Number of decertification elections and number of votes cast for participating unions during fiscal year 1950

p. 247. STRUCTURE / COLUMN MAPPING (important). This crop holds one table printed as two stacked panels with DIFFERENT column structures, not two side-by-side panels with shared headings. Panel A ("A. ELECTIONS") has 10 printed columns: stub | Number of elections | then Number+Percent pairs under the spanner "Elections won by" for A. F. of L. affiliates, C. I. O. affiliates, Unaffiliated unions, No union. Panel B ("B. ELIGIBLE VOTERS AND VALID VOTES CAST") has 13 printed columns: stub | Employees eligible to vote | Percent casting valid votes | then "Valid votes cast for" spanning Total plus the same four Number+Percent pairs | Employees in units choosing representation. Because the output grid must be rectangular, I used panel B's 13 columns and gave its headings in header_rows; the panel-A headings "Number of elections" and the spanner "Elections won by" are therefore NOT represented in header_rows (they are printed on the image but have no column of their own in this grid). Panel A's rows are mapped as follows: its four Number/Percent pairs sit in the identically labelled A. F. of L. / C. I. O. / Unaffiliated / No union columns (cols 4-11), and its "Number of elections" value is placed in the "Total" column (col 3), because that is the quantity the four pairs decompose (112 = 12+23+2+75, 46 = 12+34, etc.), exactly parallel to panel B's "Total" valid votes. Columns 1, 2 and 12 are empty for all panel-A rows. Note that this is a semantic alignment, not a positional one: on the page panel A's columns are physically offset by roughly one column width from panel B's. The two centred panel titles "A. ELECTIONS" and "B. ELIGIBLE VOTERS AND VALID VOTES CAST" are carried as label-only section rows. DECIMAL POINTS. In this 1-bit scan many decimal points inside the percent figures have dropped out and print as a gap ("10 7", "20 5", "1 8", "67 0", "26 1", "44 9", "73 9", "55 1", "87 5", "100 0", "12 2", "89 6", "87 7", "90 2", "90 5", "45 2", "56 4", "28 8", "71 2", "88 9"), while others in the same table survive ("12.5", "11.1", ".0", "36. 1"). I have transcribed all of these as one-decimal percentages, since every set sums to 100.0 exactly (10.7+20.5+1.8+67.0, 26.1+73.9, 44.9+55.1, 12.5+87.5, 12.2+36.1+2.6+49.1, 45.2+54.8, 56.4+43.6, 28.8+71.2, .0+88.9+11.1) and the "percent casting valid votes" figures reproduce the ratios (8,485/9,474 = 89.6, etc.). The cells ".0" in panel A row "A. F. of L. affiliates—C. I. O. affiliates" (No union Percent) and in panel B's last row (A. F. of L. Percent) are printed with a leading point and no zero; kept as printed. PUNCTUATION IN LABELS. The periods in "A. F. of L." and "C. I. O." are faint or missing in several stub labels and headings on this scan (e.g. panel A's third row reads "C. I. O affiliates", panel B's last stub reads "A. F of L affiliates—C. I. O. affiliates", panel B's A. F. of L. column head reads "A. F. of L affiliates"). I have written the standard fully pointed forms throughout. DASH RULES. Every "------" leader-dash cell is rendered as "". A printed 0 (panel A last row, A. F. of L. Number and No union Number; panel B last row, A. F. of L. Number) is kept as "0". ARITHMETIC. All row and column totals in both panels check out exactly; no misprints detected.