Modern Software Experience

2012-05-16

Census size is constant, but record count is not.

USA 1940 Census

This year, many American genealogists focus on the USA 1940 Census. The images were released on 2012 April 2 and multiple indexing projects are underway.
Right now, it is pays to consult each partial index. One index may already contain records other indexes do not have yet.

You might think that, once the indexing of the 1940 census is done, it will not matter much which index you consult. After all, once the indexing is done, all the sites will let you search all the records.
I've already pointed out that it does matter; Ancestry.com has both its own indexing project and, by acquiring Archives.com, the index created by the 1940 U.S. Census Community Project. Both indexes have been created through double indexing. Ancestry will be able to compare and combine them into a higher-quality double double index.
There are other differences. You may prefer the site with the best images, or the one with the smallest, fast-loading images.
Still, once the indexing is done, each index should allow you to search the entire census. Each indexing project should end up with 132.164.569 records to match the 132.164.569 people in the 1940 USA Census.

Archives.com

USA censusrecords
1790406.419
1800540.434
1810825.998
18201.222.503
18301.824.364
18402.566.554
185020.563.954
186027.119.970
187040.383.357
188050.471.636
18906.293
190076.459.965
191092.638.882
1920107.444.752
1930125.196.787

Archives.com is very much at the centre of the 1940 USA Census activity. They got the contract to host the official 1940 census site, they are participating in the 1940 U.S. Census Community Project, and they were acquired by Ancestry.com.

Archives.com offers all the USA Census. The card catalogue shows how many records each collection contains.
The table shows the size of each USA Census. Looking over the table you immediately notice the exact record count, the increasing size of successive census, broken only by the remarkably small size of the 1890 census collection.

All this is easy to understand. Producing a population count is a major purpose of the census. Moreover, when you scan and index the census, you produce one record for each individual in there.
The increasing record represents the increasing population size. The one outlier, the very small record count for 1890, represents the small remainder of the largely lost 1890 census.

WorldVitalRecords

USA censusrecords
17903.929.326
18005.308.483
18107.239.881
18209.638.453
183012.866.020
184017.069.453
185023.191.876
186031.443.321
187039.818.449
188050.189.209
1890-
190076.212.168
191092.228.496
1920106.021.537
1930122.775.046

Archives.com and WorldVitalRecords are quite similar. Both services offer very popular American databases at a competitive subscription cost. Both license their content from various third parties.

Both Archives.com and WorldVitalRecords license their American census from FamilySearch. They do not make their own scans. They do not have their own indexing teams. They merely license the USA Census from FamilySearch.
Archives.com and WorldVitalRecords offer the same databases. Yet, the Archives.com and WorldVitalRecords databases do not have the same record count.

The record counts for the USA 1900 census and 1910 are close to each other. It is odd that they aren't the same, but at least they are not very different, whereas the record counts for earlier census are wildly different.

Archives.com claims to have about half a million records in their USA 1800 census, WorldVitalRecords claims to have more than five million records. WorldVitalRecord's record count of 5.308.483 for the USA 1800 census is almost ten times Archives.com's record count of 540.434.
Surely that cannot be right.

reel versus index

USA censusreelsindex
186031.443.32125.122.104
1930122.775.046106.975.355

Not confused yet? Perhaps this helps.
For 1860 and 1930, WorldVitalRecords offers both census reels and census index. These different databases do not have the same record count.

FamilySearch

USA censusrecords
1790410.887
1800540.614
1810826.307
18201.225.304
18301.824.528
18402.566.831
18503.869.016
186027.126.196
187010.544.804
188010.263.875
18906.280
190017.014.324
191029.847.397
192030.663.476
193032.436.590

Both Archives.com and WorldVitalRecords get their database from FamilySearch. You'd think that both should have record counts identical to FamilySearch. A quick comparison with the FamilySearch numbers might reveal which numbers are right, and which ones are wrong.

Alas, neither the Archives.com nor WorldVitalRecords record counts agree with the FamilySearch record counts. For the early census, 1790 and 1800, the Archives.com numbers are close to the FamilySearch numbers. For the most recent census, 1920 and 1930, the Archives.com and WorldVitalRecords.com numbers are close to each other, but both about three or four times as large as the FamilySearch number. That does not seem right.

When you look at just the FamilySearch table, you soon notice that the record count for the USA 1860 census is much larger than the record for the USA 1850 census before it, even larger than the record count for the USA 1870 census after it. It seems a rather unnatural peak.
When you compare it with the record count for Archives.com and WorldVitalRecords, the record count does not seem far out. In comparison to these services, it seem to be the previous census and all the later census that have a relatively low count, as if records are missing.

Ancestry.com

USA censusrecords
1790419.502
1800542.070
1810830.631
18201.239.769
18301.828.796
18402.585.887
185020.053.649
186027.483.658
187040.405.477
188050.480.997
18906.304
190077.277.541
191093.627.068
1920107.634.003
1930124.962.440

The numbers for the census databases on Ancestry.com are yet different. The Ancestry.com numbers are in relatively close agreement with the Archives.com numbers. Ancestry.com claims about half a million records for both the 1790 and 1800 census, and for both 1920 and 1930 the record count is more than hundred million.

USA 1890 Census

databaserecords
Archives.com6.293
FamilySearch6.280
Ancestry.com6.304

Archives.com, FamilySearch and Ancestry.com all have different record count for the 1890 census. What's more, all three record counts are different from the official number.

The official number is 6.160, but according to the Archives.com card catalogue, the 1890 U.S. Federal Population Census contains 6.293 records.

The description for FamilySearch United States Census Population Schedules, 1890 states that The surviving records list 6,160 names which are mostly veterans schedules, yet their card catalogue claims that their collection contains 6.280 records. That is 120 records more.

The description for Ancestry.com's 1890 United States Federal Census Fragment states that The records of only 6,160 of the 62,979,766 people enumerated survived the fire., yet their card catalogue claims that their collection contains 6.304 records.

United States Census Bureau

USA censusrecords
17903.929.214
18005.308.483
18107.239.881
18209.638.453
183012.866.020
184017.069.453
185023.191.876
186031.443.321
187038.558.371
188050.189.209
189062.979.766
190076.212.168
191092.228.496
1920106.021.537
1930123.202.624
1940132.164.569

Archives.com, WorldVitalRecords.com, FamilySearch.com and Ancestry.com all provide exact record counts for their census databases, but their numbers do not agree with each other.
The small differences make you wonder what is going on already, the large differences raise even more questions. All the differences make you wonder what the right numbers are.

The United States Census Bureau has a document that list the population of the United States of America, as determined by the USA census. These are the official population counts.
Assuming one record for one person, these are the record counts the databases should have. Well, the USA 1890 census is an obvious exception. As Ancestry.com's description for the 1890 census states, only 6.160 of the 62.979.766 records remain.

The different services claim different record counts for the census databases, even though it are the same census, and most numbers do not correspond to the population counts published by the Census Bureau.
The WorldVitalRecords numbers match pretty good, with many record counts being an exact match to the Census Bureau population count.

explanations

Many of the FamilySearch record counts are considerably smaller than the Archives.com record count. Perhaps the Archives.com numbers reflect the actual size of the entire licensed database, while the smaller FamilySearch numbers reflect that only part of the database that is online at familysearch.org.

Most of the record counts for the WorldVitalRecords database match the Census Bureau's population count exactly. WorldVitalRecord licenses the same FamilySearch databases as Archives.com, yet some of their numbers are ten times as large. A possible explanation for both facts is that WorldVitalRecords is displaying the official population count instead of the actual record count. Perhaps someone entered these numbers manually back when the databases were first added to their system.

The small differences may seem less important, but are hardest to explain. How do the 6.160 records in the USA 1890 Census translate into 6.280, 6.293 or 6.304 database records?
There may be a perfectly reasonable explanation, such as extra records in support of multiple spellings of the same name, but that does not change the fact that all the services are claiming to offer more records than the database actually contains. That is odd and confusing.
The record count for a database should equal the number of logical records the database contains.

links