Summary
Contents
Subject index
Public opinion theory and research are becoming increasingly significant in modern societies as people’s attitudes and behaviors become ever more volatile and opinion poll data becomes ever more readily available. This major new Handbook is the first to bring together into one volume the whole field of public opinion theory, research methodology, and the political and social embeddedness of polls in modern societies. It comprehensively maps out the state-of-the-art in contemporary scholarship on these topics.
Survey Non-Response
Survey Non-Response
Sometimes, the information we do not collect on surveys is as important as the information we do collect. In recent years, social scientists and professional pollsters have paid increasing attention to survey non-response. In this chapter, I review current literature on this phenomenon and point to some areas of future inquiry.
When we speak of survey non-response, we are in fact, speaking of two distinct but related phenomena: unit non-response and item non-response. Unit non-response occurs when an entire observation unit is missing from our sample (Lohr, 1999). In the context of survey research, unit non-response occurs when we have no information about a respondent selected to be in our sample. Item non-response, on the other hand, occurs when some measurements are present for an observational unit, but at least one measure of interest is missing (Lohr, 1999). In survey research, item non-response occurs when we have some information about the respondent, but we are missing data for a given variable or variables of interest. For instance, we may know a respondent's education level and partisanship, but not her income.
These two forms of survey non-response both involve missing information, but have been dealt with separately in the literature. In this chapter, I will follow convention and take up each topic in turn. However, as my review of the literature will demonstrate, both types of non-response arise from similar causes and create related concerns for survey researchers, thereby implying that similar solutions may be in order. Future work should therefore take more seriously the theoretical and empirical linksbetweenunitnon-responseanditemnon- response.
Unit Non-Response
Unit non-response has become an increasingly serious problem over the last 40 years. Studies in the 1990s demonstrated that face- to-face surveys by academic organizations, such as the National Election Study (NES) and the General Social Survey (GSS), have non-response rates between 25 and 30%, up from 15 to 20% in the 1950s (Brehm, 1993; Luevano, 1994; Groves & Couper, 1998). Telephone surveys conducted by commercial polling houses, which produce the majority of polling information in circulation in the political world, are often even higher. For instance, in a study of polls conducted by the news media and government contractors, Krosnick, Holbrook, and Pfent (2005) found that the mean response rate for these surveys was 36%.1 Furthermore, indications are that unit non-response has become an even more serious problem since 2000. Since 1994, the response rate to the NES has fallen below 70%, dropping to 56% in 2002 before rebounding to 66% in 2004.2
These non-response rates are, in fact, the product of two distinct processes: (1) some respondents cannot be found by the poll's sponsors; and (2) other respondents decline to participate in the poll. In both cases potential data is lost. Unit non-response is therefore a function of both respondent contact and respondent cooperation (Groves & Couper, 1998).
Conventional wisdom holds that survey response rates are falling because potential respondents are harder to contact. In face- to-face surveys, physical impediments, such as locked apartment buildings have made it more difficult to access households in the sample (Groves & Couper, 1998). Likewise, researchers employing telephone surveys seem to have a more difficult time reaching respondents in recent years because of the rise of caller ID, answering machines, and other technological innovations. However, academic studies have shown that the rise in unit non-response is as attributable to increasing refusal rates as it is to decreasing contact rates. Take, for instance, the NES. In 1952, 6% of potential respondents refused to be interviewed by NES. By 1986, this number had risen to 26%. Though refusal rates have fluctuated somewhat in recent years—the 2004 refusal rate for the NES was 23%—they have remained high in the modern era (Luevano, 1994). Similar patterns of responsehavebeenfoundinotherface-to-face surveys. Curtin, Presser and Singer (2005) study the response rates to the University of Michigan's Survey of Consumer Attitudes and find that the response rate dropped from 1979 to 2003, with the steepest decline occurring in the 1997 to 2003 period. While non-contacts increased over this time, that increase has slowed in recent years. The majority of the decline in response rate was due to an increase in the refusal rate by nearly 1% per year from 1997 to 2003, leading to a refusal rate of almost 30% in 2003.
...
- Loading...
Get a 30 day FREE TRIAL
-
Watch videos from a variety of sources bringing classroom topics to life -
Read modern, diverse business cases
-
Explore hundreds of books and reference titles
Read next
More like this
Sage Recommends
We found other relevant content for you on other Sage platforms.
Have you created a personal profile? Login or create a profile so that you can save clips, playlists and searches