---
title: "Calculating a Birth Date From Age at Death: How Far Can You Trust the Answer?"
description: "Subtracting an age at death from a death date gives you a precise-looking birth date. The arithmetic is the easy part. Here is what the linkage studies found about the ages written on death records..."
canonical: "https://kleiobase.com/blog/birth-date-from-age-at-death-how-accurate"
last_updated: "2026-10-28"
---

# Calculating a Birth Date From Age at Death: How Far Can You Trust the Answer?

Subtracting an age at death from a death date gives you a precise-looking birth date. The arithmetic is the easy part. Here is what the linkage studies found about the ages written on death records, why ages bunch on round numbers, and how wide a range to search.


The death certificate says your great-grandmother died on 3 March 1931, aged 78. You subtract, get 1853, type it into your tree, and start looking for a baptism in 1853. You do not find one. You try 1852 and find nothing there either.

The subtraction was not the problem. A birth year worked out from an age at death can only be as good as the age, and the age on a death record was nearly always given by someone other than the person who died, often after a lifetime of the person being vague about it themselves.

**The short answer.** An age in whole years puts the birth inside a one-year window. Someone who died on 15 June 1890 aged 72 was born between 16 June 1817 and 15 June 1818. The bigger uncertainty is whether the age itself was right. When demographers linked American death certificates from 1985 back to the same people in the censuses of 1900, 1910 and 1920, ages for older white Americans agreed closely, but ages for older African Americans were too young on average, and more so for women. Ages on round numbers deserve the most suspicion, because people who did not know their age reported 40, 50 or 70. Treat a calculated birth year as the centre of a search range, and confirm it against a record made near the birth.

## The arithmetic gives you a range, not a date

**Whole years leave a year of doubt.** "Aged 72" means the person had passed their 72nd birthday and not reached their 73rd. If they died on 15 June 1890, the 72nd birthday fell at some point between 16 June 1889 and 15 June 1890, so the birth fell between 16 June 1817 and 15 June 1818. A birth year of "1818" is right only about half the time. The other half, it is 1817.

**Years, months and days give a single date.** Some gravestones and burial registers give an age such as "72 years, 3 months and 14 days". Subtract the years, then the months, then the days, and you get one date, as long as the person who supplied the age counted correctly. If only the years and months are given, the answer is a range one month wide.

**A spreadsheet will get this wrong.** Subtracting 72 from 1890 gives 1818 and silently throws away the other possible year. Our free [birth date calculator](/tools/birth-date-calculator) gives the full range and writes it as a GEDCOM date such as `BET 16 JUN 1817 AND 15 JUN 1818`, so the doubt travels with the date instead of being dropped.

All of that assumes the age was correct.

## The age on a death record came from someone else

A person cannot report their own death. Somebody else gave the age, and that person was usually grieving, sometimes a neighbour or an in-law, and may have been working from what the dead person had told them years earlier.

We made the same point about [how accurate an obituary is](/blog/obituary-accuracy-genealogy-evidence) and about [the dates carved on a gravestone](/blog/how-accurate-are-gravestone-dates): the record is only as good as its informant. For ages, there is unusually good evidence about how good that informant was, because demographers needed to know.

## Two linkage studies measured how often death certificate ages were wrong

A team of American demographers wanted to know whether very old Americans were as old as their death certificates said. Death rates at the oldest ages depend on getting the ages right, so they checked them against earlier records for the same people.

**For older white Americans, the ages held up.** Mark Hill, Samuel Preston and Ira Rosenwaike, writing in *Demography* in 2000, linked a sample of death certificates for native-born white Americans who died aged 85 or over in 1985 to Social Security Administration records and to the same people in the US censuses of 1900, 1910 and 1920. They found the inconsistencies between the sources were minimal, even beyond age 95, with no systematic bias.

**For older African Americans, the ages were too young.** An earlier study by Samuel Preston, Irma Elo, Ira Rosenwaike and Mark Hill, published in *Demography* in 1996, did the same for death certificates of African Americans who died aged 65 or over in 1985. The ages at death were too young on average, and the errors were larger for women than for men. Even so, too many deaths were registered at ages 95 and over.

Those two papers are about deaths in one year in one country, and they did not measure a general error rate you can apply to your ancestor. Our reading of them, which is a judgement rather than their finding, is that the reliability of an age depends on whether the person ever had paperwork telling them what it was. Someone who was asked their age by censuses, schools and employers all their life, and could check it against a birth record, ends up with a consistent age. Someone who never had that record may never have known their exact age, and neither did whoever filled in the form after they died.

## Ages that end in 0 or 5 are the ones to doubt

People who do not know their exact age usually round it, and they round to the numbers that feel natural. Demographers call this age heaping.

**You can see it in the 1850 census.** A course blog on the history of the US census, published at Dartmouth College in 2015, tallied the last digit of every age in a 1 per cent sample of the 1850 census from the IPUMS project. It found a marked preference for ages ending in 0 among both men and women, a weaker preference for 5 and 2, and an avoidance of ages ending in 1, 7 and 9. The pattern faded over the following century.

**It is not only a historical problem.** The National Academies of Sciences, Engineering, and Medicine reported in *Assessing the 2020 Census: Final Report*, published in 2023, that age heaping was much more common in the 2020 Census than in 2010, by an estimated factor of at least 2.5 for ages 23 to 62. The panel traced it almost entirely to the follow-up visits made to households that had not responded, and especially to proxy respondents, people answering on behalf of someone else. A proxy respondent is the modern census's version of the informant on a death certificate.

**Historians use it as a measure of numeracy.** Brian A'Hearn, Jörg Baten and Dorothee Crayen, in "Quantifying Quantitative Literacy: Age Heaping and the History of Human Capital" in *The Journal of Economic History* in 2009, treated heaping on multiples of five as evidence of how many people in a population could handle numbers. They found it correlated robustly with literacy where both could be measured, and that western Europe had reached high numeracy by 1600, ahead of eastern Europe. If your ancestors were rural, unschooled or came from a region where few people could read, expect more rounding in their ages, on any record.

So a death certificate that says "aged 70" carries less information than one that says "aged 73". Seventy may be exact. It may also mean "about seventy", and the true age could sit several years either side.

## Some records rounded ages on purpose

Before you blame an informant, check whether the record itself asked for a rounded age.

**The 1841 British census rounded adults down.** The National Archives' research guide to the censuses of England and Wales says the 1841 census recorded ages rounded down to the nearest five years for anyone aged 15 or over. Scotland's 1841 census followed the same rule, and FamilySearch's guide to it gives the examples: someone of 32 would be listed as 30, and someone of 49 as 45. So "45" in 1841 means 45 to 49, not 45. The same guide notes that some enumerators ignored the instruction and wrote down the exact age they were given, so a 1841 age that is not a multiple of five was probably not rounded. From 1851 the census asked for age at last birthday.

If you calculate a birth year from an 1841 age without knowing this, you can be up to four years out before any human error has entered the picture. Our guide to [reading old census records](/blog/reading-old-census-records) covers how the American census changed what it asked about age from decade to decade.

## A pension could make an ancestor older overnight

Rounding is innocent. Sometimes there was a reason to give a different age.

**Ireland's old age pension rewarded being 70.** Under the Old Age Pensions Act 1908, pensions began in January 1909 for people aged 70 and over on a low income. The National Archives of Ireland explains the difficulty this caused. Civil registration of births in Ireland only began in 1864, so most claimants had no birth certificate. To prove their age, claimants gave their parents' names and where the family had lived, and officials searched the 1841 and 1851 census returns for them. The surviving forms from those searches, the census search forms, contain 136,152 names, and the Archives says 98 per cent of eligible people in Ireland applied, against 45 per cent in England and Wales.

**The 1911 census shows the effect.** John Budd and Timothy Guinnane linked the same people across the 1901 and 1911 Irish census returns, and published the results in *Population Studies* in 1991. They found three kinds of discrepancy: changes that sharply reduced age heaping, which suggests some people corrected a rounded 1901 age; ages pushed up to qualify for the pension before turning 70; and some exaggeration that had nothing to do with the pension at all.

So an Irish ancestor who aged more than ten years between the 1901 and 1911 returns may have been correcting a rounded age, claiming a pension, or both. If you are working from an Irish death record after 1909, remember that the person may have spent the last part of their life with an official age that they or their family had chosen. Our comparison of [free Irish record sources](/blog/free-irish-genealogy-records-vs-rootsireland) covers where to look for the baptism that settles it.

## How wide a range to search

The ranges below are our judgement, built on the studies above. Nobody publishes a standard table, and the right width depends on what else you know about the person.

| What you have | Range to search around the calculated year | Why |
| --- | --- | --- |
| Age in years, months and days | The single date, then a year either side | The informant may still have miscounted, but someone took care |
| Age in whole years, not a multiple of 5 | The two possible years, then two either side | The arithmetic window, plus ordinary informant error |
| Age on a multiple of 5 or 10 | Five years either side | Heaping: the age may be an estimate |
| Age in the 1841 British census, multiple of 5 | The stated age to four years older | The census rounded down by instruction |
| Very old age, 90 and over | Ten years either side, older and younger | Old ages are the ones most likely to be estimated |
| Irish age recorded after 1908 | Check against 1901 and the census search forms | The pension gave a reason to adjust |

**Prefer the earliest age you can find.** An age given in a census when the person was 20 was probably reported by a parent who was present at the birth. An age on a death certificate at 85 was reported by someone who was not. Where you have both, the younger record usually deserves more weight.

**Calculate from every record, not just the death.** Work out a birth range from each census, the marriage record and the death record, and see where they overlap. A person whose census ages imply 1843, 1844, 1843 and 1850 probably has one bad record, and it is usually the last one.

**Keep the calculated date marked as calculated.** If you write "1853" into your tree with no qualifier, the next person to use it, possibly you in five years, will treat it as a date someone recorded. Write "about 1853" or "between 1852 and 1853", and note which record it came from.

We could not find a study that measures how often a birth year calculated from an ordinary death record is wrong for the population as a whole. The two *Demography* papers come closest, but they covered only older Americans who died in 1985. Treat any single percentage you see quoted as a guess.

## What KleioBase does with a calculated birth date, and what it does not

We are a tool in this space, so read this as disclosure.

KleioBase is an AI research companion that turns scattered historical records into one connected family archive. That is useful here because the answer to "how old was she?" is almost never in one record. It is in the agreement, or the disagreement, between the census, the marriage, the death certificate and the gravestone.

When you upload and confirm those records, the person's age or date from each one is kept with the record it came from, and the profile's timeline shows how old the person was at each event, so a census age that drifts by five years stands out. A birth date can be stored as "About", "Estimated" or "Between" two dates, as the [dates and calendars guide](/docs/dates-and-calendars) explains, so a calculated range does not have to pretend to be a recorded date. If you add a second birth date from another record, KleioBase keeps both and raises a Same Life Event Recorded Twice [conflict](/docs/knowledge-base#conflicts) for you to decide. The free [birth date calculator](/tools/birth-date-calculator) needs no account.

There are three honest limits.

**It does not know which age was rounded.** It records the age each document states. Deciding that a "70" is probably an estimate, or that an 1841 age was rounded down, is your judgement.

**It does not weigh informants.** It keeps the record and the date, but it does not rank a death certificate below a baptism on its own. That ranking is the method described above, and you apply it.

**It does not add the year of doubt automatically.** If a record says "aged 72" and you confirm it as a birth year of 1818, that is what is stored. Use the calculator, or enter the date as a range, if you want the range kept.

## Where to start on the ancestor whose baptism you cannot find

Pick the person whose birth year came from their death record.

List every record that gives their age: each census, the marriage, the death, the gravestone. Calculate a birth range from each one, not just the death.

Mark any age that ends in 0 or 5, any 1841 British census age, and any age over 90. Widen your search for those.

Then search the baptism or birth registers across the overlap, starting with the year the earliest records agree on.

The clerks, enumerators and relatives who wrote these ages down wrote what they were told by people who had often never seen their own birth recorded, and who answered as best they could. A rounded age is often a sign that the person never had a birth certificate, and that tells you something about where their family lived.

Canonical: https://kleiobase.com/blog/birth-date-from-age-at-death-how-accurate