An Assessment of Ancestry for Census Searching

Old Herts How To Census

I used Ancestry extensively in preparing the first stages of the tutorial - and was so impressed that I decided to do a test to see how easy it was to find the census entries of some of my ancestors. I started with John Locke, who lived in Aylesbury all his life, and ending with Francis Evered Reynolds, who frequented the East Anglian race tracks - and whose location was unknown in the 1851, 1861 and 1871 censuses. I then extended the search to include two of my wife's ancestors - Walter Pollard Rendell who spent most of his life in Devon, and the mysterious George Cox from Ireland who I expected to find in Wales..

TEST 1

As a simple initial test I started with my maternal great grandfather, John Locke, who spent all his life in Aylesbury. I already knew where he was in all the censuses and this should be a good starting test. I carried out a search using Ancestry's very simple opening form. Not only did Ancestry find him in all four relevant censuses - but gave the entries top ranking - and added the reference for his death certificate as the 6th ranked entry. This must be rated a 100% success.

TEST 2

Next I used the same approach to look for my maternal grandfather, Walter Richard Locke, as while I had good information I was unaware where he was in 1891. Three of the four census returns are ranked 1-3. The 1891 entry was erroneously indexed and at rank 51 could easily be overlooked - but at least it was found without great difficulty (I changed the search name to "Wal* Locke" and it came out as rank 1) - and without a computer index I would had had no idea where to look.

TEST 3

So what about my paternal grandfather, Harry Finch Reynolds, as while I had good information I was unaware of where he was in 1891 but suspected that he was in somewhere in London. The system reports that he is recorded in the OneWorldTree as rank 1 - with the four relevant census entries ranked 2 to 5. Again a 100% success

TEST 4

So what about my paternal great grandfather, Jacob Reynolds who lived through all six censuses. According to family tradition he was living in Hatfield and commuting to London by train to work for Sir John Bennet Lawes of Rothamsted in about 1860. Subsequent research suggests that he was involved in marketing Lawes fertilizers.  Prior to this search I had been unable to locate him in the 1861 census. 

The system reported that he is recorded in the OneWorldTree as rank 1 - with seven relevant census listings ranked 2 to 8 . A real turn up for the book. In 1861 it would seem he was recorded as visiting a Hertfordshire address on Harpenden Common close to Sir John Bennet Lawes of Rothamsted - and also in lodgings in London where he is described as an agent . These duplicated entries would never have been found without computer help - as having found one entry for 1861 I would have stopped looking. A 200% performance.

At the time this test was first done the Ancestry did not include the 1841 census. It is now available and Jacob and his father Francis (below), were mis-indexed at "Reynold" but otherwise easy to find. A look at the original shows that the transcription error was an easy one to make.

TEST 5

The real test should be my great great grandfather, Francis Evered Reynolds. He was born in Chippenham, Cambridgeshire and at the time of the 1841 census he was a farmer at Newton-cum-Castle Acre, Norfolk. By 1851 he appears to have deserted his family and when he was buried at Newton-cum-Castleacre in 1874 an obituary suggests he had spent many years frequenting the race tracks of East Anglia. I had no idea where he was in the 1851, 1861 and 1871 censuses, although I knew he died in lodgings in Kings Lynn, Norfolk.

At first sight there was a possible ambiguity because of the Rank 3 item - but this turned out to be his wife, Frances Reynolds (nee Finch) who was dead by 1861. This left the three entries given above, and while the fit on age and birth places are not perfect there are often uncertainties in the case of lodgers.

TEST 6

With such success I almost stopped testing at this point but decided to look at my wife's paternal grandfather, Walter Pollard Rendell. I had already tracked him down on micro-films for the 1861 and 1871 censuses, had looked him up on the 1881 CD set and also found him on the PRO 1901 census site - so in view of the success so far it hardly seemed worth bothering. 

The result - disaster. OK there is a OneWorldTree entry ranked at 1, and the 1881 census uses the transcript (on familysearch) that allowed me to find him before. There was no sign of the 1861, 1871 and 1901 entries I knew existed, or anything for 1891. What has gone wrong? Knowing the various ways Rendell is spelt in Devon I tried Rendle and Randell without success. I then reduced the search as far as I could, restricting the search to census records so I could include his birth place using only his first name:

Well, its an improvement, and Ancestry does allow searching when you don't know the surname. A search using a wild card for "Wal* Rendell" quickly showed that the 1901 census entry was mis-indexed as Waller L Rendell. 

While all five censuses were eventually found, four had indexing errors and would have been very difficult for a beginner to find - so only 25% on this one.

TEST 7

Finally I decided to try and find my wife's mysterious maternal grandfather, George Cox, as all that was known was that he was born in Ireland and married a Johanna Scott (also from Ireland) a few years before the 1901 census. He was known to be in Wales in the early 20th century. The following entry, on further examination of the full household, proved to be the one I wanted. Very satisfactory in the circumstances

Summary Findings

The following summaries the findings, the Ranking excluding the OneWorldTree entries:

All the tested ancestors were found in all possible censuses - giving a 100% hit rate. Of these 21 out of 28 were given the highest possible rankings, two could be easily found using wild card searches, and 3 were hard to find, in every case because of indexing errors. All the hard to find errors related to capital letter mis-transcriptions for the same individual, who came from Devon, and it may be that the indexing/checking of the Devon indexes is substandard.

I did not keep a check of the time for the search, but all were located on the same day, and I estimate that the time spent on searching and downloading the found pages was between 3 and 4 hours (over broadband) and maybe half of the time was spent trying to find the mis-indexed records. Definitely the total time, including formatting the first draft of this page, was no more that 8 hours.

Comparison with PRO index

When the 1901 census first became available online I carried out a survey My Ancestors in the 1901 Census. This identified a number of errors - and the following table looks at the differences where the error affected the person's name or their place of birth.

Because all the examples are ones where the PRO index "got it wrong" Ancestry could only do better. If one excludes the "enumerator errors" Ancestry corrected corrected four out of seven transcription errors. Unfortunately Ancestry introduced three additional transcription errors in the lines examined - and also demonstrated a bug in the software. This would suggest that the accuracy of indexing leaves little to chose between them.

Comparison with findmypast

The seven tests with Ancestry, described above, have been repeated with findmypast and the results (allowing for the limited coverage of findmypast) were comparable apart from one total mis-index. A second test, of people indexed with the rare surname Phipson showed Ancestry correct 8 times out of 8, compared with 2 out of 8 for findmypast, including another total mis-index. Two other mis-indexes were noted. Such a level of serious errors is unacceptable, and for census searching Ancestry must be considered to be greatly superior to findmypast.

Advice on Searching the Census

The experience from these tests, and the work done on in preparing the tutorial, are that transcription errors are the commonest cause of difficulty. Most of theses are in indexing, but some could have occurred at the time the census was recorded.

The following points should be noted if Ancestry does not come up with the goods on the first search:

• There is a Soundex option under the Exact option which can help if the the recorded name sounds like the name you are searching for.
• There are wildcard facilities using "?" for a single character and "*" for a character string - but can only be used after the first three characters, which must be correct.
• Try looking at the responses for a single source - such as a single census year, and if successful that might give some ideas as to how to extend the search.
• Errors in transcribing capital letters are easily made, and Ancestry does not allow wildcard facilities at the beginning of names. However if you know the first name, age and place of birth you may be able to find the individual by leaving the last name out of the search. (This method can also be used to identify possible identities for a wife before marriage when their maiden name is not known.)

See also Problems with finding census returns