|
Article Excerpt
THE CONCEPTUAL AND EMPIRICAL CENTRALITY OF occupations in many domains of sociology, along with the inclusion of occupational information in other social science fields, has fueled the development of occupation-based measures of socioeconomic status. The latter is a term for the relative position of persons, families, households, and other aggregates with respect to social and economic factors, particularly the capacity to create or consume the goods valued in postindustrial societies (Nam and Terrie 1982; Hauser and Warren 1997). In this paper, I apply a methodology used in the recent Nam-Powers-Boyd scaling of the U.S. census occupations to generate a scale for the measurement of the socioeconomic status of occupations, one that builds on the detailed occupational titles found in the recently revised Canadian census classification of occupations (the National Occupational Classification for Statistics [NOCS]). In order to justify the construction of such a measure, and to clarify the methods used, I begin with an overview of previous occupational scaling exercises. Following this overview, I turn to the construction of the new occupational scale and highlight some of its unique features for capturing dimensions of the occupational hierarchy in Canada. I conclude with a review of the more salient criticisms voiced in recent years against the use of composite occupational scales in studies of inequality and intergenerational mobility. In light of these criticisms, I also provide separate occupational educational scores and occupational earnings scores for those working in the occupations enumerated in the 2001 census.
HISTORICAL ORIGINS OF THE DEMAND FOR OCCUPATIONAL STATUS SCALES
In the early to mid-1900s, North American sociologists and political scientists created social maps of social relations and stratification hierarchies by studying small communities. In these studies, social scientists created hierarchies of social standing in two primary ways: by asking respondents to indicate the most important persons in their communities, or by asking respondents to sort a limited number of occupations into no more than 9-10 slots indicating high social standing, intermediate and lower social standing. These procedures permitted social scientists to develop profiles of occupational hierarchies that captured the social standing of these occupations in small communities.
Over time, two important events altered the emphasis placed on community-specific studies. First, the growth of large cities that accompanied twentieth-century urbanization made asking respondents to indicate important persons or to sort occupations less tenable as methods for describing the social stratification system of any given area. The capacity to know all members of a geographical community, and hence the ability to rank them, disappeared in larger towns and cities. Furthermore, the number of different occupations found in any one geographic place increased, while the familiarity of its occupants with the full range of likely occupations declined, particularly in large communities with diverse industrial structures (Reiss 1961; Nam and Terrie 1982). Second, larger surveys became possible, fueled by the increasing use and funding of social science research and by technological improvements in data capture, which today include telephone interviewing and computer-assisted interviewing. Instead of the holistic study of one community, using approaches that today would be described as akin to social anthropology, focused surveys emerged that were often national in scope.
One concomitant development in North America was the fielding of national surveys in which respondents were asked to rank the social standing of occupations. While the term "social standing" was often left undefined, the methodology sought to capture the prestige of any given occupation based on respondents' perceptions of the underlying status and power dimensions. The 1947 North-Hatt study in the United States was one of the first such studies, and was replicated in 1964 by the National Opinion Research Center (NORC). Then, in 1971, several survey-based scales were combined to produce a comprehensive prestige scale covering 203 occupations (Siegel 1971). Subsequent extensions include those by Nakao and Treas (1994) and by Treiman (1977), who developed the Standard International Occupational Prestige Scale and subsequently revised it (Ganzeboom and Treiman 1996). In Canada, early prestige studies were geographically constrained: in the late 1940s, Tuchman asked college students and job applicants to rank 25 occupations (Blishen 1958, 1967), while Rocher developed a scale for Quebec in the 1950s (Langlois 2002). Studies undertaken by Guppy and Siltanen (1977) in 1975 and by Goyder and colleagues (Goyder, Guppy, and Thompson 2003; Goyder 2005) focused on the local community of Kitchener-Waterloo in Ontario. In 1965, Pineo and Porter (1967) fielded the first national rankings of occupations; it remained the only national study until 2005, when a new SSHRC-funded survey was launched by John Goyder (see Goyder and Frank 2007).
The growth of larger and often national surveys of occupational prestige rankings occurred alongside increasing publication of detailed census-based occupational profiles. However, analysts face several problems with these census classification systems. First, these classifications have their own logics and rationales, and they do not automatically represent the monotonically increasing statuses of occupations. Thus, using a census classification often forces researchers to adhere to assumptions of nominal or, at best, ordinal data. Second, large numbers of categories (over 500 titles in recent census classifications) present statistical problems for researchers who wish to preserve the classification detail in their multivariate models. A large number of dummy variables must be created for regression analysis, and many empty cells are likely to be generated in log-linear or multinomial statistical techniques. Third, classification systems used by federal agencies tend to be quickly incorporated into other survey research designs because the classification systems carry with them the legitimating imprimatur of the state, and the rules associated with sorting data into categories often are freely available. Consequently, these rules provide a protocol for researchers on how to code their own data, and often incite use.
Taken together, these developments generated demand for ways in which detailed occupational data available at national or regional levels could be incorporated into research. Fueled by a theoretically based emphasis on social status and social class that rested on the writings of Max Weber (Blishen and Carroll 1982b; Nam 2000), a pragmatic question emerged: how could one handle the occupational detail generated by surveys and censuses? To answer this question, two main approaches developed.
THE OPTIONS: REDUCING CATEGORIES
The earliest efforts involved the development of ordinal classification schemes that parsimoniously collapsed occupational titles into a limited number of categories (generally < 20) along one or more relevant dimensions. Although earlier time points are noted for Britain (Jones and McMillan 2001), in North America the most full-blown early construction of a parsimonious ordinal classification of occupations was the Edwards classification system in the United States, published in 1917. This classification grouped census occupational titles according to "skill" (Powers 1982; also see Nam and Boyd 2004). In 1977, Pineo, Porter, and McRoberts introduced a widely used ordinal scale of 17 occupational categories that captured general educational development and skill properties of occupations (Pineo, Porter, and McRoberts 1977; Jones 1980). Other later initiatives include those by Drouilly and Brunelle (1988) and Bernard et al. (1994).
Both conceptually and theoretically, the aggregation of occupational titles into ordinal classifications is different from later initiatives that generated classifications representing social class (e.g., of the latter, see Wright 1979; Erikson and Goldthorpe 1993; Clement and Myles 1994; Goldthorpe 2000). The social class approach seeks to represent asymmetrical relationships of power that exist between groups and the individuals in them that result from economic-based interactions in the workplace. In addition to occupation, information about managerial and supervisory roles, the degree of autonomy and decision making, size of the workplace, and number of employees is often used to construct classifications of social classes. In contrast, the ordinal classification of occupations ranks individuals or the positions they hold within a hierarchy defined by social or socioeconomic status; it thus represents distributional aspects of stratification, commonly expressed as "who gets what" (Kerckhoff 1984).
THE OPTIONS: PRESTIGE-BASED SCALES AND PURE SOCIOECONOMIC SCALES
The second approach also focuses on existing occupational classifications, albeit those with numerous occupational titles, and transforms them into interval scales. It too represents the distributional aspects of stratification because it ranks occupational positions according to a socioeconomic hierarchy. In scaling occupations, two procedures dominate North American research although more can be identified (Nam and Terrie 1982; Miller 1991; Nam 2000; Miller and Salkind 2002). The first approach incorporates the results of surveys that ask respondents to rank a number of occupations according to perceived social standing, creating prestige-based scales. The second rests solely on the educational and income/earnings characteristics of occupations.
Prestige-Based Scales
Before 1950, most studies of occupational status assumed that the source of prestige of an occupation lay in the opinions of people rather than in the characteristics of the occupations, and that people could estimate and articulate the prestige levels (Treiman 1977; Powers 1982). Small community studies lent credence to the belief that prestige rankings could be readily produced by asking raters to rank a relatively small number of occupations. This methodology of ranking was employed in national surveys, such as the 1947 North-Hatt study in the United States, the 1965 Pineo-Porter (1967) study and the recent national survey by John Goyder (Goyder and Frank 2007). In these ranking studies, the occupational titles are limited in number, reflecting efforts to reduce respondent burden associated with ranking large numbers of occupations and to avoid rankings of occupations not well known to respondents. The North-Hatt study has 90 occupational titles, while the Pineo-Porter (1967) study uses 208. A subset of these titles is often similar to titles found in the census classifications. This similarity permits a procedure in which the original rankings of occupations are used to transform the much larger array of census occupational titles into "prestige" rankings.
The first step in such a procedure is to match titles found in "prestige" studies to those used in census classifications of occupations. Once a match is obtained, census data on the educational and income characteristics of the labor force are obtained for those "matched" occupational titles. Then, a data set is generated that contains the prestige rankings for the subset of occupational titles that match with the census classification titles along with the census-generated educational and earnings or income characteristics for the incumbents of each occupation. Using this database, a prediction equation is obtained for a limited number of occupations, the number determined by the possible matches that exist between the prestige study and the census occupational titles.
Y (prestige ranking) = a + [b.sub.1] education + [b.sub.2] income + error term (1)
Once the parameters of the equation are generated, the educational and income characteristics for the full array of census occupational titles can be substituted into Equation 1, thereby generating "prestige" rankings for all census occupational titles.
First produced in the 1960s, North American prestige-based occupational scales were widely used and were routinely updated for successive censuses. One reason for the popularity of the scales in Canada and the United States was their centrality in national investigations of intergenerational mobility and status attainment, such as the 1962 American Occupational Change in a Generation survey (Blau and Duncan 1967) and the 1973 Canadian Mobility and Attainment survey (Boyd et al. 1985). Today, these scales are also used in diverse disciplines that study aspects of earnings determination, health, and crime. In Canada, scales resting on the 1965 Pineo-Porter study of prestige rankings (Pineo and Porter 1967) exist for 1961, 1971, and 1981 census occupational titles (Blishen 1967; Blishen and McRoberts 1976; Blishen, Carroll, and Moore 1987), although temporal variations exist with respect to the reference population (male or total), methodologies, and number of occupations (see Boyd 2002b, Table 1). Recently, Goyder and Frank (2007) have applied the results of a national survey on the social standing of occupations to 26 categories that aggregate over 500 detailed census occupational titles. (1)
It should be noted that the early vocabulary of "prestige scores" has fallen into disuse, replaced by "socioeconomic status scores," often shortened to "socioeconomic scores" or "socioeconomic indexes" (SEIs). The reason for the changing terminology partly comes from studies conducted in the 1970s and 1980s into what was actually being captured by respondent ranking of occupations. Most concluded that the central property is the "socioeconomic goodness" of an occupation, which reflects the credentials and economic rewards associated with occupations rather than prestige in the sense of deference, power, and authority (Shils 1970; Goldthorpe and Hope 1974; Featherman and Hauser 1976).
Despite extensive use of prestige-based occupational scales, critics noted several difficulties with these scales. For one thing, if high nonresponse characterizes surveys that ask people to rank occupations, the claim that prestige scores represent the opinions of the entire underlying population may be questionable. For example, in the most recent national survey of occupational social standing in Canada, the response rate was 51 percent; further, a subset of respondents who indicated that the term "social standing" had no real meaning rated a select group of occupations somewhat differently than did other respondents (Goyder and Frank 2007). In addition, groups of people may rate occupations they know, or are close to, somewhat differently than those with which they are less familiar; also the gender of the incumbents or the sex typing of the occupation may affect rankings. In their national study of prestige, Goyder and Frank (2007) find that women are more likely than men to overrate their own National Occupational Classification (NOC) occupational group; however, a local study found that the age and the sex of the rater made no difference and that ratings were not depressed by specifying female incumbents of occupational titles. In fact, in jobs dealing with people, specifying a female incumbent increased the ratings (Goyder et al. 2003).
Second, unless prestige ranking studies are frequently repeated using titles that reflect temporal changes in the occupational structure and in classification schemes, the numbers of matches between census titles and occupations that respondents rank in a prestige study are likely to decline over time. Out of the 204 occupations ranked in the Pineo-Porter study, the following numbers were matched, respectively, with the 1961, 1971, and 1981 censuses: 88, 85, and 75. In the recreation of a Blishen prestige-based scale for the Canadian Classification and Dictionary of Occupations (CCDO) classification in the 1991 census, only 72 occupations could be matched with census titles (Boyd 2002b). The selection of specific occupational titles, along with the declining number of matches, increases the possibility that the occupations used to regress "prestige" on education and income or earnings are unusual and thus affect the coefficients in the prediction equation. Temporal alterations in census classifications of occupations can also heighten this basic difficulty of merging titles used in earlier prestige studies with those used in more recent census classifications. The new 2005 Canadian national survey of occupational prestige may remedy this difficulty, although to date, results only apply to 26 aggregated occupational groups (Goyder and Frank 2007; also see note 1).
A third problem is highlighted by the debate among U.S. scholars as to what the resultant scale actually captures. Because education and income (or earnings) from the census are used as predictors, one interpretation is that prestige ratings are being predicted, and that the scores are proxies for the prestige scores of occupations (Nam and Terrie 1982; Nam 2000; Goyder and Frank 2007). However, in the United States, the originator of the Duncan SEI scale (Duncan 1961a, 1961b) initially fluctuated in his interpretation, ultimately announcing that his SEI was not a prediction of the prestige ratings that occupations excluded from the NORC North-Hatt study would receive if included in such a study (also see Hauser and Warren 1997:213). As one commentator notes, those comments leave...
|
