The survey
The following study analyzes the truthfulness and consistency of the data entered in the questionnaires. In the study 496 questions are analyzed in total.
The information and data from the following study were obtained from three General Social Surveys (GSS).
A General Social Survey is an accomplished interview face to face which aims to measure the attitudinal and demographic change of a representative sample of US families.
Each of these studies were repeated three times remote 2 years, in order to verify the accuracy of the data entered. The first GSS questionnaire was performed in the 2006, again in the 2008 and finally in the 2010. The second in the 2008 / 2010 / 2012 and the third in the 2010 / 2012 / 2014.
The questions used were exactly repeated in the three identical distance tests of 2 years. The same possible answers and the same interview mode were formulated with the same words. The response variables used to reply were all continuous (and not categorical).
Examples of questions used are ones which are related to daily life, work, economics and family, political opinions, religious beliefs and opinions about suicide, crime and more.
Results
From the data collected and analyzed there is no theory to support the idea that multiple categories of response lead to greater truthfulness and correspondence to reality, as it is instead claimed in the information theory. In many cases, on the contrary, the results are more truthful using fewer response categories and generally there is a decline in levels of truthfulness with numbers of more and more response options.
When there is an option that appears as an "average" among others, respondents often choose this medium. For example, if a selected 3 response format is indicated in the answers:
1) LOW 2) MEDIUM 3) HIGH or 1) POSITIVE 2) NORMAL 3) NEGATIVE, the majority of respondents choose the answer number 2 which represents a sort of "average" among the others, even if it does not represent at best their real thought.
The analysis then reveals that questions that use response formats with "middle" categories lead to big errors in measurements.
In addition, a more detailed analysis was carried out regarding the possibility of choice, which may be 2, 3, 4 etc.
In the study it was found that the answer that presents itself as an "average" among the other 2 is very often chosen. This is because this option introduces ambiguity where otherwise a clear choice between two extremes would have occurred. This option is often choice because it requires less effort and provides a quick choice in case of uncertainty.
It follows that the answers to 3 options have less credibility than those with 2 and 4 options. The results of the research produce a strong support in favor of this thesis.
Furthermore, it has been found that the problem of the truthfulness of the intermediate response in the chosen 3 response model is not as frequent in the use of 5 responses.
The results of the models with 5 responses are however less truthful than templates with 4 responses.
Significant differences were found between questions with unipolar and bipolar content. A single-pole question is based on a single polarity and requires evaluating the absence or presence of what it is reported in the application: for example a question about satisfaction will have a unipolar scale 1) UNDISFACT 2) MODERATLY SATISFIED 3) VERY SATISFIED
Bipolar scales are based on two polarities, often opposite.
An example of a bipolar scale is 1) AGREE 2) DISAGREE
The questions that make use of a scale of unipolar answers are more truthful.
From the elaboration of the data it has also emerged that the evaluations and descriptions that the individual provides regarding themselves are more truthful of those concerning his attitudes.
Expectations, or the request to predict a future event, are instead the worst in terms of reliability.
Personal perceptions were measured with a reliability and truthfulness of 75%, expectations with 55%.
Moreover, as already known by many other studies, factual informations are registered with greater veracity than non-factual ones.
Conclusion
In conclusion, the answers are therefore more reliable if are used less response categories, with a use of unipolar scales.Important exceptions to this theory, however, are the answer's categories that provide the possibility of "half-way" answers, which require less effort (such as those with 3 or with 5 options).
Authors:
Duane F Alwin, Erin M Baumgartner, Brett A Beattie
