Skip to main content

Are census records reliable for genealogy research?

Over the last six months, this column has discussed the “Anatomy” of the federal census for the years from 1870-1930, not including the mostly-destroyed 1890 census. In this series of articles, each column of the census questionnaire was examined, and clues that will aid your research were discussed. If you have missed any of these articles, you can read them again using the links below:
Anatomy of the 1910 federal census
In a discussion concerning the federal census as evidence for genealogical research, however, one must also consider the question: are census records reliable sources?
First, we must understand the reason that the federal census is taken each year. The U. S. Constitution ordered a decennial (every ten years) enumeration of the population, in order to determine congressional representation. What this means is that the original purpose was, and still is, statistical by nature. The census was never designed as a source of information on individual people per seThe First Census, taken in 1790, reflected this, containing only five columns: free white males over sixteen (read: voters), free white males under sixteen (read: future voters), free white females (read: nonvoters but still counted), other free people (read: nonvoters but still counted), and slaves (read: 3/5 of a person). This is all that mattered to them at that time. The name of the head of household was listed, so that the enumerator could be sure he didn’t miss anyone or count anyone twice. Each subsequent census provided a little bit more information, as Congress realized the potential of the census to gather other statistical information regarding people of foreign birth, occupations, etc. The 1850 federal census was the first census to provide the names and exact ages, and other information, on each individual person resident in the United States. Slaves were enumerated on a separate schedule, “Schedule 2,” which is commonly called the “slave schedule.”
Secondly, we must understand how the census was taken each year. An enumerator was named in each district, usually a somewhat prominent citizen who was familiar with both the geography and the population of the district as a whole.  This enumerator would go from house to house and inquire about the make-up of the household. Generally, he would try to ask the head of household, if he was available, but may have asked whoever answered the door. If no one was home, he might either fudge the answers, if he knew the residents, or ask a neighbor. However, there was no space to report who provided the information. Remember, the purpose of the census was statistical only, so “close enough” was good enough for them, in many cases. Luckily, the enumerator did usually make an attempt to accurately record the information reported to him, and this can be assumed to be a member of the household in the majority of cases, though there are some notable exceptions.
The BCG Standards Manual prescribes the following considerations to determine the reliability of any source of primary information. Primary information is that provided by someone with direct knowledge of the event.
  • how close in time and place to the event the record was created,
  • how involved in the event the eyewitness was,
  • the age and sanity of the eyewitness and consequent extent of his/her understanding of the event’s significance and details
  • any bias … that might have affected the account.
Do federal census records meet these standards? Simply put, no. Because we cannot in any case identify the informant of a census entry, we cannot evaluate its reliability. There is no way to determine whether any given article of information is primary or secondary.
This is not, however, as many genealogists believe, to say that a census is unreliable. There is a difference between “unreliable” and “unknown reliability.” Federal census records are of “unknown reliability.” In many, if not most, cases, the information was provided by at least a member of the household who would have had direct knowledge of much of the information. Again, in many, if not most, cases, the information was provided by the head of household.
Imagine this scenario: one summer day in 1850, a government official (the census taker), who may be known to the household, comes to the door. A teenage son or daughter answers the door. Greetings may have been exchanged, but it would not have been proper for the government official to have begun questioning that teenager. Instead, he would have inquired, “Is your father at home?” If the father was not present, the mother would have been the next choice.
But there are exceptions to the rule, and this is why the federal census can only be considered of unknown reliability by themselves.
How do we, as genealogical researchers, get around this questionability? The Genealogical Proof Standard, which has been discussed in this column on several occasions, prescribes a fool-proof method: perform a “reasonably exhaustive search.” Nearly every fact in the federal census can be confirmed or denied with other records, even if only the previous or later federal census records. While later federal census records would be no more reliable on their own, independent corroborating evidence does indeed lend itself to proving the reliability of a document where the reliability is otherwise unknown. Vital records, church records, land records, probate records, among many others, can all be used to verify the information provided in a census entry. If all verifiable facts prove themselves accurate with independent corroborating evidence, then the reliability of that census record improves. It is indeed possible, through research, to judge a federal census record as reliable information.

This article has been posted on November 3, in response to the NaBloPoMo challenge of posting one article every day throughout the month of November.


  • Happy Dae 5 years ago

    Another reason, for the early census at least, was to determine an approximate number of males for military service. Whether the British, the French or any other adversary, the U.S.A. needed to know what their potential troop strength would be.

    Happy Dae·

  • Robyn 5 years ago

    Excellent article. My class is always surprised when I explain that census records are always---always---secondary information.

  • Michael Hait 5 years ago

    Robyn - That is a little misleading though. Census records are _not_ always secondary information. This would imply that the information was never provided by someone with direct/eyewitness knowledge to the fact. Obviously, this is not the case. The problem, however, is that we, as interpreters, cannot identify the source of the information, and therefore do not know whether it is primary or secondary. As such, we must assume the worst-case scenario: that it is secondary.

  • Brett Payne 5 years ago

    In my opinion, we must separate the two, quite different types of information existing in a census record, and this is not just true for the US Federal Census, but those taken in other countries too.

    I will leave aside the matter of whether the census sheet is an original schedule prepared by the enumerator or a copy made subsequently, although this is, of course, an important issue.

    First, the census is defined in my Concise OED as an "Official numbering of population with various statistics." The fact of whether a particular person was residing in a particular household on census night is, to me, the primary record. That individual's attributes, i.e. name, age, sex, marital status, etc., then form the secondary part of each record.

    The person's name is usually the means by which the individual in a census record is most clearly identifiable. It is also, of course, the attribute for which the census enumerator was most reliant on the informant/head of household (contd).

  • Brett Payne 5 years ago

    The remaining attributes were all provided by the head of household too, and therefore suffer from all the usual drawbacks of secondary data. We have no option but to treat all of this information as hearsay, as the reliability of the informant would vary considerably from one household to the next.

    However the primary attribute - or what I have defined as the primary attribute, i.e. the name - is one that has the best chance of relating to the actual individual who stayed at the house that night. One can, of course, think of many instances where names are recorded incorrectly, but it is helpful to think of it this way - if the name were omitted completely, how easy would it be to locate that person in the census? Not easy at all.

    While in a strict sense all of the information in a census record must be treated with caution, is it possible to treat the "name" and "address" columns of the census record as primary records, bearing in mind that even primary records can be wrong?