Wednesday, January 4, 2012

Finding Census Records on - Last Name Variations

In my post Finding Census Records on - General Search yesterday, I noted that First Name searches using partial names acts like a wild card (e.g., "jos" finds "joseph"but not "joe"), but the user has to use the "Exact" search box.

What about the Last Name variations - can a user use a wild card or a partial name?  Here is what I found:

1)  On the U.S. Census search form, the Last Name is a "Required" search term. 

I entered "Seaver" in the "Last Name" field, left the "Exact" box unchecked, chose "Massachusetts" in the "Location" field, and chose the "!930" in the "Census Year" field.

Here is the list of 363 Search results:

Even though I left the "Last Name" "Exact" box unchecked, the search returned only results for the surname "Seaver" - there were no variations of the surname in the 363 matches.

2)  To prove this, I checked the "Exact" box for the "Last Name" and saw 363 matches:

The search with the "Exact" box checked worked as I expected - it found matches for the exact surname.

3)  Someone might say, "well, were there any other last name variations on 'Seaver'  in the 1930 Census for Massachusettts?  One of the major spelling variations for "Seaver" is "Sever" - it has the same consonants and the same Soundex code.  Here is a search for "Sever" with the "Exact" box checked:

There are 15 matches for "Sever" with the "Exact" box checked or unchecked.  My conclusion here is that the "Exact" check box is currently useless as a way to restrict searches, and that the not "Exact" search does not use any sort of spelling variation tool for the last name.

4)  What about a partial name for the "Last Name" field?  I put "Seav" in the "Last Name" field and the "Exact" box checked (this worked with the "First Name" field!), and saw:

It says "Sorry, no results found for Seav in MA (census year: 1930)."

When I unchecked the "Exact" box for the last name, I got the same results. 

4)  A general search for all census years and all states said "Showing 2,000 most relevant Census Records for Seaver."  I'm not sure what that means - are there more than 2,000 matches?  If so, how many?  How do I find the others?  I did not scroll through these, at even 50 matches at a time, to see what was missing.

5)  The Lessons Learned here are:

* The "Exact" search results for last names are the same as the not "Exact" search matches.

*  A user has to enter a specific last name in order to obtain matches in a Census search.  A user will have to enter spelling variations if they don't find their target family.

*  There are no "wild cards," or "sounds like" matches for a last name even with the "Exact" box unchecked.

*  A user should search in a specific census year to avoid obtaining too many matches for almost any last name.

*  A user should limit their search to a specific state (or even county) for a common last name.

6)  My suggestions for are to:

*  Make a non-Exact search just that - make it at least a Soundex search so that spelling variations of a last name are provided by the search matches.

*  Implement a "wild card" system similar to that on so that spelling variations of a last name can be easily found.

*  For general searches (all years, all states) on a surname, create a list of the years and states so that users can narrow their search. 

Disclosure: provided a free subscription to their collection at the SCGS 2011 Jamboree which I appreciate. This did not influence my statements in this blog post, but it did enable them to be made!

1 comment:

bgwiehle said...

Were the index entries of the censuses different from those at ancestry or familysearch? You need quality indexing, especially when there is no flexibility in surname search. "Creatively" entered and indexed entries require creative search processes.