Number of elections and pay-roll checks and number of votes cast for participating unions, 1940-41

p. 44. 20 columns: stub, "Number of elections or pay-roll checks conducted", four Number/Percent pairs under "Elections or pay-roll checks won", four Number/Percent pairs under "Votes cast", "Against unions", "Total valid votes". Empty body cells are printed as dash rules (------) and are transcribed as "". Doubtful digits — the scan degrades some 9s toward 0 at this size, so three cells were resolved against the printed totals, which all reconcile exactly: (a) row "A. F. L. affiliates^1 ^2", column "By A. F. of L. affiliates / Percent of total" reads as either 70.6 or 79.6; 733/921 = 79.6 and every other percent in the table is internally consistent, so 79.6 is transcribed — this is the least certain cell in the grid. (b) Total row, "Number of elections or pay-roll checks conducted" reads as 2,508 or 2,568; the elections column sums to 2,568 and all four Total-row election percents (36.0, 38.6, 2.2, 5.9) are computed on 2,568. (c) Total row, "Total valid votes" reads as 720,915 or 729,915; the column sums to 729,915 and the Total-row vote percents (24.8, 46.0, 3.1, 7.0) are computed on it. Suspected misprint in the source, transcribed as printed: in row "C. I. O. affiliates-Local unaffiliated unions" the "Total valid votes" entry 66,325 does not equal its components (35,921 + 25,735 + 4,687 = 66,343), an 18-vote shortfall. The printed 66,325 is the figure the rest of the table is built on — it reproduces that row's 54.2 and 38.8 percents and is the value needed for the column to total 729,915 — while each component is confirmed by its own column total (335,619; 50,849; 140,012), so the discrepancy is in the source, not in the reading. Every other row and column reconciles with the printed totals and percents.