Data Collection Methods & Ethics: Tools, Choices & Rules

TAKEAWAYS

• Data collection methods (surveys, interviews, observations, experiments) each serve distinct research goals—choose based on your objective, budget, and participant accessibility.
• Ethical principles (autonomy, beneficence, justice, non-maleficence) are non-negotiable; violations can invalidate research and harm participants.
• Digital tools (Khalti’s SMS surveys, Pathao’s ride-sharing data) demonstrate how technology reshapes data collection while raising new ethical dilemmas (privacy, consent).
• Pilot testing is critical: flawed instruments (e.g., ambiguous survey questions) waste resources and bias results.
• Qualitative vs. quantitative methods differ in depth, sample size, and analysis—mix them strategically (e.g., surveys + interviews for triangulation).
• Nepali context matters: Cultural norms (e.g., reluctance to disclose income) may require adapted methods (e.g., anonymous surveys for sensitive topics).

1. Core Data Collection Techniques

Data collection is the backbone of research. Each method has strengths, weaknesses, and ethical trade-offs. Below, we classify techniques by purpose (exploratory vs. confirmatory) and data type (quantitative vs. qualitative).

A. Surveys/Questionnaires

Definition: Structured tools to collect standardized responses from a sample. Can be closed-ended (multiple-choice) or open-ended (text responses).

How it works:

1. Design: Questions must be clear, unbiased, and relevant. Avoid:
   • Leading questions ("Don’t you agree poverty is worsening?").
   • Double-barreled questions ("How often do you exercise and eat healthy?").
   • Jargon ("Assess your socio-economic stratum").
2. Pilot test: Administer to 5–10 participants to check for confusion or bias.
3. Administration:
   • Self-administered (online via Google Forms, paper).
   • Interviewer-administered (face-to-face, phone).
   • Digital (Khalti’s SMS surveys, WhatsApp polls).

Worked Example: Khalti’s Customer Satisfaction Survey Khalti uses closed-ended Likert-scale questions (1–5) to measure user satisfaction after transactions. Example:

"How satisfied were you with Khalti’s customer support during your last issue?" 1 (Very Dissatisfied) → 5 (Very Satisfied)

Why it works:

• Quantifiable: Easy to analyze with statistics (mean scores, chi-square tests).
• Scalable: Reaches 10,000+ users instantly.
• Ethical: Anonymous responses protect privacy.

Limitations:

• Low response rates (e.g., only 20% of Daraz users complete surveys).
• Social desirability bias: Participants may overreport "good" behaviors (e.g., volunteering).
• Cultural bias: Questions may not translate well (e.g., "agree/disagree" scales confuse some Nepali respondents).

B. Interviews

Definition: One-on-one or group conversations to gather in-depth, contextual data. Can be:

• Structured (fixed questions, like a survey).
• Semi-structured (guided topics with flexibility).
• Unstructured (open-ended, exploratory).

How it works:

1. Recruitment: Use purposive sampling (e.g., interviewing 10 homeless youth in Kathmandu for a shelter study).
2. Consent: Explain risks/benefits verbally and in writing (even for audio-recorded interviews).
3. Conduct:
   • Face-to-face: Builds trust (critical for sensitive topics like domestic violence).
   • Phone/Video: Useful for remote areas (e.g., interviewing herders in Dolpa).
4. Recording: Always ask permission to record; transcribe verbatim.

Worked Example: Ncell’s Customer Feedback Interviews Ncell trains interviewers to ask:

"Can you describe a time when our network failed you? How did it affect your day?" Analysis: Themes like "unreliable calls during emergencies" emerge, guiding service improvements.

Advantages:

• Rich data: Captures emotions, nuances (e.g., "I felt invisible when the bank ignored my loan application").
• Adaptability: Follow-up questions clarify responses.

Disadvantages:

• Time-consuming: 1–2 hours per interview; hard to scale.
• Interviewer bias: Tone or body language may influence answers.
• Ethical risks: Participants may disclose trauma (require debriefing resources).

C. Observations

Definition: Systematically watching and recording behavior in natural or controlled settings. Types:

• Participant observation: Researcher joins the group (e.g., living with a Dalit family to study caste discrimination).
• Non-participant observation: Researcher observes from outside (e.g., timing traffic delays at Kathmandu’s Thapathali intersection).

How it works:

1. Define focus: What behaviors? (e.g., "How often do teachers use positive reinforcement in class?")
2. Tools:
   • Checklists (e.g., ✔️ "Teacher smiled" / ✖️ "Teacher yelled").
   • Field notes: Descriptive (e.g., "At 10:15 AM, a student cried when denied bathroom access").
3. Ethics: Obtain informed consent from participants/gatekeepers (e.g., school principal).

Worked Example: Pathao’s Rider Behavior Study Researchers observed 50 riders in Pokhara to study:

• Safety: 60% didn’t wear helmets (despite app reminders).
• Efficiency: Riders took 2–3 minutes to locate passengers in crowded areas. Outcome: Pathao added helmet reminders and GPS passenger location sharing.

Advantages:

• Unbiased: Captures "real" behavior (vs. self-reported surveys).
• Dynamic: Reveals patterns (e.g., "Traffic jams spike at 5 PM near Ncell HQ").

Disadvantages:

• Reactivity: Participants may change behavior if observed (e.g., students study harder when a researcher visits).
• Subjectivity: Interpretations vary (e.g., "Was that a smile or a grimace?").
• Ethical dilemmas: Can’t observe private spaces (e.g., homes) without consent.

mindmap
  root((Observation Study))
    Design
      Focus: "What to observe?"
      Tools: Checklists, cameras, notes
    Conduct
      Participant vs. Non-participant
      Natural vs. Controlled setting
    Ethics
      Consent
      Anonymity
    Analysis
      Themes: Patterns, frequencies
      Triangulation: Cross-check with interviews

D. Experiments

Definition: Manipulating an independent variable to measure its effect on a dependent variable, while controlling other factors.

How it works in social work:

1. Random assignment: Participants are randomly assigned to treatment (e.g., a new counseling program) or control (business-as-usual).
2. Pre-test/post-test: Measure outcomes before/after (e.g., depression levels in a support group).
3. Quasi-experiments: No random assignment (e.g., comparing two existing shelters).

Worked Example: NTC’s "Wait-Time Reduction" Pilot Goal: Reduce customer wait times at NTC offices. Method:

• Treatment group: Offices with digital queues (like Daraz’s order tracking).
• Control group: Traditional first-come-first-served. Result: Wait times dropped by 40% in treatment offices.

Advantages:

• Causality: Can prove "X causes Y" (e.g., "Group therapy reduces PTSD symptoms").
• Control: Isolates variables (e.g., testing a new teaching method while keeping class size constant).

Disadvantages:

• Artificiality: Lab settings may not reflect real life (e.g., role-playing interviews).
• Ethics: Withholding treatment (e.g., no counseling for the control group) is unethical unless the treatment is proven safe.
• Cost/time: Requires large samples and resources.

2. Choosing the Right Method: A Decision Guide

Not all methods fit every research question. Use this table to decide:

3. Ethical Considerations: Protecting Participants

Ethics are not optional. Violations can lead to:

• Legal consequences (e.g., fines for unauthorized data collection).
• Research invalidation (e.g., biased data from coerced participants).
• Harm to participants (e.g., retraumatizing survivors of abuse).

A. Core Ethical Principles

1. Autonomy: Participants must freely choose to participate.
   • How: Informed consent (written/oral), right to withdraw.
   • Example: A Daraz survey offering a discount for participation coerces responses.
2. Beneficence: Maximize benefits, minimize harm.
   • How: Debriefing (e.g., referring trauma survivors to counseling).
   • Example: If interviewing child laborers, provide meals/safe transport.
3. Justice: Fair selection and treatment of participants.
   • How: Avoid exploiting vulnerable groups (e.g., prisoners, orphans).
4. Non-maleficence: Do no harm.
   • Example: Avoiding questions like "Have you ever been arrested?" if it could jeopardize a participant’s job.

B. Ethical Challenges in Digital Data Collection

Worked Example: eSewa’s Data Ethics eSewa collects transaction data for research. To comply with ethics:

• Anonymizes user IDs (replaces with random codes).
• Limits data retention (deletes after 6 months).
• Offers opt-out: Users can request their data be deleted.

C. Special Cases in Nepal

• Caste/ethnicity: Avoid labeling participants by caste (use self-identified terms).
• Religion: Respect religious practices (e.g., pause interviews during prayer times).
• Rural areas: Provide transport/food if participants travel long distances.

4. Triangulation: Combining Methods for Rigor

No single method is perfect. Triangulation (using multiple methods) strengthens validity.

Example: Studying Child Labor in Nepal

Result: Surveys show 30% of children work; interviews reveal debt as the primary cause; observations confirm hazardous conditions.

## In the Real World

1. Khalti’s SMS Surveys

   • Method: Closed-ended Likert-scale questions via SMS.
   • Why it works: Reaches 80% of users in rural areas (where internet is limited). Example question:

     "How likely are you to recommend Khalti to a friend?" (1–5 scale).

   • Ethics: Anonymous; users opt in via a consent message.

2. Pathao’s Rider Safety Study

   • Methods: Observation (filming 100 rides) + interviews (50 riders).
   • Finding: 70% of accidents occurred at night due to poor lighting.
   • Outcome: Pathao added night-time route warnings and helmet reminders.

3. Nepal Police’s Traffic Experiment

   • Method: Quasi-experiment comparing fine rates before/after installing speed cameras at 10 intersections.
   • Result: Fines increased by 45%, but public complaints about "harassment" led to a review of camera placement.

4. UNICEF’s Child Marriage Research

   • Methods: Interviews with 200 married girls + surveys of 1,000 families.
   • Ethical twist: Provided cash incentives ($5) to participants to reduce coercion, but ensured this didn’t pressure families into interviews.

## Exam Tip

1. Method Selection Questions:

   • Examiners love justified choices. If asked "Which one method would you pick?", structure your answer like this:

     *"For a study on homelessness in Pokhara, I’d choose participant observation because:

     • Depth: Surveys miss nuances like shelter conditions.
     • Trust: Homeless individuals may lie to avoid stigma.
     • Ethics: Observing (not interviewing) reduces coercion. However, I’d triangulate with key informant interviews (e.g., NGO workers) to cross-validate findings."*

2. Ethics Short Answers:

   • Do NOT list principles without applying them. Example:

     *"In a study on teen pregnancy, I’d ensure:

     1. Anonymity: Use codes, not names.
     2. Debriefing: Provide helpline numbers for participants.
     3. Justice: Include teens from both urban/rural areas to avoid bias."*

3. Worked Examples:

   • Always tie to Nepal. Example for a survey question:

     *"To measure food insecurity in Sindhupalchowk, I’d ask:

     • 'How often did you skip meals last month?' (1) Never → (5) Always Why?:
     • Valid: Directly measures the construct.
     • Reliable: Easy to quantify.
     • Culturally adapted: Avoids jargon like 'malnutrition.'"*

4. Common Pitfalls to Avoid:

   • ❌ Saying "I’d use a survey because it’s easy." → Always justify (e.g., "It’s scalable for 1,000+ participants").
   • ❌ Ignoring ethics → Every method must address consent, anonymity, and potential harm.
   • ❌ Mixing qualitative/quantitative without explaining why (e.g., "I’d use both because surveys give numbers and interviews give stories").

Final Checklist for Full Marks: ✅ Define the method clearly. ✅ Link it to your research question. ✅ Discuss advantages/disadvantages. ✅ Address ethical considerations. ✅ Use a Nepali example (e.g., Ncell, Khalti, NGOs). ✅ Suggest improvements (e.g., "To reduce bias, I’d pilot-test questions with 10 participants first.").