In this study, a nationally representative stratified random sample of 877 first and fourth grade reading teachers were randomly assigned to one of two modes of questionnaire delivery (paper or web) to examine the relative costs and benefits of web versus mail administration (in terms of cost and time) and to examine how the mode of administering a teacher questionnaire affects (a) coverage error, (b) nonresponse error, and (c) measurement error in the context of educational research. Web surveys have the

potential to achieve economies of scale, allowing greater sample size within a fixed budget. However, concerns about the logistics and quality of collecting data via the web have made many educational researchers wary of switching modes. Initial results from this study indicate that these concerns may be warranted. While email addresses were obtained from all but 4% of the sample and mailing addresses were obtained for the entire sample, approximately 12% of the emails obtained were undeliverable even after attempts to correct email addresses were made. After three contacts, the response rate for the web

group was only 43% while the response rate for the mail group was 59% 9 (based on AAPOR response rate 1). Individual item non-response was also higher among the web respondents. The problem was most pronounced for items that asked individuals to “mark all that apply” or that were near the end of the questionnaire. Differences in the characteristics of responders and non-responders will be explored. Initial findings also suggest that switching modes may introduce measurement error. While the costs for the

web survey were approximately 40% lower than for the mail survey, the time it took respondents to return their surveys did not differ substantially for the two groups.