Most important issue
IssuesIn your view, what is the single most important problem facing people in your area today? (Open, then coded.)
Guide · Election research
For research teams, media partners and campaign analysts
This guide explains how to conduct election survey in India the way field researchers do it: define the question, draw a sample across booths, write a neutral questionnaire in the local language, interview face to face, verify every record, and report estimates with their margin of error. It also covers the Election Commission rules that limit when exit poll results can be published.
- 8 numbered steps
- Sample size worked example
- Questionnaire outline you can adapt
- Legal and ethical limits
Booth-wise samplingFace-to-face interviewsHindi and regional languagesGPS and time stampsBack-checksMargin of errorSection 126A awarenessNo voter-roll import
Quick answer
An election survey interviews a random, representative sample of adults across a constituency or state, asks a short neutral questionnaire about issues, leaders and voting intention, and weights results to the population. Its value comes from sampling and verification, not from the number of interviews alone. It measures opinion at a point in time and does not predict results.
Opinion poll, tracker or exit poll
An opinion poll is a one-time reading before polling day. A tracker repeats the same questions over weeks to see movement. An exit poll interviews people leaving polling stations on the day. Each needs a different sample and a different schedule, so decide which one you are running before you write a single question.
Rules to know before fieldwork
Section 126A of the Representation of the People Act, 1951 prohibits conducting an exit poll and publishing its result during the period the Election Commission of India notifies, which typically runs from the start of the first phase until half an hour after the last phase closes. Election Commission notifications for each election set the exact window, so read the current one.
Decide first
- Geography: seat, region or state
- Type: opinion poll, tracker or exit poll
- Publication plan, if any
- Languages of fieldwork
- Report date
The process
These steps are the same whether you cover one assembly seat or a whole state. The scale changes; the discipline does not.
- 01
1. Write the research questions
List the decisions the survey must inform, such as issue priorities by region or leader familiarity. Cut every question that does not serve one of them.Research leadShift onHello, Ram KumarToday's workAssignmentGram Panchayat BarauliTarget 40 homes14 done15 done16 done17 doneShiftOn · 2:14 hSync6 phone · 8 serverToday's homes- Home 15 · RamlalDone
- Home 16 · SunitaDone
- Home 17 · LockedRevisit
- Home 18 · MohanFilling…
✓ #BS-0413 synced to serverStart new survey - 02
2. Build the sampling frame
Use public polling station lists and census geography to select booths at random, weighted by electorate, then choose households at random within each booth area.Statistician123456789101112131415161718192021222324252627282930A strongB fairC weakD very weakNo report today - 03
3. Draft and translate the questionnaire
Keep it under 15 minutes. Translate into the spoken language, then back-translate to catch shifted meanings.Research leadSurvey 3/6 · AmenitiesOfflineHow is drinking water in your area?- Comes regularly
- Comes sometimes
- Rarely / never
- Don't know
✓ Section saved on phone · autosaveBackNext - 04
4. Pilot in two or three booths
Watch where respondents hesitate, where interviews run long and which options nobody picks. Fix, then freeze the form.Field supervisorHome #17 · Intro & consentHello, I am Ram Kumar. This survey is run by Sample Organisation. Purpose: to learn about amenities in your area. It is voluntary, and you can say no at any time.
Consent v2 · 09:44Agreed — start surveyDeclined - 05
5. Train and deploy interviewers
Train on neutral reading of questions, consent, and the household selection rule. Assign booths so each interviewer covers a manageable daily quota.Field managerGram Panchayat BarauliShiftLocation will be sharedOnly during your shift, with your supervisor. Stops at ‘End shift’.Start shiftArrival check26.8467 N · 80.9462 E · ±12 mApproximately in area - 06
6. Interview with verification
Each record carries GPS location, start and end time and duration. Records outside the assigned area or implausibly short are flagged the same day.InterviewersLocation evidenceSurvey #BS-0412- In areaInside approx. boundary
- Accuracy±12 m
- Time09:42 · fresh fix
- Mock GPSNot detected
- Interview length11 min
Added to accepted count - 07
7. Back-check and clean
Supervisors revisit or call back a share of respondents to confirm the interview happened. Remove records that fail checks and document why.Quality teamOutside area · 14.2 km away · in review - 08
8. Weight, analyse and report
Weight by age, gender and region to match census or electoral totals. Report estimates with the margin of error and the fieldwork dates.Analyst3,000Target homes2,184Visits1,962Completed1,871AcceptedDaily accepted surveysLast 12 daysQualityAccepted1871In review64Outside area19Weak signal8Top issues · by panchayatWater62%Roads48%Power31%Drains24%
Hello, Ram Kumar
Today's work
Assignment
Gram Panchayat Barauli
Target 40 homes14 done15 done16 done17 done
Shift
On · 2:14 h
Sync
6 phone · 8 server
Today's homes
- Home 15 · RamlalDone
- Home 16 · SunitaDone
- Home 17 · LockedRevisit
- Home 18 · MohanFilling…
✓ #BS-0413 synced to server
Start new survey1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
A strongB fairC weakD very weakNo report today
Survey 3/6 · AmenitiesOffline
How is drinking water in your area?
- Comes regularly
- Comes sometimes
- Rarely / never
- Don't know
✓ Section saved on phone · autosave
BackNext
Home #17 · Intro & consent
Hello, I am Ram Kumar. This survey is run by Sample Organisation. Purpose: to learn about amenities in your area. It is voluntary, and you can say no at any time.
Consent v2 · 09:44
Agreed — start surveyDeclined
Gram Panchayat Barauli
Shift
Location will be shared
Only during your shift, with your supervisor. Stops at ‘End shift’.
Arrival check
26.8467 N · 80.9462 E · ±12 m
Approximately in area
Location evidence
Survey #BS-0412
- In areaInside approx. boundary
- Accuracy±12 m
- Time09:42 · fresh fix
- Mock GPSNot detected
- Interview length11 min
Added to accepted count
Outside area · 14.2 km away · in review
3,000
Target homes
2,184
Visits
1,962
Completed
1,871
Accepted
Daily accepted surveysLast 12 days
Quality
Accepted1871
In review64
Outside area19
Weak signal8
Top issues · by panchayat
Water62%
Roads48%
Power31%
Drains24%
Sampling
For a simple random sample, about 1,070 interviews give a margin of error near plus or minus 3 percentage points at 95% confidence for a 50% estimate. Booth clustering raises the needed number, often by 1.5 to 2 times. Every region you want to report separately needs its own adequate sample.
The worked calculation
The usual starting formula is n = z² × p × (1 − p) ÷ e². With z = 1.96 for 95% confidence, p = 0.5 as the most cautious assumption and e = 0.03, n = 3.8416 × 0.25 ÷ 0.0009, which is about 1,067. At e = 0.05 the same formula gives about 384.
- ±5 points: about 384 interviews before design effect
- ±4 points: about 600
- ±3 points: about 1,067
- ±2 points: about 2,401
Why clustering changes the number
Interviews in the same booth area are more alike than interviews spread across a seat. That similarity is captured by the design effect. If you assume a design effect of 1.5, the ±3 point sample becomes roughly 1,600. More booths with fewer interviews each keeps the design effect lower than a few booths with many interviews.
Sub-groups need their own samples
A state estimate of ±3 points says nothing reliable about one district inside it. If the client wants results for each region, age band or urban and rural split, size each of those groups separately. This is the most common reason a survey budget grows after the first conversation.
Inputs to agree
- Target margin of error
- Confidence level
- Sub-groups to report
- Assumed design effect
- Expected non-response
Fieldwork
Booth-level work means sampling polling station areas and interviewing households inside them. It does not mean using the voter list to find named people, and it never means tracking how individuals vote.
Select booths, then households
Draw booths with probability proportional to their electorate from the public polling station list. Inside each booth area, start at a random landmark and follow a fixed interval rule, such as every fifth house, so interviewers cannot pick friendly doors.
- Fixed start point per booth
- Interval rule written on the brief
- Replacement rule for locked houses
Interviewer at a household during a booth level election survey
Balance gender and age
Men are more often at the door in daytime. Use a respondent selection rule within the household, and schedule evening visits so women and working adults are not under-represented. Track the gender and age mix per booth every evening.
Home #17 · Visit outcome
- Complete
- Partial
- Declined
- No contact
- Home locked
- Revisit
Verify location and duration
An offline Android survey app records GPS and time for each interview even without signal and syncs later. Supervisors can see a booth's interviews on a map and query any that sit far from the assigned area.
- Works without network
- Syncs when signal returns
- Flags short interviews
Ram Kumar · shift on
26.8467 N · 80.9462 E · ±12 m
14
Received
12
Accepted
1
Review
1
Outside
Our view on survey numbers
Template
An outline you can adapt. Keep the order: neutral questions first, voting intention later, so earlier answers do not lead later ones.
01Screening and consentA
Screening and consent
Confirm the respondent is 18 or older and lives in the area. Read a short consent statement explaining who is conducting the survey, that answers are confidential and that they may stop at any time.
- Age and residence check
- Consent recorded in the app
02Local issuesB
Local issues
Open with unprompted issues: what is the biggest problem in your area? Then rate specific services such as water, roads, power, health and jobs on a simple scale.
03Government performanceC
Government performance
Satisfaction with the state and central governments and with the local representative, each asked on the same five-point scale for comparability.
04Leaders and familiarityD
Leaders and familiarity
Awareness of named leaders first, then favourability only among those aware. Rotate the order of names across interviews to reduce order bias.
05Voting intentionE
Voting intention
Likelihood of voting, then intended choice, with a do not know and a refuse option. Never pressure for an answer.
06DemographicsF
Demographics
Age, gender, education, occupation and household type at the end, asked in a neutral way. Community or caste questions, if included, should be optional, justified by the research purpose and handled under your data protection policy.
Question bank
Examples written to be neutral. Replace placeholders with your region's issues and translate carefully.
Service satisfaction
ServicesHow satisfied are you with drinking water supply in your area: very satisfied, somewhat satisfied, somewhat dissatisfied or very dissatisfied?
Direction of the area
MoodCompared with five years ago, would you say conditions in your area have improved, stayed the same or got worse?
Likelihood of voting
TurnoutHow likely are you to vote in the coming election: certain, likely, unlikely, or will not vote?
Deciding factor
ChoiceWhen you decide whom to vote for, what matters most: the candidate, the party, the leader, local work done or something else?
Information source
MediaWhere do you mostly get news about politics: television, newspaper, mobile phone, people around you or other?
Mistakes
| Mistake | What it does to results | Fix |
|---|---|---|
| Interviewing at markets and tea stalls | Over-represents men and people with free time | Household sampling with a selection rule |
| Leading question wording | Pushes answers toward one side | Neutral wording, pilot, back-translation |
| Asking voting intention first | Colours every later answer | Issues first, intention near the end |
| No location or time data | Fabricated interviews go unnoticed | GPS, timestamps and back-checks |
| Reporting sub-groups with tiny samples | Large hidden error | Size sub-groups or do not report them |
| Publishing exit poll data in the notified window | Legal violation under Section 126A | Check the ECI notification before release |
Not the right fit if
- Import voter-roll personal data or build lists of named voters
- Track how any individual votes
- Promise a result or a seat count
- Publish exit poll findings inside the ECI-notified window
A good fit if
- Neutral opinion polls and issue trackers
- Booth-sampled household fieldwork with verification
- Hindi and regional-language questionnaires
- Transparent methodology notes with margin of error
Political survey solutions
How we support opinion research and booth-level fieldwork for campaigns and researchers.
Learn moreराजनीतिक सर्वे (Hindi)
The same political survey solutions page in Hindi.
Learn moreOpinion poll survey
Sample design, fieldwork and reporting for pre-election opinion polls.
Learn moreMedia poll partner
Field and methodology support for news organisations running polls.
Learn moreExit poll survey
Polling-day fieldwork planned around the Election Commission's notified restrictions.
Learn moreFAQ
Straight answers to the questions people ask most about this topic.
Is it legal to conduct an election survey in India?
Opinion surveys are conducted regularly in India. The main statutory restriction is Section 126A of the Representation of the People Act, 1951, which bars conducting exit polls and publishing their results during the period notified by the Election Commission. Check the current ECI notification and any Model Code of Conduct guidance before fieldwork or publication.
How many interviews does an assembly constituency need?
It depends on the margin of error you want and whether you need separate results for parts of the seat. Around 1,070 interviews give roughly ±3 points under simple random sampling, and booth clustering usually pushes that higher. Many teams accept a wider margin for a single seat and say so in the report.
Can a survey predict who will win?
No survey can promise that. A survey estimates vote intention among a sample at a specific time, with a margin of error. Turnout, late decisions and the way votes convert to seats all add uncertainty. Good reports state fieldwork dates, sample size and margin of error, and avoid seat predictions presented as certainties.
Do you use the voter list to select respondents?
We do not import voter-roll personal data. Public polling station lists can guide which booth areas to sample, and households inside those areas are selected by a random rule. Respondents are interviewed with consent, and the survey never records how a named individual has voted.
How long does an election survey take?
A single-seat survey can move from questionnaire to report in a couple of weeks once the form is agreed, while a multi-district study takes longer because of translation, piloting and interviewer training. The timeline depends on geography, languages and sample size, so we set it after seeing the brief.
India Nationwide, Hindi-first
Share the geography, languages and the decisions the survey must inform. We will reply with a sample design note and field plan.