2005 Accessibility Diagnosis on the Government Web Sites in Taiwan, R.O.C.
Yui-Liang Chen, Yen-Yu Chen, Monica Shao · 2006 · Proceedings of the 2006 International Cross-Disciplinary Workshop on Web Accessibility (W4A) · doi:10.1145/1133219.1133243
Summary
This study assessed the accessibility of 117 Taiwanese government websites in 2005 using the Freego Stand-Alone Validation Tool, developed by Taiwan's Research, Development and Evaluation Commission (RDEC) under the Executive Yuan. Taiwan had established a four-tier conformance system (A, A+, double-A, triple-A) modeled on WCAG 1.0's three priority levels, and the RDEC had been actively pushing government agencies toward compliance since 2002. The researchers applied 24 machine-checkable accessibility checkpoints across three priority levels to all 117 sites, comparing results against a 2004 baseline of 74 sites. The study also compared the Freego validation tool against Bobby (the widely-used Western accessibility checker from Watchfire Corporation) by running both tools against the same web page to identify discrepancies in their checkpoint implementations. Taiwan's Web Accessibility Regulations adapted WCAG 1.0 into 90 checkpoints across 14 guidelines, but the mapping was not identical — checkpoint numbering, priority level assignments, and scope all differed from the W3C original, reflecting local regulatory interpretation.
Key findings
Of the 117 sites tested, 35 (29.9%) passed Priority 1 Level Validation, of which 28 also passed the Priority 2 checkpoint for mouse-independent event handlers, reaching Conformance Level A+. Only 13 sites (11.1%) achieved double-A and just 4 (3.4%) reached triple-A. Compared to 2004, the average number of Priority 1 checkpoint errors dropped from 3.45 to 1.98 per site, showing meaningful improvement. The most persistent errors across both years were: missing table summaries (checkpoint 5.5, failing 92.3% of sites), using spaces to separate adjacent links (10.6, 89.7%), missing language identification (4.3, 88.9%), absolute rather than relative sizing (3.5, 86.3%), missing DOCTYPE declarations (3.3, 76.9%), missing image alt text (1.1, 67.5%), and mouse-dependent event handlers (9.3, 63.3%). The Freego-versus-Bobby comparison revealed that of 58 shared checkpoints, 49 (85%) produced consistent results, but 6 checkpoints needed revision in Freego and 5 had different priority level assignments. Critically, Bobby's checkpoint 13.1 (meaningful link phrases out of context) had no equivalent in Freego's 90 checkpoints at all, representing a gap in Taiwan's testing coverage.
Relevance
This paper provides a valuable case study of how a non-Western government adapted international accessibility standards into national policy and the practical challenges that emerged. The discrepancies between Freego and Bobby highlight a problem that persists today: different accessibility testing tools interpret the same guidelines differently, producing inconsistent results that can give organizations false confidence in their compliance. The finding that the most common errors were nearly identical in 2004 and 2005 — despite an active government push — illustrates how certain accessibility failures (table summaries, language declarations, relative sizing) are structurally embedded in web development practices and resist policy pressure alone. For practitioners working in international accessibility, the paper demonstrates that localizing WCAG into national regulations inevitably introduces interpretation differences, and that validation tools built to national rather than international specifications can miss important checks. Taiwan's rapid rise in global e-government accessibility rankings during this period offers a model for how sustained government commitment can drive measurable improvement even when individual site quality remains uneven.
Tags: government accessibility · automated testing · accessibility evaluation · WCAG compliance · accessibility policy · East Asian accessibility · conformance levels
Standards referenced: WCAG 1.0 · Section 508