MEMO 121 Using Data on IQ and High School Rank from the WLS (With additional notes on allocation flags and recoded items from the 1957 Questionnaire) November 1996 It has come to our attention that some key materials are missing from the machine-readable versions of the codebooks for the 1957-77 WLS data. We are adding the following notes to the machine-readable index and codebook, and they will be available in the near future. However, for full information, users of those data must obtain paper copies of the appendices and of the codebook for the 1957-77 WLS data. They are available from the Data and Program Library Service at the University of Wisconsin-Madison and from the Inter-University Consortium for Political and Social Research at The University of Michigan. We have no plans to put all of that information into machine-readable form. The machine-readable versions are useful tools for searching and building extracts but are not a substitute for the complete documentation in hard copy. (This differs from the 1992-94 round of the WLS, for which all documentation is available in ASCII or WP5 files.) 1 - ALLOCATION FLAGS Allocation flags indicate whether some method other than "straight" coding was used to assign a code to a variable. For occupation and industry coding there is a standard method which is explained in Appendix P on occupation coding. For the century month allocations, if a month was missing, the month of June was assigned unless it would produce inconsistencies in the data. Any case which had a month assigned for purposes of creating century months was flagged. See Century Month Allocation Flags in Appendix C and note that these flags reflect original source variables. (This note also appears on p.1475 of the paper copy of the 1957-77 WLS codebook.) 1957 QUESTIONNAIRE DATA - RECODED The variables in columns 101-186 were added to the file in 1/83. These are recoded versions of variables from the original April 1957 Questionnaire of Wisconsin High School Seniors (cols. 101-147) and from a supplementary questionnaire mailed to the parents of Kenneth Little's original 1/6 sample in October, 1957 (cols. 148-186). All recoded variables or those coded for the first time for this update have variable names which end in "Q". Some of the questions on the April 1957 questionnaire were coded for the earliest files but the variables were not carried forward on the master files. This set of variables appeared for the first time on the WLS22 Master File and is described in detail in COR #336 - Appendix G. 2 - EXPANDED CODEBOOK DESCRIPTIONS OF IQ VARIABLES. (SOME OF THIS INFORMATION IS CONTAINED IN APPENDIX G - COR 336.) RECOMMENDED BEST MEASURES TO USE: IQHNQ OR IQHNSCRQ (As explained below, we are planning to update these measures with raw scores on the Henmon-Nelson Test of Mental Ability and with nationally normed IQ scores, as specified in the 1954 revision of the manual for the H-N test. When the updated scores are available, they may be preferable to IQHNQ or IQHNSCRQ. The following discussion pertains to Wave 1 WLS data as released from 1983 to 1996.) 63 - 65 IQHNQ IQ - BEST MEASURE of Henmon-Nelson Centile Rank (1957Q) Updated version of IQ. This variable contains IQJCR if available else the recoded (1983) value from the original 1957 questionnaires - Recoded and added to file 1/83 as per CORS #332 and #336. If Henmon-Nelson Score was missing from the original 1957 questionnaire, then IQHNQ = IQ. See IQSOURCE (col. 2138) for source flag. See Appendix G for difference between this and other IQ variables. See IQHNSCRQ for IQHNQ-Normalized. IQHNQ From 1933 to the late 1950s or early 1960s, Wisconsin's Cooperative State Testing Program regularly administered, scored, and recorded the performance of Wisconsin secondary school students on forms A, B, or C of the Henmon-Nelson Test of Mental Ability. During that period, coverage was nearly universal. While national norms were occasionally updated -- the test itself was not revised (until 1957), and it is not clear when, if ever, the revised form was introduced in Wisconsin. (After the revision, both editions of the test were still published.) Before 1949, tests were typically administered in the sophomore and senior years, and later, tests were typically administered in the freshman and junior years. Records of the testing service are organized by calendar year, grade level, and school, and after the mid-1930s, each student record includes a raw test score and a (per)Centile Rank (CR). Until about 1950, the CRs were based on the students taking the test in each year. After that, the CRs were based on the performances of freshman (in 1949 and 1950) or juniors (in 1950 and 1951). Thus, depending on the source of the test score (freshman or junior), the post-1950 CRs were based on different populations as well as different age-grade performance levels. Each form of the Henmon-Nelson test is a 30 minute test consisting of 90 items. The items are presented in order of increasing difficulty. According to a review by Ann Anastasi (Buros' Mental Measurement Yearbook, 2:1394), the test includes "vocabulary, sentence completion, disarranged sentences, classification, logical selection, series completion, directions, analogies, anagrams, proverb interpretation, and arithmetic problems. Spatial, as well as verbal and numerical materials, are employed. The different types of items are not segregated, but are arranged in a 'scrambled' sequence." However, at least in Form A, 52 of the 90 items are verbal. H.M. Fowler (Buros 4:299) reports that, "these tests are patently measures of that highly verbal component sometimes called 'general intelligence' but perhaps more aptly labelled 'scholastic aptitude.'" When H-N test data were originally collected by Kenneth J. Little and supplemented by W.H. Sewell and associates, only the CRs were recorded, not the raw scores. In almost all cases, the scores for WLS graduates were obtained either in the fall of 1953 (freshman year) or in the spring of 1956 (junior year). No flag was used to indicate whether the CR had been obtained in the junior year, which was preferred, in the freshman year, which was used when a junior year score was unavailable -- or whether the test score had been obtained from some other source. Neither was the age of the student at testing recorded. The H-N manual reports a translation table from raw scores to IQ by quarter of birth up to age 16 and for all persons aged 16 or over. Thus, to use the national norms, one should want to know age at administration for individuals who took the test in the freshman year. (In many cases, it is possible to guess whether the test score was probably obtained in the junior year or in the freshman year because certain CRs could appear only in the scores for one or the other year. However, that indication is not definitive, and in many other cases the CR could appear for tests administered in either year.) In more than 7000 cases, IQHNQ is based upon CRs that were entered on the original 1957 WLS questionnaires, as recoded in 1983 (see discussion of IQ and IQSCOR below). However, in connection with the selection and interview of approximately 2000 randomly selected siblings of WLS graduates, in 1978 a new search for H-N test scores was conducted both for the siblings and for their graduate brothers and sisters. Records were found for 2151 junior year scores and an additional 132 freshman year CRs and raw scores. In the former cases, the CR from the junior year was entered in IQHNQ. In an additional 748 cases, there was no indication of H-N performance on the 1957 questionnaire, and the original "IQ" variable was entered as IQHNQ (See Appendix G, COR336, Addendum III). Presumably, these were cases in which IQ had been ascertained directly by letter from schools in 1964. It is perhaps unfortunate that no separate search of Testing Service records was conducted for those cases in 1978, but in the great majority of cases where a freshman score, but no junior score was found, the freshman CR agrees with the original entry in "IQ." 66 - 68 IQHNSCRQ IQ SCORES- BEST MEASURE of Henmon-Nelson Percentile Rank Normalized on Wisconsin High School Juniors (1957Q) 3-digit transform of source code IQHNQ. Added to file 1/83 as per COR #336 - see Appendices D and G. IQHNSCRQ In the absence of age at testing and raw scores on the test, and despite the mixture of scores normed on Wisconsin juniors and freshman, the most appealing procedure was to translate the percentile ranks into nominal IQ scores, as if all of the scores had been obtained using a single set of test norms in a population reflecting the full range of intellectual performance. This translation from CRs to IQ is reported in Appendix G, COR 336, Attachment VII. The implication of this procedure, assuming that some attrition related to test score occurred before the freshman and/or junior years, is that the upper tail of the distribution if IQHNSCRQ is too thin and the lower tail of IQHNSCRQ is too dense, relative to the distributions that would have been obtained using national norms for the same cohorts. Aside from the mixture of norms in the CR data, another problem with IQHNQ is that the CRs fail to discriminate scores in the extreme upper and lower tail of the distributions. That is, at the extremes, a number of raw scores would enter each percentile rank, and it is not possible to distinguish among these scores when only the CRs are available. Note that, although the codebook entry refers to norming on high school juniors, the CRs were in some cases normed on freshmen, rather than juniors. NOT RECOMMENDED: IQ AND IQSCOR - VARIABLES USED IN ALL PUBLICATIONS PRIOR TO 1983 RECODING. 69 - 70 IQ Henmon-Nelson Percentile Rank IQ (1957) Original version of this variable used for all publications prior to 1983. Has obsolete coding conventions described on COR #336 - see Appendix G for difference between this and IQHNQ. See also IQSCOR for IQ-Normalized. IQ For half the sample, Kenneth J. Little's 1/6 sample, CR for the junior year was coded as reported on the 1957 questionaire. In the other half-sample, William H. Sewell's supplementary 1/6 sample, CR was rounded to the nearest digit ending in "2" or "7," apparently as a convenience for punchcard tabulation. These data were recoded in 1983, and the full detail is included in IQHNQ. Note that, although the codebook entry refers to norming on high school juniors, the CRs were in some cases normed on freshmen, rather than juniors. This variable is retained in the current file only because it was used in published papers before 1983. 71 - 73 IQSCOR IQ Scores--Henmon-Nelson Percentile Rank Normalized on Wisconsin High School Juniors (1957) Original version of this variable used for all publications prior to 1983. 3-digit transform of source code IQ. Has obsolete coding conventions described on COR #336 - see Appendices D and G for difference between this and IQHNSCRQ. IQSCORE This is the transformation of "IQ" into nominal IQ scores, based on the transformation reported in Appendix G, COR336, Attachment VII. Note that, although the codebook entry refers to norming on high school juniors, the CRs were in some cases normed on freshmen, rather than juniors. This variable is retained in the current file only because it was used in published papers before 1983. IQ DATA WERE COLLECTED A SECOND TIME FOR FRESHMAN AND JUNIOR TEST SCORES FOR THE 2438 INDIVIDUALS WHOSE SIBLINGS WERE SELECTED FOR THE 1977 SIBLING SURVEY. THIS ADDITIONAL INFORMATION AND THE CORRECTION OF THE OBSOLETE CODING CONVENTIONS USED IN THE ORIGINAL VARIABLES (IQ AND IQSCOR) WERE USED TO CREATE THE NEW "BEST MEASURES" IN IQHNQ AND IQHNSCRQ. 74 - 76 IQFRS IQ-Freshmen Henmon-Nelson Raw Score (1983) Added to file 1/83 as per CORS #332 and #336 for N=2438 (sample members whose siblings were selected for 1977 sibling survey.) See also IQFCR for Centile Rank and IQFSCOR for Centile Rank-Normalized. 77 - 79 IQFCR IQ-Freshmen Henmon-Nelson Centile Rank (1983) Added to file 1/83 as per CORS #332 and #336 for N=2438 (sample members whose siblings were selected for 1977 sibling survey.) See also IQFRS for Raw Score and IQFSCOR for Centile Rank-Normalized. 80 - 82 IQFSCOR IQ-Freshmen Henmon-Nelson Centile Rank-Normalized (1983) Added to file 1/83 as per COR #336- Attachment VII, p.1 for N=2438 (sample members whose siblings were selected for 1977 sibling survey.) 3-digit transform of source code IQFCR. See also IQFRS for Raw Score. 83 - 85 IQJRS IQ-Junior Henmon-Nelson Raw Score (1983) Added to file 1/83 as per CORS #332 and #336 for N=2438 (sample members whose siblings were selected for 1977 sibling survey.) See also IQJCR for Centile Rank and IQJSCOR for Centile Rank-Normalized. 86 - 88 IQJCR IQ-Junior Henmon-Nelson Centile Rank (1983) Added to file 1/83 as per CORS #332 and #336 for N=2438 (sample members whose siblings were selected for 1977 sibling survey.) See also IQJRS for Raw Score and IQJSCOR for Centile Rank-Normalized. 89 - 91 IQJSCOR IQ-Junior Henmon-Nelson Centile Rank-Normalized (1983) Added to file 1/83 as per COR #336- Attachment VII, p.1 for N=2438 (sample members whose siblings were selected for 1977 sibling survey.) 3-digit transform of source code IQJCR. See also IQJRS for Raw Score. IQFRS, IQFCR, IQFSCOR, IQJRS, IQJCR, IQJSCOR In connection with the 1977 sibling survey, WLS project staff searched records of the Wisconsin Cooperative State Testing Service for junior or freshman test scores of 2438 members of the WLS graduate sample whose siblings had been selected for possible interview. Records were found for 2151 junior test scores and 1498 freshman test scores. The latter included 132 freshman scores for individuals who were not listed in records for the junior year. IQFSCOR and IQJSCOR are based on the same nominal transformation into the IQ metric that was applied to IQHNQ and IQ (See Appendix G, COR336, Attachment VII). However, because the raw scores have also been reported in each of these cases, it is possible to apply other available norms for the Henmon-Nelson test. Published norms are available from editions of the H-N manual published in 1946 and 1954. 92 - 100 GEOG57 Address of High School Attended in 1957 (1975) Coversheets (not available on public release) Includes state, county and place. 3 - HIGH SCHOOL RANK NOTE: HSRANKQ and HSRSCORQ are the "best measures" for high school rank. The other variables (HSR and HSRNRM) were used for all publications prior to 1983. 52-54 HSRANKQ High School Grades Percentile Rank (1957Q) updated version of HSR. Recoded and added to file 1/83 as per COR #336-see Appendix G. See HSRSCORQ for HSRANKQ-Normalized. 000-099 Lowest to highest 999 Not ascertained NOTE: To calculate percentile rank: HSRANKQ = 100 - (rank in class/# of students in class X 100) 55-57 HSRSCORQ High School Grades Percentile Rank-Normalized (1957Q) 3-digit transform of source code HSRANKQ. added to file 1/83 as per COR #336-see Appendix G. 061-139 Lowest to highest 999 Missing data HSRANKQ and HSRSCORQ The source variable is HSRANKQ, and it was transformed into HSRSCORQ using the same mapping as from IQHNQ to IQHNSCRQ, Appendix G, COR 336, Attachment VII. Thus, HSRANKQ is a percentile rank, normed separately within each school, and HSRSCORQ is a re-expression of that rank in the same metric as IQHNSCRQ. 58-59 HSR High School Grades Percentile Rank (1957) Original version of this variable used for all publications prior to 1983. Has obsolete coding conventions described in COR #336-see Appendix G for difference between this and HSRANKQ. See HSRNRM for HSR-Normalized. 00 Missing data 01-99 Lowest to highest NOTE: To calculate percentile rank: HSR = 100 - (rank in class/# of students in class X 100) NOTE: 108 of these respondents whould have been coded the lowest legal code "01" and not missing data. This was discovered during the 1983 recoding of the April, 1957 questionnaire. See HSRANKQ for the updated version of this variable. 60-62 HSRNRM High School Grades Percentile Rank-Normalized (1957) Original version of this variable used for all publications prior to 1983. Percentile calculated within each high school. Has obsolete coding conventions described in COR #336-see Appendix G for difference between this and HSRSCORQ. 3-digit transform of source code HSR. 067-139 Lowest to highest 999 Missing data NOTE: 108 of these respondents should have been coded the lowest legal code "061" and not missing data. This was discovered during the 1983 recoding of the April, 1957 questionnaire. See HSRANKQ for the updated version of this variable. HSR and HSRNRM These are obsolete versions of high school rank (percentile) and the normalized transform of that variable. In half the sample, HSR was rounded to the digits "2" or "7" for tabulation purposes, and -- as noted above -- some low ranking individuals were coded as if they had missing data. These variables are retained in the current file only because they were used in published papers before 1983.