Dataviz logo
Oxford Consultants for Social Inclusion logo

Case study: How is data visualised in key local strategy documents?

Introduction

In this case study, we look at three key local strategy documents required by central government, and how they use visualisation to present data. We identify the level of visualisation used in each type of document, and highlight good practice visualisation examples used in different areas.

Key local strategy documents

We have reviewed three flavours of local strategy document:

  • Sustainable Community Strategies (SCS): The long term framework that influences all other plans and strategies produced by local partner organisations;
  • Local Area Agreements (LAA): The LAA document sets out the priorities for each local area, agreed between central government and the local strategic partnership;
  • Joint Strategic Needs Assessments (JSNA): An assessment of the needs of the local population for children and adult services, shared between the Director of Public Health (Primary Care Trust) and the Directors of Children / Adult Services (local authority).

Sampling coverage

To ensure representative coverage, a sample of JSNA, LAA and SCS documents were reviewed, representing each of the nine Government Office Regions and a mix of upper- and lower-tier authorities:

  • 94 of the 388 Sustainable Community Strategies were reviewed (25%) - including those from 53 upper-tier Local Authorities and 41 lower-tier Districts;
  • The majority of the 149 LAA documents were reviewed;
  • 63 of the 149 JSNAs (40%) were reviewed.

How is data visualised in key local strategy documents?

Overview

The level of visualisation used varied widely between the three different types of strategy document, reflecting their different roles and intended audiences. The table below sets out key findings from our review of actual local documents.

Strategy document
Visualisations used
Sustainable Community Strategies (SCS)
  • Sustainable Community Strategies set out the key long-term vision local partners. The documents are typically intended for general audiences rather than researchers.
  • More than 60% of the Community Strategies reviewed did not visualise data in any way. A further 20% only visualised data using using tables. The most common form of visualisation other than tables were thematic maps.
  • Reflecting the role of communicating with general audiences, Community Strategies tended to show simpler visualisations - less than 20% used thematic maps, while only 3 of the 94 strategies we reviewed presented relationships between two variables using scatterplots (scatterplots were used by Tufte as an indicator of visual sophistication when looking at different types of publication).
  • The overall level of visualisation was far below that used in Joint Strategic Needs Assessments (see below), reflecting that data visualisations are seen as less appropriate for general audiences,
Local Area Agreements (LAA)
  • The sample of LAAs we reviewed did not visualise data in any way except in table format, with the tables used to present priority indicators, current indicator values and targets for future indicator values. This reflects the role that LAAs play, as an agreement on priorities and targets between local and national partners.
  • However, the lack of visualisation raises an interesting point. The format makes the documents rather a difficult read for those users who are not used to wading through masses of data tables. Given the importance of the LAA document, it is perhaps surprising that no local and/or national partners incorporated summary visualisations to help understand and communicate the key agreements in the LAA.
Joint Strategic Needs Assessments (JSNA)
  • JSNAs provide analysis on health and social care needs, for both strategic and operational purposes. Reflecting the need for readers to understand key messages from the data, every single JSNA we sampled provided some form of visualisation.
  • The JSNAs were relatively sophisticated in the way they visualised data, with: nearly half showing (at least some) data with error limits, 80% showing time series/ trend analysis, and 80% presenting some form of small area data. Nearly one-third presented relationships between two variables using scatterplots, again reflecting the relative sophistication of the audience.
  • In other words, JSNAs provide an exploration of the key data, albeit limited by having little user interaction (although data and analysis for some JSNAs have been made available on Local Information Systems as a "living document", so enabling more interaction and exploration of the data by users).

What are the most common visualisations used?

The figure below identifies the most common visualisations used in the Sustainable Community Strategies and Joint Strategic Needs Assessments (LAAs only used tables to present data).

Image for this case study

Perhaps unsurprisingly, data tables are the most common form of presentation of data, used by all but two of the JSNAs reviewed. More than half of all JSNAs reviewed contained Bar Charts, Thematic Maps, Line Charts, Stacked Bar Charts, Population Pyramids and Pie Charts.

By contrast, more than 80% of the Community Strategies reviewed did not visualise data in any way, except using tables.

Good practice visualisation examples

Below we highlight some interesting examples from our review of key strategy documents:

  1. Score-cards: comparing performance between areas
  2. Treemaps: exploring relationships between many variables
  3. Wall-charts: highlighting performance of many areas on many indicators
  4. Boxplots: identifying where differences are likely to be significant
  5. Quadrant plots: classifying areas on two indicators

1. Score-cards: comparing performance between areas and indicators

Nearly one-third of all JSNAs use some form of score-card, to compare a range of areas or indicators against a mean or target. Some Community Strategies also present similar data, enabling comparisons between areas on specific indicators.

Image for this case study
Source: Oxfordshire (2008) JSNA

The figure above shows County, District and ward level all-age all-cause mortality rates for areas in Oxfordshire. Each row shows a (horizontal) boxplot distribution (see below for more information on boxplots), with the ward score shown as the vertcial blue line, the county average as the central black dotted line, and the ward distribution across the county shown as the fainter grey horizontal bar. It is very quick to identify which wards are significantly above or below the average for the County.

Image for this case study

Source: Bexley Sustainable Community Strategy

A variant on this is shown above, from the Bexley Community Strategy. A stacked bar-chart is used instead of a boxplot, with the stacked bars showing the best and worst performing 20% of all trusts. Performance for all GP surgeries across the Borough is shown on several indicators, with error limits around performance - enabling users to quickly see how performance on a range of issues compares with national averages for all trusts.

Key features and advantages of score-cards for visualising data include:

  • Present range of data in single diagram, for example showing how all local areas fare on a particular indicator (with targets and comparisons against other areas also shown)
  • Users can quickly identify which areas are above or below average (or targets), and/ or how performance compares against national averages
  • Can be developed in standard applications, including Excel (based on standard bar-charts, or stocks-and-shares charts)

Click here for examples of score-cards in the visualisation gallery.

2. Treemaps: exploring relationships between many variables

Treemaps were invented by Ben Shneiderman, originally as a way of visualising the way that computer hard-drives were taken-up by different types of files (some background and history is here). During the early 1990s, Shneiderman and researchers at University of Maryland explored using treemaps for portraying all sorts of structured hierarchical data, including sports datasets, stock and shares portfolios, and satellite management systems.

Shneiderman and other researchers identified that roughly 15 minutes training was needed before users could clearly understand - and successfully use - the treemaps. But once understood, they allow users to easily and intuitively explore datasets in many different dimensions.

Treemaps represent data in two ways: the size of the boxes and the colour of the shading. Larger boxes and darker colours usually represent higher numbers, and the boxes are usually ordered by size so that the biggest rectangles are shown in the top-left corner. For example, this allows users to visualise how a particular population is broken down into sub-groups, and how those sub-groups compare on some measure

Image for this case study

Source: Staffordshire (2008) Joint Strategic Needs Assessment

The treemap figure above is taken from the Staffordshire Joint Strategic Needs Assessment, and shows the population of Staffordshire arranged by BME group and gender (the size of the boxes is proportional to the size of BME group), and the percentage of people within each group with a limiting long-term illness (LLTI) who are not in good health (colour of the boxes). For example, the treemap shows that the proportion of Irish females with a LLTI who are not in good health is higher than for Irish males (the Irish females box in the treemap is darker than the Irish males box, identifying a higher level of LLTI in the group). An interactive version of the treemap could be used to explore how LLTI and other health variables varied across different age, gender, and ethnic sub-groups.

Interactive treemaps allow users to explore different variables of interest - for example, see the Map of the Market which shows stocks organised by industry groups, size-coded by market capitalisation, and colour-coded to show rise or fall. The New York Times Treemap of infant deaths enables you to compare infant death rates between any two countries since 1960, with data for one country represented by colour and data for the other represented by rectangle size.

Key features and advantages of treemaps for visualising data include:

  • Present two different aspects of a dataset at once, for example the number of people in some population group (BME groups by gender in the example above), and the % of that group experiencing some issue (limiting long-term illness in the case above)
  • Once users are familiar with the concept and use of treemaps, they provide a powerful and intuitive interface to explore complex datasets. But treemaps are less straightforward to interpret for new users.
  • Interactive treemaps (for example web-enabled) allow very powerful analysis of complex datasets in real-time.
  • More difficult to develop, requiring non-standard applications

Click here for examples of treemaps in the visualisation gallery.

3. Wall-charts: highlighting performance of many areas on many indicators

Wall-charts are essentially summary tables, typically showing how different areas compare on a range of different indicators - allowing users to view a great deal of information at once. Colour-coding is used to highlight where data values are above or below a baseline.

The example below is taken from the Oxfordshire Joint Strategic Needs Assessment, showing wards across the county against key health indicators. The colours show whether performance is significantly worse (red and yellow) or better (green) than the county average. Of the 63 JSNAs we reviewed, 4 summarised information in this way.

Image for this case study

Source: Oxfordshire (2008) Joint Strategic Needs Assessment

Key features and advantages of wall-charts for visualising data include:

  • Present very large range of data at once, for example all local neighbourhoods against all key indicators (or range of key indicators over time)
  • Colour-coding is used to highlight where data values are worse or better than some target
  • Easy for users to identify general patterns, eg which areas/ services are performing well or badly across a range of indicators (but by contrast, more difficult to pick-out specific data values)
  • Easy to develop in standard applications, including Word and Excel

4. Boxplots: identifying where differences are likely to be significant

When comparing different areas on some indicator, it is common to use a bar-chart, with the height of the bar representing the data value. Where the data is taken from a sample, a bar-chart can still be used to show the average value (for example the average Index of Multiple Deprivation score across all Super Output Areas in a Local Authority can be plotted).

However, a straight bar-chart does not help users identify whether differences between data values across areas are likely to be significant - this is where box-plots (also known as box-and-whisker plots) come in.

Image for this case study
Source: Surrey (2008) Joint Strategic Needs Assessment

The figure above shows a box-plot of variation in ward-level under-18 conception rates across the 34 English shire counties. For each county, the box and line show five data values:

  • The smallest and largest values are shown at the top and bottom of the lines. In the circled example, this shows that under-18 conception rates across all wards in Surrey vary from 15 to 30 per 1,000
  • The size of the box shows the interquartile range - from 25% of the sample to 75%. In Surrey this varies from 22 to 28 per 1,000
  • The line in the box shows the median value of the data sample. In Surrey this is 25 per 1,000

Box-plots can be used to show whether differences between areas are likely to be meaningful. Where there is a good deal of overlap between the boxes (which show the 25-75% interquartile range), differences in the mean or median values are unlikely to be significant.

Key features and advantages of boxplots for visualising data include:

  • Summarise a distribution of values, for example (as above) showing for different shire Counties how teenage pregnancy rates vary at ward level
  • Provide more information on the variation in a group of values than simply showing the Confidence Intervals
  • Help users understand whether differences between values for different areas are likely to be significant (in the example above, showing whether differences between the County teenage conception rates are likely to be meaningful)
  • Users can quickly identify which areas are significantly above or below average
  • Can be developed in standard applications, including Excel (based on bar-charts, or stocks-and-shares charts)

Click here for examples of boxplots in the visualisation gallery.

5. Quadrant plots: classifying areas on two indicators

Essentially scatter-plots, quadrant plots are used to classify areas, groups or issues into four categories depending on their scores on two indicators. Often, the scale of the problem (eg, numbers or proportion of cases) is plotted against change over time (eg, increase over 12 months). Alternatively, quadrant plots can be used to compare the importance of an issue as measured by (a) objective data such as recorded offences or unemployment levels; and (b) subjective views such as priorities of local citizens. 3 of the 94 of the Sustainable Community Strategies we reviewed use Quadrant Plots.

Image for this case study
Source: Redcar and Cleveland Sustainable Community Strategy

In the quadrant plot shown above, priorities highlighted by residents in Redcar and Cleveland are shown against two axes:

  • Importance (shown along the x-axis) - based on the question "how important is this issue?"
  • Improvement (y-axis), based on "in need of most improvement"

The different issues are then categorised into the four quadrants:

  • Bottom-left quadrant: Lesser importance and lesser need of improvement
  • Upper-left quadrant: Lesser importance but greater need of improvement
  • Upper-right quadrant: Greater importance and greater need of improvement
  • Bottom-right quadrant: Greater importance but lesser need of improvement

Key features and advantages of quadrant plots for visualising data include:

  • Present two different aspects of a dataset at once, for example the importance of an issue as measured by (a) objective data such as recorded offences or unemployment levels; and (b) subjective views such as priorities of local citizens
  • Relatively easy to intepret for general users
  • Can be easily developed in standard applications, including Excel (as a scatterplot with labels)

Click here for examples of quadrant plots in the visualisation gallery. Scatterplots can also be used as quadrant plots, if the axes can be meaningfully divided into quadrants.