Research Article
Web Site Accessibility and Web Site Development in Malaysia
Department of Marketing and Management, School of Business, Curtin University of Technology, CDT 250, 98009 Miri, Sarawak, Malaysia
The number of web sites grew 1758% in 1994 and doubled in size every 53 days by 1995 (The Economist, 1995). It has increased more than six times in just one year from January 1994 to January 1995 (Levy, 1996). New web sites on the internet have been appearing at the rate of one per minute (Schwartz, 1997). A few studies found that the number of the domain name and web site has increased day by day. This can be referred to from the following studies:
• | In the subsequent six years, the web grew from 130 separate web servers to more than 7 million servers (Zakon, 1999) |
• | The average number of total bytes at a random sample of web sites grew from about 3.5 million to 6.3 million in one year. Individual web sites seem to be almost doubling in size each year (Koehler, 1999) |
• | The publicly indexable web contains an estimated 800 million pages as of February 1999, encompassing about 15 terabytes of information or about 6 terabytes of text data on about 3 million servers after removing HTML tags, comments and extra whitespace (Lawrence and Giles, 1999) |
• | The number of web pages on the Internet is 2.1 billion and pages added per day is 7.3 million (Murray and Moore, 2000) |
• | By early 1999, the number of registered domain names was 5.3 million and by February 2000, there were about 11 million sites (Tschong, 2000) |
• | The total number of Generic Top Level Domains (gTLD) domains registered worldwide in April 15, 2009 is 107,235,401 from five main primary domains (Biz, Info, Org, Net and Com) (Zooknic, 2009) |
• | Internet domain survey found that the total domain in January 2009 was increased to 625,226,456 compared to January 1993 which was just 1,313,000 (Internet Systems Consortium, 2009) |
Refer to Malaysia context, acceptance of web site by Malaysians is increasing year by year. Very few com.my domain names were registered prior to 1995 under MYNIC that was only 100 registrations. However, the new domain name registrations were increased to 64841 registrations in November 2009. The total number registration of my, com.my, net.my, org.my, gov.my, edu.my, mil.my and name. my in November 2009 is 89798 (MYNIC, 2009).
Several underlying forces coming together have caused web site explosion of utilisation. The four reasons that caused web site explosion are mentioned below (Kiani, 1998):
• | The development of graphical and user-friendly browsers based on point-and-click like Mosaic and the Netscape |
• | The development of software and hardware tools that can be used to create rich content; the emergence of open standards in development tools and at the network protocol level |
• | The growth in support services such as web business design, hosting and gateway services that help accelerate adoption |
• | The development of critical processes such as ordering, billing and payment |
The quality of web site designs depends on task, performance and development (Brajnik, 2001). It is important to understand what skills (technical, business and analytical) and knowledge are required for successful web site development to work, how such skills and knowledge are used in actual practice and how such skills and knowledge can be improved (Taylor et al., 2001). However, web-based business models are still in the nascent stage as is research into the design and utilisation of commercial web sites. Given the early stage of research in web site development for business purposes, there are no obvious criteria to evaluate the effectiveness of commercial web sites.
The first objective is mainly to identify the major problems of web site content in Malaysia. By using bobby web accessibility tool, the study is focusing on two aspects of web site:
• | To identify number and percentages of failed and approved web sites |
• | To identify number of errors of Priority 1, 2 and 3 for web sites of each industry |
The second objective is to study perception of marketers on web site development in Malaysia based on four factors (planning and preparation; development and design; management and maintenance and security, privacy and trust).
The Web Content Accessibility Guidelines (WCAG) are the most important single resource for web developers seeking to make their sites accessible (Slatin, 2001). Accessibility evaluation studies carried out during the last decade or so significantly widened the literature-base on accessibility. Researchers from different fields carried out small and large scale studies to find out the level of web accessibility of web sites of their choice (Wijayaratne and Singh, 2010).
Sullivan and Matson (2000) found that if content accessibility is defined in a continuous, rather than dichotomous manner, 29 of 50 of the webs most popular sites can still be classified as inaccessible. Most of the studies as shown in Table 1 have found that the accessibility is very less, even less than 50%. Most of the studies are related to education and library (Schmetzke, 1999, 2001a, b; Rowland and Smith, 1999; Flowers et al., 1999, 2001; Lilly and Van Fleet, 2000; Guitierrez and Long, 2001; Zaphiris and Ellis, 2001; Spindler, 2002; Kelly, 2002; Anonymous, 2003; Opitz et al., 2003; Alexander, 2004; Zaparyniuk and Montgomerie, 2005; Anonymous, 2006; Wijayaratne, 2008; Green and Huprich, 2009; Klein et al., 2003), most popular web sites (Sullivan and Matson, 2000), Fortune 100/500 (Romano, 2002a, b; Loiacono, 2004), health and aging (Zaphiris and Kurniawan, 2001), government (Potter, 2002; Hackett and Parmanto, 2005; Hong et al., 2008; Kuzma, 2010), geographical-based (Zaphiris and Zacharia, 2001), difference mixed industries (Jackson-Sanborn et al., 2002; Lewis et al., 2007; Freedman, 2007a, b) and non-profits web sites (Loiacono and McCoy, 2004). In Hackett et al. (2004) study, they have found that the accessibility problems seem to have worsened in the last few years.
The issue at the heart of web accessibility is that many web sites are not designed with equal access in mind. In other words, lack of web accessibility is more a result of faulty design rather than inadequate technologies. The most common accessibility barriers mainly related to images without alternative text, misleading use of structural elements on a web page, uncaptioned audio or undescribed video, tables that are difficult to decipher when linearised, sites with poor colour contrast, poorly written code underlying the web design, poor navigational design, missing headings or titles and alternative text for images.
Brown et al. (2010) suggested that to achieve the accessibility, web site need to be equitable use (does it disadvantage any of our user groups), flexibility (the design accommodates a wide range of individual preferences and abilities), simple and intuitive use (design is easy to understand, regardless of users experience, knowledge, language skills or concentration levels), perceptible information (design communicates necessary information effectively to the user, regardless of ambient conditions or users) sensory abilities, tolerance for error (the design minimises hazards and adverse consequences of accidental or unintended actions) and low physical effort (the design can be used effectively and comfortably and with a minimum of fatigue).
A review of previous studies (Anonymous, 2006; Wijayaratne, 2008; Loiacono, 2004) have found that most of the problems of web sites related to text, frame, image, type of button in form, broken link, relative size and position and site often used JavaScript, which sometimes fails to work with certain assistive technology (Table 2).
Smith et al. (2004) highlighted three key issues underpinning usable web site development in a global context are considered in detail, which are requirements for design, tools for design (interpreting requirements) and strategies for evaluation. Lu and Yeung (1998) stated that the most critical factor for web site usefulness is functionality.
Web site design starts with the planning and preparation that refer to the corporate objectives as the guideline (Clyde, 2000). In the web site development, two important characteristics need to be focussed, content and design.
Table 1: | Previous study on bobby test |
Both characteristics were measured by means of features (objectively) and perceptions (subjectively) (Huizingh, 2000).
Table 2: | Previous study on checklist errors |
Table 3: | Development model, tasks and usability methods for web sites |
Cunliffe (2000) |
Web site quality has a direct and positive impact on customer satisfaction and that customer satisfaction has a direct and positive impact on purchase intentions (Bai et al., 2008).
The management and maintenance process that needs to be focussed are ensuring that new pages meet the quality and usability requirements, indexing and full maintenance (Bevan, 1999).
The study of web site development models has found that web developer has focusing on planning and preparation; development and design; management and maintenance and security, privacy and trust. A study by Cunliffe (2000) found that an informal web site development model covers establish the need before the web site is adopted as a solution; gather information before any web site development takes place; develop and evaluate before creating the complete site; implementation should be done once all design decisions have been finalised and maintain (Table 3).
The analysis will be based on two parts: All web sites were analysed with web site evaluation tool, bobby and all data from qualitative study were analysed with SPSS and AMOS.
Data collection: Three hundred and seventy two web sites have been selected and tested for accessibility with bobby. The selections of web sites are based on search engine (Google, Yahoo and Cari.com.my), yellow pages and online directories. Only the active web sites are selected. Three months were needed to observe all these web sites. The sampling selection is based on eight industries, which are Internet and ICT; tourism and hospitality; manufacturing; retailing; construction and real estate; printing and publishing; banking and finance and education. Fifty web sites from each industry were selected, excepted banking and finance industry with 22 web sites, due to the small size of industry in Malaysia. The selection of these industries was based on two reasons, performance and income, as well as active use of web site for business purposes.
For the second part of the study, questionnaire is use to study perception of marketers on web site development in Malaysia. One hundred marketers were randomly selected and involved in this study. Questionnaires have been distributed to marketers who are involved in eight industries.
Data analysis
CAST bobby test: CAST bobby bases its accessibility analysis on the World Wide Web Consortiums (W3C) Web Accessibility Initiative (WAI) Web Content Accessibility Guidelines 1.0 (WCAG). WCAG contains 14 as shown in Table broadly phrased guidelines that are translated into 91 specific checkpoints that explain how the guidelines should be applied to specific content development scenarios.
Table 4: | Web content accessibility guidelines 1.0 |
These checkpoints are organised into three levels of Priority: Priority 1 contains 29 checkpoints must be satisfied; Priority 2 contains 40 checkpoints that should be satisfied and Priority 3 contains 22 checkpoints that may be satisfied (Table 4).
Bobby divides the accessibility errors into three sections to be tested:
• | Priority 1: Errors are problems that seriously affect the pages usability by people with disabilities, in accordance with Priority 1 of WCAG. A bobby approved rating can only be granted to a site with no Priority 1 errors. Bobby approved status is equivalent to Conformance Level A for the WCAG |
• | Priority 2: Errors are secondary access problems. If all items in this section including relevant user checks passed the test, it meets conformance level AA for the WCAG |
• | Priority 3: Errors are third-tier access problems. If all items in this section including relevant User Checks passed the test, it meets Conformance Level AAA for the WCAG |
SPSS and AMOS: For the second part of study, all data were analysed with SPSS and AMOS. SPSS is used for statistical analysis, mainly mean score and correlation of coefficient. AMOS is one of the Structural Equation Modelling (SEM) software. In this study, AMOS enables to build models that more realistically reflect complex relationships with the ability to use observed variables. At the same time, model fitness will be tested.
To determine the number of the sampling units, literature suggests ten to fifteen of participants in the case of homogeneous group. The experience indicates that few new ideas are generated within a homogeneous group once the size exceeds 30 well-chosen participants . The size of the samples is also determined according to literature by the size of samples would be 10 to 20 samples. Malhotra suggested that the minimum sampling size for problem solving research is 200 samples. Based on these literatures, the size of the sampling unit is between 10 and 200. Thus, the total samples for this research will be 100 samples.
Bobby test: The total of web sites that passing the bobby test is 65, which equals to 17.47% of total sample of 372 web sites. From the total evaluated web sites, 307 web sites (82.53%) failed outright, which means that they pose serious accessibility problems. The highest number of failed web sites is from tourism and hospitality industry.
Table 5: | Numbers and percentages of failed and approved web sites |
This follows by printing and publishing, Internet and ICT, education, manufacturing, construction and real estate, retailing and banking and finance. Banking and finance industry web sites had the highest percentage passing the bobby test, which covers 41%. From the total evaluated web sites (22 web sites), 9 web sites passed the bobby approved test. Looking at the categories of web sites, retailing industry web sites had the best bobby approval rate with 12 web sites or 24% passing and tourism and hospitality industry web sites were the worst with 5 web sites or 10% passing. Beside tourism and hospitality industry web sites that had worse passing rate of 10%, the other industries that had passing rate less than 20% were the Internet and ICT (6 web sites), manufacturing (9 web sites), printing and publishing (6 web sites) and education (8 web sites). On the other hand, beside banking and finance and retailing industry web sites passed bobby approval rate of above 20%, other industry passing that percentage was construction and real estate industry with 20% or 10 web sites passing the bobby approval (Table 5).
The analysis with bobby test will be focused on accessibility errors and user checks. User checks 1are triggered by any specific feature on certain page and user checks 2 are not triggered by any specific feature on certain page, but are still important for accessibility and are required for bobby approved status. The total of Priority 1, 2 and 3 errors from 372 samples of web sites consisted of 76360 errors (Table 2). Priority 2 had the highest percentage facing the errors; consist of 40309 errors (52.8%). This is followed by Priority 1 with 26953 errors (35.3%) and Priority 3 with 9098 errors (11.9%) from the total errors (Table 6).
Checkpoint errors: Violations of just ten checkpoints accounted for as many as 59188 (77.5%) of the total number of errors (76360 errors) (Table 7). First three major problems are related to Checkpoint 2.2. Check that the foreground and background colours contrast sufficiently with each other (11228 errors or 14.7%), Checkpoint 2.1. If you use colour to convey information, make sure the information is also represented another way (11227 errors or 14.7%) and Checkpoint 3.4.
Table 6: | Total of priority 1, 2 and 3 error |
Three levels of conformance: Conformance Level A: All Priority 1 checkpoints are satisfied; Conformance level AA: All Priority 1 and 2 checkpoints are satisfied; Conformance Level AAA: All Priority 1, 2 and 3 checkpoints are satisfied |
Table 7: | Checkpoint most reported errors |
Table 8: | Mean score for web site development factors |
Use relative sizing and positioning (% values) rather than absolute (pixels) (11055 errors or 14%). This follows by other seven major errors, Checkpoint 11.2. Avoid use of obsolete language features if possible, Checkpoint 1.1. If an image conveys important information beyond what is in its alternative text, provide an extended description, Checkpoint 7.3. Do not create a blinking effect with animated gif images, Checkpoint 1.1. Provide alternative text for all images, Checkpoint 5.5. Provide a summary for tables, Checkpoint 5.2. If a table has two or more rows or columns that serve as headers, use structural mark-up to identify their hierarchy and relationship and Checkpoint 5.5. If this is a data table (not used for layout only), provide a caption. Thus, major problems of web sites are related to text, colour, animation, image, pixel and table.
Web site development: One hundred questionnaires were returned from marketers. 96% of the questionnaires are from retailing (43), manufacturing (29) and ICT (24). The rest of the questionnaires are from printing and publishing (2), construction and real estate (1) and banking and finance (1). No questionnaire returned from education and tourism and hospitality.
All factors of web site development are important. From the feedback of respondents proved that the mean score for all four factors is above 3.00. The highest mean score is security, privacy and trust (4.4900). This followed by development and design (4.1340); planning and preparation (3.9067) and management and maintenance (3.6133), as shown in Table 8.
Using pearson correlation coefficients, the study shows that all factors are significant at the 0.01 and it is between moderate positive correlation (0.5) and strong positive correlation (0.75) (Table 9).
Table 9: | Correlations-web site development factors |
**Correlation is significant at the 0.01 level (2-tailed) |
Table 10: | Correlations-all variables |
Note: PC = Pearson correlation; **Correlation is significant at the 0.01 level (2-tailed). *Correlation is significant at the 0.05 level (2-tailed) |
Refer to variables for each factors (Table 10), all correlations, except variable 3.3 and 2.2 are significant either at the 0.01 or 0.05. Majority of these variables are between moderate positive correlation and strong positive correlation (46.2%), as well as weak positive correlation and moderate positive correlation (44%). 5.5% are between no correlation and weak positive correlation. It shows that only correlation between security on online activities (4.1) with privacy on online activities (4.2) and trust on online activities (4.3) and privacy on online activities (4.2) and trust on online activities (4.3) are in the category of strong positive correlation and perfect positive correlation.
The Chi-square value (CMIN) is 360.746, which is highly significant (p≤0.000). However, that this does not mean the model is good. In fact it is the opposite, from the point of view of statistical significance. We may say that the model is badness-of-fit.
Goodness-of-fit are based on fitting the model to sample moments, which means to compare the observed covariance matrix to the one estimated on the assumption that the model being tested is true. These measures thus use the conventional discrepancy function. The Chi-square value should not be significant if there is a good model fit, while a significant Chi-square indicates lack of satisfactory model fit. That is, Chi-square is a badness of fit measure in that a finding of significance means the given models covariance structure is significantly different from the observed covariance matrix. If model Chi-square <0.05, the model is rejected. Hoelters critical N is the size the sample size must reach for the researcher to accept the model by Chi-square, at the 0.05 or 0.01 levels. This throws light on the Chi-square fit indexs sample size problem. Hoelters N should be greater than 200 (Garson, 2006). Carmines and McIver (1981) state that relative Chi-square should be in the 2:1 or 3:1 range for an acceptable model. Kline says 3 or less is acceptable.
Fig. 1: | Path model |
Some researchers allow values as high as 5 to consider a model adequate fit, while others insist relative Chi-square be 2 or less. Hoelter, at the 0.05 or 0.01 levels is 26 or 10, which is less than 200 as suggested by Garson (2006) and relative chi square (CMIN/df) less than 5, which is 4.942.
For GFI and AGFI, coefficients closer to unity indicate a good fit, with acceptable levels of fit being above 0.90 (Marsh et al., 1988). AGFI can yield meaningless negative values. AGFI >1.0 is associated with just-identified models and models with almost perfect fit. AGFI <0 is associated with models with extremely poor fit. The closer the RMR to 0 for a model being tested, the better the model fit (Garson, 2006). Garson (2006) also agree that for CFI, IFI and TLI, coefficients closer to unity indicate a good fit, with acceptable levels of fit being above 0.90. NFI, TLI, CFI and RFI are varies from 0 to 1. NFI, TLI, CFI and RFI close to 1 indicate a very good fit. The fit indices of GFI and AGFI were 0.666 and 0.519, respectively, suggesting that this model not provides a good fit. NFI (0.708), RFI (0.638), IFI (0.752), TLI (0.686) and CFI (0.748) were less than 0.90, which can be considered as not a good-fit model.
Parsimony measures are used in goodness-of-fit measures. The higher parsimony measure represents the better fit. For RMR and RMSEA, evidence of good fit is considered to be values less than 0.05; values from 0.05 to 0.10 are indicative of moderate fit and values greater than 0.10 are taken to be evidence of a poorly fitting model. The closer model is to the saturated model, the more PNFI and PCFI is penalised. There is no commonly agreed-upon cut-off value for an acceptable model (Garson, 2006). The value of RMR is 0.061, which is between 0.05 and 0.10. It can be categorised under moderate fit. For RMSEA, the value is 0.200, which is greater than 0.10. Thus for this part, it is can be considered as a poorly fitting.
PNFI and PCFI were 0.568 and 0.600. The closer model is to the saturated model (<0.001), the more PNFI and PCFI is penalised (Garson, 2006). The result shows that GFI (0.666), RMSEA (0.200), AGFI (0.519) and NFI (0.708), which does not achieve the recommended values.
Finally, the study of regression weights found that all factors (planning and preparation; development and design; management and maintenance and security, privacy and trust) are significant at the 0.05 significance level. This study has proved that all four factors have a statistically impact on web site development (Fig. 1, Table 11 and 12).
Even the analysis found that each of the factors have a statically impact on web site development, but, an overall of the model, which is combination of all 4 factors have shows that it is not a good-fit model. This study has agreed that web-based business models are still in the nascent stage as is research into the design and utilisation of commercial web sites as stated. Each marketer has their own point of view on web site development.
Table 11: | Regression weights |
Table 12: | Testing of model fit |
Thus, there are no obvious criteria to evaluate the effectiveness of commercial web sites.
Out of 372 of the total evaluated web sites with bobby, 307 web sites (82.53%) failed outright. The total of Priority 1, 2 and 3 errors from 372 samples of web sites consisted of 76360 errors. The highest number of failed web sites is tourism and hospitality industry. This follows by printing and publishing, Internet and ICT, education, manufacturing, construction and real estate, retailing and banking and finance. Top errors are check that the foreground and background colours contrast sufficiently with each other, if you use colour to convey information, make sure the information is also represented another way, use relative sizing and positioning (% values) rather than absolute (pixels), avoid use of obsolete language features if possible, if an image conveys important information beyond what is in its alternative text, provide an extended description, do not create a blinking effect with animated gif images, provide alternative text for all images, provide a summary for tables, if a table has two or more rows or columns that serve as headers, use structural mark-up to identify their hierarchy and relationship and if this is a data table (not used for layout only), provide a caption.
The study shows that all four factors of web site development are important and significant at the 0.01. All factors are significant between moderate positive correlation (0.5) and strong positive correlation (0.75). However, the study of model fitness based on four factors shows that it is not a good fit model. This can be referred to GFI (0.666), RMSEA (0.200), AGFI (0.519) and NFI (0.708), which is less than recommended values (0.900).
This research hopefully will help Internet marketer and web developer to identify the major problems of the web site and development processes at the first stage. Further research should be carried out in the future to solve and to focus more on text, colour, animation, image, pixel and table aspects.