Size of Units in Representation Election Cases Closed, Fiscal Year 1977

p. 321. Crop: the image ends immediately below the "3,000 to 9,999" row. Section "A." implies at least one further panel (and possibly further size classes such as "10,000 or more") below the crop; no footnotes are visible on this image. Decimal points: this 1-bit scan drops many decimal points, so a large share of the percent cells print as e.g. "22 9", "100 0", "0 3". Every value in the four percent columns ("Percent of total", "Cumulative percent of total", and each "Percent by size class") is a one-decimal figure, and the point is clearly present in other instances of the same values, so the decimal point has been written in each such cell. No digits were altered. "180 to 189", Cumulative percent of total: the separator prints as a comma, "93,5"; transcribed as 93.5 (it continues the monotone cumulative series 92.9 -> 93.5 -> 94.1 and equals the previous cumulative plus this row's 0.6). Heading of the first pair of "won by" columns: the scan reads like "AFO-CIO unions"; the third letter is a degraded "L". Transcribed as "AFL-CIO unions". Section line: the scan shows "A Certification elections (RC and RM)" with a wide gap after the "A"; the period is assumed lost with the other dropped points and has been written as "A. Certification elections (RC and RM)". Dash rules: wherever a count cell is 0, the matching "Percent by size class" cell carries a dash rule (------); those cells are "". The "Cumulative percent of total" cell on the "Total RC and RM elections" line is likewise a dash rule and is "". The dash rules sit low in their row and can appear to belong to the row below; each one here has been assigned to the row whose count is 0. Arithmetic checks (all passed, no misprints found in the counts): "Number eligible to vote" sums to 528,866; "Total elections" to 8,635; AFL-CIO 2,417; Teamsters 1,194; Other national 236; Other local 312; no representative 4,476; and 2,417+1,194+236+312+4,476 = 8,635. The "Percent of total" column sums to 100.0 and agrees cell-by-cell with the differences of the cumulative column. Percent-by-size-class columns: each sums to exactly 100.0 as printed, because in each column one cell is adjusted off its computed value to force the total. Printed vs. computed: "Under 10" AFL-CIO 24.4 (585/2,417 = 24.2); "Under 10" Teamsters 39.1 (469/1,194 = 39.3); "10 to 19" Other national 18.4 (43/236 = 18.2); "10 to 19" Other local 20.4 (63/312 = 20.2). All four are transcribed as printed; every other cell in those columns matches its computed value, which is what confirms these readings rather than casting doubt on them.

p. 322. STRUCTURE. The crop shows only the right-hand panel of a two-panel table; panel A (RC elections) is on the preceding page and is not on this image. Fifteen columns: stub + 14 data columns, counted from the column rules. The only printed heading is the spanner "B. Decertification elections (RD)" running over all 14 data columns, so header_rows has a single level and the stub heading cell is empty; the per-column sub-headings that would name each column are not printed on this crop and were not reconstructed. The 4th data column (the one left blank/dashed in the "Total RD elections" row) is a cumulative-percent column: its values are the running sum of the 3rd column (31.9, 31.9+23.4=55.3, ... 100.0), which confirms the column assignment. DECIMAL POINTS. In many percent cells the printed period did not survive the 1-bit scan and shows as a blank gap ("31 9", "100 0", "7 7", "23 1", "0 0"). I restored the decimal point, because the identical columns print it clearly in other rows ("23.4", "0.6", "1.3", "87.2", "98.9"). No value was otherwise altered. DOUBTFUL CELLS (scan quality; 0/1, 3/5/6 are confusable here): - Stub "Under 0" (row 2): reads "Under 0" in the scan, but the series ("10 to 19", "20 to 29", ...) and the data (1,503 employees over 270 elections, mean ~5.6) imply the printed label is "Under 10" with the leading 1 lost. Transcribed as seen. - Row "Under 0", 8th data column: read as 20.1, but 12/60 = 20.0 exactly; may be a misread of "20 0". - Row "10 to 19", 6th data column: read as 20.6, but 24/118 = 20.3; may be a misread of "20 3". - Row "Under 0", 14th (last) column: read as 37.3, but 242/645 = 37.5; may be a misread of "37 5". Every other percent cell in the table reproduces its count/total ratio exactly, which is why these three stand out. APPARENT PRINT DISCREPANCY (transcribed as printed, not corrected): row "Under 0", 3rd column 31.9 — 270/849 = 31.8. The cumulative column independently prints 31.9 and 31.9+23.4 = 55.3, so 31.9 is what the report set, and the 270 is confirmed because the 2nd column sums exactly to the printed total 849. CHECKS THAT PASS. All five count columns sum exactly to their printed totals: 1st = 41,850; 2nd = 849; 5th = 118; 7th = 60; 9th = 13; 11th = 13; 13th = 645. OTHER. Cells printed as a dash rule ("--------") are transcribed as ""; printed zeros are kept as 0. The heavy black streak across the top edge is a scanning artifact, not table content. The running head "Appendix" and the page number 307, set sideways in the right margin, are omitted per instructions, as is the caption (on the previous page).