Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
7. Accessibility of data visualizations :
An overview of European statistics
institutes
Mikael Snaprud and Andrea Velazquez
Abstract
Access to public data is important for people to stay informed. Access
to visualizations of national statistics can be essential in order to take
part in political discussions and so to shape a democratic society. In this
chapter we investigate accessibility for people with disabilities to data
visualizations from a selection of European National Statistics Institutes
(NSIs). We outline related practices and approaches to accessibility
improvements and propose a way to evaluate and compare accessibility
aspect s of data visualizations. The ndings indicate t hat in contrast to the
recently harmonized European legal requirements, the degree to which
the data visualizations meet the requirements, and the approaches to
meet them, are very diferent among the NSIs across Europe.
Keywords: Data visua lization; Accessibility; Web Accessibility Directive;
National Statistics Institutes; NSI.
Introduction
Data visualizations can inform citizens about political topics, and access
to them for all citizens, regardless of ability and related technology use, is
essential for democratic processes. The United Nations Convention on the
Rights of Persons with Disabilities requires that appropriate measures are
taken to ensure access for persons with disabilities, on equal basis with
others, to information and communication technologies, including the
internet. The European Web Accessibility Directive (WAD), transposed into
Engebret sen, M. and H. Kenne dy (eds.), Data Vi sualiza tion in Societ y. Amsterda m: Amsterd am
University Press, 2020
10.51 17/9789463722902_07
MIKAEL SNAPRUD AND ANDRE A VELAZQUEZ
national law for all the EU member states from September2018, makes web
accessibility a legal obligation. It requires that public sector bodies provide
accessibility statements, including a list of content that is not accessible,
the reasons for the inaccessibility and accessible alternatives to it, and a
feedback mechanism for users to report accessibility problems on all of their
websites, including in relation to data visualizations. National Statistics
Institutes (NSIs) are a key source of such visualizations. Therefore, in this
chapter, we focus on the accessibility of data visualizations produced and
provided by European NSIs. We present results from our evaluation of the
accessibility of data visualizations on NSI websites and from research with
NSIs regarding their preparations to conform with the Directive.
In order to be accessible, the data visualizations (DVs), like all other web
content, need to be perceivable, operable, understandable, and robust. For
this chapter we leave out understandable, as an evaluation of such would
require many more resources than we had available for our study. However,
we add ndability, the ease by which a piece of information on a website can
be found (Jacob & Loehrlein, 2009; Wikipedia, Findability, 2018), since it can
have an impact on the ability of citizens to participate in democratic discourse.
Perceivability and operability of DVs are related to general website ac-
cessibility issues. For example, menus that cannot be used via keyboard
navigation or input elds that are not properly labelled can cause problems
for web users with disabilities. We used Tim Berners-Lee’s 5-star scheme,
described below, to assess the robustness of data formats (5-star data, 2015).
We used an automated accessibility checker tool, WTKollen, to test
the accessibility of the websites of 44 out of 59European NSIs (WTKollen,
European NSI sites, 2018), the results from which are presented as scores.1
The results presented below therefore refer only to the 44 websites we tested,
and not to others which, for various reasons, it was not possible to test prior
to publication of this chapter. We also carried out expert testing of data
visualizations found on these websites, and conducted email surveys and
semi-structured interviews with appropriate staf within the NSIs.
A limitation of our work is that we did not carry out user testing of the
websites and DVs with web users with disabilities. This is widely deemed to
be the most appropriate way of evaluating the accessibility of websites (e.g.
Coyne & Nielsen, 2001), but it is resource-intensive, especially in the case of
EU-wide research such as ours. Automated checker tools, like accessibility
measures more generally, also tend to privilege the needs of people with
certain disabilities, such as visual impairment, and ignore the needs of
1 The WTKollen projec t is suported by the Swedish Post and Telecom autorithy.
ACCESSIBILITY OF DATA VISUALIZATIONS
others, such as intellectual disabilities (Kennedy, Thomas, & Evans, 2010). It is
also the case that some tests cannot be automated—for example, automatic
image processing may not be able to distinguish a cat from a dog in a blurr y
picture to determine whether an alternative text for the picture is helpful
or not—and this is another limitation of automated accessibility testing.
Despite these limitations, we think that our research can provide valuable
insights into the extent of the accessibility of DVs across European NSIs.
Data collection
Automated accessibility testing
The European Web Accessibility Directive is based on the Web Content Ac-
cessibility Guidelines (WCAG) from the World Wide Web Consortium (W3C).
The guidelines are intended to cover any online content, including DVs, for
people with disabilities, such as visual, auditory, physical, speech, cognitive,
language, learning, and neurological disabilities. The WCAG 2.0 (W3C, 2008)
was replaced by WCAG 2.1 in June2018 (W3C, 2018).2 Following these guidel ines
will often make web content more accessible as well as serving other purposes.
Proper use of alternative text descriptions on images, for example, can enable
search engines to provide accurate search results. To guide the testing process,
the W3C published the WCAG Evaluation Method ‘WCAG-EM’ in 2014 (W3C,
2014). This methodology ofers guidance on the expertise required to test web
accessibility, how to select webpages from a website, and how to report the
ndings. WCAG2.0 and WCAG-EM are therefore the basis for a range of web
accessibility testing methods, tools, and legislation in Europe.
We carried out website accessibility evaluations with t he WTKollen checker
tool (WTKollen Page checker, 2018), which is based on WCAG 2.1 and WCAG-
EM 1.0. Whereas the WCAG-EM guidelines indicate what to look for to design
the accessibility tests, they do not specify exactly how to implement tests.
The applied tests are listed online (GitHub, 2018)—not all of them are equally
relevant to all people with disabilities. For example, colour contrast can be
important for a person with visual impairment while irrelevant for a blind
user. Hence diferent user groups would assign di ferent weights to the same
test. For the score calculation, we needed to have one weight only for each
test. Therefore, we decided to let all tests have the same impact on the score.
2 To better meet t he needs for three m ajor groups: user s with cogn itive or learn ing disabil ities,
users w ith low vision, and users w ith disabilities on mobile devices.
MIKAEL SNAPRUD AND ANDREA VELAZQUEZ
To select pages from websites, we used a crawler, to nd up to 6,000 pages. A
random sample of 600 of these pages was used to represent the site to be tested.
The score for a webpage was computed as the ratio of passed tests to applicable
tests for each success criteria where associated tests applied. Similarly, the score
for a website was aggregated over the test results for all the success criteria.
Manual test procedure
We uses a DV grouping proposed by Kirk (2012, p.76) for the analysis of the DVs:
Exploratory visualizations which aim to allow readers to discover
features by interrogating the data themselves;
Explanatory visualizations which aim to convey specic information
to readers, based on a prede ned narrative;
Exhibitory visualizations which are also based on data, but contain an
artistic element.
Further, we also grouped the DVs by ways of interacting with them:
Static visualizations, such as a PNG image;
Dynamic visualizations which move, but without users activating them;
Navigable visualizations, which change based on user interaction;
Congurable visualizations, which enable users to select graph types
or numbers, to move levels, or to select variables.
Our evaluation proceeded according to the following steps:
1. Locate the selected DV through a search on Google and local search,
record rank (automatic accessibility check of the website as indica-
tion for how easy it is to navigate is done earlier)
2. Examine data presentation and downloadable data formats (i.e.
how the data are provided, discussed below)
3. Group DV (according to Kirk’s groupings discussed above)
4. Group DV (according to mode of interaction commented above)
5. Carr y out manual accessibility tests for keyboard navigation, zoom,
and textual description of the image (discussed below)
6. Look for accessibility feedback option (i.e. verify if there is an ac-
cessibility feedback mechanism on the page)
7. Look for supplementary ser vices that may help the user understand
the DV:
7.1. FAQ: Is there an FAQ available from the page? What kind of
FAQ (general info, speci c content)?
7.2. Languages: Is there support for multiple languages?
7.3. Glossary: Is there a glossary to explain terms used in the
statistics on the page?
ACCESSIBILITY OF DATA VISUALIZATIONS
The outcomes of the tests were recorded in a spreadsheet together with
screenshots and URLs so that a test can be repeated if needed and for ac-
countability purposes.
An important accessibility feature for people with motor impairments
is the ability to navigate by using the keyboard or other input device
instead of a mouse. If the page with the DV is designed to enable this,
then users can reach all elements on the page and navigate forward and
backward through the elements in the browser window with the tab and
shift keys. Keyboard navigability can also allow users to select data and
con gure DVs.
The zoom feature is essential to magnify both text and images for people
with visual impairments. Text zoom should render the text so that there is
no need to scroll sideways. To test zoom features, it is necessary to explore
whether the page has an option to enlarge the content or not. We evaluated
the ability of the page to present the screen content enlarged.
There are two ways to provide text alternatives to images on webpages,
thus making them accessible to people with visual disabilities: a short
alternative text (alt-text) and a longer description: longdesc. The purpose
of the longdesc is to provide more elaborate information when a short
alt-text does not adequately convey the function or information provided
in a non-text element on a webpage (W3C, 2016).
Our test recorded whether
the longdesc was used for complex DVs.
In many cases the data behind a DV are provided for download from the
NSI site. Data formats have a strong impact on whether users can access
and reuse the data. If a person is not able to use the DV, then a reusable
data format can be more accessible and thus enable users to understand
the data. Reusable data formats are also in line with the intention of the
European Public Service Information directive, ‘PSI’ (EUR-Lex, 2005). The
5-star scheme proposed by Tim Berners-Lee is a practical way of evaluating
the extent to which a given dataset can be reused (5-star data, 2015).
1-Star: the data are open; however, they are locked-up in a document
making it hard to get them out of the document, e.g. in a PDF or JPEG.
2-Star: the data are accessible on the web in a structured way; however,
they are still locked-up in a document depending on proprietary soft-
ware, such as Microsoft Excel.
3-Star: the data are available on the web and can be manipulated in
any way, without the need to own any proprietary software package,
e.g. in CSV format.
4-Star: as above, and the data items have a URL and can be shared on
the web, for example via links.
MIKAEL SNAPRUD AND ANDREA VEL AZQUEZ
5-Star: as above, and it is also possible to link data to other data to
provide context, to discover more related data while consuming the
data, thus beneting from the network efect, e.g. through a link to a
Wikipedia article.
More stars means more reusable and also to some extent more accessible.
For example, a screen reader, used by blind users, will not be able to read
data in a JPEG image. In addition, if access to data is only possible with
proprietary software, then users without the software in question will
be unable to access it. The context provided in the 4- and 5-star levels
does not really matter for accessibility, but is helpful for automated
assessment of what the data are about. For our test, we detected if the
data could be downloaded, and recorded the data format mapped to
this 5-star scheme.
Finally, for interactive DVs, we also recorded the following two properties.
Comparability: number of variables that could be represented in the
same graph.
Number of representations available: di ferent kinds of charts available
for representing the data, such as bar charts, linecharts, maps, pie
charts.
Web accessibility for the NSI websites
The automated evaluation of website accessibility was carried out in the
period from October25 to November13, 2018. The NSIs with the 12 highest
scores are listed in Table7.1. The highest score is awarded to the Irish NSI,
followed by a group of 8 NSIs with a score of 99. At the lower end of the
list, we nd the NSIs from Greenland (score 67), Cyprus (69), and Iceland
(71). To view the full list and the details about the detected accessibility
issues, visit the webpage (http://axe.checkers.wtkollen.se/en/benchmark ing/
testrunresults/d235468e-65a4-43b2-8428-5908f061f9).
The above list was up-to-date at the time of publication. For the manual
testing 14 NSIs were selected, based on high accessibility scores (in alphabetic
order): Czech Republic (CZ), Denmark (DK), Germany (DE), Ireland (IE),
Luxembourg (LU), The Netherlands (NL), Norway (NO), Poland (PL), Spain
(ES), Sweden (SE), Switzerland (CH), and United Kingdom Visual ONS
(UK-visual), and on a suggestion from Statistics Norway that they contain
interesting DVs: Portugal (PT) and Slovenia (SI).
ACCESSIBILITY OF DATA VISUALIZATIONS
DV ndability on the NSI websites
We searched for statistics about national population as a case to obtain
an indicator of the DV ndability on the NSIs websites. Population is well
covered across all NSIs and it also seems to be a popular search topic. In
a rst attempt we used Google to search for ‘population’ and the name of
the NSI. For Norway the search phrase was then ‘population SSB’. For a
corresponding search for each of the selected NSIs, all but one NSI appeared
as the rst item in the search results list. For thirteen out of fourteen NSI
sites the local search returned the relevant page in rank one. Both the Google
analytics data from Statistics Norway and the experience from Eurostat
indicate that it is more common to search just for ‘population’ without any
NSI portal name. For some searches the rst result in Google is data from
the World Bank. For Norway these data originate from Statistics Norway NSI.
A DV included in the Google search results list (Google public data, 2018),
such as the one we found from SSB in Norway, can be convenient for the
user, who may not need to look any further for the requested statistics.
The DV we found had colour contrast issues, but otherwise it was quite
accessible, using SVG graphics and ofering the ability to present the data
in several languages.
Our initial approach to test ndability was to use similar English search
phrases to nd population DVs across all NSIs (e.g. ‘population SSB’ for Norway).
In the course of the study we noted that the content on the NSI websites is
Table7.1 Overview of NSI websites and accessibility score from the WTKollen
checker tool
Rank Score Country NSI short name URL
1100 Ireland Nisra UK https://www.nisra.gov.uk/
299 Spain INE ES http://ine.es/
399 Sweden SCB SE http://www.scb.se/
499 Denmark DST http://www.dst.dk/
599 Germany Statistikportal DE http://www.statistik-portal.de/
699 Switzerland BFS CH http://www.bfs.admin.ch/
799 Norway SSB NO http://www.ssb.no/
899 Luxembourg Statistiques http://www.statistiques.public.lu/
998 United Kingdom ONS UK https://www.ons.gov.uk/
10 98 Czech Republic CZSO CZ https://www.czso.cz/
11 98 Poland Stat PL http://stat.gov.pl/
12 97 The Netherlands CBS NL https://www.cbs.nl/
MIKAEL SNAPRUD AND ANDREA VEL AZQUEZ
mostly prepared for the national audience, to be searched in a national lan-
guage. Therefore, a search in English across diferent NSI may produce results
that are not relevant for the targeted national users. To rene this result we
could do a new search for population in the local language. We also note that
the search eng ine result can depend on who is doing the search and from where.
Data presentation analysis
The data presentation was assessed for the fourteen evaluated NSIs. The
accessibility score is the result obtained in the period from September2018 to
November2018. The accessibility scores of the population data presented are:
Score of 95 to 99, a few tests failed: DE
Score of 85 to 95, some tests failed: ES, LU
Score of 70 to 85, many tests failed: CZ, IE, NL, NO, PT, SE, UK-visual
Score below 70, most tests failed: CH, DK, PL, SI
The data presented of twelve NSIs are congurable in tables in which it is
possible to select the variables to show; only one is navigable (DE) allowing
movement or arrangement of the data presented; and one static (UK-visual).
For nine out the fourteen NSIs the keyboard navigation is enabled. Only
four NSIs (CZ, ES, NL, NO) have the zoom feature; and in all but one (DE),
the option to download the data is possible: CZ is 1-star; 10 NSIs are 3-star
(that is, non-proprietary open format), NL 4-star and LU 5-star.
Exploratory/Interactive visualization analysis
Twelve out of fourteen NSIs have interactive tools to graph the data which
enable users to produce their own visual representations of the available
data. CZ and UK-visual do not have interactive DVs. PT graphs need Adobe
ash player which is not accessible for people with disabilities, and the DE
tool is provided only in German and therefore not evaluated. Therefore,
only ten DVs were manually tested.
Only four out of ten DVs could be checked automatically: DK, NL, NO, and
ES, mostly because the generated graphics do not have an explicit link to
enter into the checker tool. Single page applications will present the same
URL independent of user conguration of the DV. The accessibility scores are:
ES scored 85 to 95, some tests failed
NO and DK scored 70 to 85, many tests failed
NL scored 65, most tests failed
ACCESSIBILITY OF DATA VISUALIZATIONS
We note that all of the exploratory DVs tested have data to download, but
only in a format not necessarily accessible and not suited for machine
processing, mainly JPG and PNG format. Zooming is only supported
by the DK, ES, NL, and NO examples and keyboard navigation is only
supported in four out of ten cases (CH, ES, NL, PL). None of the DVs has
longdesc enabled.
The DVs can be presented in diferent graphs depending on the nature of
the data selected. For example, if the data selected do not include territories
the map visualization is not a valid option. Some of the options available
include bars, pie, lines, points, pyramid, and map. The number of graphs
available is variable among NSIs, with the maximum fteen (CH) and
the minimum two (ES and PL). All the tools tested can compare multiple
variables. By their nature, all the DVs are congurable.
We can identify six diferent tools in use among the 10 NSIs assessed.
The rst one used in Norway, Ireland, Slovenia and Denmark. The second
in Sweden and Switzerland. While Spain, the Netherlands, Luxembourg,
and Poland all seem to use diferent tools.
Explanatory visualization analysis
A total of twelve NSIs were evaluated for explanatory DVs. For the Luxem-
bourg and Slovenia NSIs we did not nd any explanatory visualizations.
The accessibility scores obtained are the following:
Score of 95 to 99, a few tests failed: CH, CZ, DE, DK, ES, NL, NO, PL, SE
Score of 85 to 95, some tests failed: IE, UK-visual
Score of 70 to 85, many tests failed: PT
Eleven of the explanatory DVs tested are static and one is dynamic (NL).
Only ve out of the twelve have the option to download data:
1-star: CZ, IE, NO, PL
3-star: PT
Only two support keyboard navigation (NO, PT), ve have the zoom ability
(DE, ES, NL, NO, PL) and two support the longdesc (PT, UK-visual). The
accessibility scores are higher for this category than for other DVs. This
is because the content is typically simpler, mainly consisting of text and
numbers. Even though the explanatory visualizations are simple, the
option to download the data is not common. This makes the data harder
to reuse.
MIKAEL SNAPRUD AND ANDRE A VELAZQUEZ
Exhibitory visualization analysis
Ten NSIs were assessed for their exhibitory visualizations; for the remaining
four (DE, DK, LU, SE) we did not nd any exhibitory DVs. Exhibitory DVs are
produced like artistic posters, in diferent formats mainly in PDF formats and
in PNG. The scores were also calculated with the PDF checker when necessar y:
NO and PL: 95 to 99, a few tests failed
CZ, IE, NL, and UK-visual: 85 to 95, some tests failed
SI PDF checker score: pass 6, fail 3
CH, ES, and PT PDF checker score: pass 5, fail 2
Seven exhibitory DVs are static, two dynamic (CZ, NL), and one navigable
(UK-visual). None has longdesc; and only NL support keyboard navigation.
Five out of the nine exhibitory DVs have zoom ability, and seven have data
to download, all are 1-star, making it hard to reuse the information.
Services to support users to understand the DVs
None of the fourteen evaluated NSIs has an accessibility feedback form,
although most of them have a general feedback form to comment on the data
or the page. One NSI page has a phone and email address for accessibility
feedback and two NSIs have accessibility statements. By September2020,
all NSIs will need to have an accessibility feedback mechanism in place on
their websites to conform to the Web Accessibility Directive.
FAQs are found on nine out of the fourteen NSIs. We found four FAQs cont ain-
ing general information about the page and ve FAQs about specic content
like consumer prices, wages, or summer prices. The option to select national
language or English is provided by twelve out of fourteen NSIs. The UK-visual
content is provided in English, and does not support any ot her language, possibly
because this website is focused on DVs. The German NSI is only in German.
Glossary access to explain terms used on pages or in DVs is provided by
nine out of fourteen NSIs. The glossary entries are found directly on the
page, provided as references to an internal glossary or external ones like
the ones from the OECD or from Eurostat’s Statistics Explained (Eurostat,
Statistics Explained, 2017).
NSI practices relating to DV accessibility
To supplement the analysis discussed above, we also used surveys and
interviews. Statistics Norway helped us to distribute two surveys to the
ACCESSIBILITY OF DATA VISUALIZATIONS
network of European NSIs. Both surveys had only three questions each, to
keep them simple and to increase the response rate. The rst survey was
intended to get an idea about how the NSIs are preparing for the WAD, raise
awareness about detected barriers with accessibility evaluation results, and
to collect some input on how evaluation results can be shaped to enable the
NSIs to understand them and to use them to repair the reported barriers. The
second survey was designed to capture good DV examples and developments
in the WAD preparations. The rst survey, sent out by Statistics Norway in
October2017 to about 80 representatives from 39 NSIs in Europe, received
a response from about 20percent. The second survey, sent in April2018 to
the same group of respondents, had a lower response rate of 13percent. One
possible reason for the lower response rate may be a focus on the General
Data Protection Regulation (GDPR) at this time. Interviews were carried
out with Statistics Norway and with Eurostat.
The practical responsibility to make sure that DVs are accessible lies
with software developers or with communications departments within
the NSIs. Useful accessibility input has in several cases been obtained from
colleagues with visual impairments. External consultants are sometimes
contracted to audit overall websites. This is a costly operation and therefore
not carried out regularly. Advanced and regular usability testing has been
in place for a long time across a number of NSIs, to ensure that the statistics
can be found and used. It seems that accessibility is an emerging topic to
be included in regular testing activities for NSI online content.
Several diferent automatic tools are used by the NSIs to evaluate acces-
sibility. Commonly used tools are Site Morse (see https://sitemorse.com/)
and aXe Core (see http://deque.com/). One of the NSIs also reported that
they intend to build a new tool. Such tools are helpful and cost-efective
to operate, but not always straightforward to use. One important caveat is
that these tools do not cover all conceivable tests.
From the rst survey we found that respondents planned to pursue mainly
two diferent approaches to improving their website and DV accessibility.
They planned to invest in human resources which include staf training
programmes and hiring accessibility expertise consultants, and facelifts
or complete redesigns of their website.
Together with the survey we provided a benchmarking list similar to
Table7.1. The respondents were asked to comment on the results form. The
checker tool (WTKollen, European NSI sites, 2017) was perceived as useful by
respondents, and the ndings veried that older or more complex webpages
are more likely to have accessibility barriers than newer or simpler pages.
For future tool development, respondents said that they would like to have a
readability test and image evaluations. There was also a suggestion to group
M IKAEL SNAPRUD AND ANDREA V ELAZQUEZ
webpages according to complexity for better comparisons. To improve the
presentation of the results in the checker report, the following suggestions
were made:
– The results could have example images for better and faster under-
standing.
– The results could be sorted (e.g. by error status or importance)
– The page could be responsive
– It would be nice to be able to export all the errors to .csv/.pdf/.xlsx, as
this would help the organization of corrections.
Some respondents raise concerns about tools since they do not always return
the same result for the same content on a webpage. Such diferences may
even have prevented people from using checker tools at all. We also noted
that some NSIs expected an ocial tool to be prescribed by the European
Commission or their national ministry. However, the WAD is prepared in
a way that is tool independent and there is no tool mentioned in the WAD
implementation act.
In the second survey we requested users to provide examples of accessible
DVs. Most respondents declined and indicated that they were working on
this now. In our view, there is great potential in using DV templates from
Eurostat to spread good accessible practices. We also asked about NSIs’
preparations for the provision of accessibility statements and feedback
mechanisms. Several NSIs have accessibility statements, sometimes linked
from the page footer. However, in general, such statements do not list known
deviations from accessibility requirements. Mostly they provide information
about accessibility features of the sites. One NSI has had an accessibility
statement since 2005 and regularly performs tests in cooperation with
external experts. Many respondents aim to collect the information for the
accessibility statement from the accessibility reports produced by ocially
adopted verication tools, or from user feedback.
In terms of preparations for the provision of feedback mechanisms, we
recorded two approaches. One is to use a dedicated email to receive reports
about accessibility problems. This approach can make it dicult for a user
to remain anonymous, and it can also become hard to manage responses
and task assignments for large volumes of feedback. The second approach
is to use a general feedback mechanism already existing on the website.
This may meet the formal requirements, but such feedback mechanisms
are not designed to (automatically) collect data on accessibility problems,
or to export reports about them to share good practices in terms of xes
or repair approaches.
ACCESSIBILITY OF DATA VISUALIZATIONS
Sharing good practices
There are at least three current approaches to reusing good DV examples.
The Digicom project is an initiative to share good practices among the NSIs.
As part of this project, Eurostat has developed templates to present DVs of
statistics. These templates are designed so that they are easy to translate
and connect with an API to the data from Eurostat.
There are also a range of DV libraries that can be used to reuse good
DV examples, like D3, Google charts, or Highcharts. In more accessible
solutions, DVs are scalable to allow for zooming, and have functionality
to encourage or force the developer to describe the non-textual elements.
Such encouragement could raise developers’ awareness of inaccessibility
impacts for users with disabilities. A simple export of the data can also be
helpful to enable users to explore the data with a tool of their own preference.
Presentation through aggregators like Google can also be e cient. The
Google Public Data Explorer (see https://support.google.com/publicdata)
provides large, public-interest datasets from sources like Eurostat and the
World Bank in a common presentation format. With this service, the user can
nd datasets and explore them with diferent chart types. Two important
advantages of this approach are, rst, that users will be familiar with the
user interface and, second, that they will easily nd it, since Google has over
90percent of the European search market share (StatCounter GlobalStats,
2018). However, such intermediary access can also be used to track users
and to prevent the user from nding the original data source with more
updated data or further information about the dataset.
We have not been able to identify a reference library of DVs. The Internet
Archive is a valuable resource for longtime references for a large portion
of the online content. Unfortunately, this archive does not have all the
relevant pages from the NSIs and it cannot store the dynamic features of
most dynamic DVs. The Internet Archive also will not have direct access
to the static databases often serving the ‘live’ data to the dynamic DVs.
Conclusions
There are good examples of DVs where we found few accessibility barriers.
However, despite the Web Accessibility Directive, there is still lot of room
for improvement. There are several diferent accessibility testing tools in
use among NSIs to test the accessibility of their websites and their DVs. In
our survey we were not able to nd examples of NSIs who systematically
MIK AEL SNAPRUD AND ANDREA VELAZQUEZ
apply user testing approaches to uncover accessibility issues. Several report
that they occasionally ask a colleague with disabilities to test content. For
the WAD preparations, we saw very limited work towards design of an
accessibility statement or to organize a feedback mechanism.
In general, NSIs are aware of accessibility issues. Still, three factors seem
to have hindered focused progress towards comprehensive accessibility
provisions, and to prepare for the WAD. Several NSIs indicated that they
would wait for the WA D implementation act to be nalized before they
would take action. The General Data Privacy Regulation seemed to demand
more attention, as there are high nes associated with a breach compared to
accessibility problems which breach the WAD. The third reason is diferences
in accessibility checker tool reports for the same element on a webpage. Some
NSIs expect that an ocial tool will be named. The draft implementation
act for the WAD does not refer to any named accessibility tool, and there
seems to be no intention to use a particular tool for the implementation
from the regulators as far as we have been able to nd out.
There are several diferent accessibility testing tools in use among NSIs
to test the accessibility of their websites and their DVs. For exploratory,
interactive visualization we have found six diferent tools in use. Given
this relatively small number of tools, targeted improvements of them can
have a large efect for many users. Whatever approach is used, the central
role of the NSIs and their DVs in national democratic discourse calls for
particular awareness of accessibility.
Acknowledgements
To the INDVIL team for all the fruitful discussions and inputs, to the W TKol-
len project for providing the accessibility checker tools to test websites,
webpages with DVs, and PDF documents. And last but not least, many
thanks to Helen Kennedy for her helpful comments and substantial support
to shape this chapter.
References
5-sta r data. (2015). 5 star open data . Retrieved October22 , 2018 from https://5stardat a.
info/en/
Coyne, K., & Nielsen, J. (2001). Beyond A LT text: Making the web eas y to use for users
with disabilities. Fremont, CA: Nielsen Norman Group.
ACCESSIBILITY OF DATA VISUALIZATIONS
EUR-Lex. (2005). Decision No 456/2005/EC of the European Parliament and of the
Council of 9March2005 establishing a multiannual community programme to
make digital content in Europe more accessible, usable and ex ploitable. Retrieved
from https://eur-lex.europa.eu/legal-content/EN/TXT/?qid=1535296253785&u
ri=CELEX:32005D0456
Eurost at—Statistics Expla ined. (2017). Welcome to statistics explained. Retrieved
October22, 2018 from http://ec.europa.eu/eurostat/statistics-explained/index.
php/Main_Page
GitHub. (2018). Axe Core tests. Retrieved October22, 2018 from https://github.com/
dequelabs/axe-core/blob/master/doc/rule-descriptions.md
Google public data. (2018). Public data. Retrieved October22, 2018 from https://
www.google.com/publicdata/explore
Jacob, E. K., & Loehrlein, A. (2009). Information architect ure. Annual Re view of Infor-
mation Science and Technology, 43(1). https://doi.org/10.1002/aris.2009.1440430110
Kennedy, H., Thomas, S., & Evans, S. (2010) Can the web be accessible for people
with intellectual disabilities? The Information Society, 27(1), 29-39. https://doi.
org/10.1080/01972243.2011.534365
Kirk, A. (2012) Data visualization: A successful design process. Birmingham: Packt
Publishing Ltd.
StatCounter GlobalStats. (2018). Search engine market share Europe. Retrieved
October22, 2018 from http://gs.statcounter.com/search-eng ine-market-share/
all/europe
W3C. (2008). Web content accessibility guidelines (WCAG) 2.0. Retrieved October22,
2018 from http://www.w3.org/TR/WCAG20/
W3C. (2014). Website accessibility conformance evaluation methodology (WCAG-
EM) 1.0. Retrieved October22, 2018 from https://www.w3.org/TR/WCAG-EM/
W3C. (2016). Using longdesc. Retrieved October22, 2018 from http://www.w3.org/
TR/WCAG20-TECHS/H45.html
W3C. (2018). Web content accessibilit y guidelines (WCAG) 2.1. Retrieved October22,
2018 from https://www.w3.org/TR/WCAG21/
WTKollen—European NSI sites. (2017). Checked sites from European NSI sites.
Retrieved October22, 2018 from http://checkers.wtkollen.se/eu-nsi
WTKollen—European NSI sites. (2018). Checked sites from European NSI sites.
Retrieved October22, 2018 from http://checkers.wtkollen.se/en/benchmark-
ing/testrunresults/d235468e-65a4-43b2-8428-5908f061f9?Sector&ags.
fulltestrunresult
WTKollen—Page Checker. (2018). Check the accessibility of a web page. Ret rieved
October22, 2018 from http://checkers.wtkollen.se/
Wikipedia, Findability. (2018). Findability. Retrieved October22, 2018 from https://
en.wikipedia.org/wiki/Findability
MIKAEL SNAPRUD AND ANDREA VELA ZQUEZ
About the authors
Mikael Snaprud is the CEO of Tingtun AS. He has managed several projects
in the area of e-Inclusion with European and national research grants. He
has 20 years of experience from research and teaching, and has co-authored
reports about ICT policies for the European Commission and the United
Nations.
Andrea Velazquez works as scientic adviser at Tingtun AS. She has been
part of the team of several e-Government projects funded by the European
Commission and the Research Council of Norway. In the past she worked
in processes benchmarking for quality engineering, and productivity
programmes in the industry.
