Another example is my great-great grandmother Virginia Raymond Chicoine's census in 1880. I had found her in every census in US census from 1860 to her death after 1930 but the 1880 one had eluded me. I knew she lived in Highgate, Vermont and was twice widowed - with two children from her first marriage to Antoine Bouvier and four from her second to Paul Chicoine. She appears as Vergine Reymo with 6 children all with the surname Reymo. As a French woman, if she was the one speaking with the census taker she would have used her maiden name to identify herself - did the census taker assume that all the children shared the same last name? Or was the informant one of the older children who identified her mother this way and just gave the first names of her siblings? I certainly didn't search the indices for Raymond/Reymo. It is obvious that the census taker wrote down what he heard for the last name. This is true of many of both my French and Irish ancestors where Leary becomes Lary and Shea becomes Shay or even Chay.
Other errors may be deliberate. In researching a friend's family I discovered two sisters one older and one 2 years younger with ages correctly noted in all census until 1900 when the elder had gotten married. The newly weds were living with the younger single sister. The married sister had recorded her age as 2 years younger so her sister had to adjust her own birth date as to not give away her sister's fudged birth date. Since the husband would have been 2 years younger than his wife if the real date had been recorded I assume that was why she had adjusted her age.
Birth places can be in error - one granduncle who is listed as born in Vermont and by family tradition was born there was really born in Ireland A baptism record from Limerick proved that. In census where he probably gave the information he is listed as from Ireland but in others where his family probably were the informants it is Vermont. Why?
In a Kansas 1870 census a boarding school with about 1/2 Indian students and 1/2 white has the column denoting race with Ind crossed out. Why did the census taker or a later official do this? A native woman in the same town married to a white man is correctly identified in all the census but her children, originally designated as MB or mixed blood appear to get whiter with the passing decades. In another case the opposite is true. A man named George Washington, born in Virginia in 1841, one of the few persons listed as "colored" in the census in Swanton, VT in 1870 appears with his white wife and white children. By 1880 his children have all become mulatto.
In the census for the 1860 census in Fall River, MA my Corbett ancestors and all Irish in the town appear in one index as from Idaho. Since that state was not yet in existence in 1860 it appears the census taker got creative and wrote Id for place of birth Ireland instead of the usual abbreviation IR or IRE.
It is very possible the census taker missed some people. Thanks to ancestry.com's census search I have been looking for a family who should be in the 1860 census. My great grandaunt Mary Leary Bresnahan and her child Cornelius Andrew Bresnahan appear in the 1870 and later census without her husband Andrew Bresnahan. They had married in 1854 in Manchester, NH and Cornelius was born in Fishville, near Concord NH in 1856. I have searched the US census nationwide for a child of that age and found none that match. Have even tried searching for child by just age and place of birth or first name only. Where was he? Where was his mother? What happened to his father? (disappeared? lost in war? civil war records doesn't turn up an Andrew that matches) He was probably deceased at least by 1872 when his wife remarried but no death record has yet to surface. A mystery!
