Developing survey questions - PowerPoint Presentation

Slide1

Developing survey questionsSlide2

Problems that careful questionnaire design can alleviate

Consistency effect – respondents attempt to make their later answers fit the earlier ones

Fatigue effect – questions toward the end of the survey are not carefully answered

Redundancy effect – all questions in the same style are answered in the same waySlide3

Question types

There are two major types of questions

Closed-ended questions

Open-ended questionsSlide4

Closed-ended questions

The respondent is provided with a limited number of response options

Each response option has a numerical value or code it representsSlide5

Closed-ended questions

Examples:

Are you:

18-34 years of age (1)

35-54 years of age (2)

55+ years of age (3)Slide6

Closed-ended questions

If you were about to purchase a digital TV, which source of information, if any, would you consult first?

1. Magazines

2. Advertising

3. Salesperson

4. Product brochure

5. Friends

6. Other

7. Would not consult information sourceSlide7

Closed-ended questions

Please say whether you

Strongly agree (5),

agree somewhat (4),

neither agree nor disagree (3),

disagree somewhat (2)

or strongly disagree (1)

with the following statement:

“People who spend over 400 dollars on a cell phone are vain.”Slide8

Open-ended questions

The question is standardized, but the respondent answers in her own words.

The interviewer usually ‘probes’ at least once for expansion and/or specificity after the initial response.

Example:

“What features would the ideal science fiction movie have?”Slide9

Closed-ended v. open-ended

Closed-ended questions take less time to complete

Closed-ended questions take less time to enter into the database

Closed-ended questions allow the respondent to interpret her position in relation to the questionSlide10

Closed-ended v. open-ended

Open-ended questions allow the respondent to elaborate and to make connections you may not have anticipated

Answers to open-ended questions must be ‘coded’ in order to be interpreted in a quantitative manner

Respondents often prefer open-ended questions and are bored and frustrated by too many closed-ended questionsSlide11

Closed-ended v. open-ended

Because they take longer and are more difficult to deal with, open-ended questions should be used sparingly

Mix open-ended questions through the survey instrument, breaking it up and adding interestSlide12

Question wording

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?Slide13

Question wording

Avoid abbreviations

In your view, does

USC

provide a liberal arts education worth its yearly tuition?

In your view, does the

University of South Charleston

provide a liberal arts education worth its yearly tuition?Slide14

Question wording

Avoid slang

and colloquial expressions

Should president Obama take his economic advisers behind the woodshed?

Avoid jargon

and technical terms

Should a

summative evaluation

of Head Start be commissioned by the U.S. government?

Better: Should the U.S. government commission a history of Head Start to review its activities and accomplishments?Slide15

Question wording

Avoid biasing

words and phrases

The

left-wing

student group protesting U.S. presence in Sumatra

claims

that American presence there is unwarranted. Do you

agree?

?

An exception:

If necessary

,

use

loaded questions

to prompt response, but be careful, respondents may be offended

Parents get really angry at their children sometimes.

In the past week, have you been really angry at your son?Slide16

Question wording

Avoid

double-barreled

questions

Do you think the University of Kentucky should subsidize student tuition

and

staff benefits?

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?

You could think it would be a boon but not careSlide17

Question wording

Avoid negative wording

Double negatives are especially bad

“Please indicate whether you strongly agree, 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.

Either

emphasize NOT

when asking the question or else

reword the question

in a positive mannerSlide18

Proper scale construction

Response options on scales should be

1) Mutually exclusive

2) Equivalent

3) ExhaustiveSlide19

Mutual exclusivity

A given respondent or item should not fit into more than one category

Example:

Are you:

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”Slide20

Category equivalency

Categories in a scale must differ on a

single dimension

Example: Would you say that the service in the shoe department is:

a) fast

b) slow

c) polite

d) knowledgeableSlide21

Exhaustiveness

All respondents or items must be classifiable

in one of the categories

Example: Does your Herald-Leader arrive on your doorstep by 6 AM:

a) Always

b) Seldom

c) Never

What about “usually”?Slide22

Scales used to measure knowledge, attitudes, beliefs

Developing measures of knowledge, attitudes, and/or beliefs is a challenging task.

While anyone can come up with a measure, it is hard to develop a truly

valid

measure of KAB’sSlide23

Main types of attitude measures

Semantic differentials

Likert

scales

Thurstone

scalesSlide24

Semantic differentialGood __ __ __ __ __ __ __ Bad

Weak __ __ __ __ __ __ __ Strong

Beautiful __ __ __ __ __ __ __ UglySlide25

Likert scales

Example: The Employment Self Esteem Scale

INSTRUCTIONS:

Please rate how strongly you agree or disagree with each of the following statements by placing a check mark in the appropriate box.

(Strongly Disagree/Somewhat Disagree/ Somewhat Agree/Strongly Agree)Slide26

I feel good about my work on the job.

On the whole, I get along well with others at work.

I am proud of my ability to cope with difficulties at work.

When I feel uncomfortable at work, I know how to handle it.

I can tell that other people at work are glad to have me there.

I know I'll be able to cope with work for as long as I want.

I am proud of my relationship with my supervisor at work.

I am confident that I can handle my job without constant assistance.

I feel like I make a useful contribution at work.

I can tell that my coworkers respect me. Slide27

Guttman scale

Statements are ordered so that they represent increasing agreement with or acceptance of one position on a dimension

If respondent accepts one statement higher on the scale, he should also accept those below it on the scaleSlide28

A type of Guttman scale:Bogardus

social distance scale

I am willing to permit immigrants to live in my country

I am willing to permit immigrants to live in my community

I am willing to permit immigrants to live in my neighborhood

I am willing to permit immigrants to live next door to me

I am willing to have immigrants for friends

I am willing to have an immigrant marry my son/daughter

Agreement with item 3 implies agreement with items 1 and 2, and so onSlide29

Thurstone scales

Thurstone

scales include a number of statements that have been developed to represent their position or magnitude on some dimension

The average of the scores of the statements agreed with by the respondent provides her placement on the scaleSlide30

People with AIDS are like my parents.

Because AIDS is preventable, we should focus our resources on 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.

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.

I can't get AIDS if I'm in a monogamous relationship

.Slide31

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 validSlide32

“Neutral” option

Should scales have a middle option (odd-numbered) or no neutral point (even-numbered)?

Even-numbered scales force respondents to take a standSlide33

Indexes

Indexes combine scores from multiple items to generate a single score for each respondent on some variable of interest

Thurstone

scales are indexes

Some variables are complex, needing several items to provide a valid measure

The items could be organized in a single dimension (

unidimensional

) or in several dimensions (multidimensional)Slide34

Unidimensional indexes

Summated index

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 Slide35

Multidimensional indexes

If the construct you are measuring has more than one significant

subconcept

, you probably would gain by using a multidimensional index

Example -- communicative credibility is composed of three

subconcepts

:

Authoritativeness

Trustworthiness

Dynamism

To measure communicative credibility you would need to ask questions related to each of the

subconceptsSlide36

So:

Questionnaire design is the most important part of survey analysis

Developing valid measures of ‘black box’ phenomena is especially difficult

Writing good questions is probably the most important part of questionnaire design

A number of common problems in question wording existSlide37

Closed-ended questions are simpler and faster but open-ended questions allow respondents to provide more personalized information and are more interesting for them

A number of forms of attitude/belief measures have been developed over the years

Commonly used types are

Likert

scales, semantic differentials,

Thurstone

scales,

Guttman

scalesSlide38

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