Response bias is a general term for a wide range of tendencies for participants to respond inaccurately or falsely to questions. These biases are prevalent in research involving participant self-report, such as structured interviews or surveys. Response biases can have a large impact on the validity of questionnaires or surveys. Response bias can be induced or caused by numerous factors, all relating to the idea that human subjects do not respond passively to stimuli, but rather actively integrate multiple sources of information to generate a response in a given situation. Because of this, almost any aspect of an experimental condition may potentially bias a respondent. Examples include the phrasing of questions in surveys, the demeanor of the researcher, the way the experiment is conducted, or the desires of the participant to be a good experimental subject and to provide socially desirable responses may affect the response in some way. All of these "artifacts" of survey and self-report research may have the potential to damage the validity of a measure or study. Compounding this issue is that surveys affected by response bias still often have high reliability, which can lure researchers into a false sense of security about the conclusions they draw. Because of response bias, it is possible that some study results are due to a systematic response bias rather than the hypothesized effect, which can have a profound effect on psychological and other types of research using questionnaires or surveys. It is therefore important for researchers to be aware of response bias and the effect it can have on their research so that they can attempt to prevent it from negatively impacting their findings.
History of research Awareness of response bias has been present in psychology and sociology literature for some time because self-reporting features significantly in those fields of research. However, researchers were initially unwilling to admit the degree to which they impact, and potentially invalidate research utilizing these types of measures. Some researchers believed that the biases present in a group of subjects cancel out when the group is large enough. This would mean that the impact of response bias is random noise, which washes out if enough participants are included in the study. However, at the time this argument was proposed, effective methodological tools that could test it were not available. Once newer methodologies were developed, researchers began to investigate the impact of response bias. From this renewed research, two opposing sides arose. The first group supports Hyman's belief that although response bias exists, it often has minimal effect on participant response, and no large steps need to be taken to mitigate it. These researchers hold that although there is significant literature identifying response bias as influencing the responses of study participants, these studies do not in fact provide empirical evidence that this is the case. They subscribe to the idea that the effects of this bias wash out with large enough samples, and that it is not a systematic problem in mental health research. These studies also call into question earlier research that investigated response bias based on their research methodologies. For example, they mention that many of the studies had very small sample sizes, or that in studies looking at social desirability, a subtype of response bias, the researchers had no way to quantify the desirability of the statements used in the study. Additionally, some have argued that what researchers may believe to be artifacts of response bias, such as differences in responding between men and women, may in fact be actual differences between the two groups. Several other studies also found evidence that response bias is not as big of a problem as it may seem. The first found that when comparing the responses of participants, with and without controls for response bias, their answers to the surveys were not different. Two other studies found that although the bias may be present, the effects are extremely small, having little to no impact on dramatically changing or altering the responses of participants. The second group argues against Hyman's point, saying that response bias has a significant effect, and that researchers need to take steps to reduce response bias in order to conduct sound research. They argue that the impact of response bias is a systematic error inherent to this type of research and that it needs to be addressed in order for studies to be able to produce accurate results. In psychology, many studies are exploring the impact of response bias in many different settings and with many different variables. For example, some studies have found effects of response bias in the reporting of depression in elderly patients. Other researchers have found that there are serious issues when responses to a given survey or questionnaire have responses that may seem desirable or undesirable to report, and that a person's responses to certain questions can be biased by their culture. Additionally, there is support for the idea that simply being part of an experiment can have dramatic effects on how participants act, thus biasing anything that they may do in a research or experimental setting when it comes to self-reporting. One of the most influential studies was one which found that social desirability bias, a type of response bias, can account for as much as 10–70% of the variance in participant response. Essentially, because of several findings that illustrate the dramatic effects response bias has on the outcomes of self-report research, this side supports the idea that steps need to be taken to mitigate the effects of response bias to maintain the accuracy of research. While both sides have support in the literature, there appears to be greater empirical support for the significance of response bias. To add strength to the claims of those who argue the importance of response bias, many of the studies that reject the significance of response bias report multiple methodological issues in their studies. For example, they have extremely small samples that are not representative of the population as a whole, they only considered a small subset of potential variables that could be affected by response bias, and their measurements were conducted over the phone with poorly worded statements.
Types
Acquiescence bias
… excerpt ends here. Continue reading the full article.
