summer 2020 (lockdown restrictions eased) occupancy ratios by property. A number of properties have high occupancy in both periods (top right quadrant), with a noticeable number of properties exhibiting 100% occupancy during the first national lockdown (spring 2020). There are, however, a number of properties with either:

  1. low occupancy during lockdown and high occupancy during the summer (top left), or;
  2. high occupancy during lockdown and lower occupancy during the summer (bottom right).

Figure 1 - comparison of dwelling level occupancy - Spring and Summer 2020

Whilst we have no data capturing the actual status of any of these properties during the study period, those properties with non-standard occupancy patterns are of particular interest to our study and could enable us to infer properties with non-standard usage patterns, including short-term tourist rental properties (which we would expect to have low occupancy during lockdown periods and higher occupancy during staycations) or second/holiday homes which may have more complex and individualised non-standard occupancy patterns during this period.

Based on their occupancy ratios during the five time periods of interest, these properties cluster into four distinct groups as shown in Figure 2. The largest group (694 households) exhibits the profile typically expected of residential dwellings, with near-complete occupancy in all time periods, especially during the two national lockdowns. Cluster group 2 (42 households) is also likely to represent predominantly residential dwellings, with highest occupancy in lockdown periods and lower occupancy at other times of the year, representing periods away from home for work, study, leisure or other activities.

Figure 2 – classification of dwellings according to their occupancy trends in our five periods of interest

Cluster groups 3 and 4 are likely to represent dwellings that have occupancy patterns not associated with traditional residential usage. We infer that the 35 properties in cluster group 3 are second homes or associated with tourism (e.g. short term self-catering holiday lets), exhibiting low occupancy during the first national lockdown and peak occupancy during the summer 2021 staycation period. Higher rates of occupancy among these properties in winter 2020 could reflect greater occupancy of second homes during this period or longer-term lets in holiday accommodation amidst a growing trend for remote working in this region. The smaller group of 13 properties in cluster 4 may represent under-occupied dwellings including residential properties that have been empty for a longer period of time, tourist lets with low occupancy rates (including weekend-only lets) or properties with alternative occupancy trends.

Value of the research

Whilst we do not know the true occupancy pattern or status of the dwellings used in our study – and are therefore unable to validate these findings – we are incredibly encouraged by these insights. Further work using smart water meter derived occupancy rates for the year 2022 has been compared to hotspots of tourism activity drawn from AirDNA data, with very encouraging results. Using non-intrusive approaches, we have been able to identify plausible trends and cluster properties into logical groups based on their inferred occupancy rates. The period in question coincides with unique occupancy trends associated with our response to the pandemic and our ability to observe these trends within our analysis highlights the tremendous potential these data offer in inferring dwelling occupancy, extending far beyond the pandemic. Throughout the project we have shared our approaches and findings with the Office for National Statistics (ONS) Methodology and Quality directorate and our ongoing work seeks to relate a larger subset of these data to underlying tourism trends at a neighbourhood level.

These data could present a new opportunities to identify dwelling occupancy trends and status (including the presence of occupancy patterns associated with tourism) at scale and in near real-time, with applications in capturing area-based housing-stock and tourism statistics, for example as part of the ONS’ Future of  Population and Social Statistics work package, or their Review of Travel and Tourism Statistics.

Quote from project partner

It has been exciting working on this ground breaking project which uses data not yet fully exploited for statistical insight. The results from this project, will prove valuable for our ongoing research at ONS in making greater use of administrative data for statistical purposes.  

Charlie Wroth-Smith, Head of the Methodological Research Hub, Office for National Statistics.

Insights

  • Ongoing roll-out of domestic smart water metering means these data will be routinely collected at a dwelling level.
  • We highlight that these data can capture a range of dwelling-level occupancy patterns observed during Covid-19 lockdown and staycation periods in residential dwellings and in properties inferred to be associated with tourism.
  • They could offer new, non-intrusive, near-real time insights into dwelling occupancy characteristics, including those associated with tourism.

Research theme

  • Societies

Programme theme

  • Statistical Data Science

People

Owen Hibbert – Data Scientist, Leeds Institute for Data Analytics

Andy Newing – Associate Professor in Applied Spatial Analysis, University of Leeds

Alan Smith – Lecturer in Environmental Management, Plymouth University

Stuart Ellaway – formerly Post-doctoral Research Assistant, Plymouth University, subsequently Data Scientist, South West Water.

Charlie Wroth-Smith, Head of the Methodological Research Hub, Office for National Statistics.

Partners

Office for National Statistics - Methodology and Quality directorate.

South West Water (data provider).

Funders

Funded by the Economic and Social Research Council (ESRC) Secondary Data Analysis Initiative (SDAI). Grant number: ES\T005904/1 – ‘WatPop: understanding seasonal population change’