Developing survey questions Problems that careful questionnaire design

Developing survey questions

Problems that careful questionnaire design can alleviate l Consistency effect – respondents attempt to make their later answers fit the earlier ones l Fatigue effect – questions toward the end of the survey are not carefully answered l Redundancy effect – all questions in the same style are answered in the same way

Question types l There are two major types of questions u Closed-ended questions u Open-ended questions

Closed-ended questions l The respondent is provided with a limited number of response options u Each response option has a numerical value or code it represents

Closed-ended questions l Examples: l Are you: u 18 -34 years of age (1) u 35 -54 years of age (2) u 55+ years of age (3)

Closed-ended questions l If you were about to l 1. Magazines purchase a digital TV, l 2. Advertising which source of l 3. Salesperson information, if any, would you consult first? l 4. Product brochure l 5. Friends l 6. Other l 7. Would not consult information source

Closed-ended questions l Please say whether you Strongly agree (5), u agree somewhat (4), u neither agree nor disagree (3), u disagree somewhat (2) u or strongly disagree (1) u with the following statement: “People who spend over 400 dollars on a cell phone are vain. ”

Open-ended questions l The question is standardized, but the respondent answers in her own words. l The interviewer usually ‘probes’ at least once for expansion and/or specificity after the initial response. l Example: u “What features would the ideal science fiction movie have? ”

Closed-ended v. open-ended l Closed-ended questions take less time to complete l Closed-ended questions take less time to enter into the database l Closed-ended questions allow the respondent to interpret her position in relation to the question

Closed-ended v. open-ended l Open-ended questions allow the respondent to elaborate and to make connections you may not have anticipated l Answers to open-ended questions must be ‘coded’ in order to be interpreted in a quantitative manner l Respondents often prefer open-ended questions and are bored and frustrated by too many closed-ended questions

Closed-ended v. open-ended l Because they take longer and are more difficult to deal with, open-ended questions should be used sparingly u Mix open-ended questions through the survey instrument, breaking it up and adding interest

Question wording l Concrete questions are best a. Are you happy with the service at Radio Shack? b. Better: Were you served within 5 minutes the last time you went to Radio Shack?

Question wording l Avoid abbreviations u In your view, does USC provide a liberal arts education worth its yearly tuition? u In your view, does the University of South Charleston provide a liberal arts education worth its yearly tuition?

Question wording l Avoid slang and colloquial expressions u Should president Obama take his economic advisers behind the woodshed? l Avoid jargon and technical terms u Should a summative evaluation of Head Start be commissioned by the U. S. government? u Better: Should the U. S. government commission a history of Head Start to review its activities and accomplishments?

Question wording l Avoid biasing words and phrases u The left-wing student group protesting U. S. presence in Sumatra claims that American presence there is unwarranted. Do you agree? ? l An exception: If necessary, use loaded questions to prompt response, but be careful, respondents may be offended u Parents get really angry at their children sometimes. In the past week, have you been really angry at your son?

Question wording l Avoid double-barreled questions u Do you think the University of Kentucky should subsidize student tuition and staff benefits? l Would you say that the decision to increase President Lee Todd’s salary by 67% will be a boon to the University of Kentucky or don’t you care? u You could think it would be a boon but not care

Question wording l Avoid negative wording u Double negatives are especially bad l “Please indicate whether you strongly agree, neither agree nor disagree, disagree or strongly disagree with the following statement: The United Nations should not have more authority to intervene in a nation’s military affairs. u Either emphasize NOT when asking the question or else reword the question in a positive manner u

Proper scale construction l Response options on scales should be u 1) Mutually exclusive u 2) Equivalent u 3) Exhaustive

Mutual exclusivity l A given respondent or item should not fit into more than one category l Example: u Are you: u a) Single; b) Married; c) Divorced; d) Widowed; e) Separated, or f) Living in a couple relationship but not married • In one interview I heard, a 70+ year old woman said “All of the above”

Category equivalency l Categories in a scale must differ on a single dimension l Example: Would you say that the service in the shoe department is: u a) fast u b) slow u c) polite u d) knowledgeable

Exhaustiveness l All respondents or items must be classifiable in one of the categories l Example: Does your Herald-Leader arrive on your doorstep by 6 AM: a) Always b) Seldom c) Never l What about “usually”?

Scales used to measure knowledge, attitudes, beliefs l Developing measures of knowledge, attitudes, and/or beliefs is a challenging task. l While anyone can come up with a measure, it is hard to develop a truly valid measure of KAB’s

Main types of attitude measures l Semantic differentials l Likert scales l Thurstone scales

Semantic differential l Good __ __ Bad l Weak __ __ Strong l Beautiful __ __ Ugly

Likert scales l Example: The Employment Self Esteem Scale l INSTRUCTIONS: Please rate how strongly you agree or disagree with each of the following statements by placing a check mark in the appropriate box. l (Strongly Disagree/Somewhat Disagree/ Somewhat Agree/Strongly Agree)

1. I feel good about my work on the job. 2. On the whole, I get along well with others at work. 3. I am proud of my ability to cope with difficulties at work. 4. When I feel uncomfortable at work, I know how to handle it. 5. I can tell that other people at work are glad to have me there. 6. I know I'll be able to cope with work for as long as I want. 7. I am proud of my relationship with my supervisor at work. 8. I am confident that I can handle my job without constant assistance. 9. I feel like I make a useful contribution at work. 10. I can tell that my coworkers respect me.

Guttman scale l Statements are ordered so that they represent increasing agreement with or acceptance of one position on a dimension u If respondent accepts one statement higher on the scale, he should also accept those below it on the scale

A type of Guttman scale: Bogardus social distance scale l I am willing to permit immigrants to live in my country l I am willing to permit immigrants to live in my community l I am willing to permit immigrants to live in my neighborhood l I am willing to permit immigrants to live next door to me l I am willing to have immigrants for friends l I am willing to have an immigrant marry my son/daughter l Agreement with item 3 implies agreement with items 1 and 2, and so on

Thurstone scales l Thurstone scales include a number of statements that have been developed to represent their position or magnitude on some dimension l The average of the scores of the statements agreed with by the respondent provides her placement on the scale

l People with AIDS are like my parents. l Because AIDS is preventable, we should focus our resources on l l l l l prevention instead of curing. People with AIDS deserve what they got. Aids affects us all. People with AIDS should be treated just like everybody else. AIDS will never happen to me. It's easy to get AIDS doesn't have a preference, anyone can get it. AIDS is a disease that anyone can get if they are not careful. If you have AIDS, you can still lead a normal life. AIDS is good because it helps control the population. l I can't get AIDS if I'm in a monogamous relationship.

Response option considerations �“Don’t know” option � Researchers debate whether the interviewer should explicitly ask whether the respondent doesn’t know an answer �Fear that the DK option gives the respondent an easy out without thinking about the question If an explicit “Don’t Know” option is provided, you’ll get more DKs but your conclusions may actually be more valid

�“Neutral” option � Should scales have a middle option (oddnumbered) or no neutral point (even-numbered)? � Even-numbered scales force respondents to take a stand

Indexes l Indexes combine scores from multiple items to generate a single score for each respondent on some variable of interest u Thurstone scales are indexes l Some variables are complex, needing several items to provide a valid measure l The items could be organized in a single dimension (unidimensional) or in several dimensions (multidimensional)

Unidimensional indexes l Summated index u measured by a set of indicators that can be added together to derive a single, overall score • Items are related among themselves (correlated), (if you score high on one item you should score high on the other ones) • Make sure the items are not too highly correlated

Multidimensional indexes l If the construct you are measuring has more than one significant subconcept, you probably would gain by using a multidimensional index l Example -- communicative credibility is composed of three subconcepts: Authoritativeness u Trustworthiness u Dynamism u l To measure communicative credibility you would need to ask questions related to each of the subconcepts

So: l Questionnaire design is the most important part of survey analysis l Developing valid measures of ‘black box’ phenomena is especially difficult l Writing good questions is probably the most important part of questionnaire design u. A number of common problems in question wording exist

l Closed-ended questions are simpler and faster but open-ended questions allow respondents to provide more personalized information and are more interesting for them l A number of forms of attitude/belief measures have been developed over the years u Commonly used types are Likert scales, semantic differentials, Thurstone scales, Guttman scales

l Using multiple items to measure a single construct and then combining them increases measurement validity