Skip to Main Content
Main site homepage

IM Course Guide

Surveys and Questionaires

The questionnaire is the most commonly used research tool in the social sciences.

Essentially, it is a list of questions that does not require too much time nor money, and easily maintains the privacy of the respondent.

5 Steps to writing a questionnaire

1. Defining objectives of survey      
2. Determining the sampling group
3. Writing the questionnaire                       
4. Administering the questionnaire
5. Interpreting the results


Defining the objectives of a questionnaire

 Defining the objectives of a questionnaire

What is the goal of your research?  Define the hypothesis and the parameters of your survey.

For example:

1.  Is the data measured going to be qualitative (ex: opinions, assumptions & beliefs) or quantitative (observable, countable & factual)? 
2.  Is the interview going to be anonymous (easy to administer, time efficient & answers not influenced by presence of interviewer) or face-to-face (allows elaboration on answers)?

Determining the sample group

To determine the sample group it is helpful to ask:

1. Who are the group of interest for your research?
2. What sampling method will you use?
3. How will you contact them?

Writing the questionnaire

All questions should reflect back on the main goal of your questionnaire.

How? For each question ask yourself:   what’s the point?  And, what do I want to know?”


1. Start with non-invasive questions on demographics. It puts respondents at ease.
2. Ask only questions that are necessary.
3. Order your questions logically, and group themed questions together.
4. Avoid splitting questions across pages.
5. Always number questions.

Survey rationale, or “Why are you doing this?”

State the reason for your survey before starting to ask questions. It gives respondents context for their answers and helps them in providing relevant answers.

Question writing

1. Video: Choose between open-ended and closed questions

Open-ended question

Closed question

What is your gender?
Are you male or female?

Closed questions have clearly defined categories, allowing us to enter and analyse data easily.

In some cases, closed format questions exclude possible answers, like in the question on gender, above. 

2. Avoid leading questions

 Leading questions influence responses through the question itself or through the answer categories.


Leading question
Better question
How much do you enjoy the cafeteria food?
Rate the quality of the cafeteria food from 1-5.


Leading answer categories
Better answer categories
1 – Excellent
2 – Very good
3 – Good
4 – Enjoyable
5 – Edible
1 – Very good
2 – Good
3 – Neither good nor bad
4 – Bad
5 – Very bad
In this case the answer categories for the leading question are skewed toward having more positive responses. 


3. Avoid double-barrelled questions
Video: Double-barrelled questions force a respondent to answer multiple questions with one answer.  Answers to these questions can inaccurately portray respondent opinions, and they may lower respondent trust in the survey!
Double-barrelled question
Better question
Do you agree the government should reduce taxes and improve health care?
Should the government:
a. Reduce taxes?
b. Improve health care?
In this case, the respondent may agree that healthcare should be improved, but not that we should lower taxes; or vice versa. 
4.  Using a rating scale
Two commonly used rating scales (there are more!) are linear numeric scales and Likert scales.
Linear numeric scales rate the answer to a question on a numeric scale, often of 1-10.  At least the end points are labelled.  They are used to measure strength of (for example) importance, agreement or likelihood.
Not at all important
Extremely important
Likert scales use an odd-numbered point scale (often 7-pts, but sometimes 5, or 9 or more) to measure strength of agreement. They are often used to gauge attitudes or reactions.
Somewhat agree
Neither agree nor disagree
Somewhat disagree
5.   Avoid hypothetical questions
Hypothetical questions have hypothetical answers. They do not represent real opinions, and often generate hard-to-process data.