E-rolls (electoral rolls) – noisy but brutally useful

---

Tamil Ancestry Research

---

In this article:

1. What e-rolls (electoral rolls) actually are
2. Why e-rolls are powerful for family history (despite the mess)
3. What you really get from a voter list entry
4. Typical errors and distortions in e-rolls
5. How to pull structured data from e-roll PDFs and portals
6. Modelling e-roll data in TamizhConnect (household, address, age)
7. A minimal, sane workflow for using e-rolls without drowning

---

1. What e-rolls (electoral rolls) actually are

Forget the jargon. An electoral roll / e-roll is:

A government-maintained list of people eligible to vote in a given constituency, at a given time, usually grouped by house/door number.

Key points:

• Compiled and updated periodically (revisions, additions, deletions).
• Published as PDFs or searchable lists: India, Sri Lanka and many other places now have them online.
• Organised by:
  • polling station,
  • part / section,
  • “house number” / “building name” / “street”.

They are not:

• a full census,
• a clean genealogy,
• proof of citizenship for all purposes (though they get used like that),
• guaranteed accurate for age, spelling or relationships.

They’re just one repeated, semi-structured government snapshot of who lives where and is old enough to vote.

---

2. Why e-rolls are powerful for family history (despite the mess)

If you use e-rolls properly, you get:

• Addresses over time
  • Who was at which house in which year / revision.
• Approximate age
  • Year of birth or age at compilation for millions of people.
• Family clusters
  • Voters listed under the same house/door number are usually household members or close neighbours.
• Migration traces
  • Sudden disappearance from one constituency and appearance in another.
• Name variants
  • Different spellings, initials, and orderings that show up only in election data.

For many ordinary people, especially post-1970s:

• e-rolls will be the only recurring official record that tracks them as adults.
• passports, pattas, employment records may cover only a fraction.

If you ignore e-rolls in TamizhConnect, you’re throwing away:

• decades of address data,
• a cheap way to cross-check ages,
• and clear evidence of when branches left the native place.

---

3. What you really get from a voter list entry

Exact fields vary by country and year, but typical Indian-style e-roll entry has:

• Serial number (within part)
• Voter ID / EPIC number (not in earliest lists, but standard now)
• Name of voter
• Father’s / husband’s / mother’s name (relationship field)
• Sex / gender
• Age / Year of birth (sometimes only age, sometimes year-of-birth)
• House number / address fragment
• Part number / polling station
• Status markers in some revisions:
  • new addition, deleted, shifted, etc.

From this, for TamizhConnect you can extract:

• A person (with one name variant, one age-at-date, one relationship field).
• A household cluster (who else shares the house number).
• An approximate location at a concrete point in time.
• A government-issued voter ID you can use to link later e-rolls.

That’s plenty, if you don’t treat it like the Bible.

---

4. Typical errors and distortions in e-rolls

Stop assuming “government list = clean data”. It isn’t.

Common problems:

4.1. Age lies and approximations

• Ages are often:
  • rounded (30, 35, 40),
  • copied from earlier revisions with no update,
  • just guessed during revision.
• Year-of-birth fields can be off by several years.

So:

• use e-roll age as approximate,
• cross-check with other documents,
• don’t build exact birthdates from it unless corroborated.

4.2. Name spelling chaos

• Muthusamy, Muthuswamy, MUTUSAMI, MUTHU SAMI all for same person.
• Initials misplaced or expanded randomly.
• Women often under husband’s name or with shortened/pet forms.

You have to treat these as name variants, not separate individuals, and match via:

• house number,
• relationship field,
• neighbours,
• other records.

4.3. Out-of-date deletions and additions

• Deceased voters may stay on rolls for years.
• Migrants may show up in both old and new places until revisions catch up.
• Some eligible people never get added at all.

So a name appearing in a 2010 e-roll doesn’t guarantee they were alive and present on the exact date – it just means the system hadn’t removed them.

4.4. Mis-assigned house numbers

• Revise squads guess or copy house numbers incorrectly.
• Entire hamlets get odd numbering patterns.
• Building name changes; rolls lag or use old name.

Treat house number as a strong hint, not absolute truth.

---

5. How to pull structured data from e-roll PDFs and portals

If you’re serious, you don’t just screenshot e-roll pages. You actually extract.

5.1. Identify relevant parts

For each known family village/town/city:

• find the Assembly / Parliamentary constituency,
• list likely polling stations (streets, wards, villages),
• download e-roll PDFs for:
  • multiple revisions across years (e.g. 2004, 2010, 2015, 2020).

5.2. Manual vs semi-automatic extraction

You have two options:

• Manual (slow, safer for small data):

  • read relevant pages,
  • type entries into a spreadsheet:
    • name, relation name, age, sex, house number, part, year, voter ID.

• Semi-automatic (for larger sets):

  • run OCR on PDFs,
  • clean up output,
  • standardise column headers,
  • then manually correct obvious errors for known families.

Either way, your goal is a table like:

| Year | Part | Serial | Name | RelationName | RelationType | Age | Sex | HouseNo | VoterID |

Once it’s in that shape, you can import into TamizhConnect / your data model.

5.3. Respect privacy and law

Basic sanity:

• Don’t dump full modern e-rolls publicly; they contain living people’s data.
• Inside your family system, limit access to sensitive details if needed (voter IDs, exact addresses for living people).
• Use e-rolls for genealogy and history, not surveillance.

---

6. Modelling e-roll data in TamizhConnect (household, address, age)

You want three main things from each e-roll entry:

1. Person evidence
2. Household cluster at a given time
3. Address history

6.1. Person evidence: “this human existed here in year X”

Per e-roll entry, create a PersonEvidence object linked to a person profile:

• sourceType: "e-roll"
• year: e.g. 2011
• constituency: "X Assembly Constituency"
• partNumber: "Part 123"
• serialNumber: "456"
• nameAsWritten: "R. MUTHUSAMY"
• relationNameAsWritten: "Ramasamy"
• relationType: "father" | "husband" | "mother" | "other/unknown"
• ageRecorded: 45
• sexRecorded: "M" / "F" / other
• houseNoRecorded: "12A"
• voterId: "ABC1234567" (if present)
• confidence: "high" | "medium" | "low" match to your person.

This lives as a source for that person, not as the person itself.

6.2. Household cluster: who shared the address

For each unique combo of:

• year,
• partNumber,
• houseNoRecorded,

group the entries. That’s your voter-household:

• householdId: internal
• year: e.g. 2011
• addressLabelRaw: e.g. "12A, East Street"
• personLinks[]: list of e-roll persons at that address.

In TamizhConnect, you can:

• show who was together at that time,
• track which siblings/relatives split from which household,

and map this onto your internal household model.

6.3. Address timeline

From multiple years:

• 2004: Person X at Old Village
• 2011: Person X at Town Y
• 2019: Person X not on roll; children appear in City Z.

Construct an AddressHistory for each person:

• addressEvents[]:
  • year: 2004, place: Old Village, source: "e-roll 2004"
  • year: 2011, place: Town Y, source: "e-roll 2011"
  • year: 2019, place: City Z, source: "e-roll 2019 (children)", note: “Parent likely migrated/died.”

Don’t guess exact move years – use ranges:

• “between 2004 and 2011”.

---

7. A minimal, sane workflow for using e-rolls without drowning

Here’s a straightforward approach that doesn’t require you to become an election-office archivist.

Step 1: Pick 1–2 core locations

Start with:

• native village / ooru,
• main city area where your branch concentrated.

Download 2–3 e-roll years for those places (spaced 5–10 years apart).

Step 2: Extract only known family names first

Instead of indexing entire rolls:

• search within PDFs (if possible) for your surnames / stems,
• or manually scan where you know the street/ward.
• Pull out rows that obviously belong to your people.

Enter into a spreadsheet with:

• Year, Part, Serial, Name, RelationName, RelationType, Age, Sex, HouseNo, VoterID.

Step 3: Map to existing profiles in TamizhConnect

For each row:

• match to a person profile based on:
  • name pattern,
  • father/husband name,
  • approximate age,
  • known address.

If there’s ambiguity:

• flag as unresolved; don’t force it.

Add each as a source record to that person.

Step 4: Build address and household snapshots

From that extracted subset:

• group by year + house number to form voter-households,
• see which relatives appear together,
• note address changes across years.

Overlay on your migration timeline:

• “This is roughly when this branch shifted from ooru to town / city.”

Step 5: Use e-rolls to check myths

Once you have a small sample in TamizhConnect, use it to challenge stories like:

• “We only came to Chennai after 2000.” (But roll shows them in 1993.)
• “So-and-so never left the village.” (But roll says otherwise.)
• “X never lived separately from parents.” (Multiple addresses say they did.)

Log corrections as notes and timeline updates.
Don’t cling to stories when the data contradicts them.

---

If you treat e-rolls as “just boring government PDFs”, you’re ignoring a cheap, dense timeline of your family’s adult lives.

If you treat them as what they actually are:

• frequent, messy, honest snapshots of who was old enough to vote, living roughly where, with whom,

and model them properly in TamizhConnect, they become one of the sharpest tools you have for:

• pinning addresses to years,
• tracking migration step-by-step,
• and catching gaps in your oral narratives long before they become permanent myths.