I agree with CaroleW. He probably thought he had to name all children on the form, living or dead, and then realised his mistake. I wouldn't say that it supports the child having died between filling out and submitting the form, unless other evidence of the child's date of death supports that.
One of my ancestors did something similar in the 1911 census, listing three children, one of whom had died several years previously. But having either realised his mistake or having had it pointed out, ho compounded it further by crossing out and marking as deceased one of the two surviving children, leaving the deceased one still apparently alive.
