Web Survey Bibliography

Title Reducing Error Associated with Non-Probability Sampling through Propensity Scores: Evidence from Election 2000
Author Terhanian, G., Marcus, S., Bremer, J., Smith, R.
Source JSM 2001. Statistical Science for the Information Age
Year 2001
Access date 20.04.2004
Abstract The primary aim of this paper is to explain how and why it is possible to produce trustworthy information through Internet-based research; specifically, through Internet-based research that relies upon the cooperation of respondents recruited by means other than probability sampling. The approach described here benefits from the conscientious thinking of statisticians who have developed sturdy statistical substitutes for the randomization of random assignment. To support our contention, we compare the accuracy of 54 pre-election statewide forecasts made by Harris Interactive through surveys of non-probability samples of Internet users in the 2000 election with 118 pre-election forecasts made by polling organizations through random-digit-dialing procedures. We also compare the Harris Interactive nationwide forecast with those of ten polling organizations. The evidence presented suggests that the Harris Interactive forecasts were approximately 47% more accurate than the telephone forecasts. A secondary aim is to explain how public opinion researchers might exploit propensity score to improve responses elicited through probability samples.
Access/Direct link Homepage - conference (abstract)
Year of publication2001
Bibliographic typeConferences, workshops, tutorials, presentations
Print

Web Survey Bibliography - Terhanian, G. (30)