You are on page 1of 5

26 Interest Assessment 527

inappropriate operationalisation of the construct. Translating items from one


measure into another language or context may not produce equivalence because the
items are not the most appropriate for the second culture. The construct still may
apply but better items, more in line with the specific culture may be needed. So it
is very possible that construct equivalence can be established with very different
items and scales, each uniquely suited to a particular culture (i.e., emic issues).
Constructs can be equivalently represented with very different measures.
These issues make it especially difficult to establish the construct equivalence of any
measures. In this chapter the focus will be mostly on the simpler aspect of construct
equivalence, that of structural similarity. If the intra-measure relations (i.e., structural
relations) are similar across cultures, one has some support for the equivalence of the
construct. It should be noted again that construct equivalence across cultures is never
something that is concluded but it is something that is ever approximated.
After this very brief introduction to issues of cross-cultural examination of construct
equivalence, attention is given to interest assessment. Most of the interest measures,
models, and constructs have been developed and researched in the United States.

Holland’s Typology

Holland’s (1973, 1985, 1997) theory of career choice is the most widely studied and
influential theory of vocational interests in history (Borgen, 1986; Swanson &
Gore, 2000; Tracey & Rounds, 1993), and it remains a cornerstone of theory devel-
opment and career counselling practice. Holland’s is considered a person-environ-
ment typology (Rounds & Tracey, 1990), which articulates some broad, defining
principles. First, people and environments can be classified according to six catego-
ries or types: Realistic, Investigative, Artistic, Social, Enterprising, and Conventional.
Second, individuals, for whom these measured categories are said to constitute
work personalities, tend to seek work environments that capitalise on these traits.
Lastly, work performance is a product of the interaction between a person’s work
personality and the demands and rewards of the workplace environment (Swanson
& Gore, 2000). Among several extensions of this theory is the notion of congru-
ence, which is an index of the degree of fit between a person’s work personality
type and the demands and rewards of the work environment. The aim of career
counselling in the vocational decision-making process is to guide exploration of
careers that correspond to a person’s vocational type. This is presumed to facilitate
career satisfaction (Tranberg, Slane, & Ekeberg, 1993) and achievement (Assouline
& Meir, 1987). However, the validity of these assumptions has not been free of
controversy. Notably, Spokane (1985) touched off a debate with the assertion that
measures of congruence (for a review see Brown and Gore, 1994) were rather poor
predictors of academic and career outcomes. Several years later the debate contin-
ued in a special issue of Journal of Vocational Behavior (Tinsley, 2000) devoted to
various reviews and theoretical and empirical analyses of the person-environment
fit model. Notwithstanding the doubts of some researchers, the person-environment
528 T.J.G. Tracey, S. Gupta

model and congruence constitute the foundation upon which the contemporary
practice of vocational guidance is built. With this basic framework in mind, atten-
tion can now be turned to the various structural models that represent distinct uses
of Holland’s typology, with associated theoretical and practical implications.

Holland’s Hexagonal Structure

Holland (1973, 1985) articulated a structural hypothesis, which takes the shape
of hexagon, with the six RIASEC types located at the vertices of the polygon as
seen in Fig. 26.1. It is often interchangeably referred to as a circular structure

Social

Enterprising Artistic

Conventional Investigative

Realistic

Holland’s circular ordering of RIASEC types

R I A S E C

Gati’s three group partition of RIASEC types

Fig. 26.1 Spatial representation of Holland’s and Gati’s models of RIASEC types
26 Interest Assessment 529

Social
Facilitating Helping
S

E
Managing A Artistic

Business C I
Detail Nature / Outdoor

R
Data Mechanical
Processing

Fig. 26.2 Circular arrangement of PGI octant’s along with RIASEC types

because, as can also be seen in Fig. 26.2, a circle can be super-imposed onto the
equilateral hexagon, with the types distributed equally (every 60° apart) around
it. Holland’s hexagonal structure has meaning for the theoretical relations among
the variables in the model. The relative proximity among the types in the model
reflects the strength of their relations. This also implies a specific ordering of
variable relations in the model. For example, the correlations between variables
adjacent to one another in the model (R-I, I-A, A-S, S-E, E-C and R-C) should
be greater than variables in alternate relation to one another, or one step removed
(R-A, I-S, A-E, S-C, E-R and C-I). Further, variables in alternate relation to one
another should be more highly correlated than those variables that oppose one
another on the structural model (R-S, I-E and A-C). Finally, it follows that all
adjacent relations should be greater than all opposite relations. This circular
ordering is important with respect to the concepts of profile differentiation and
consistency and in determining person-environment match. It provides informa-
tion about what interests are similar and different and this can prove useful in
career decision-making.
The circular arrangement of RIASEC scores has received good support in U.S.
contexts. In a structural meta-analysis, Tracey and Rounds (1993) found support for
the claim that the RIASEC scales were fitted by the circle across a variety of scales
for U.S. samples. They found that there was no difference in fit of the circle across
gender or age (with the lowest ages being greater than 14 years old). They did
however demonstrate that the fit of the circular ordering to non U.S. samples and
ethnic U.S. samples was less strong, raising doubts about the construct equivalence
across cultures. Subsequent examinations on large representative samples of U.S.
ethnic groups (Day & Rounds, 1998) provided some support for the construct
equivalence of the RIASEC circular structure in the U.S. ethnic groups.
530 T.J.G. Tracey, S. Gupta

In another structural meta-analysis of RIASEC scales but focused on cross


cultural structural equivalence, Rounds and Tracey (1996) found that with the
exception of Iceland, no country showed a better a better fit for Holland’s hexag-
onal structure than that found in the United States. However, Japan and Israel
demonstrated roughly comparable levels of fit as compared to U.S. benchmark
samples. Notably, the authors were surprised to find that the fit of the circular
order model of the hexagon was significantly less even in countries that were
linguistically similar to the United States and with similar occupational structures
(Australia and Canada). In general, Rounds and Tracey (1996) found, “that
regardless of the level of analysis, the cross-culture structural equivalence of
Holland’s circular order model was not supported” (p. 324). The model fit differ-
ences were particularly pronounced when the U.S. samples were compared to the
international samples as a whole. On this score, the authors found that 94% of the
U.S. benchmark samples displayed a better fit than the international samples. In
seeking to account for this rather unexpected finding, the exploration of potential
moderator variables such as the gross national product or certain cultural values
(individualism-collectivism) of a country failed to provide any viable explana-
tions. A final meta-structural analysis of Chinese data by Long and Tracey (2006)
also revealed a poor fit of the RIASEC data to a hexagon. These results indicate
that construct equivalence has not been established and that care should be taken
in using and interpreting these RIASEC measures in non U.S. contexts. The scales
are measuring different things in the across cultures as indicated by the varying
degrees of fit to the circular model.

Gati’s Partition Model

Gati (1979, 1991) introduced his three-group partition model (see Fig. 26.1) for the
structure of vocational interests as an alternative to the reigning hexagonal structural
model posited by Holland (1973). Gati cited what he considered to be several con-
ceptual and empirical problems with the better-established hexagonal and circular
models and offered the partition model as an option that better accounted for the
relations among the variables in the model. Although the focus of this chapter pre-
cludes extensive discussion of the empirically based assertions about the fit of com-
peting models to data, it should suffice to say that the advent of more sophisticated
methods of structural analysis (Hubert & Arabie, 1987; Rounds, Tracey, & Hubert,
1992; Tracey & Rounds, 1993) cast real doubt about Gati’s (1979, 1991) claims.
The structure of interests articulated by Gati has both theoretical implications
and meaning for the career guidance process. In brief, the structure, which looks
like an inverted tree, indicates that the variables paired together at the bottom are
more highly related than any other two variables, as would be reflected in a correla-
tion matrix. For example, according to the arrangement of the variables in the
model, R and I are more highly correlated than R and A, R and S, R and E, R and
C and even I and A.
26 Interest Assessment 531

For the purposes of career guidance in selecting a desired occupation, one can
start at the top where no specific choice preference is indicated. Then the decision-
maker may select from one of the three clusters. After selection of a cluster is made,
the decision-maker may then select from among one of the broad interest types
within the cluster. Finally, a more detailed exploration of specific careers within the
selected interest type may take place. However, if the career decision-maker finds no
career in which she/he is interested, she/he may take a step back up the “tree” or
hierarchy and explore careers within the other type in the same cluster. Alternatively,
after the administration of an interest assessment measure, the respondent’s domi-
nant RIASEC type may be identified. Careers representing the dominant type may
be explored. If this perusal of possible careers fails to pique the interest of the
respondent, then the other type in the same cluster can be explored further.
The structural meta-analyses of Tracey and Rounds (1993) demonstrated that
Gati’s model did not fit U.S. samples as well as the Holland hexagon but it did fit
the data better in international contexts although there was still a good deal of vari-
ability. Subsequent structural meta-analyses by Rounds and Tracey (1996) and
Long and Tracey (2006) also supported this conclusion. These results indicate again
that there is questionable construct equivalence of RIASEC measures across cul-
tures, but that there may be some equivalence across non U.S. cultures with respect
to Gati’s partition model. This model certainly cannot be assumed valid as there
still was a good deal of variability with its fit varying across countries. Gati’s model
is much simpler than Holland’s hexagon and it is this simplicity that better fits
across non U.S. cultures. However the simplicity also means that it makes few
predictions (see Tracey & Rounds, 1993 for a discussion).

Spherical Structure of Vocational Interest

The spherical model is associated with the Personal Globe Inventory (PGI, Tracey,
2002), which is a three-dimensional, spherical model, with 18 scales and assesses
avocational activity and occupational preference as well as perceived competence,
or self efficacy. Two rather notable findings served as precursors in the development
of the PGI: (a) that the six RIASEC types typically used in most contemporary
interest measures were arbitrary and (b) that support was found for a third substan-
tive dimension in mapping interests. With regard to the first finding, Tracey and
Rounds (1995) found that people responded to occupational titles such that an
analysis of the placement of the titles revealed that they were uniformly distributed
around the circle, rather than clustering around the six RIASEC vertices. As a
result, the number of types is arbitrary; there is no natural clustering around the six
RIASEC types. So any number of types can be used to represent interests, not just
the six RIASEC types. Tracey and Rounds proposed an eight-type octant model and
found that it fitted the data at least as well as the more traditional six-type model.
The eight types are presented in Fig. 26.2 (with the RIASEC types superimposed
for comparison purposes). Tracey (2002) has found that the octant type model fitted

You might also like