The employmeny data for the period of 1900 - 1947 may be found in Bicentennial Edition: Historical Statistics of the United States, Colonial Times to 1970
The methodology for the collection of this data is not identical to U3 in 1948, but does appear to be fairly close. So for the following charts The unemployment data from the Bicentennial Edition: Historical Statistics of the United States, Colonial Times to 1970 will be treated as a U3 analogue.
This raises the question of the U3/U6 ratio. if the ratio of 1.77 applies for the period of 1948 to present can we apply it to 1900 - 1947? The implication is that unemployment was much higher in the depression then is currently accepted ( a peak of 44%).
I suspect that the ratio would become a much looser fit the further back we go but it should still provide us with some sort of approximation, the usefulness of which is in question without a better understanding of the underlying methodological implications.
(Click to enlarge)