www.frederikfunke.de


home

 

Enhancing Web surveys with new HTML5 input types
Frederik Funke

Paper presented at the Internet Survey Methodology Workshop
November 14–16, 2012 in Ljubljana (Slovenia)
 
Travel grant from the COST Action IS1004. www.webdatanet.eu


 
Abstract
 
A new revision of the HTML standard is forthcoming. HTML5 offers many new features relevant for Web-based survey research, some of which are already implemented in many current browsers. The current experimental study focused on two new input types that could be especially relevant for mobile Web surveys.
 
On mobile devices like smart phones or tablets, there are different types of virtual keyboards. Due to space constraints, some of these keyboards either result in small buttons or require switching between layouts. Reducing the keyboard to relevant keys only (e.g., numbers for numerical input) may have two beneficial effects. Firstly, customized keyboards lead to larger buttons that may improve usability, resulting in fewer typing errors. Additionally, the number of clicks (required to change between keyboard layouts) can be reduced. Secondly, a specific keyboard illustrates that a certain type of answer is expected, which could improve question understanding. On the flipside, restricting respondents’ input to certain characters may have a negative effect on data quality and respondent motivation. Some new input – especially those with pre-selected answers – could even lead to biased answers or to break- off.
 
In a Web experiment two new HTML5 input types have been evaluated. The first input type (element form type="time") is used to collect time information. In desktop browsers two control buttons that can be clicked to enter a certain time of day are added to a text field. In mobile browsers a scroll wheel is presented instead of an alphanumeric keyboard. The second HTML5 input type (element form type="tel") does only affect mobile devices. Instead of offering a full keyboard, a numeric keypad is presented. The impact on different aspects of data quality (e.g., validity, incorrect answers, item non-response, and break-off) considering client-side paradata will be presented and recommendations for survey design will be discussed.