WLS MEMO 122 Mapping 1970 Census occupation-industry-class of worker combinations into the Featherman, Sobel, and Dickens (FSD) update of Duncan SEI and Siegel (NORC) Prestige scores Robert M. Hauser, Taissa S. Hauser, and Jennifer Sheridan February 18, 1998 As noted in COR315, there are discrepancies in the original (1976-77) mapping of 1970-basis Census occupation-industry-class of worker combinations into the Featherman, Sobel, and Dickens (FSD) update of Duncan SEI and Siegel (NORC) Prestige scores. There are three types of discrepancies. * First, non-split occupation lines were mapped into SEI and Prestige scores for the "total" work force in the FSD update rather than into the scores for the male labor force. This is contrary to our prior understanding (as reported in COR226 of 12/13/76 and reproduced in Appendix P to SWL20 (p. P3), March 1977, Appendices, A User's Guide to the Machine-Readable Data File: Wisconsin High School Student Panel Study of Social and Psychological Factors in Status Attainment 1957, TAX, 1964, 1975, and 1977), which was that the non-split occupation lines were mapped into the FSD update for the male labor force. While this error affects a large number of variables, the overall level of agreement between "total" and "male" scores is quite high, and thus the effects of this problem should be minimal. * Second, certain occupation lines were split by class-of-worker and/or industry. Except in the case of two occupational aspiration variables (see below), these were mapped into Duncan SEI and Siegel (NORC) Prestige scores using modifications of Duncan's and Siegel's original detailed mappings, which was prepared by Hauser and Massagli. These supplementary mappings were reported in COR315 (March 1979) and have been updated in COR612C (which is adapted from COR521C). The nominal discrepancy occurs because the scores for the split lines pertain to the original populations for which Duncan and Siegel prepared the scores, which were men and (presumably) the total work force, respectively. In a few of the split occupation lines, e.g., 245, the score for the "residual" lines -- for which there are no specific scores for industry or class-of-worker groups or for which industry or class-of-worker were not reported -- is not the same as that for the original non-split line because of differences between "total" and "male" scores. Very few cases in the split categories fell through to the "residual" lines, and therefore the effects of this discrepency should be minimal. * Third, COR226 (Appendix P) reports that the source of the score mappings for non-split occupation lines was Appendix B of Featherman, Sobel, and Dickens, CDE WP 75-1, "A Manual for Coding Occupations and Industries into Detailed 1970 Categories and a Listing of 1970- Basis Duncan Socioeconomic and NORC Prestige Scores." This source was also appended to L. Sheehy, D. Netkin, and J. Grant, "Social Factors in Aspirations and Achievements: Occupation-Industry Coding Handbook, 1974-1975," and it was published (with some typographic errors) in R.M. Hauser and D.L. Featherman, The Process of Stratification: Trends and Analyses (New York: Academic Press, 1977, pp. 59-80, 319-329). However, there are minor discrepancies between the listing in Appendix B and the mapping into "total" SEI and Prestige scores. In general, where the the 1976-77 WLS mapping into "total" scores differs from Appendix B, the mapping is correct. (But see the discussion of allocation codes, below.) Again, the effect of these errors is minimal, because of the relatively few cases affected. For these reasons, we have preserved the original mappings, which have been widely used in analyses of the WLS data, and we are issuing new versions of the 1975-77 WLS data with modified SEI and Prestige scores. (The same, modified mapping rules are also being used for all 1970-basis Duncan SEI and Siegel (NORC) Prestige scores in the 1992-94 WLS data. They are documented in COR521C.) In the rest of this memorandum, we describe these three sets of discrepancies and their consequences in detail. "Total" vs. "Male" SEI and Prestige Scores In the 1970s and 1980s, several researchers constructed separate socioeconomic scales for occupations held by men and for occupations held by all workers. These have been based upon weighted averages of the educational attainment and income or earnings of men and of all workers, respectively. There are substantial differences between "male" and "total" scores so constructed, and there are major differences in the statistical properties of those scores. There are fewer and less substantial differences between the "male" and "total" SEI and Prestige scores in the Featherman, Sobel, and Dickens update from 1960-basis to 1970-basis Census occupation lines. In that case, the "male" and "total" scores only differ by dint of gender differences in the composition of 1970-basis Census occupation lines in terms of their component 1960-basis occupation lines. That is, both sets of SEI scores are based solely on the characteristics of male occupational incumbents, and both sets of Prestige scores are based on ratings of gender-typed occupational titles. Briefly, FSD used a Census tabulation of a large sample of workers from the 1960 Census, whose jobs had been coded both to 1960 and 1970 standards and then cross-classified by 1960-basis occupation and 1970-basis occupation (J.A. Priebe, J. Heinkel, and S. Green, "1970 Occupation and Industry Classification Systems in Terms of Their 1960 Occupation and Industry Elements," U.S. Bureau of the Census, Technical Paper 26. Washington, D.C. U.S. Government Printing Office, June 1972, Table 1). To construct the "male" scores, FSD assigned each 1960-basis occupation line its SEI or Prestige score as reported by Duncan or Siegel. They took a weighted average of the scores within each 1970-basis occupation line, where the weights were the numbers of men from each 1960-basis line in that 1970-basis line. Correspondingly, in constructing "total" scores, the weights were the numbers of men and women in each component 1960-basis occupation line. For several reasons, the differences between FSD's "total" and "male" scores were relatively small. There was a high level of occupational segregation in 1960. The changes in the composition of occupation lines between the 1960 and 1970 classification systems were usually small. Thus, the so-called "total" SEI scores are not based on the socioeconomic characteristics of all occupational incumbents; rather, they are based on the characteristics of male occupational incumbents in 1950, as updated by Duncan to the 1960 Census classification and standardized by the composition of the total labor force by constituent 1960-basis occupation lines within each 1970- basis occupation line. Similarly, the "total" Prestige scores pertain to weighted averages of ratings of gender-typical occupational titles. In Table 1, we have listed the occupation lines for which "total" and (revised) "male" SEI or Prestige scores differ. Note that the list is based on "total" scores actually used in mapping the 1975-77 WLS occupation data into SEI and Prestige, while the "male" scores incorporate corrections made in the January 1996 revision of the data. There are 447 lines in the two sets of mappings, and a difference occurs between the original ("total") and revised ("male") scores in 148 lines. In most cases these differences are very small -- note that the scores are given here to three digits, although one can scarcely claim two digit accuracy in the source materials. However, there are occasionally large differences, e.g., 172 between total and male SEI for occupation 212 (Health administrators) and 142 between total and male Prestige for occupation 922 (Health aides, exc. nursing). To validate these findings, we have rechecked the FSD calculations for all four entries in each occupation line where one of the score differences is greater than 40 in absolute value, that is, where the difference in original scores exceeds 4 points on the SEI or Prestige scale. These are occupation lines 145, 194, 212, 264, 444, 672, 922, and 982. In each case, we have exactly reproduced the differences observed in Table 1, and in each case, excepting line 672, the validated scores were identical to those in Appendix B of Featherman, Sobel, and Dickens. We have also validated the FSD calculations in several other lines where there were discrepancies between Appendix B of Featherman, Sobel and Dickens and our machine readable listings of the maps from 1970-basis Census lines to SEI and Prestige scores. These are occupation lines 545, 551, 672, 711, 714, and 962. In each of the latter cases, the validation showed that the listing in Appendix B of Featherman, Sobel, and Dickens was in error, and the original machine-readable listing used in the WLS was correct. The validation procedure is reproduced in Table 2. For each selected 1970-basis occupation line, the table shows each 1960-basis component, along with the population estimates of men and women from U.S. Bureau of the Census Technical Paper 26 and the total population estimates. Duncan SEI scores were obtained from O.D. Duncan, "Socioeconomic Scores for Detailed Occupations [in the 1960 Census]," University of Chicago, Population Research and Training Center, October 1961. Siegel Prestige scores were obtained from Paul M. Siegel, Prestige in the American Occupational Structure (University of Chicago: Unpublished doctoral thesis, 1971:Table 5). Despite the occasionally large differences between the "total" scores assigned in 1976-77 and the revised "male" scores assigned in December 1995, the overall level of agreement between the two sets of scores is very high. For example, even in the 148 cases where there is disagreement between SEI or Prestige scores, the unweighted correlations between the two sets of scores are 0.996 (SEI) and 0.994 (Prestige). Split and Non-Split Occupation Lines When O.D. Duncan (1961) developed the Socioeconomic Index for Occupations, he split several large residual categories of occupation by industry and/or class of worker and calculated the SEI separately for the component lines. In the 1970 Census classification, these were categories 245 (Managers and administrators, n.e.c.), 441 (Foremen, n.e.c.), 452 (Inspectors, n.e.c.), 640 (Mine operatives, n.e.c.), 690 (Machine operatives, miscellaneous specified), 692 (Machine operatives, not specified), 694 (Miscellaneous operatives), 695 (Not specified operatives), 780 (Miscellaneous laborers), 785 (Not specified laborers), and 964 (Policemen and detectives). For a smaller set of titles, Siegel (1971) specified differences in prestige scores by industry (but not by class of worker). These were categories 245, 441, 640, 690, 692, 694, and 695. Subsequent updates and revisions of the SEI (and of prestige scores) by Featherman, Sobel and Dickens, Stevens and Featherman (1981), Stevens and Cho (1985), and Nakao and Treas (1994) have not incorporated any of these "splits," even though Duncan's and Siegel's split categories include a substantial fraction of the workforce. To fill this gap in the Featherman, Sobel, Dickens work, WLS investigators adapted Duncan's and Siegel's splits of 1960-basis Census occupation lines for use with the 1970-basis Census occupation lines, as reported in COR315 (March 29, 1979). Because Technical Paper 26 does not subdivide 1960-basis lines by industry, it was not possible to follow the updating procedure employed by Featherman, Sobel, and Dickens. For this reason, the mappings for "split" 1970-basis Census lines pertain only to the original SEI and prestige scores, without regard to change in the composition of lines by occupation, industry, or class of worker. When jobs were ascertained in the WLS, most question series ascertained industry and class of worker as well (See COR315). In all of those cases, scores for the split categories that had been assigned from the list for 3-digit occupations were reassigned on the basis of industry and class of worker. Consequently, because the 1976-77 mapping of the WLS data into SEI and prestige scores was inadvertently made using the "total" list rather than the "male" list from FSD, in most cases the resulting arrays of scores combined "total" scores (for non-split lines) with "male" scores (for split lines). In a few combinations of occupation and industry, notably, occupation 245 (Managers and administrators, n.e.c.) with industry codes 907-937 (public administration), the split codes were the same as the FSD "male" codes, so no further transformation was made. Since the scores had been mapped into "total," rather than "male" scores, this procedure introduced additional small inconsistencies. The "male" SEI for occupation 245 was 620, while the "total" score was 617, and the "male" prestige for occupation 245 was 508, while the "total" score was 507. Occupational Aspirations and Allocation Codes In series where occupational aspirations or expectations (for self, past or future, or for children) were ascertained, the questions were not as detailed, and the procedures for assigning SEI and prestige scores were modified. The source codes for these aspiration items are OCZ (detailed occupation for the job respondent expected to have when respondent was a senior in high school), OCR10 (detailed occupation for respondent's preferred job 10 years from now), OCASSK (detailed occupation of respondent's aspirations for child), OCEXSK (detailed occupation of respondent's expectations for child), XOCZ (detailed occupation for the job sibling respondent expected to have when respondent was about 16 years old), and XOCR10 (detailed occupation for sibling respondent's preferred job 10 years from now). In the cases of OCR10 and XOCR10, the Respondent or Sibling Respondent was asked whether he or she wanted to be self employed or not, and this variable was used in place of class of worker to map into the SEI and prestige scores. In the cases of OCASSK and OCEXSK, neither industry nor class of worker was ascertained and, for these variables, the original 1976-77 mapping into SEI and prestige codes used "total" scores from the FSD update. In the cases of OCZ and XOCZ, occupation and industry were ascertained, but not class of worker. This did not affect the mapping into prestige scores, but the procedures for mapping into the SEI were modified in the cases of codes 245 and 964 by taking the average SEI for self- employed and salaried workers. In several hundred cases, because the quality of the aspiration data was not as good as that of the reports of actual occupations, coders assigned Census "allocation" codes for some major groups (or subgroups) of occupations. (The more typical use of allocation codes in the 1970 vintage Census and Current Population Survey was to fill in missing data at random using hot-deck techniques.) These allocation codes were never used in coding actual occupations. According to COR226 (Appendix P), the allocation codes, SEI, and prestige are as follows: Code SEI Prestige Occupation title -------------------------------------------------------------------------- 196 750 600 Professional, technical, and kindred workers -- allocated 246 570 500 Managers and administrators, except farm -- allocated 296 490 340 Sales workers -- allocated 396 450 390 Clerical and kindred workers -- allocated 586 310 390 Craftsmen and kindred workers -- allocated 696 180 290 Operatives except transport -- allocated 796 070 180 Laborers, except farm -- allocated 806 140 410 Farmers and farm managers -- allocated 846 090 190 Farm laborers and farm foremen -- allocated 976 170 250 Service workers, except private household -- allocated Note this list fails to include two of the 1970-basis allocation lines, 726 (Transport equipment operatives -- allocated) and 986 (Private household workers -- allocated). Moreover, a review of the actual codes appearing in the 1975-77 WLS occupational aspiration items shows 200 entries of 196, 152 entries of 246, 131 entries of 296, no entries for codes 696, 806, or 846, and a handful of entries for the other four categories. Thus, only in the cases of allocation codes 196, 246, and 296 does the assignment of SEI and prestige codes have any substantive import. In the course of our efforts to improve the consistency and quality of the mapping from 1970-basis occupation lines into the SEI and prestige scores, we reviewed the sources of the scores for the allocation codes given above and of alternatives to them. COR226 (Appendix P) cites as its source for these codes D.L. Featherman, F.L. Jones, and R.M. Hauser, "Assumptions of Mobility Research in the U.S.: The Case of Occupational Status" (Social Science Research 4 (1975): 341). There, the sources of the codes are given as Duncan's 1961 chapter (in A.J. Reiss, Jr., Occupations and Social Status, Glencoe: Free Press) and Siegel's 1971 doctoral thesis. Duncan (1969) also prepared a supplementary listing, "Scores on Socioeconomic Index for Extended Major Occupation Groups, 1950," which gives scores for an extended, 18-category classification similar to the 17-categories used by Blau and Duncan (The American Occupational Structure. New York: Wiley, 1967). From Duncan's 1969 list, we can add a score of 80 for code 986 (Private household workers -- allocated); the score of 170 for code 976 (Service workers, except private household) remains unchanged. Duncan's list does not distinguish transport equipment operatives from other operatives, so both of codes 696 and 726 are assigned SEI=180. Siegel's (1971) list of prestige scores for all occupations also includes scores for various aggregates of occupations, and this is the source of the scores used by Featherman, Jones, and Hauser. However, their list actually rounds scores given by Siegel. From Siegel's list, it is possible to obtain a prestige score of 196 for code 986 (Private Household Workers). The following list assembles Duncan SEI and Siegel prestige scores for the full set of 12 allocated occupation categories from the 1970 Census, plus an alternative set of prestige scores that are used by the National Opinion Research Center (NORC) in its General Social Survey (GSS). The principle used, both in the present work and by Featherman, Jones, and Hauser has been to assign the score for an occupational aggregate to the allocated codes within that aggregate. We think that is appropriate because the allocation codes are used when assignment (at random or because of a vague response) cannot be made below the major group level and might in fact belong to any constituent line. However, the principle used by NORC in assigning its prestige scores has been to give each allocation category the score for jobs in the major group that could not be classified into any other detailed title in the group (n.e.c., for "not elsewhere classified"). Note that the GSS scores thus differ substantially from the Siegel scores for major groups in some instances. Code SEI Prestige GSS Occupation title ------ --------------- -------------------------------------------------------- 196 750 596 510 Professional, technical, and kindred workers -- allocated 246 570 504 500 Managers and administrators, except farm -- allocated 296 490 336 340 Sales workers -- allocated 396 450 394 360 Clerical and kindred workers -- allocated 586 310 388 470 Craftsmen and kindred workers -- allocated 696 180 287 320 Operatives except transport -- allocated 726 180 287 290 Transport equipment operatives -- allocated 796 070 184 170 Laborers, except farm -- allocated 806 140 407 410 Farmers and farm managers -- allocated 846 090 187 190 Farm laborers and farm foremen -- allocated 976 170 252 250 Service workers, except private household -- allocated 986 080 196 180 Private household workers -- allocated Since the only allocation codes that have been used with any frequency in the WLS data are the first three, 196, 246, and 296, and since their scores remain essentially unchanged relative to the scores used in the 1976-77 mapping, revision of the mapping to include these, our preferred scores for the 1970- basis allocation categories, has few if any substantive implications. Revised Lists of SEI (Duncan) and Prestige (Siegel) Scores for 1970-Basis Occupations For the record, we have created full listings of the updated and revised "male" and "total" SEI and prestige scores for 1970-basis Census occupation lines. The scores are shown in Table 3. These pertain only to three-digit occupation lines; see COR612C (a revision of COR521C incorporating the different Wave 1 nonresponse codes) for a mapping of split as well as non-split lines based on the updated "male" scores of Featherman, Sobel, and Dickens. The updated lists correct errors in Appendix B of Featherman, Sobel, and Dickens, and they included our preferred mappings of allocation codes. They do not include codes for missing data, which have sometimes been treated differently from one wave to another of the WLS data. In addition to the male and total SEI and Prestige scores, the list also includes the NORC update of prestige scores that was used in data released by the General Social Survey from 1972 to 1990. Note that the GSS prestige codes sometimes differ from those in the "male" or "total" lists based on the update by Featherman, Sobel, and Dickens. As noted above, there are differences in the treatment of allocation codes. Other procedures used in the NORC GSS scoring are explained in NORC General Social Surveys, 1992-94: Cumulative Codebook (Chicago: NORC-Roper, November 1994: Appendix F and Appendix G). A Note on SEI and Prestige Scores for Military Occupations Military occupations appear in two places in Table 3. First, code 580 pertains to Former Members of the Armed Forces, and it is assigned missing data codes on SEI and prestige, while GSS NORC assigns it a score of 47. Second, the WLS established its own coding procedures for current members of the Armed Forces. See the 1974-75 WLS Occupation-Industry Coding Handbook, 1974-75 and COR476 (August 1994) for details. The following codes were used both in the 1975-77 data and in the 1992-94 data. AFC (970) Armed Forces Commissioned Officer AFN (971) Armed Forces non-Commissioned Officer AFE (972) Armed Forces Enlisted AFX (973) Armed Forces no information on rank On the lists in Table 3, these codes have been assigned missing data codes for SEI and Prestige, and they have been assigned a score of 47 in the GSS list. There are codes for current or former members of the Armed Forces, given by Duncan, SEI=180, and by the GSS, prestige = 47(0). We have three reasons for not using these. First, in the WLS, we have consistently aimed to describe civilian occupations. Second, the SEI is based not on characteristics of persons in the military, but on "Former members of the Armed Forces" circa 1950. That is, it describes civilian men who had been in the military, but who had not held a civilian job since leaving military service. I doubt its validity in characterizing actual members of the Armed Forces. Third, the prestige score is not from Siegel's list of scores for 1960-basis occupation lines, but part of a list that others at NORC, probably James Davis and colleagues, produced at a later date and that appears regularly in the GSS codebook with a flag to indicate that it is "adapted" from Siegel. In the 1994 Cumulative GSS Codebook, Appendixes F and G, the "47(0)" that the GSS assigns to current or former members of the Armed Forces appears to come from the 46.6 that Siegel assigned to "Craftsmen and Kindred Workers, n.e.c." Since the rated titles were Skilled Craftsman on a Construction Crew (49.6), Skilled Craftsman in a Factory (45.8), and Skilled Craftsman in a Metal Working Shop (44.4), there is no special reason to take it seriously as a prestige rating for military occupations. In fact, Siegel reports (p. 90) that in the prestige studies on which his scale was based, there were ratings of some military titles: Captain in the Regular Army (63.3) Colonel in the Army (70.8) Corporal in the Regular Army (34.9) Enlisted Man in the Army (40.3) For whatever it may be worth, the average of these four scores is 52.3, but it is perhaps more important that they vary greatly between officers and enlisted ranks. It is not clear why the GSS did not use this information or how one might adapt it for use with the four categories used in the WLS. Perhaps, one might average the prestige scores for Captain and Colonel to represent officers in the WLS list, and one might average the scores for Corporal and Enlisted Man to represent Enlisted Men and "Armed Forces no information on rank" in the WLS list. However, no rated title corresponds to "non-Commissioned Officer" in the WLS list, and there does not appear to be any reasonable strategy for assigning an SEI code to the WLS categories for 1970-basis occupations.