Group 3 1. Chow Siew Wai 2. See Wei Chan 3. Leow Chai Chin 4.

Catherine

CGA110026 CGA110030 CGA110049 CGA

Framework
I.
II. III. IV.

Introduction
Empirical Results Conclusion Q&A

Responding to survey question involves many, frequency iterative, steps of complex information processing. Responding to survey consider as a complex communication process whereby the product of the interaction between researches and respondents lead to the sharing and creating of meaning. In this article, a study into questions and the various possibilities of their encoding and decoding is reviewed. Followed by a studies investigating the impact of different forms of encoding and decoding responses.

Dr Petra Lietz

Questions

Responses

Survey Communication Framework

1. Question lengths

o o

To keep question or statement as short as possible Specific a maximum number of 16 woks or 20 words per sentence To keep grammatical complexities to a minimum Question should employ the active rather than passive voice Repeat nouns instead of using pronouns Avoid possessive forms

2. Grammar

o o o o

3. Specificity and simplicity

o o o o o

Use specific rather than general terms Breaking down complex questions into simple ones Providing behavioral illustrations of certain concepts Avoid words such as Probably, May be or Perhaps Against the use of hypothesis questions instead use vignettes or alternative scenario

4. Social desirability (SD)

o o o o

Questions shall form as indirect questions What do you believe other people think about Use introductory phrase Do you happened to know Use words as neutrally as possible Propose value on a certain topic in different directions Suggest the normalcy of social deviant behavior To avoiding question or statement that contain two different verbs or two different concepts Do you have time to read the newspaper every day?

5. Doublebarreled questions

o o

6. Negatively worded questions

o X

To avoid negatively worded questions or statement What is your review about the statement that conservationists should not be so uncooperative with the government? What is your view about the statement that conservations should cooperate with the government?

7. Adverbs of frequency

o o o

Offer respondents specific quantifies in the response options Should included numeric reference points for a specified time period How many hours do you watch TV on a weekday? with providing options such as <0.5hours, 0.5hours to < 1 hour, 1 hour to <1.5 hours. Effect of part-whole combination General question shall placed before the specific questions Effect of part-part combination To consider the sequence of question if they are asked at the same level of specificity Effect as a result of salience Demographic questions such as age, education, income and marital status should come at the end of the questionnaires rather than at the beginning

8. Question order

1. Dont Know (DK) Option

o o o

Response scale should avoid to offer DK option Percentages of respondents choosing DK increased by between 22% to 25% when it was explicitly offered About 20% of respondents shifted from DK option to substantive response option if DK option was not offered

2. Opinion Floating

o o

Less educated respondents tended to give more DK responses than more educated respondents Whether a filter or standard question in a questionnaires is depend on respondent is interested in informed opinion on a issue

3. Opinion Filtering

o Do you have an opinion on this or not? Respondent self-identification is being ignorant

Have you already thought sufficiently about XYZ so that you could form an opinion? Instead of Do you have an opinion on XYZ? Used stronger filter questions

4. Number of Response Scale Option

5-point scales are preferable in situation where respondents are asked for absolute judgment 7 to 9 point scales to be more appropriate in situation where more abstract judgments are sought from respondents

5. Odd or Even Number of Response Scale Option

o Odd or even number of response scale options are implying a decision on whether or not to offer a usually neutral middle option that allow respondents not to commit themselves to a direction in their opinions or attitude. o From analysis reported, response scale without the middle point had lower validity and higher random error variance, indicating that people randomly choose other available response options when the middle options was not available. o Research with having middle point to increase the reliability and validity of response scales has been confirmed by Saris and Gallhofer (2007). (I really have no idea how to present this?!!!)

6. Labeling of Response Scales Option

o Numbered scale (unipolar) which a scale from 0-10 or (bipolar) which a scale from -5 to 5 o Verbal scales (agree, slightly agree, either agree nor disagree, slightly disagree, disagree) o Multiplying adverbs (slightly, rather, extremely) o Direction of response option is negligible in most situation o Lowest numerical value should be attached to the disagree option o Questionnaires should convey less desirable responses

7. Order and Direction of Response Option

Dont Know (DK) Option

o o Opinion Floating o

In conducting research, the first central issue that need to be addressed is whether all the respondents should answer all the questions or whether those respondents with little or no knowledge should be filtered out and not be asked certain questions. Trometer (1996) reported that percentages of respondents choosing DK increased by between 22% to 25% when it was explicitly offered. About 20% of respondents shifted from DK option to substantive response option ( i.e. agree or disagree ) if DK option was not offered. Respondents give substantive response when DK option is not offered but respondents choose opinion floating when offered. Floaters as these people seem to vary their responses depending on the response options on offer. From experiments conducted, less educated respondents tended to give more DK responses. A more explicit way of filtering out respondents that questions are advocated by some as means of excluding anyone who is ignorant on a particular issue. 2 things that very important: Respondents self-identification as being ignorant might vary systematically as a consequence of question topic as well as respondents characteristics such as gender and age. A serious consequence of filtering out respondents is the impact on the representatives of the sample, in particular where stronger filter questions are used (e.g. Have you already thought sufficiently about XYZ so that you could form an opinion? Instead of Do you have an opinion on XYZ?) that lead to the overestimation of people without an opinion (Hippler et al. 1987). Evidence that runs counter to the hypothesis that less well-informed people have no attitudes on certain issues stems from such people being consistent in their response behavior over time. For the group of people with non-attitudes if could be anticipated that half of them would favor an issue and the other half would oppose an issue.

o Opinion Filtering o o o

Number of Response Scale Option

o 5-point and 7-point scale response options are the most commonly used due to 5 and 7-point scale can easily be rescaled in order to facilitate comparisons. o The 7-point scale has been shown to be more reliable, it allows for greater differentiation of responses that the 5-point scale. o The 5-point scales are preferable in situations where respondents are asked for absolute judgments. While 7 to 9-point scales to be more appropriate in situation where more abstract judgments are sought from respondents. o Odd or even number of response scale options are implying a decision on whether or not to offer a usually neutral middle option that allow respondents not to commit themselves to a direction in their opinions or attitude. o From analysis reported, response scale without the middle point had lower validity and higher random error variance, indicating that people randomly choose other available response options when the middle options was not available. o Research with having middle point to increase the reliability and validity of response scales has been confirmed by Saris and Gallhofer (2007). o Decisions regarding the labeling response scale option include whether to use numbered scale that are unipolar ( a scale from 0-10), bipolar ( a scale from -5 to 5), or verbal scales ( agree, slightly agree, either agree nor disagree, slightly disagree, disagree)

Odd or Even Number of Response Scale Option

Labeling of Response Scales Option

Order and Direction of Response Option

o Foddy (1993) has outlined a number of response options effects including primacy effect, recency effect and effect of shifting frames of reference. o Primacy effect refer to assumption that respondents will select earlier alternatives more frequently than later alternatives. o Recency effect refers is said to apply when respondents select the later alternatives and is thought to apply mainly apply respondents only hear the alternatives. o Shifting frames of reference effect to possibility that the selection of a certain alternative depends on whether the more favorable alternatives are presented earlier or later.

Questions should be clear, simple, specific

Demographic questions should be put at the end of the questions

Questions should focus on current attitude and very recent behavior

Disagree options should have lower numeric values attached than agree options Numerical scale should be unipolar with matching verbal label at both end of the scales

Likert-type response scale length ranges from 5 to 8 is desirable