Address-based surveys with web components have been seen as replacements for telephone surveys due to potential advantages in cost, coverage, and non-response. Web surveys are known to have higher response propensities among certain groups, despite coverage limitations for households lacking internet access or proficiency. As a result, it can be challenging to recruit general population households to a web survey from a random sample of addresses. Our paper presents results from multiple studies at the national and state levels where households were contacted via physical mail to answer web instruments. Our analysis focuses on spatial modeling within geographic information systems (GIS) to understand how the kinds of people who respond to web surveys at as recruited by either mode of contact may differ from other modes as well as the population at-large by location and socioeconomic environment. In so doing, we consider the types of households respond to a given mode, their area-level characteristics, in addition to the relationship between geography and coverage or non-response bias. Our paper continues a program of research that uses GIS to approach issues in survey methodology from a different perspective, and is useful to multi-mode ABS and web practitioners.