Response bias

From Wikipedia, the free encyclopedia
Jump to: navigation, search
Example Likert Scale.
A survey using a Likert style response set. This is one example of a type of survey that can be highly vulnerable to the effects of response bias.

Response bias is a general term for a wide range of cognitive biases that influence the responses of participants away from an accurate or truthful response. These biases are most prevalent in the types of studies and research that involve participant self-report, such as structured interviews or surveys.[1] Response biases can have a large impact on the validity of the questionnaire or survey to which the participant is responding.[1][2] This bias can be induced or caused by a number of 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.[3] Because of this, almost any aspect of an experimental condition may be able to bias a respondent in some form or another. For example, 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 bias the response of the participant in some way.[1][2][3][4] All of these "artifacts" of survey and self-report research may have the potential to damage the validity of a measure or study.[2] Compounding this issue is that surveys affected by response bias still often have high reliability.[5] This insidious combination can lure researchers into a false sense of security about the conclusions they draw.[5] The effect of this bias means that 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.[5] It is therefore important for researchers to be aware of response bias and the effect can have on their research so that they can attempt to prevent it from impacting their findings in a negative manner.

History of research[edit]

Awareness of response bias has been present in psychology and sociology literature for some time because self-report usage features significantly in those fields of research. However, while these types of biases have been known about, researchers were initially unwilling to admit the degree to which they impact, and potentially invalidate research utilizing these types of measures, questionnaires surveys.[5] This is because some researchers believe that the biases present across a group of subjects cancel out when the group of subjects is significantly large.[6] Essentially, this argument states that the impact of response bias in research is random noise and that the biases present in experimental participants would wash out if enough participants included in the study.[5] However, at the time this argument was proposed, effective methodological tools that could test this hypothesis were not available.[5] Once more modern methodologies were developed, researchers began to investigate the impact of response bias.[5] From this renewed research, two opposing sides arose.

Arguments against response bias[edit]

The first argues and supports the hope belief espoused by Hyman, believing that no large steps need to be taken to mitigate response bias, since it does not contribute a significant effect[5][7][8] These researchers hold that although there is significant literature identifying that response bias may play a major role in influencing the responses of study participants, these studies do not in fact provide empirical evidence that this is the case.[5] They ascribe 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.[5][7] These studies also call into question earlier research that investigated response bias on basis of 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.[5] 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 just be actual differences between the two groups.[7] Several other studies also found evidence that response bias is not as big of a problem as it seems. 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.[7] Two other studies found that although the bias may be present, the effects are extremely small, having little to no impact towards dramatically changing or altering the responses of participants.[8][9] In sum, this side argues that although response bias exists, it often has minimal effect on participant response.

Arguments for response bias[edit]

The second group argues against Hyman's point, saying that response bias has a significant effect, and that researchers need to take steps to mitigate and reduce response bias in order to conduct sound research.[1][2] 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. There are many studies that represent this argument in psychological literature, exploring the impact of response bias in multiple different settings and with many different variables. For example, some studies have explored and found effects of response bias in the reporting of depression in elderly patients.[10] 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 some responses to certain questions can be biased by the Culture of the participant in the study.[2][11] 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-report.[3] One of the most influential findings for those arguing that response bias is a significant factor was a study conducted by Nederhof, where he found that social desirability bias, a subset of response bias, can account for as much as 10-70% of the variance in participant response.[2] 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.

Conclusion[edit]

While both sides have support in the literature, there appears to be greater empirical support for the dangers of response bias.[1][2][3][11][12][13] 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.[5][7]

Types of response bias[edit]

Acquiescence bias[edit]

Main article: Acquiescence bias

Acquiescence bias, which can also be referred to as yea-saying, is a category of response bias in which respondents to a survey have a tendency to agree with all the questions in a measure.[14][15] This bias in responding may represent a form of dishonest reporting because the participant automatically endorses any statements, even if it results in contradictory responses.[16][17] For example, a participant could be asked whether they endorse the following statement, "I like to spend time with others" but then later on in the survey also endorses "I like to spend time alone," which are contradictory statements. This is a distinct problem for self-report research because it does not allow a researcher to understand or gather accurate data from any type of question that asks for a participant to endorse or reject statements.[16] Researchers have approached this issue by thinking about the bias in two different ways. The first deals with the idea that participants are trying to be agreeable, in order to avoid the disapproval of the researcher.[16] A second cause for this type of bias was proposed by Cronbach, when he argued that it is likely due to a problem in the cognitive processes of the participant, instead of the motivation to please the researcher.[13] He argues that it may be due to biases in memory where an individual recalls information that supports endorsement of the statement, and ignores contradicting information.[13]

Researchers have several methods to try and reduce this form of bias. Primarily, they need to make balanced response sets in a given measure, meaning that there are a balanced number of positively and negatively worded questions.[16][18] This means that if a researcher was hoping to examine a certain trait with a given questionnaire, half of the questions would have a "yes" response to identify the trait, and the other half would have a "no" response to identify the trait.[18]

Nay-saying is the opposite form of this bias. It occurs when a participant always chooses to deny or not endorse any statements in a survey or measure. This has a similar effect of invalidating any kinds of endorsements that participants may make over the course of the experiment.

Demand characteristics[edit]

Demand characteristics refer a type of response bias where participants alter their response or behavior simply because they are part of an experiment.[3] This arises because participants are actively engaged in the experiment, and may try and figure out the purpose, or adopt certain behaviors they believe belong in an experimental setting. Martin Orne was one of the first to identify this type of bias, and has developed several theories hoping to address their cause.[19] His research points to the idea that participants enter a certain type of social interaction when engaging in an experiment, and this special social interaction drives participants to consciously and unconsciously alter their behavior[3] There are several ways that this bias can influence participants and their responses in an experimental setting. One of the most common relates to the motivations of the participant. Many people choose to volunteer to be in studies because they believe that experiments are important. This drives participants to be "good subjects" and fulfill their role in the experiment properly, because they believe that their proper participation is vital to the success of the study.[3][20] Thus, in an attempt to be good, the participant may attempt to gain knowledge of the hypothesis being tested in the experiment and alter their behavior in an attempt to support that hypothesis. Orne conceptualized this change by saying that the experiment may appear to a participant as a problem, and it is his or her job to find the solution to that problem, which would be behaving in a way that would lend support to the experimenter's hypothesis.[3] Alternatively, the participant may be volunteering for the experiment as part of a requirement for a college course, or have some other reason. In this case, the participants may try to discover the hypothesis simply to provide faulty information and wreck the hypothesis.[20] Both of these results are harmful because they prevent the experimenters from gathering accurate data and making sound conclusions.

Outside of participant motivation, there are other factors that influence the appearance of demand characteristics in a study. Many of these factors relate to the unique nature of the experimental setting itself. For example, participants in studies are more likely to put up with uncomfortable or tedious tasks simply because they are in an experiment.[3] Additionally, the mannerisms of the experimenter, such as the way they greet the participant, or the way they interact with the participant during the course of the experiment may inadvertently bias how the participant responds during the course of the experiment.[3][21] Finally, prior experiences of being in an experiment, or rumors of the experiment that participants may hear can greatly bias the way they respond.[3][20][21] Outside of an experiment, these types of past experiences and mannerisms may have significant effects on how patients rank the effectiveness of their therapist.[12] Many of the ways therapists go about collecting client feedback involve self-report measures, which can be highly influenced by response bias.[12] Participants may be biased if they fill out these measure in front of their therapist, or somehow feel compelled to answer in an affirmative matter because they believe their therapy should be working.[12] In this case, the therapists would not be able to gain accurate feedback from their clients, and be unable to improve their therapy or accurately tailor further treatment to what the participants need.[12] All of these different examples, may have significant effects on the responses of participants, driving them to respond in ways that do not reflect their actual beliefs or actual mindset, which negatively impact conclusions drawn from those surveys.[3]

While demand characteristics cannot be completely removed from an experiment, there are steps that researchers can take to minimize the impact they may have on the results.[3] One way to mitigate response bias is to use deception to prevent the participant from discovering the true hypothesis of the experiment.[22] Researchers use deception so that it is more difficult for participants to discover the true hypothesis of the experiment, and then debrief the participants.[22] For example, research has demonstrated that repeated deception and debriefing is useful in preventing participants from becoming familiar with the experiment, and that participants do not significantly alter their behavior after being deceived and debriefed multiple times.[22] Another way that researchers can attempt to reduce demand characteristics is by being as neutral as possible, or training those conducting the experiment to be as neutral as possible.[20] For example, studies show that that excessive one-on-one contact between the experimenter and the participant makes it more difficult to be neutral, and go on to suggest that this type of interaction should be limited when designing an experiment.[18][20] Finally, a last way to prevent demand characteristics would be to use blinded experiments with placebo or control groups.[3][18] This prevents the experimenter from biasing the participant, because the researcher does not know in which way the participant should respond. Although not perfect, these methods can significantly reduce the effect of demand characteristics on a study, thus making the conclusions draw from the experiment more likely to accurately reflect what they were intended to measure.[20]

Extreme responding[edit]

Extreme responding is a form of response bias that drives respondents to only select the most extreme options or answers available.[1][17] For example, in a survey utilizing a Likert scale with potential responses ranging from 1 to 5, the respondent may only give answers as 1's or 5's. Another example would be if the participant only answered questionnaires with "strongly agree" or "strongly disagree" in a survey with that type of response style. There are several reasons for why this bias may take hold in a group of participants. One example ties the development of this type of bias in respondents to their cultural identity.[17] This explanation states that some cultures are more likely to respond in an extreme manner as compared to others. For example, research has found that those from the Middle East and Latin America are more prone to be affected by extremity response, whereas those from East Asia and Western Europe are less likely to be affected by extremity response.[17] A second explanation for this type of response bias relates to the education level of the participants.[17] Research has indicated that those with lower intelligence, measured by an analysis of IQ and school achievement, are more likely to be affected by extremity response.[17] Finally, one other way that this bias can be introduced is through the wording of questions in the survey or questionnaire.[1] Certain topics or the wording of a question may drive participants to respond in an extreme manner, especially if it relates to the motivations or beliefs of the participant.[1]

The opposite of this bias occurs when participants only select intermediate or mild responses as answers.[1] In this type of bias, a participant would highly prefer to select a 3 in a scale ranging from 1 to 5, or the most mild response in a survey asking for their feelings on a topic, such as "Neutral," "slightly agree," or "slightly disagree."

Social desirability bias[edit]

Social desirability bias is a type of response bias that influences a participant to deny undesirable traits, and ascribe to traits that are socially desirable.[2] In essence, it is a bias that drives an individual to answer in a way that makes them look more favorable to the experimenter.[1][2] This bias can take many forms. Some individuals may over-report good behavior, while others may under-report bad, or undesirable behavior.[1] A critical aspect of how this bias can come to affect the responses of participants relates to the norms of the society in which the research is taking place.[2] For example, social desirability bias could play a large role if conducting research about an individual's tendency to use drugs. Those in a community where drug use is seen as cool or popular may over-inflate their own drug use, whereas those from a community where drug use is looked down upon may choose to under-report their own use to better conform to the societal norm. This type of bias is much more prevalent in questions that draw on a subject's opinion, like when asking a participant to evaluate or rate something, because there generally is not one correct answer, and the respondent has multiple ways they could answer the question.[4] Overall, this bias can be very problematic for self-report researchers, especially if the topic they are looking at is controversial.[1] The distortions created by respondents answering in a socially desirable manner can have profound effects on the validity of self-report research.[2] Without being able to control for or deal with this bias, researchers would be unable to determine if the effects they are measuring are due to individual differences, or from a desire to conform to the societal norms present in the population they are studying. Therefore, it is important for researchers to employ strategies aimed at mitigating social desirability bias so that they can draw valid conclusions from their research.[1]

Several strategies exist to limit the effect of social desirability bias. In 1985, Anton Nederhof compiled a list a of techniques and methodological strategies that researchers can use to mitigate the effects of social desirability bias in their studies.[2] Most of these strategies involve deceiving the subject, or are related to the way questions in surveys and questionnaires are presented to those in a study. A condensed list of seven of the strategies are listed below:

  • Forced-choice items: This technique hopes to generate questions that are equal in desirability to hopefully prevent a socially desirable response in one direction or another.[2]
  • Neutral questions: The goal of this strategy is to use questions that are rated as neutral by a wide range of raters so that socially desirable responding does not apply.[2]
  • Randomized response technique: This technique allows participants to answer a question that is randomly selected from a set a of questions. The researcher in this technique does not know which question the subject responds to, so subjects are more likely to answer truthfully. Researchers can then use statistics to interpret the anonymous data.[2]
  • Self-administered questionnaires: This strategy involves isolating the participant before they begin answering the survey or questionnaire to hopefully remove any social cues the researcher may present to the participant.[2]
  • Bogus-pipeline: This technique involves a form of deception, where researchers convince a subject through a series of rigged demonstrations that a machine can accurately determine if a participant is being truthful when responding to certain questions. After the participant completes the survey or questionnaire, they are debriefed. This is a rare technique, and does not see much use because of the cost, time commitment and because it is a one-use only technique for each participant.[2]
  • Selection interviewers: This strategy allows participants to select the person or persons who will be conducting the interview or presiding over the experiment. This method hopes that with a higher degree of rapport, subjects will be more likely to answer honestly.[2]
  • Proxy subjects: Instead of asking a person directly, this strategy questions someone who is close to or knows the target individual well. This technique is generally limited to questions about behavior, and is not adequate for asking about attitudes or beliefs.[2]

The degree of effectiveness for each of these techniques or strategies differs depending on the situation and the question asked.[2] In order to be the most successful in reducing social desirability bias in a wide range of situations, it is suggested that researchers utilize a combination of these techniques to have the best chance at mitigating the effects of social desirability bias.[1][2]

Related terminology[edit]

  • Non-response bias is not the opposite of "response bias" and is not a type of cognitive bias: it occurs in a statistical survey if those who respond to the survey differ in the outcome variable.
  • Response rate is not a cognitive bias, but rather refers to a ratio of those who complete the survey and those who don't.

Highly vulnerable areas[edit]

Some areas or topics that are highly vulnerable to the various types of response bias include:

See also[edit]

External links[edit]

Notes[edit]

  1. ^ a b c d e f g h i j k l m n Furnham, A (1986). Response bias, social desirability and dissimulation. Personality and individual differences 7. 385-400
  2. ^ a b c d e f g h i j k l m n o p q r s t u Nederhof, A (1985). Methods of coping with social desirability bias: a review. European Journal of Social Psychology 15. 263-280
  3. ^ a b c d e f g h i j k l m n Orne, M (1962). On the social psychology of the psychological experiment: with particular reference to demand characteristics and their implications. American Psychologist 17. 776-783
  4. ^ a b Kalton, G, Schuman, H, (1982). The effect of the question on survey responses: a review. J. R. statist. Soc. 154. 42-73
  5. ^ a b c d e f g h i j k l Gove, WR; Geerken, MR (1977) Response bias in surveys of mental health: and empirical investigation. American journal of Sociology 82. 1289-1317
  6. ^ Hyman, H; 1954. Interviewing in Social Research. Chicago: University of Chicago Press.
  7. ^ a b c d e Clancy, K; Gove, W (1974). Sex differences in mental illness – analysis of response bias in self-reports. American Journal of Sociology 80. 205-216
  8. ^ a b Campbell, A. Converse, P. Rodgers; 1976. The Quality of American Life: Perceptions, Evaluations and Satisfaction. New York: Russell Sage.
  9. ^ Gove, W. McCorkel, J. Fain, T. Hughes, M (1976). Response bias in community surveys of mental health: systematic bias or random noise?" Social Science and Medicine 10 497-502
  10. ^ a b Knauper, B; Wittechen, HU (1994). Diagnosing major depression in the elderly – evidence for response bias in standardized diagnostic interviews. Journal of Psychiatric Research 28. 147-164
  11. ^ a b Fischer, R (2003). Standardization to account for cross-cultural response bias – a classification of score adjustment procedures and review of research in jccp. Journal of cross-cultural psychology 35. 263-282
  12. ^ a b c d e Reese, R; Gillaspy, AJ; Owen, J (2013). The influence of demand characteristics and social desirability on client's ratings of the therapeutic alliance. Journal of Clinical Psychology 69. 696-709
  13. ^ a b c Cronbach, LJ (1942). Studies of acquiescence as a factor in the true-false test. Journal of Educational Psychology 33. 401-415.
  14. ^ Watson, Dorothy. (1992). Correcting for Acquiescent Response Bias in the Absence of a Balanced Scale An Application to Class Consciousness Sociological Methods Research August 1992 vol. 21 no. 1 52-88. doi:10.1177/0049124192021001003
  15. ^ Moss, Simon. (2008). Acquiescence bias
  16. ^ a b c d Knowles, E. S., & Nathan, K. T. (1997). Acquiescent responding in self-reports: Cognitive style or social concern. Journal of Research in Personality, 31, 293-301.
  17. ^ a b c d e f Meisenberg, G., & Williams, A. (2008). Are acquiescent and extreme response styles related to low intelligence and education? Personality and Individual Differences, 44, 1539-1550.
  18. ^ a b c d Podsakoff, P. M., MacKenzie, S. B., Lee, J., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88, 879-903.
  19. ^ Orne, Martin T. Demand Characteristics and the concept of Quasi-Controls. in Artifacts in Behavioral Research: Robert Rosenthal and Ralph L. Rosnow's Classic Books, beginning with page 110
  20. ^ a b c d e f Nichols, AL; Maner, JK (2008). The good-subject effect: investigating participant demand characteristics. Journal of General Psychology 135. 151-165
  21. ^ a b Cook, TD; Bean, JR; Calder, BJ; et al (1970). Demand characteristics and three conceptions of frequently deceived subjects. Journal of Personality and Social Psychology 14. 185
  22. ^ a b c Cook, TD; Bean, JR; Calder, BJ; et al (1970). Demand characteristics and 3 conceptions of frequently deceived subject. Journal of Personality and Social Psychology 14. 185-194
  23. ^ Babor, TF; Stephens, RS; Marlatt, GA (1987). Verbal report methods in clinical research on alcoholism – response bias and its minimization. Journal of studies on alcohol 48. 410-424
  24. ^ Embree, BG; Whitehead, PC (1993). Validity and reliability of self-reported drinking behavior – dealing with the problem of response bias. Journal of studies on Alcohol 54. 334-344