The Weak Suffer What They Must — A Natural Experiment in Thought and Structure
The etic grid defines members, stewards, and reps as the “union side” as versus supervisors and managers on the “company” side. Furthermore, stewards are on the same level of management as supervisors while reps are equal to managers and vice presidents. As part of the larger study of the local that we were embarked on, I wanted to determine to what extent members agreed with this etic grid. The triads test was a way of measuring union consciousness. It arranged the terms “steward,” “rep,” “manager,” “supervisor,” “worker in the same department (or line),” and “worker in a different department (or line)” in all possible combinations of three. The instructions asked each respondent to select the item in each line that was most different from the other two. Weller and Romney (1988) and Bernard (1988) discuss this procedure and its history in anthropology. For more recent discussions of the triads multidimensional scaling representations, and further citations to the methodological and substantive literature on the topics see Romney et al 1997 and Romney and Moore 1998.
For the triads test I designed, there are three “pure” models. If the triad were “supervisor, co-worker, union rep” and a person selected “supervisor” as the most different, indicating similarity between workers and reps, it would imply a “union model.” If the person elected “coworker” it would indicate a conceptual scheme based on hierarchy as “co-workers” are less powerful than supervisors and reps. The choice of “rep” would indicate a workplace proximity schema as that is the feature that supervisors and co-workers share.
Tabulation of responses in a matrix of all possible relationships among items determines the similarity, “closeness,” or proximity of any two items. Every time a respondent does not select two items in single row of 3, every time a respondent selects the other item, the cell for the unselected pair in the matrix increases one point for similarity. Each row of each respondent’s responses thus adds one point to some pair. In this design, since all pairs occur 4 times, the maximum score for any pair is 4, indicating the greatest degree of similarity between the two items. All individual matrices are summed into an aggregate matrix to show the strength of similarity for the population. Then the counts are converted into percentages to develop a measure of proximity from 0 to 1.
As an example, consider a “perfect” model of union consciousness that the “etic grid” represents. All respondents would agree and always select the item that indicates that they are thinking in terms of union relationships. Clearly, they would select “manager” or “supervisor” as most different whenever they occur in triads with “steward,” “rep” and/or “co-worker.” Thus, manager-supervisor would score 100% and supervisor-steward, supervisor-rep, supervisor-same or different line worker would score 0. Likewise for Manager in all of these pairs. Steward-rep would be 100% and different line worker-same line worker would be 100%. Worker-steward and worker-rep cells would be 50% or .5 because in those 2 (of 4) triads where workers, stewards, and reps all occur, union membership does not distinguish among the items, but either workplace proximity or hierarchy does. Thus the highest score could be .5 in the steward-worker cells and the rep-worker cells. Table 1 shows this similarity matrix.
I used Anthropac (Borgatti 1996a) to make and score the triads tests. I then used Anthropac to test the individual proximity matrices for consensus. The aggregate matrices for each worksite indicate the strength of similarity between each pair of items. I used these aggregate proximity matrices in two further ways. First, I correlated them to one another to determine the similarity between any two groups (see table 2). For this, I eliminated the relationships which were universally deemed to be similar–manager-supervisor, worker-worker, and steward-rep–roughly, management with itself; members with themselves, and union officers with themselves. Eliminating these reduces the correlations from the full matrix of relationships and highlights differences rather than similarities, which are artifacts of the instrument. Groups with highly correlated matrices share ideas about the similarity of roles while those that are not correlated do not. Some of the correlations are negative. I can offer no interpretation for these except that they indicate difference. Second, I used Anthropac’s multi dimensional scaling program to develop a visual representation of the “distances” among the items. Those items that are similar cluster together; those that are not are dispersed in the diagram thus allowing a visual representation of similarities and differences that the aggregate proximity matrices measure. Table 2 shows the correlations among the “perfect model” and the union local’s staff (n=18), a public hospital (n=9), a group of industrial workers (n=11), and Schwab members in 1996 (n=6) and 1998 (n=25). To further test the reliability of this method, I administered the triads test to the staff of a different local in the same international union in Chicago (n=25) and Table 2 shows that their similarity matrix is highly correlated to the one for the staff of the local in question as well as the “perfect” model.
I used the triads test as a means to the end of determining the relationship between union consciousness and other variables. A similar measure administered to the stewards of the local gathered at their annual convention in 1996 indicated that there was no relationship between union activism and union consciousness (Durrenberger 1997).
There was a relationship between union consciousness and the structure of worksites. The aggregate similarity matrix for the triads tests of the union local’s staff was very close to the etic grid. I therefore used this as the measure of union consciousness. To the extent that a group of members showed a similar cognitive pattern, I concluded that they exhibited union consciousness. Only two worksites fit this pattern. One was a public hospital (Pearsons r=.89) where for various reasons the union is very strong and significant (Durrenberger 1997). The other was Schwab (r=.64). I concluded that this was because of the small size of the membership and the strength of the chief steward who had been in office for several decades and cultivated a vast network of relationships not only among members but management. Furthermore, just before I administered the triads tests, there had been a union victory which all of the members had celebrated. I argued that without time depth it was not possible to know whether such configurations were stable through time or whether they respond to such episodes as a victory or a defeat for the union at a site (Durrenberger 1997). As the milti-dimension scaling diagram in Figures 2 and 3 show, the one difference between Schwab (Figure 2) and the public hospital (Figure 3) was that the members at Schwab departed from the etic grid to place their steward hierarchically above rather than below the rep so their steward was on a par with the manager and the rep with the supervisor. Another, whch represents the relative power of the union in the two work places is the placement of the public hospital’s stewards and rep at a more higher level than supervisors or managers (Durrenberger 1997). The reversal of the steward/rep relationship at Schwab was an accurate reflection of the chief steward’s long association and developed networks with the management of the hospital. At the worksite, she in fact had more power than the rep, though the rep could reinforce it by putting the authority of the local behind her.
Other hospitals had neither the extensive union structure of the public one nor the history of recent victories or intricate networks of relations that connected the leadership of the unit of Schwab. I concluded (Durrenberger 1997) that union consciousness seemed to be related to features of structure and perhaps history or recent events.