Direction, Magnitude and Implications of Non-response Bias in Mail Surveys

Henry C. K. Chen,
The University of West Florida

INTRODUCTION

Mail questionnaire surveys have been used widely to collect research data. Walker Research Inc. (1984, 1988) estimated that about 10% of the 90 million telephone-owning households in the US have participated in mail surveys. The validity of a mail survey depends on the response rate and the representativeness of the sample to its parent population. Many well-designed mail surveys achieved a return greater than 50% (Danbury 1977), yet Yu & Cooper (1983) found that the average response rate of mail surveys was only 47.3%. An earlier study recorded non-response rate as high as 70-90% (Moser & Kalton 1971). A low response rate in a mail survey may produce a sample mean with substantial non-response error. Thus, the validity of sample estimates based solely on the sample mean obtained from the portion responding is suspicious.