This is our final report on item nonresponse analysis of the Spanish Forms Availability Test or SFAT (also known as the 1993 National Census Test III). In an earlier report we provided preliminary results from item nonresponse analysis. In that report we examined overall item nonresponse rates for demographic and housing questions by panel. The first panel is the control panel and consists of an English language census form. The second panel is the dual forms panel and contains an English language census form, identical to the one in the first panel and a Spanish language census form. The third panel is the bilingual forms panel and it contains forms identical to those in the second panel but the English and Spanish language census forms were converted into a back to back version. We then examined item nonresponse rates by stratum, and by language of census form (de la Puente and Wobus 1994). Each stratum consisted of roughly equal half of the test universe. 1 There were two strata: 1) areas with between 15 and 30 percent linguistically isolated households and 2) areas with 30 percent or more linguistically isolated households.

We concluded that in order to provide more definitive statements concerning the desirability of mailing out Spanish language census forms to areas containing high concentrations of linguistically isolated households, additional information on the item nonresponse rates of Hispanics versus non-Hispanics taking into account language of form, panel and strata is required. This final report provides this information.