Assessing accuracy and confidence in different types of records?

Question

My assumption is that an official country census is probably at the highest level of accuracy as although individuals may report details incorrectly, there is a fair amount of cross checking, but where there are other records is there a way to measure the level of assurance we can gain from them?

For example in my tree I have some information from:

• passenger transport manifests and prison records (for some of my Australian ancestors)
• birth, wedding and death certificates
• censuses
• newspapers
• personal journals
• word of mouth

Is there a scale of accuracy which is generally used for these?

Answer 1

You cannot assume that a census has a high level of accuracy. For the UK census, for example, little or no checking was done of the information provided by householders. There are instances where the householder has clearly provided "joke information" -- such as an occupant recorded as Peter Tabby with nationality Persian and occupation mouser. The census enumerator picked up on this and annotated the return to show that this was a cat, but there will be many other instances of deliberate misinformation or inadvertent errors.

To assess the degree of reliance you can put on any source, you have to determine whether it is an original source (in its first oral or recorded form) or a derivative source (created by copying, transcribing, abstracting or otherwise manipulating the original). A derivative source may not contain all the information in the original (for example, pages missed or rendered illegible during scanning, facts omitted to strengthen an argument being made by the creator of the derivative source), so an original source is preferable, if it is available.

You then need to consider the information (or content) of the source. Each piece of information can be primary information (detail provided by somebody with firsthand knowledge of the information) or secondary (detail provided by somebody with secondhand or worse knowledge). Primary information can be more reliable, but the person providing it might have been mistaken or lying about the facts, so you also have to consider why and when the information was provided by that person.

Answer 2

Determining the level of confidence of any particular source is at best a compromise as a source may be extremely accurate in some claims, and less so in others. Take census records as an example. Although much of the information in census records is very accurate, I've found in my own database that census ages are one of the least certain claims. In an effort to help genealogists in this matter, I wrote a small set of guidelines (http://timforsythe.com/blog/guidelines-for-categorizing-sources/) that you may find useful. These guidelines should help you to categorize your sources in 3 areas: record authority, record concurrency and author association. These are just a few of the many possible categories that you could choose to use; others, such as legibility, bias, etc. may provide additional value. I personally categorize all my sources, and then use these to determine a certainty assessment (level of confidence) for EVERY claim in my database. I then present these CAs along with my source references for each claim so that viewers are able to tell at a glance which data I am presenting is the most accurate and which suffers in the area of reliability.

Answer 3

Don't try to assume one type of record is more accurate than other. There is no way to assess beforehand the accuracy of any specific type of record. Each is fraught with its own possible problems.

You cannot gain confidence from a single record. It may be the only information you have so you normally would assume it is likely correct. However it may be wrong or right.

If you have two records and both agree. Then you gain confidence. However both may still be wrong. If they disagree, then you know at least one is wrong, maybe both and your confidence is low.

If you have three records and all agree, your confidence will grow some more. Add some non-direct but collaborating evidence and your confidence grows even more. Find a complication, and your confidence shrinks.

Just gather all the records you can and use them together to come to your conclusion. Generally, the more agreement you have from many different records, the more confidence you will have.

But is your data correct? You never really know for sure.

Answer 4

Is there a scale of record group/collection accuracy?

When viewed as a whole (as a statistical mass) there may be some scientific basis for assessing the validity of historical record group or collection information. This might in part draw on the approaches in methodology/science that are used to develop and assess the reliability of research surveys and the like.

Such assessment would not be nearly so cut and dry or meaningful in terms of the work of genealogists and family historians, especially because

• We are not dealing with the norm, but with the individual records and the bits of information found in the record groups and collections. Statistically speaking, these individual entries/bits of information vary widely as to their degree of "accuracy"; any one entry/bit has greater probability of error and omission.
• Only sometimes do genealogists set out to use information in a manner consistent the original intent of that information." In other words, it is only is my folly to believe the US government took the 1850 census so that I would know how old my John Miller then really was.

While I am light on the science to back it up, I believe there is some general qualitative assessment at work in genealogy—probably among record/source types rather than between them. For example, take an 1850 era baptismal record entered by a traveling preacher from Pennsylvania and the same era Norwegian parish record. Whether is it bias or not, I would initially place more weight or give more credence to the Norwegian record. Between a traveling preacher a and a traveling preacher b, though, all my bets are off lack material study about the two record sets.

Generally speaking, and in my own words, our science is different because our interest is unique. The Genealogical Proof Standard talks about a reasonably exhaustive search including review/exposure to a wide variety to high quality sources.* What this has meant to me is seeking diversity in the records I'm using, the quest for something closer to a 360 degree perspective on the information/evidence.

*Genealogists categorize sources as to whether each is an original or derivative. A common comparison could be a marriage record (original) vs an abstracted or indexed entry about that same marriage record (derivatives). Even working with an original, there may be more value to viewing that original in the context of the original record book. Take the example of a parish register entry or a New England birth/marriage/death recorded in the town books or records--viewing the entries in sequence often provides better contextual information than viewing the a single, isolated record.

Answer 5

RE: My assumption is that an official country census is probably at the highest level of accuracy

One of my favourite genealogical recreations is to use images of the Census returns to calculate how much people in Ireland aged between 1901 and 1911. The range I have found is from 2 to 16 years. A fascinating feature is when everybody in a household ages at a different rate.

Introducing the Old Age Pension (in 1908) clearly had an impact on the age of seniors but there is room for speculation on all the other circumstances behind the changes.

I am convinced one ancestor put her age up when she married then spent the rest of her life trying to bring it back down. So we talk about "conflicting data".