The Skill Content of Recent Technological Change: An Empirical Exploration

• p. 34 Equation (18)

  (σ −1) ln(w_cτ / w_hτ)

Only the authors can say what is right here.

Judge's note: My own working from eq. (17): σ ln(w_c/w_h) = D − ln(N_c/N_h), so D_t = ln(w_ct N_ct/(w_ht N_ht)) + (σ−1) ln(w_ct/w_ht). Taking first differences gives ΔD = Δln(wage-bill ratio) + (σ−1)Δln(w_c/w_h). On the p034 image the Δ applies only to the bracketed wage-bill term, which closes before '+', and the second term carries τ subscripts with no Δ. The text below describes the term as depending on 'the change in the log college wage premium', so the formula as printed does not match the derivation or the text. This is probably a dropped Δ.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 48 Figure 3, HS Grads bars

  HS Grads

• p. 55 Table 7 Panel B

  -2.297

Only the authors can say what is right here.

Judge's note: I read the bars from the image of p. 48. HS Grads bars: RC ≈ -0.22, RM ≈ -0.08, NRM ≈ +0.015, analytic ≈ +0.08, interactive ≈ +0.20. These match Panel A × 0.198 (-0.223, -0.080, 0.016, 0.079, 0.200). They do not match Panel B × 0.198 (-0.455, -0.165, 0.011, -0.002, 0.158). For the other three groups the bars match Appendix Table 4 × 0.198: dropouts RC -0.103, interactive 0.135; some college RC -0.284, interactive 0.148; college RC +0.078, interactive 0.098. The text (p. 25) says Figure 3 summarizes results for all education groups, so the HS Grads series appears to plot the wrong panel.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 54 Table 6 note

  -0.91 in 1990-97

• p. 54 Table 6 notes

  and -0.91 in 1990-97

• p. 58 Table 10 Panel A, interactive, 90-98 NIPA

  0.266

Judge's note: In weighted OLS with decade dummies, mean ΔY in each decade = intercept + decade effect + β × mean x. Check with the printed means: col 1 1960-70 0.152-0.228×0.480 = 0.043 (0.042), 1970-80 0.039 (0.038), 1980-90 0.081 (0.080). For 1990-98, col 1 needs x = (0.096-0.117)/0.228 ≈ -0.092, col 3 needs (0.255-0.332)/0.854 ≈ -0.090, and col 5 needs ≈ -0.090. With -0.91, col 1 would give -0.090 against the printed 0.096. Table 10's 90-98 NIPA entries also fit -0.091 scaled by 0.8 for the 8-year period (interactive 0.266, analytic 0.071, RC -0.210, RM -0.102). Reviewer note 2's alternative (-0.169) is not supported by Table 6.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 56 Table 8, row 'HS graduates'

  0.050 33%

• p. 55 Table 7 Panel B, ∆ Computer row

  -0.011

Only the authors can say what is right here.

Judge's note: Compared Table 8 (p56) with Table 7 (p55) and Appendix Table 4 (p62). Each of the dropout, some-college and college rows is a constant multiple of its own group coefficients: about 0.050, 0.088 and 0.089, which looks like employment share × 0.198 × 1.8. The HS row is exactly 0.125 × the Panel A coefficients in all five columns. Against Panel B the implied ratios vary, and in column 1 the sign is opposite (-0.011 gives +0.050). The 'within education groups' sums (0.069 = 0.003+0.050+0.019-0.003) carry this row, and so do the text's '45 to 91 percent' (p28). Nothing in the text or notes explains the use of Panel A coefficients for this row. Only the authors can say which coefficients were meant.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 58 Table 10 Panel B

  0.87

• p. 58 Table 10 Panel B, rows 1-3

  1. Routine tasks 0.13 0.39 0.39 0.76 2.46 3.36 3.22 / 2. Non-routine tasks 0.15 0.44 0.45 0.99 0.81 1.85 1.80 / 3. All tasks 0.28 0.87 0.88 1.75 3.27 5.29 5.02

• p. 58 Table 10 Panel B, columns 80-90, 90-98 and Sum 70-98

  3. All tasks 0.28 0.87 0.88 1.75 3.27 5.29 5.02

• p. 58 Table 10 Panel B

  2. Non-routine tasks 0.15 0.44 0.45 0.99 0.81 1.85 1.80

Only the authors can say what is right here.

Judge's note: Checked in the p58 image. The 70-80, CPS 80-98, intensive and Sum 80-98 columns add exactly. The 80-90 and 90-98 columns are 0.04 short, and the gap carries into the 70-98 sum (row-wise, 0.28+0.87+0.88+3.27 = 5.30). Footnote 41's interaction term applies to the intensive margin, not to these NIPA columns. Footnote 43 says non-routine manual is a small offsetting contribution, which cannot produce a positive excess. Rounding cannot produce gaps of this size. The text's '5.3 percentage points over 1970-1998' uses row 3. It is not clear which entries are wrong.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 61 Appendix Table 3, panel 4, column B

  -0.010

• p. 61 Appendix Table 3, row 4 (Routine Manual), column B

  -0.390

Only the authors can say what is right here.

Judge's note: In p61, every other checked cell of Appendix Table 3 satisfies intercept + β×0.252 ≈ mean within about 0.003. Examples are row 1 A-C, row 4 A and D, and row 5 B. The 0.252 comes from the Table 5 note, p53. This cell misses by about 0.05. Table 5 RM 1970-80 shows mean -0.010, consistent with its own 0.062 - 0.290×0.252. So the Appendix Table 3 entry may have been carried over from Table 5. Whether the intercept or the mean is misprinted is for the authors to say.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 37 References

  Berman, Eli and Stephen. 2000.

• p. 3 Footnote 1

  Berman and Machin (2000)

Judge's note: The p037 image shows the entry as quoted. The preceding Berman, Bound and Machin (1998) entry gives 'Stephen Machin', and footnote 1 cites Berman and Machin (2000).

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 5 Footnote 7

  cf. Babbage, 1888

• p. 37 References

  Babbage, Henry P. 1883.

• p. 5 Footnote 7

  (cf. Babbage, 1888)

• p. 37 References

  Babbage, Henry P. 1883. “The Analytical Engine.”

Judge's note: Checked footnote 7 (p. 5) and the reference list (p. 37). The years 1888 and 1883 differ, and a text search finds no other Babbage entry. The Bartel–Lichtenberg and Hounshell mismatches also listed here are judged separately as C003 and C007. Which year is correct can only be settled from outside the files.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 5 Footnote 4

  Bartel and Lichtenberg (1989)

• p. 37 References

  Bartel, Ann P. and Frank R. Lichtenberg. 1987.

Judge's note: Checked footnote 4 (p. 5) and the reference list (p. 37). Only one Bartel–Lichtenberg entry exists, dated 1987.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 6 Footnote 7 continuation

  As Acemoglu (1998 and 2000) argues

• p. 6 Footnote 7 continuation

  As Acemoglu (1998 and 2000) argues forcefully

• p. 37 References

  Acemoglu, Daron. forthcoming.

• p. 6 fn 7

  Acemoglu (1998 and 2000)

Only the authors can say what is right here.

Judge's note: Checked the footnote 7 continuation on p. 6 and the Acemoglu entries on p. 37. No Acemoglu 2000 item exists, so it is unclear whether the 1999 AER paper or the forthcoming JEL survey is meant. Only the authors can say.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 7 Section 1

  Hounshell, 1985

• p. 39 References

  Hounshell, David A. 1984.

Judge's note: Checked the p. 7 crop and the p. 39 reference entry. There is only one Hounshell entry, dated 1984.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 9 Footnote 14

  Bresnahan, Brynjolfsson and Hitt (1999)

• p. 38 References

  Bresnahan, Timothy F., Erik Brynjolfsson, and Lorin M. Hitt. Forthcoming.

Only the authors can say what is right here.

Judge's note: Checked footnote 14 (p. 9) and the reference list (p. 38). No 1999 entry by these authors exists. The 1999 date may refer to a working-paper version, which only the authors can confirm.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 24 Footnote 34

  Berndt, Morrison and Rosenblum (1995)

• p. 37 References

  Berndt, Ernst R. and Catherine J. Morrison. (1995)

Judge's note: The citation is in footnote 34 (p.24) and the entry is on p.37 (Journal of Econometrics 65, 9-43). No Rosenblum entry exists. The two author lists disagree. Which list is correct (the published article includes Rosenblum) is outside knowledge, but the mismatch between citation and entry is visible on the pages.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 44 Data Appendix A.5

  are found in Panel B of Table 2 and in Appendix Table 5

• p. 24 Section 4c

  (Appendix Table 3)

• p. 56 Table 8 note

  Table 7 and Appendix Table 6

• p. 44 Data Appendix A.5

  Panel B of Table 2 and in Appendix Table 5

Judge's note: Searching the text finds only Appendix Tables 1–4 (pp. 59–62). Appendix Table 3 is the composite-measure table, and the main text (p. 24) cites it as '(Appendix Table 3)'. The education-group regressions behind Table 8 are in Appendix Table 4, which footnote 39 cites, so 'Appendix Table 6' most likely means Appendix Table 4. No supplement in the files contains Tables 5 or 6.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 29 Section 5b

  represented by the 1973 CPS file (National Academy of Sciences, 1981)

• p. 42 Data Appendix A.2

  a special version of the April 1971 CPS Monthly File

• p. 29 Section 5b

  represented by the 1973 CPS file

• p. 42 A.2

  April 1971 CPS Monthly File

Only the authors can say what is right here.

Judge's note: P.29 cites National Academy of Sciences (1981) for the '1973 CPS file'. On p.42, A.2 ties the same source to the April 1971 CPS ('represented in the 1971 CPS'), and A.4 (p.44) says 'our 1971 CPS file'. No 1973 CPS file is described anywhere. The year is inconsistent, and only the authors can say which is right.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 32 Section 6, last paragraph

  As is visible in Panel A, task changes on the extensive margin attributable to computerization explain 1.8 percentage points

• p. 32 Section 6, paragraph after caveats

  As is visible in Panel A, task changes on the extensive margin attributable to computerization

• p. 58 Table 10 Panel B, All tasks, CPS 80-98

  1.75

Judge's note: The preceding sentence on p032 says Panel A tabulates the task changes ΔT and Panel B converts them into percentage points of college employment. In Table 10 (p058), Panel A has only task-unit changes (e.g. 0.359), while Panel B has CPS 80-98 All tasks = 1.75 and NIPA 0.28, 0.87 and 0.88, which sum to about 2.0 (the figure on p033). The pointer should be Panel B.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 34 Text before equation (17)

  relative supplies (w_ct/w_ht)

• p. 34 Text introducing eq. (17)

  relative wages (w_ct/w_ht) and relative supplies (w_ct/w_ht)

• p. 34 text before Eq. (17)

  relative supplies (wct/wht)

Judge's note: The p034 image shows w_ct/w_ht printed twice. Eq. (17) right below uses ln(N_ct/N_ht) for relative supplies, so the second ratio is mislabelled.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 43 Data Appendix A.4

  (available in electronic form as National Academy of Sciences, 1981)

• p. 40 References

  Fourth Edition Dictionary of DOT Scores for 1970 Census Categories

• p. 43 Data Appendix A.4

  available in electronic form as National Academy of Sciences, 1981

Only the authors can say what is right here.

Judge's note: The p043 image shows the citation as quoted. In the reference list (crop from p040), NAS 1981 is the Fourth Edition (1977) DOT scores for 1970 Census categories (ICPSR #7845), and a 1981 document cannot contain the 1991 revision. The reference list also has a separate U.S. Department of Labor 1991 DOT entry, so this citation looks wrong. Only the authors can say what the intended electronic source is.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 50 Table 2 note

  n is 140 consistent 3-digit CIC occupations

Judge's note: Confirmed in the image of p. 50. The dependent variable is industry computer use, the same note refers to 'DOT industry means', and the text (p. 14), A.3 and the Table 5/7 notes all say 140 consistent CIC industries.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 56 Table 8 title and note

  Dependent Variable: 10 x (Annualized Within-Industry Change in DOT Task Measure)

• p. 55 Table 7 Panel A, Mean ∆

  0.082

Only the authors can say what is right here.

Judge's note: Table 8 'Overall change 80-98' is 0.152, 0.388, -0.321, -0.143, -0.045. Table 7 Panel A Mean Δ is 0.082, 0.212, -0.177, -0.079, -0.026, so the ratio is about 1.8 in each column. The 'Due to computerization' entries equal β × 0.198 × 1.8 exactly in all five columns. The percentage columns are unaffected. The unit label in the title and note does not match the scale of the entries.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 56 Table 8 note

  using regression estimates from Table 7 and Appendix Table 6

• p. 56 Table 8 note

  regression estimates from Table 7 and Appendix Table 6

• p. 62 Appendix Table 4

  Appendix Table 4. OLS Estimates of the Impact of Industry Computerization on Industry

• p. 56 Table 8 footnote

  Table 7 and Appendix Table 6

Judge's note: The appendix tables in the file are 1-4 (pp59-62). Footnote 39 (p28) says the estimates for the other education groups 'are found in Appendix Table 4', and Table 8's values reproduce from Appendix Table 4's coefficients. The data appendix also cites a non-existent 'Appendix Table 5' (text line 2473), which suggests the numbering was left over from an earlier draft.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 19 Section 4a

  while both measures of non-routine cognitive tasks trended upward during the 1960s

• p. 51 Table 3 Panel A, column 2

  2.40

Judge's note: Table 3 Panel A (p.51): interactive is 2.40 in 1960 and 2.40 in 1970; analytic rises from 3.61 to 3.72. In Table 4 (p.52) the 1960-70 overall shift for interactive is 0.005, against 0.107 for analytic. Only the analytic measure trended upward in the 1960s, so the sentence misdescribes the table it cites.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send

• p. 24 Section c

  In all cases, the relationship between computerization and the composite task measure is not statistically significant during the 1960s and becomes statistically and economically significant in subsequent decades.

• p. 61 Appendix Table 3

  0.157

Only the authors can say what is right here.

Judge's note: Appendix Table 3 (p.61), 1980-90: NR analytic 0.157 (0.130), NR interactive 0.112 (0.120), routine cognitive -0.106 (0.097), NR manual 0.138 (0.093); 1990-98: routine cognitive -0.181 (0.116). All have |t|<2. The next sentence concedes inconsistencies only for the 1970s, so 'in all cases' overstates the 1980s and 1990s results as well.

Second check: another model agreed

Report a problem with this finding

This finding is

wrong right unclear or overstated already corrected (erratum, later version)

I am an author of this paper Send