A questionnaire is a structured instrument consisting of a series of questions designed to gather information from respondents systematically. It is the most widely used data collection tool in survey research, and its quality directly determines the quality of the data and the validity of research conclusions.
Developing a good questionnaire is a systematic process that involves several stages:
A well-designed questionnaire typically has the following sections:
A researcher designs a questionnaire to study gender discrimination in the private vs. government sector. The questionnaire structure is:
Section A — Screening: (1) Are you currently employed? (Yes/No) (2) Which sector do you work in? (Private / Government / Other)
Section B — Workplace Experience: (3) How would you rate your overall job satisfaction? (1–5 Likert scale) (4) Have you ever felt that you were treated differently at work because of your gender? (Yes/No) (5) If yes, in which of the following areas? (Checklist: salary, promotion, assignment of responsibilities, training opportunities, other) (6) "Men and women receive equal pay for equal work in my organisation" (Strongly agree to Strongly disagree)
Section C — Career Progression: (7) How many promotions have you received in the last 5 years? (8) What is your current designation level? (Entry / Middle / Senior / Executive) (9) "Promotion decisions in my organisation are based purely on merit" (5-point scale)
Section D — Demographics: Age, gender, education, years of experience, annual income range.
Each question maps to a specific research objective: Q4 and Q5 measure perceived discrimination, Q7 and Q8 measure career progression, and demographics allow subgroup analysis.
A researcher investigates the impact of unemployment on mental health and financial stability. The questionnaire includes:
Section A — Employment Status: (1) Current employment status (Employed / Unemployed / Underemployed) (2) Duration of current unemployment (if applicable) (3) Reason for unemployment (Layoff / Resignation / Contract ended / Fresh graduate / Other)
Section B — Financial Impact: (4) Has your household income decreased since unemployment? (Yes/No) (5) By what percentage has your monthly income changed? (0% / 1–25% / 26–50% / 51–75% / 76–100%) (6) Have you taken any loans to meet expenses? (Yes/No)
Section C — Psychological Impact: (7) PHQ-9 depression screening scale (9 items, standardised instrument) (8) "I feel hopeful about finding employment soon" (5-point Likert)
Section D — Coping Strategies: (9) What steps have you taken to find work? (Checklist: job portals, networking, skill development, relocation, others) (10) Have you sought professional mental health support? (Yes/No)
Section E — Demographics: Age, gender, education, previous occupation, location.
Using a validated instrument (PHQ-9) for depression measurement ensures reliability and comparability with other studies.
Survey questions are classified by the type of response they elicit. The choice of question type depends on the nature of the information sought, the desired level of measurement, and the analytical methods to be applied.
Respondents answer in their own words without pre-set options. Useful for exploratory research, capturing unexpected responses, and rich qualitative data. However, they require more effort from respondents and are difficult to code and analyse quantitatively.
Example: "What do you think is the biggest challenge facing the education system today?"
Dichotomous questions: Only two response options.
Example: "Do you own a smartphone? (Yes / No)"
Multiple-choice questions: Select one from several options. Options must be mutually exclusive and exhaustive.
Example: "What is your highest educational qualification? (No formal education / Secondary / Higher Secondary / Graduate / Postgraduate)"
Checklist questions: Select all that apply. Used when multiple responses are valid.
Example: "Which of the following social media platforms do you use? (Facebook / Instagram / Twitter / LinkedIn / YouTube / None)"
Ranking questions: Order options by preference or importance.
Example: "Rank the following job attributes from 1 (most important) to 5 (least important): Salary / Job security / Work-life balance / Career growth / Work environment"
Questions that apply only to a subset of respondents, determined by a preceding screening question.
Example: Q5: "Do you smoke? (Yes/No)" → Q5a (if Yes): "How many cigarettes per day?" → Skip to Q6 (if No).
A restaurant designs a customer feedback form using multiple question types:
The mix of question types captures both quantitative ratings (for statistical analysis) and qualitative suggestions (for actionable feedback). The dichotomous screening question allows analysis by first-time vs. repeat visitors.
The National Family Health Survey uses a complex structure of filter questions:
Q201: "Are you currently married? (Yes/No)"
→ If Yes: Q202: "How long have you been married? (years)"
→ If Yes: Q203: "How many children have you given birth to?"
The wording of a question can dramatically affect responses. Poorly constructed questions introduce measurement error — the difference between the true value and the recorded value — and can invalidate the entire study. The following principles guide the construction of reliable, valid questions.
1. Simplicity: Use simple, everyday language. Avoid technical jargon unless the target population is familiar with it.
2. Specificity: Be precise about the time period, quantity, or context being asked about.
3. Neutrality: Do not lead the respondent toward any particular answer.
4. Avoid double-barrelled questions: Each question should ask about only one thing.
5. Avoid loaded or presumptive questions: Do not assume facts not in evidence.
6. Ensure mutually exclusive and exhaustive response categories:
7. Avoid double negatives:
A student designs a questionnaire on the impact of subsidy removal. The original questions are problematic:
A survey asks: "Did you vote in the last election?" In reality, the voter turnout was 58%, but 82% of survey respondents claim they voted. This is social desirability bias — voting is seen as a civic duty, and non-voters are embarrassed to admit it. Techniques to mitigate this include:
Scaling is the process of assigning numbers to qualitative responses so that they can be analysed quantitatively. A scale is a set of related items that together measure a construct (such as satisfaction, attitude, or perception) more reliably than any single item could.
The most widely used attitude scale. Respondents indicate their level of agreement with a statement on a symmetric agree–disagree scale.
| Code | Response |
|---|---|
| 5 | Strongly Agree (SA) |
| 4 | Agree (A) |
| 3 | Neutral (N) |
| 2 | Disagree (D) |
| 1 | Strongly Disagree (SD) |
Some researchers use a 7-point scale for finer discrimination.
Respondents rate a concept on a series of bipolar adjective pairs using a numerical scale (typically 7 points).
Example: Rate your workplace: Modern ①②③④⑤⑥⑦ Traditional; Friendly ①②③④⑤⑥⑦ Hostile
A set of items ordered by increasing intensity. Agreement with a stronger item implies agreement with all weaker items. Used to measure unidimensional attitudes.
Example: (1) Immigrants should be allowed in the country. (2) Immigrants should be allowed to work. (3) Immigrants should be allowed to vote. A person who agrees with (3) should also agree with (1) and (2).
Judges assign scale values to statements representing different levels of an attitude. The respondent selects statements they agree with, and their attitude score is the median of the scale values of those statements.
The most common measure of internal consistency (reliability):
\[ \alpha = \frac{k}{k-1}\left(1 - \frac{\sum_{i=1}^{k} s_i^2}{s_T^2}\right) \]
where \(k\) = number of items, \(s_i^2\) = variance of item \(i\), and \(s_T^2\) = variance of the total score. Values: \(\alpha \geq 0.70\) is acceptable, \(\alpha \geq 0.80\) is good, \(\alpha \geq 0.90\) is excellent.
A researcher measures job satisfaction using a 10-item Likert scale. Each item is a statement (e.g., "I feel valued at my workplace," "My opinions are considered in decision-making") rated from 1 (Strongly Disagree) to 5 (Strongly Agree). The total score ranges from 10 to 50.
For 150 respondents, the item variances are: \(s_1^2 = 1.2, s_2^2 = 0.9, s_3^2 = 1.1, s_4^2 = 1.3, s_5^2 = 1.0, s_6^2 = 1.4, s_7^2 = 0.8, s_8^2 = 1.2, s_9^2 = 1.1, s_{10}^2 = 1.0\). Sum of item variances: \(\sum s_i^2 = 11.0\). Total score variance: \(s_T^2 = 85.5\).
\[ \alpha = \frac{10}{9}\left(1 - \frac{11.0}{85.5}\right) = 1.111 \times (1 - 0.1287) = 1.111 \times 0.8713 = 0.968 \]
Cronbach's alpha of 0.968 indicates excellent internal consistency — the items are measuring the same underlying construct (job satisfaction) very reliably.
A market researcher uses a semantic differential scale to compare perceptions of two smartphone brands (Brand X and Brand Y) on 6 bipolar pairs:
A pilot test (pre-test) is a small-scale trial run of the questionnaire and survey procedures before the main study. It is indispensable for identifying problems with question wording, response options, skip patterns, timing, and field procedures. No questionnaire should ever be fielded without a pilot test.
The pilot sample should be 20–50 respondents drawn from the target population. It need not be a probability sample but should include diverse types of respondents (different ages, education levels, genders) to test the questionnaire across subgroups.
A researcher pilots a questionnaire on dietary habits with 30 respondents. Several problems emerge: (a) The question "How many servings of fruits and vegetables do you consume daily?" confuses respondents because "serving" is not clearly defined — some count a whole fruit as one serving, others count a bite. The question is revised to: "How many cups of fruits and vegetables do you eat per day?" with visual aids showing what a cup looks like. (b) The income categories "₹0–10,000 / ₹10,000–20,000 / ₹20,000–30,000" create confusion because the boundaries overlap. A respondent earning exactly ₹10,000 doesn't know which category to choose. The categories are revised to "Less than ₹10,000 / ₹10,000–19,999 / ₹20,000–29,999 / ₹30,000 and above." (c) The average completion time is 25 minutes, longer than the target of 15 minutes. Three low-priority questions are removed to reduce respondent burden.
A university pilots its online course evaluation survey with 50 students. The pilot reveals: (a) The skip logic for the question "Did you attend the lab sessions?" is faulty — students who select "No" are still shown lab-related questions. The branching logic is corrected. (b) On mobile devices, the 7-point Likert scale wraps to two lines, making it confusing. The scale is changed to a 5-point version that fits on one line. (c) The open-ended question "Any suggestions for improvement?" has a 100-character limit that is too restrictive. The limit is increased to 500 characters. (d) The average completion time is 8 minutes (within the target of 10 minutes). After these fixes, the main survey is launched with confidence.
Fieldwork encompasses all activities involved in collecting data from respondents in the field — recruiting and training interviewers, supervising data collection, ensuring quality control, and managing logistics. Well-organised fieldwork is essential for achieving high response rates and data quality.
Interviewers are the face of the survey. Their competence and professionalism directly affect response rates and data quality. Training should cover:
The NSSO conducts the Periodic Labour Force Survey (PLFS) across India. Fieldwork involves: (a) 1,200 trained investigators covering 12,800 sample villages and urban blocks each quarter. (b) Each investigator is assigned 8 households per village/block and completes the schedule in 2–3 days. (c) Supervisors conduct spot-checks on 25% of completed schedules. (d) Regional offices review 100% of schedules for completeness and consistency. (e) Data entry is centralised with double-key verification. The multi-layer quality control system keeps the item non-response rate below 2% and the unit non-response rate below 5%, which is exceptional for a large-scale survey in a developing country.
A student researcher conducts a survey on the impact of the removal of government subsidy on university students. The fieldwork plan is: (a) Train 5 student volunteers as interviewers in a 2-hour session covering questionnaire content, interview technique, and ethical guidelines. (b) Assign each interviewer to a different campus location (library, cafeteria, hostel, science block, arts block) to ensure diversity. (c) Use systematic sampling — approach every 5th student passing through the location during a 2-hour time slot. (d) Target: 50 completed questionnaires per interviewer = 250 total. (e) Each interview takes approximately 10 minutes. (f) The researcher (supervisor) floats between locations, observing interviews and providing real-time feedback. (g) Completed questionnaires are checked at the end of each day. (h) The response rate is 68% (250 out of 368 students approached). The 32% who declined are predominantly students rushing to class or exams. Non-response bias is assessed by comparing the demographic profile of respondents with the known student body profile.
The syllabus specifically requires students to develop questionnaires, collect data, and interpret results for research problems such as: gender discrimination in private vs. government sectors, unemployment rates, removal of subsidy, and impact on service class vs. unorganised sectors. This section outlines how to approach these applied problems.
Research Question: Is there a significant difference in the perception and experience of gender discrimination between employees in the private sector and government sector?
Study Design: Cross-sectional survey with stratified sampling (private vs. government), equal allocation. Key variables: perceived discrimination score (Likert scale), salary, designation level, years of experience, gender.
Analysis: Two-sample t-test or Mann-Whitney U test for comparing discrimination scores; chi-square test for comparing proportions of respondents reporting discrimination; logistic regression with discrimination (Yes/No) as dependent variable and sector, gender, experience as predictors.
Research Question: How has the removal of a government subsidy (e.g., on cooking gas or food grains) differentially affected the service class and unorganised sector workers?
Study Design: Before-and-after comparison (if pre-removal data are available) or cross-sectional survey with recall questions. Key variables: change in monthly expenditure, change in consumption patterns, coping mechanisms, perceived impact.
Analysis: Paired t-test (before vs. after); two-sample t-test (service vs. unorganised); chi-square test for comparing proportions using coping strategies; regression with change in expenditure as dependent variable.
A researcher collects data from 200 employees (100 private, 100 government) on a gender discrimination perception scale (scored 10–50, higher = more perceived discrimination). The results show:
| Sector | Mean Score | SD | n |
|---|---|---|---|
| Private | 32.5 | 8.2 | 100 |
| Government | 26.8 | 7.5 | 100 |
Two-sample t-test: \(t = \frac{32.5 - 26.8}{\sqrt{\frac{67.24}{100} + \frac{56.25}{100}}} = \frac{5.7}{\sqrt{1.2349}} = \frac{5.7}{1.111} = 5.13\) with 198 df.
Since \(t = 5.13 \gg 1.972\) (critical value at \(\alpha = 0.05\)), we reject \(H_0\). Private sector employees perceive significantly more gender discrimination than government sector employees. The 95% CI for the difference is \(5.7 \pm 1.972 \times 1.111 = [3.51, 7.89]\). Cohen's d = \(\frac{5.7}{\sqrt{(8.2^2 + 7.5^2)/2}} = \frac{5.7}{7.86} = 0.725\), which is a medium-to-large effect. A logistic regression further reveals that being female and working in the private sector are both significant predictors of reporting personal experience of discrimination (OR = 3.2 for female, OR = 2.4 for private sector).
A researcher studies the impact of LPG subsidy removal on 300 households: 150 from the service class and 150 from the unorganised sector. The key outcome is the percentage increase in monthly cooking fuel expenditure after subsidy removal.
| Sector | Mean % Increase | SD | n |
|---|---|---|---|
| Service class | 12.3% | 5.1 | 150 |
| Unorganised | 28.7% | 9.8 | 150 |
Two-sample t-test: \(t = \frac{28.7 - 12.3}{\sqrt{\frac{96.04}{150} + \frac{26.01}{150}}} = \frac{16.4}{\sqrt{0.814}} = \frac{16.4}{0.902} = 18.18\). Highly significant (p < 0.001).
The unorganised sector experienced a disproportionately larger burden from the subsidy removal. While the service class saw a modest 12.3% increase (manageable for most), the unorganised sector saw a 28.7% increase — a severe impact on household budgets. The effect size (Cohen's d) is \(d = \frac{16.4}{\sqrt{(5.1^2 + 9.8^2)/2}} = \frac{16.4}{7.84} = 2.09\), which is very large. Additionally, 68% of unorganised sector respondents reported switching to cheaper but more polluting fuels (firewood, kerosene), compared to only 12% of service class respondents. This highlights the unintended environmental and health consequences of subsidy removal on vulnerable populations.