Sampling, Measurement & Data Collection: Methods, Tools & Ethics

TAKEAWAYS:

• Sampling determines who/what you study: probability methods (random, stratified) ensure representativeness, while non-probability (purposive, snowball) target specific groups.
• Measurement scales (nominal, ordinal, interval, ratio) dictate how data can be analyzed—ratio data (e.g., income) allows advanced statistics, while nominal (e.g., gender) only permits frequencies.
• Data collection tools must match the research question: structured surveys for large samples, in-depth interviews for qualitative insights, and participant observation for cultural contexts.
• Ethics in data collection include informed consent, anonymity, and avoiding harm—critical for vulnerable populations like child laborers or survivors of gender-based violence.
• Pilot testing tools (e.g., a survey or interview guide) before full deployment catches flaws and improves reliability.
• Digital tools (e.g., Khalti’s transaction data or Pathao’s driver surveys) now dominate data collection, requiring IRB approval and cybersecurity safeguards.

---

1. Sampling: Selecting Your Study Population

Sampling is the process of choosing a subset of a population to represent the whole. Poor sampling = biased results. In social work, you might study child labor in Kathmandu but can’t survey every child—so you sample.

Key Sampling Methods

Real-World Example: eSewa’s User Satisfaction Survey

Problem: eSewa wants to know why some users abandon transactions mid-process. Solution: They use stratified sampling to ensure representation across:

• Age groups (18–30, 31–50, 50+),
• Transaction types (bill payments, remittances, top-ups),
• Geographic regions (Kathmandu Valley, hills, Terai). Tool: Online survey with systematic sampling (every 100th user who logs in on a given day). Ethics: Anonymous responses, no personal data collected beyond email for incentives.

---

2. Measurement: How to Quantify Social Work Data

Measurement turns abstract concepts (e.g., "poverty," "social support") into data. The scale of measurement determines what statistical tests you can use.

Levels of Measurement

Worked Example: Measuring "Social Support" in Elderly Care

Concept: Social support is abstract—how do you measure it? Tool: Oslo-3 Social Support Scale (ordinal data):

• Items: "How many people can you count on to help you if you are sick?" (0=none, 1=1 person, 2=2–5 people, 3=5+ people).
• Analysis: Median social support score in senior citizen homes in Lalitpur. Why Ordinal?: The intervals between "1 person" and "2–5 people" aren’t mathematically equal, but the order matters.

---

3. Data Collection Tools: Choosing the Right Method

The tool must match the research question, population, and resources. Common tools:

A. Surveys/Questionnaires

• Structured: Fixed questions, closed-ended (e.g., Likert scales).
  • Example: Nepal Demographic and Health Survey (NDHS) uses structured questions like: "In the last 12 months, how many times did you experience domestic violence?" (0–5+).
• Semi-structured: Fixed questions + open-ended probes.
  • Example: Interviewing female-headed households in Sindhupalchowk about resilience after earthquakes.
• Unstructured: Open-ended, exploratory.
  • Example: Qualitative study on "youth perceptions of mental health" in TU.

Advantages:

• Standardized, easy to analyze quantitatively.
• Can reach large samples (e.g., NTC’s customer satisfaction surveys).

Disadvantages:

• Low response rates (e.g., only 30% of Pathao drivers complete surveys).
• May miss nuanced cultural contexts.

B. Interviews

• Structured: Same questions in same order (like a survey but verbal).
  • Example: Ncell’s customer service training evaluation—interviewing 50 agents on stress levels.
• Semi-structured: Guided conversation with probes.
  • Example: Interviewing child laborers in brick kilns about working conditions.
• Unstructured: Conversational, emergent themes.
  • Example: Studying "how Dalit women negotiate caste discrimination" in daily life.

Advantages:

• Deeper insights into why people behave a certain way.
• Can clarify ambiguous survey responses.

Disadvantages:

• Time-consuming, expensive.
• Interviewer bias possible (e.g., social desirability bias—respondents lie to appear "good").

C. Observations

• Participant observation: Researcher joins the group (e.g., living with a family to study child marriage dynamics).
• Non-participant observation: Researcher watches from outside (e.g., observing traffic police corruption at busy intersections).
• Structured observation: Checklist of behaviors (e.g., counting how many times a day a home-based care worker checks a patient’s blood pressure).

Example: Nepal Police’s internal study on gender-based violence response times by observing 999 calls in Kathmandu.

Advantages:

• Captures real behaviors, not just self-reported data.
• Useful for hidden or sensitive topics (e.g., drug use, domestic abuse).

Disadvantages:

• Reactivity: People may change behavior if observed (e.g., Hawthorne effect—workers perform better when watched).
• Ethical concerns (e.g., informed consent for participant observation).

D. Secondary Data

Using existing data (e.g., NPC census reports, UNICEF child welfare data, bank transaction records).

• Example: Analyzing Khalti’s transaction data to study remittance patterns from Gulf countries.

Advantages:

• Saves time/money.
• Large, reliable datasets (e.g., Nepal’s Civil Registration System).

Disadvantages:

• May not fit your research question exactly.
• Ethical issues: Some data is confidential (e.g., Ncell’s customer call logs).

---

4. Data Collection in Digital Platforms

Nepal’s tech boom (eSewa, Khalti, Daraz, Pathao) creates new data sources—but raises ethical challenges.

Worked Example: Pathao Driver Stress Study Research Question: How does gig economy work affect mental health? Data Source:

• Primary: Surveys of 200 Pathao drivers (semi-structured interviews on app-based stress).
• Secondary: Pathao’s internal data on trip cancellations (proxy for driver frustration). Findings:
• Drivers in Kathmandu’s busy areas (e.g., Thapathali) had 30% higher cancellation rates than in Pokhara.
• Qualitative theme: "The app never stops—even when I’m sick, it keeps pinging for rides."

---

5. Ethics in Data Collection

Social work research prioritizes do no harm. Key ethical principles:

A. Informed Consent

• Participants must know:
  • Purpose of the study.
  • How data will be used/stored.
  • Right to withdraw.
• Example: Before interviewing child laborers, explain: "This study helps NGOs push for better laws. Your name won’t be used, and you can stop anytime."

B. Anonymity vs. Confidentiality

• Anonymity: No one (including researchers) can link data to participants.
  • Example: Coding survey responses as P001, P002 instead of names.
• Confidentiality: Researchers know identities but won’t disclose them.
  • Example: Storing Ncell call logs with only researcher access.

C. Avoiding Harm

• Physical harm: Avoid dangerous fieldwork (e.g., observing drug dens without safety protocols).
• Psychological harm: Trauma triggers in interviews (e.g., asking survivors of incest about abuse).
• Social harm: Stigma (e.g., labeling sex workers in reports).

Example: Red Cross’s earthquake survivor study in Gorkha:

• Risk: Retraumatizing victims by asking about losses.
• Solution: Trained social workers, debriefing sessions, and referral to counseling.

D. Plagiarism and Data Fabrication

• Plagiarism: Copying others’ data or ideas without citation.
  • Example: Using UNICEF’s child malnutrition report without permission or proper referencing.
• Fabrication: Making up data (e.g., inflating survey responses to meet donor expectations).
• Solution: Always cite sources, use raw data, and audit trails (documenting data collection steps).

---

6. Practical Steps: Designing Your Data Collection Plan

Use this checklist to avoid common mistakes:

1. Define your population (e.g., "migrant workers in Malaysia").
2. Choose a sampling method (e.g., snowball sampling for hard-to-reach groups).
3. Select measurement scales (e.g., ordinal for "sense of belonging").
4. Pick tools (e.g., semi-structured interviews for qualitative depth).
5. Pilot test (e.g., try your survey on 5 social work students first).
6. Address ethics:
   • Get IRB approval (Institutional Review Board).
   • Ensure informed consent.
   • Anonymize data.
7. Train data collectors (e.g., NGO volunteers interviewing slum dwellers).
8. Plan for digital tools (e.g., Google Forms for surveys, audio recording for interviews).

---

In the Real World

1. eSewa’s Financial Inclusion Study

   • Idea Used: Stratified sampling (urban vs. rural users) + secondary data analysis (transaction logs).
   • How: eSewa partnered with Central Bank of Nepal to study how women in remote districts use digital payments. Found that illiteracy was a bigger barrier than internet access—leading to SMS-based payment tutorials.

2. Pathao’s Driver Wellbeing Initiative

   • Idea Used: Mixed methods (surveys + app data) + participant observation (riding with drivers).
   • How: Pathao’s internal team used systematic sampling of 1,000 drivers to find 30% reported sleep deprivation. They then introduced flexible shift options and mental health hotlines.

3. Nepal Police’s Domestic Violence Hotline Evaluation

   • Idea Used: Non-participant observation (call center monitoring) + structured interviews with survivors.
   • How: Researchers recorded 999 call durations and found calls from rural areas took 2x longer to connect. This led to regional call center expansions.

---

Exam Tip

This unit is heavily tested in TU/PU exams with:

1. Short notes (e.g., "Levels of measurement," "Snowball sampling").
2. Case studies (e.g., "How would you sample for a study on child marriage in Kavrepalanchowk?").
3. Tool comparisons (e.g., "When would you use an interview vs. a survey?").
4. Ethical dilemmas (e.g., "A participant in your study on drug addiction wants to withdraw—how do you handle it?").

High-scoring strategies:

• Always link theory to Nepal: "In Nepal, stratified sampling is ideal for studying caste-based disparities because..."
• Use real examples: "Like eSewa’s transaction data, secondary data can reveal..."
• Compare methods: "Surveys are better for large samples, but interviews uncover..."
• Ethics is non-negotiable: Any answer without consent, anonymity, or harm reduction loses marks.

Common pitfalls to avoid:

• Confusing ordinal (ordered categories) with interval (equal intervals).
• Forgetting pilot testing in data collection plans.
• Ignoring cultural sensitivity (e.g., asking personal questions in group interviews).