WFH / Telecommuting

• From a niche means of accessing work to a widespread way of life
• From \(5\)% to \(47\)% in the UK (ONS 2020a, 2020b)
• \(37\)% WFM in European April \(2020\)
• \(60\)% in Finland (Eurofound 2020)
• Almost \(50\)% in the US, \(37\)% could be permanently moved online (Brynjolfsson et al. 2020; Dingel and Neiman 2020)

WFH / Telecommuting

• WFH = economic resilience during Covid-19

• Capability to WFH:

  • occupations
  • quality digital infrastructure

• Spatial clustering and dependence

Digital divides

• 1st level: availability and quality of internet connectivity

• 2nd level: skills to effectively utilise digital technologies

• 3rd level: returns of internet usage

• Cross-level interactions

Our approach

• Intersection of digital and material divides

• Economic resilience of places as digital technologies became an essential tool of productivity during the pandemic
  
  internet services + occupations \(\Rightarrow\) WFM \(\Rightarrow\) economic resilience

Our approach

• UK Local Authorities Districts (LAD) as a case study
• Quality of internet service: experienced internet speeds during the spring \(2020\) lockdown
• Time-series clustering to create clusters of local authorities with similar temporal signatures of internet speeds
  • \(\Rightarrow\) 1st level of digital divide
• Regress cluster membership against their socioeconomic and geographic characteristics
  • \(\Rightarrow\) how the 1st level of digital divide intersects with the 2nd and 3rd

Experienced Broadband Speeds

• Experienced vs. advertised internet speeds
• \(\Uparrow\) internet demand  usage \(\Downarrow\) experienced internet speed
• www.broadbandspeedchecker.co.uk/
• Riddlesden and Singleton (2014) & Nardotto, Valletti, and Verboven (2015)
• Stream data, anonymised individual speed tests
• Upload speeds and the frequency of speed tests March to May \(2020\)
  
  

Experienced Broadband Speeds

• An overall trend of increased testing from March to April and then a slight reduction from April to May in 2020, compared to less testing as Spring progresses in 2019

• \(2019\) evening peak of testing, leisure activities, download speeds

• \(2020\) new morning peak \(\Rightarrow\) flatter hourly temporal profile

Aggregation

• Relatively scarce data to analyse as stream

• Composite week time series

• Mon-Friday, hourly \(6:00\) – \(24:00\), excl. bank holidays

• \(18\) hourly data points \(x\) \(5\) weekdays (\(90\)) for each local authority (\(382\))

• Focus on upload speeds

Time series clustering

• Aim: create clusters of local authorities with similar temporal signatures of internet speed

• Shape-based approach

• Match two separate time-series objects based on the similarity of their shapes

• Distances between the shapes

• Simplicity:

  • k-means
  • medoids, time-series objects with a minimal distance to all other cluster objects

Time series clustering

Post-clustering regression analysis

• Explain the characteristics of clusters

• How the 1st level of the digital divide intersects with the 2nd and 3rd?

• Multinomial logit regression: cluster membership against:

  • population
  • pop. and job density
  • distance to nearest met. area and to London, South of the UK dummy
  • % of managerial, tech, skilled trade, professional, administrative, leisure, machine operation jobs
  • earnings, n. business est. per hab.
  • % furloughed pop., 2020
  • AM broadband tests per hab.
  • % of Virgin Media connections

Cluster 6

• Workers are more likely to hold managerial, tech and professional jobs
• Least likely to be furloughed
• Slow and less reliable speeds
• High job density

Clusters 6 and 9

• Although speeds were slow and not as reliable during the morning peak as in cluster 9, the skills that enabled telecommuting also enabled greater returns from doing so (\(\Downarrow\) furloughed)

• LADs in cluster 9 were able to benefit both from reliable internet connections and populations able to work from home to capitalise on their digital infrastructure

• But lower returns in the pandemic \(\Rightarrow\) greater numbers furloughed

Being on the right side of digital divides

• Clusters 3, 7 and 9 are on the right side of the 1st layer of digital divide

• Internet resilience supports, but also frustrates urban economies in different geographies:

  • \(\Uparrow\) morning peak tests per capita run in cluster 6

• Being on the right side of the 2nd level digital divide had a greater impact on economic resilience (3rd level digital divide), than having quality internet connectivity

  • \(\Rightarrow\) cluster 1 and 6

• Seven of our nine clusters had slower upload speeds in the morning than in the evening

• Evidence of widespread telecommuting and other daytime internet use (e.g. education) which changed the temporal profile of internet activity throughout the UK

• Upload speeds have not previously been seen as integral to universal service, considering there has never before been such extreme demand for telecommuting and operations such as video calls

• Τelecommuting from the viewpoint of the complex web of digital divides

• Quality of internet infrastructure

• Places may depend upon good internet reliability and connectivity to achieve economic resilience in a period like the current pandemic

• Νot every place hosts individuals with the necessary skills and in occupations to effectively use the internet to telecommute