These general notes duplicate some of the information that is available on the CAMSIS web pages. Further details can be obtained from these or, in the unlikely event that they prove unhelpful, the authors. Details that are specific to this version are in CAPITALS. These notes relate to the GB91SOC2000 PRELIMINARY version. THIS IS AN APPROXIMATE VERSION ALLOWING ACCESS TO CAMSIS SCORES FOR UK SOC2000 UNITS BUT BASED UPON 1991 DATA IN SOC90 SCORES. THIS VERSION WAS RELEASED IN FEB2003 PRIOR TO AN ANTICIPATED RELEASE OF A FORMAL VERSION DERIVED USING 2001 CENSUS DATA IN SOC2000 UNITS. The following files should be found in the SE90ISCO88.zip file: GB91SOC2000.README.TXT - this file! GB91SOC2000.SAV - spss system file GB91SOC2000.POR - a portable version of the above GB91SOC2000.DAT - the table index in plain text format UPDATES NOTE FOR GB91SOC2000 RELEASE : CURRENT PRELIMINARY RELEASE IS VERSION 0.1 FROM FEBRUARY 2003, UPDATED FROM 0.0 (AVAILDABLE FOR DONWLOAD BETWEEN AUGUST 2002 AND FEBRUARY 2003). NEW RELSEASE IS BASED UPON IMPROVED TRANSLATION OF GB91 VERSION IN LIGHT OF ADDITIONAL INFORMATION ON SOC1990 TO SOC2000 RELATIONS. 1) CAMSIS Scales The CAMSIS scale scores represent an occupational unit's relative position within the national order of social interaction and stratification. Different versions of the scale are produced for different countries and, in some cases, time periods. It is a standard outcome of the methods used that different scores are obtained for men and women. The CAMSIS scores supplied are BASED UPON GB91SOC90 SCORES. THESE ARE re-scaled representations of the 'raw' scores generated by 'RC-II' association models applied to patterns of social interaction between occupational unit groups, as defined in the relevant classification AND DATASET (SOC90 AND 1991 CENSUS). We use the same re-scaling parameters for every CAMSIS version, namely that a nationally representative population's distribution of scale scores should have a mean of 50 and standard deviation of 15. Higher scores reflect, in our interpretation, greater advantage along the stratification dimension which the scores represent. Additionally, we impose the constraint that all re-scaled scores which would fall outside the range 1-99 are 'cropped', to a minimum value of 1 and maximum value of 99. (In practice this constraint affects very few occupational unit scores.) Although, strictly speaking, the CAMSIS scores reflect only the differentiation of occupational unit groups within a specific version, this re-scaling does ensure a reasonable degree of comparability between versions; that is, they all represent the same order of distance from the version-specific mean. This property is especially relevant in the comparison of scale scores between gender groups, as discussed below. NOTE, HOWEVER, THAT FOR THE SOC2000 PRELIMINARY RELEASE, THE RESCALING TOOK PLACE IN TERMS OF THE 1991 SOC90 VERSION; MATCHED TO DATA IN SOC2000 UNITS, IT IS UNLIKELY THAT THE EXACT SAME DISTRIBUTION PROPERTIES WILL HOLD, ALTHOUGH EQUALLY THERE SHOULD NOT BE A DRAMATIC DEVIATION. In the first instance, many researchers may be interested in a descriptive review of the qualitative structure of the CAMSIS scores for the version relevant to them. This is most easily done by looking at the ranked order of occupational unit groups by their within-gender CAMSIS scale scores. In a few cases, such tables are provided on the associated web pages, but it is a simple matter to sort the SPSS index files by scale score. Using SPSS, switching on the 'view value labels' option will usually show the text labels for each occupational unit group; remember, though, to sort back to the original order (ascending occupation and status in employment) before attempting to match files. 2) Occupational Unit and Status in Employment Groups CAMSIS versions are created using the data available for a particular country at a particular time, using the occupational classification scheme to which the data are coded. In some cases it is possible to use datasets that are close in time, but have been coded to an alternative classification - most usually, the International Standard Classification of Occupations. In other cases, we have made use of a conversion scheme to convert the scores from one classification to another. Such conversions are rarely wholly satisfactory, but they are probably adequate. This version uses the UK'S STANDARD OCCUPATIONAL CLASSIFICATION 2000, LINKED FROM STANDARD OCCUPATIONAL CLASSIFICATION 1990 DATA BY CONVERSIONS OBTAINED FROM THE OCCUPATIONAL INFORMATION UNIT (OCCUAPTION.INFORMATION@ONS.GSI.GOV.UK) BASED ON DATA FROM THE 1991 UK CENSUS. Although we recommend using the fine detail of any occupational classification whenever possible, there are situations where data are only available at a more aggregate level, coded, say, into minor or even just major group. Where appropriate, the index files include scores for groups at these higher levels of aggregation. Wherever possible, we have tried to use what appears to be the standard way of coding these aggregated groups. AGGREGATED SCORES FOR 1- AND 2-DIGIT LEVEL GROUPS ARE PROVIDED USING THE CODES IN THEIR ORIGINAL 1- OR 2-DIGIT FORM. In addition to occupation, strictly defined, many countries collect information on status in employment, for example, whether someone is self-employed or an employee. Differences in employment status of this kind could be regarded as, in effect, defining different occupations, and this is how they are treated in CAMSIS. The files supplied provide for both the standard national classification of employment status and a standardised form that can be used in international comparative analysis. The tables of scores allow users to make use of more and less detailed employment status categories, including one for no information, that is, an occupational title only version. THIS VERSION HAS TWO EMPLOYMENT STATUS VARIABLES, THE STANDARDISED VERSION (STDEMPST) AND THE NORMAL UK 1991 VERSION (UKEMPST). It is possible for users to match their data with appropriate CAMSIS scores using information on occupational title alone or occupational title in combination with employment status. The occupational title variable should be numerically equivalent to the occupational title units used for the relevant CAMSIS version (STANDARD OCCUPATIONAL CLASSIFICATION 2000). If it is relevant for the particular country, an employment status variable will also be needed, coded to match the index files (SEE BELOW). If there is no information on employment status, this will need to be set to '0'; this is necessary even if employment status is missing for every case in order to avoid duplications in the look-up. In some cases, coding the employment status variable may require cross-classifying information from more than one source variable. (Normally, users should take care to code as 'not known' (0) any cases where employment status situation as defined in the CAMSIS scheme is not completely clear.) EMPLOYMENT STATUS CODES USED ON THE GB90SOC2000 FILE STDEMPST 0 "Status unknown (UKEMPST 1,2,3,4,5,6,7)" 1 "Self-employed (all) (NOT KNOWN FOR UK)" 2 "Self-employed (principals) (UKEMPST 1,2,3)" 3 "Own account (ICSE-93 3) (UKEMPST 3)" 4 "Employer (ICSE-93 2) (UKEMPST 1,2)" 5 "Family worker (ICSE-93 5) (not available for UK)" 6 "Employee (ICSE-93 1) (UKEMPST 4,5,6,7)". UKEMPST 0 "Unknown" 1 "EMPLOYERS, LARGE ORGANISATIONS (25+ EMPLOYEES) 2 "EMPLOYERS, SMALL ORGANISATIONS (LT 25 EMPLOYEES)" 3 "SELF-EMPLOYED, NO EMPLOYEES" 4 "MANAGERS, LARGE ORGANISATIONS" 5 "MANAGERS, SMALL ORGANISATIONS" 6 "SUPERVISORS" 7 "EMPLOYEES" . NOTE HOWEVER THAT THE DISTINCTION BETWEEN UKEMPST CATEGORIES 4 AND 5 IS MAINTAINED FOR THE CONVENIENCE OF MATCHING WITH OTHER STANDARDISED FILES ONLY. IN PRACTICE THE DISTINCTION WAS NOT USED IN THE CAMSIS SCALE CONSTRUCTION SO THERE SCORES OF EITHER UNIT SHOULD BE EQUIVALENT. Matching to Data Files The CAMSIS index files are normally made available only in SPSS format or as ASCII (plain text) files. Versions as Excel worksheets can be provided on request. The SPSS version is the most convenient and is likely to be the most commonly chosen. Within SPSS, it is possible to link files between the users own data and the CAMSIS index file. Firstly, the CAMSIS file should be sorted by the occupational and employment status variables which will be used, and, though not essential, only the relevant employment status variable retained. (When released, the files are usually sorted by one of the two employment status variables already, but this step is worth taking just in case it is not the variable which will be used) : . ** Example SPSS syntax for preparing UK90SOC2000 file for linking with own data. ** and using the employment status variable UKempst : . get file="UK90SOC2000.sav". sort cases by SOC2000 UKEMPST. sav out="UK90SOC2000mtch1.sav" /keep=SOC2000 UKEMPST mcamsis fcamsis. Then, assuming that the user has a data set in which each record has one or more variables relating to occupation (e.g. respondent's present job, previous job, spouse's job, father's job), the index file functions as a look-up table using the SPSS Merge files/Add variables function. Assuming two variables "occ" (SOC2000) and "empst" (UKEMPST or stdempst) relating to a particular occupation, all that is needed is, first, to sort by "occ" and "empst", then to match with the index file: sort cases by occ (a) empst (a). match files /file=* /table="UK90SOC2000mtch1.sav" /rename SOC2000=occ UKEMPST=empst /* or stdempst /by occ empst . execute. The occupational index files supplied have a unique record for each possible combination of occupational title "occ" (SOC2000) and employment status "empst" (UKEMPST OR STDEMPST). Associated with each are the relevant CAMSIS scale scores for men "mcamsis" and women "fcamsis". The most convenient way to deal with the different scores is first to set a variable, say "camscor" equal to "mcamsis", then to change the value to equal "fcamsis" where the respondent is female. (There may be situations, though, where it is thought desirable to retain both values for each occupation. In particular, it could be argued that it would be more appropriate to assign the male score to women's occupations if men and women are being compared as to the 'desirability' of their occupational locations - see below.) Remember that for each occupational variable for which CAMSIS scores are being added the file will have to be re-sorted. Remember, also, that for those countries where employment status is not included (or where it is included but in categories which cannot be compared with either of the variables on the CAMSIS index files) an arbitary employment status value of 0 (for 'unknown') should be calculated and used as the employment status index variable. Other Variables THERE ARE SEVERAL OTHER VARIABLES SUPPLIED WITH THIS FILE WHICH LINK SOC2000 - BY - EMPLOYMENT STATUS UNITS WITH OTHER RELEVANT CLASSIFICATIONS. SOME OF THE VARIABLES ARE ONLY RELEVANT AS INTERIM VARIABLES FOR INFORMATION ON HOW THE FINAL LINKAGE WAS DERIVED. THE VARIABLE NAMES OF MOST RELEVANCE TO USERS ARE: NS_SEC : THE NATIONAL STATISTICS SOCIO-ECONOMIC CLASSIFICATION, SEE http://www.statistics.gov.uk/methods_quality/ns_sec/ RCSC : AN APPROXIMATION TO THE REGISTRAR GENERAL'S SOCIAL CLASS CATEGORIES (USED WIDELY IN THE UK THROUGHOUT THE 1990'S) SEG : AN APPROXIMATION TO 'SOCIO-ECONOMIC GROUPS' AS USED WIDELY IN THE UK OVER THE 1990'S. Gender Groups It is important to remember that the CAMSIS scales are derived within the context of gender groupings; the male scales represent the ranking of the occupations males in a hierarchy of social interaction, and the female scales a ranking of those of females. Thus, for example, there is no necessary relationship between the values of an occupation on its male and female scales (although they are likely to share similar relative locations). An implication is that the analysis of social stratification through occupations graded by the CAMSIS measure is ideally appropriate only within gender groups. Indeed, it is a little misleading, albeit a commonly made mistake, to analyse a mixed gender population through CAMSIS scores which are the male scale scores for the men and the female scale scores for the women. The occupational scale indexing used for men and women is invariably the same, further giving the impression of equivalent meanings. However the CAMSIS methodology assumes different systems of relative positions prevail within the male and female occupational structures, and hence implicitly that equivalent titles are not necessarily the same between genders. There are however two data arrangements where, if the use of a mixed gender population is a substantive priority, use of CAMSIS scale scores for that population can be supported. If either are adopted, we would also recommend that researchers using modelling techniques consider terms which represent the interaction between the CAMSIS scale values and gender, since the patterns associated with such terms may shed light on the relative influence of gender differences. Firstly, it may be plausible to represent the occupational position of both men and women through their scores on the scale derived solely for male occupations. This position would represent a 'conventional' view on occupational locations, that among the working population the advantages of occupations for all workers follow the hierarchy of the predominantly full-time, lifelong career, male occupational structure. Secondly, an alternative account would note the fact that we have re-scaled all CAMSIS scores within versions to have the same mean and standard deviations. Because of this, male and female scores have some equivalence, in that they represent each occupation's average position with respect to the population of men for men, and women for women. Because of this, the CAMSIS scale representation alluded to above, namely the commonly used attribution of male scores for men and female scores for women, whilst imperfect, is defensible. Implications of Pseudo-Diagonals In almost all versions we treat a number of husband-wife occupational combinations as 'diagonals' or 'pseudo-diagonals' (that is, as combinations which we consider occur much more often than would be expected from their respective general locations in an order of social interaction / stratification, for some identifiable, but relatively trivial, economic or institutional reason). For the majority of occupational units analysed, the proportion of cases within that unit which are excluded on the grounds of being a pseudo-diagonal are relatively small, for instance less than 1% of all cases representing the unit group. However, it is also typically the case that for a small number of distinctive occupations, many cases, indeed the majority of cases from that occupational unit, are excluded as being pseudo-diagonals. This issue is perhaps best illustrated by comparing two examples. In many versions, two specific pseudo-diagonal combinations are those where husband 'farmers' are married to wife 'agricultural workers', and those where husband 'doctors' are married to wife 'nurses'. In both cases, by treating the combination as a pseudo-diagonal we effectively remove its ability to influence the general scale score location assigned to each occupation. For the latter example, we therefore evaluate the position of male doctors from the pattern of their wives' occupations, excluding those cases where their wives are nurses, and we evaluate the position of female nurses from the patterns of the husbands' occupations, but excluding those cases where their husbands are doctors. In this example, as indeed is the case with most pseudo-diagonals used in the scale construction, only a small proportion of the wives of doctors or the husbands of nurses are excluded, and therefore we are reasonably confident that the occupational units remain sensibly represented. However, in rarer cases, such as the former example, when we exclude from the evaluation of male farmers all cases where their wives are agricultural workers, and exclude from the evaluation of female agricultural workers all cases where their husbands are farmers, we actually exclude a high proportion of cases from each occupational unit group. In such circumstances we are less confident that the two occupations have been appropriately represented in our scale derivation procedure. The issue is significant because our treatment of pseudo-diagonals implies a theoretical position that, whilst a general structure of social interaction (and stratification) between occupational units exists, it is overlain by more specific structures of social interaction which are determined by economic and institutional factors. This implies that all individuals in marital combinations determined by such 'pseudo-diagonal' reasons are distinguished from those in combinations determined by the general structure of interaction / stratification, and therefore their occupational location has (or at least could have) a different meaning for their interaction / stratification position than does the occupational location of a non-pseudo-diagonal individual. For instance, the occupation of a female agricultural worker means, or could mean, something different about her general stratification position if her husband is or is not a farmer. At present, the derived CAMSIS unit scores tell us only about the relative location of occupations whose incumbents are not part of pseudo-diagonal combinations, and moreover only define that location within the structure of closeness or distance to other, non-pseudo-diagonal, occupational combinations. A solution in principle, therefore, would be to assign different scale scores to individuals' occupations depending on whether or not their partner's occupation places them in a pseudo-diagonal. One possibility would be the ad hoc identification and reassignment of particular scores. In particular, it may be thought preferable to assign to female agricultural workers whose husbands are farmers the derived CAMSIS score of female farmers, rather than that of female agricultural workers. However, such ad hoc solutions are difficult to implement consistently when applied to a wide range of other data sets.