The state of corporate website accessibility

• Authors: Eleanor T. Loiacono;Nicholas C. Romano, Jr.;Scott McCoy
• Venue: Communications of the ACM - The Status of the P versus NP Problem
• Year: 2009

Web accessibility continues to have important social, legal and economic implications for ecommerce. Over 50 million Americans have disabilities and so do around 600 million world-wide (www.census.gov/prod/2003pubs/c2kbr-17.pdf; www.usdoj.gov/crt/ada/busstat.htm). Disabilities include a vast range of issues: cognition, vision, motor skills, and hearing. The disabled comprise 19.3% of the U.S. population, more than any other minority group, including the next largest group---Hispanics (14.9%). This growing population commands significant discretionary funds. The U.S. Department of Labor estimates those with disabilities control $175 billion (www.dol.gov/odep/pubs/ek98/provide.htm) exceeding twice that of teens and 17 times that of tweens (8--12 year-olds), currently the most sought after demographic groups (www.census.gov/prod/2003pubs/c2kbr-17.pdf). This untapped, growing market exceeds most company's estimations. In this study, we extend a previous CACM paper that surveyed accessibility at a snapshot in time with historical and additional perspectives on accessibility of Fortune 100 (F100) Web sites. The initial study revealed that over 80% of the F100's websites were potentially inaccessible to people with visual disabilities. Companies have become more aware of accessibility in recent years, which leads one to wonder whether the predominance of inaccessible websites continues or if companies have actually begun to seriously address website accessibility. Researchers have called for accessibility reviews over time. Following these suggestions, in addition to the initial study sample this study adds three additional data sets for a total of four: 2000, 2002, 2004, and 2005. The F100 Web sites were chosen for the usual reasons this population is studied, but also because we expect the largest and most profitable companies to be the most likely to have the resources and personnel to ensure website accessibility. The unit of analysis was the top-level home page for each Web site. This is an optimistic approach as companies may put their best foot forward here and then fail to consider accessibility for lower level pages. Deeper level analyses are possible but beyond the scope of this study. All samples were analyzed with the Watchfire® Bobby™ 5.0 Web accessibility validation tool. The Web Accessibility Initiative (WAI) guidelines were used to review all error types: Priority 1 (developers must satisfy), Priority 2 (developers should satisfy), and Priority 3 (developers may satisfy) (Table 1). WAI guidelines are a good starting point from which to evaluate Web accessibility and they are quickly becoming the global standard. WAI guidelines focus predominantly on problems encountered by blind users and therefore following only these guidelines will not always ensure accessibility for all disabled people. Given the breadth of existing disabilities it is difficult to evaluate all possible limitations to website accessibility. For this reason and to remain consistent with the previous study, this research concentrates on the WAI guidelines and thus issues that primarily impact visually impaired users. Automatically identified errors are important, yet many Web site accessibility problems must be checked manually by users or developers. For example, as the first criteria under Priority 1 suggests in Table 1, an accessible website should "provide alternative text for all images." Automated checkers cannot determine if the alternative text provided is meaningful to the user. Similarly, blind users often encounter problems with unlabeled "forms" that would allow them to enter information that could be searched, such as an author or title to a book. Further, 'user checks' may be more important than automatically validated errors because they can hide more subtle, yet potentially more problematic issues. We ran an automated user check analysis to determine the types and numbers of user checks for all sample years (Table 2).