Complexity of Organizational Identification : Measuring Ambivalent Identification

Organizational identification is the root construct of organizational phenomena. Many researchers have examined this concept in terms of identification (positive) and disidentification (negative). However, recently some researchers began focusing on ambivalent identification, which refers to the simultaneous appearance of both identification and disidentification. This new concept is likely to shed light on organizational phenomena that cannot be demonstrated by identification or disidentification alone. This study focuses on ambivalent identification. We started by creating a concrete definition for this concept. Further, we tried to develop a valid measurement for it through pilot and main studies. Consequently, we defined this concept in terms of cognitive and affective aspects and observed the complexity associated with ambivalent identification.


Introduction
Innovation in products, services, and the organization itself is required for achieving sustainable growth.The source of innovation is creativity (Amabile, 1997).In other words, innovation in the organization depends on how the organization manages and applies its members' creativity.
Organizational identification (OID) is the root construct in organizational phenomena (Albert, Ashforth, & Dutton, 2000).OID is related to the self-identity and decision-making of individuals (Tompkins & Cheney, 1985).Consequently, OID influences organizational phenomena because every organization is an aggregation of individuals.Assuming that innovation is an organizational phenomenon, it is likely that OID influences innovation when the organization actively promotes it.However, if one considers what is called "the dark side of OID" (c.f., Dukerich et al., 1998;Elsbach, 1999), the following contradictions appear.According to Pratt (2000), an organization that employs members who identify strongly with it would have increased homogeneity, which would consequently lead to groupthink.Thus, it becomes difficult for such an organization to change its strategy in order to adapt to environmental changes.Innovation is likely to occur with the existence of diverse viewpoints (Verganti, 2011).If creativity results from having a different viewpoint, then OID, which approximates members' personal-identity to organizational identity, may hinder creativity.However, because OID is thought to be a requirement for organizational phenomena, organizations should consider it seriously.That is, organizations may experience dilemma about whether to stimulate creativity or support OID.However, are these concepts necessarily incompatible?
Many researchers discuss the phenomenon of OID only in terms of positive and negative vectors (i.e., OID and organizational disidentification).However, recently some researchers have identified not only two vectors of identification but also ambivalent identification (AID), which is the simultaneous appearance of both OID and disidentification (Dukerich, Kramer, & Parks, 1998;Kreiner & Ashforth, 2004;Pratt, 2000;Sluss & Ashforth, 2007).According to Dukerich et al. (1998), a member who identifies ambivalently with the organization perceives conflicting feelings over it and engages in whistle-blowing and behavior change in order to resolve this conflict.While studying ambivalence, which is a super-concept of AID, Fong (2006) applied informational theories of emotion.He demonstrated that emotional ambivalence led to the interpretation that the person was in an unusual environment, and accordingly led to uncommon reactions to the environment.These uncommon reactions were based on creative thinking.As mentioned above, it is likely that research on AID could overcome the problem that previous OID studies had.
However, as a concept, AID is in the early stage of development.For example, although Pratt (2000) argued that AID is a partial failure of OID, Dukerich et al. (1998) argued that AID occurs when both OID and disidentification appear strongly.Their disagreement exemplifies the inconsistency in the conceptualization of AID.The reasons for this inconsistency may be (1) few studies have investigated this issue, (2) no definition has been established, (3) empirical research on this topic is difficult because scales to measure AID are underdeveloped (Note 1), and (4) inductive research is difficult because it is difficult to comprehend the phenomenon of AID through observation and interview (Note 2).The present study focuses on the resolution of points ( 2) and (3), and accordingly aims to contribute to (1).
Most studies on OID have treated identification as a cognitive concept.However, Tajfel (1978) defined social identity as "that part of an individual's self-concept which derives from his knowledge of his membership of a social group (or groups) together with the value and emotional significance attached to that membership," essentially stating that social identity consists of both cognitive and affective dimensions.Therefore, studies on OID in which social identity theory (Tajfel & Turner, 1979) is central, should consider both cognitive and affective dimensions (Harquail, 1998;Johnson, Morgeson, & Hekman, 2012;Van Dick, 2001).Accordingly, the present study considers AID as a cognitive and affective construct.Bleuler (1911) defined ambivalence as "the simultaneous existence of strong feelings of love and hate toward the same object" (cf.Eidelberg, 1968;Luescher & Pillemer, 1998).Similarly, Alexander (1951), Argyris (1957), andFreud (1913), defined ambivalence as the co-existence of conflicting feelings directed toward the same object.However, Sincoff (1990), defined ambivalence as "overlapping approach-avoidance tendencies, manifested behaviorally, cognitively, or affectively, and directed toward a given person or experience."Although the former four researchers treated ambivalence affectively, the latter treated it both cognitively and affectively.

Definition of Ambivalent Identification
Ambivalence has been measured through subjective and objective approaches (Constarelli & Colloca, 2004).Subjective approach refers to the direct measurement of ambivalence by using a scale.On the other hand, objective approach refers to the development of an algorithm that factors in positive and negative attitudes to calculate ambivalence toward a particular object.Studies that use this objective approach are based on the view of Scott (1966), who stated that the greater and more equal the opposite tendencies, the higher the degree of ambivalence (e.g., Costarelli & Colloca, 2004;Kaplan, 1972;Moore, 1972;1973;Tompson, Zanna, & Griffin, 1995).These studies that use objective approaches defined ambivalence as an attitudinal concept.That is, they treated ambivalence as manifold concepts because attitude consists of behavior, affect, and cognition (Kelman, 1958).
In addition, Scott (1966) suggested considering the weight of each component.The present study synthesized these suggestions and defined ambivalence as "the simultaneous and equal existence of the opposite feeling or thinking toward the same object."This definition of ambivalence should then operate on the definition of AID.Although some studies have examined AID (e.g., Dukerich et al., 1998;Pratt, 2000;Sluss & Ashforth, 2007) (Note 3), its definition has not been explicitly stated.According to Kreiner and Ashforth (2004), AID refers to the idea that "one can simultaneously identify and disidentify with one's organization (or aspects of it)."Their definition is vague because OID and disidentification were not clearly defined.In fact, because they considered only the cognitive aspect of OID, it is more likely that this definition does not accurately reflect the concept of ambivalence.
We examined various definitions of OID, which confirmed that most researchers defined the cognitive or affective dimensions of OID.For instance, researchers who considered the cognitive aspect, defined OID as "a perceived oneness with an organization and the experience of the organization's successes and failures as one's own" (Mael & Ashforth, 1992); "the degree to which a member defines him-or herself by the same attributes that he or she believes define the organization" (Dutton, Dukerich, & Harquail, 1994); "a psychological state wherein individuals perceive themselves to be part of a larger organization" (Rousseua, 1998); and "an individual's belief about his or her organization becomes self-referential or self-defining" (Pratt, 1998).On the other hand, the researchers who considered the affective aspect defined OID as "the degree to which an individual values having a specific organizational identity" (Harquail, 1998); "the feelings individuals experience about themselves in relation to the social referent and the value they place on that social identity" (Johnson & Morgeson, 2005); and "an individual's positive feelings about being one with a group" (Johnson et al., 2012).Observing the difference between both dimensions plainly, OID seems to be the perception and feeling of individuals as organizational members (Harquail, 1998), the perception and feeling of oneness with the group (Johnson et al., 2012), and the emotional attachment one has to this group and the knowledge of a certain group membership ( Van Dick, 2001).Assimilating these ideas, the present study thus defined cognitive OID as "the degree to which a member perceives himself or herself as part of an organization" and affective OID as "the degree to which a member holds positive feelings toward the organizational identity." Based on these definitions of both cognitive and affective OID and ambivalence, AID is therefore defined cognitively as "the process in which a member simultaneously and equally perceives and does not perceive himself or herself as part of an organization," and affectively as "the process in which member simultaneously holds both positive and negative feelings towards organizational identity in equal parts."

Measuring Ambivalent Identification
We first searched the existing literature for published scales that measured OID and disidentification, by using "word identification," and found 26 relevant articles (Note 4).The items in these scales were then classified into cognitive and affective OID based on the above definitions and the study of Johnson et al. (2012).Finally, 22 items were derived for the cognitive dimension and 13 items were derived for the affective dimension.

Scale Development
After developing a scale to measure AID by using the above items, we attempted to establish its validity.In this case, this study followed Hinkin's procedure (Hinkin, 1995;1998;Hinkin & Tracey, 1999).The procedure is composed of the following three steps: (1) content validity assessment, (2) item reduction, and (3) confirmatory factor analysis.
The assessment of content validity followed the procedure of Hinkin and Tracey (1999).Respondents first rated each of the 35 AID items according to the extent to which they believed the items were consistent with each of two dimensions' definitions of AID.Response choices ranged from 1 (not at all) to 5 (completely).The definition of one of the two dimensions was presented at the top of each page of the questionnaire.The results were then used for explanatory factor analysis.
Prior to conducting factor analysis, we examined the inter-item correlations of the variables and any variable that correlated at less than .40 with all other variables were excluded from the analysis.In this case, respondents rated the 35 AID items obediently.Thus, the items used here were different from the items used to assess content validity.Afterwards, explanatory factor analysis with varimax rotation was conducted.The criterion used to support the theoretical distinction was an eigenvalue greater than 1 (Kaiser criterion) and a scree test of the percentage of variance explained.Finally, to assess internal consistency reliability, Cronbach's alpha was calculated for each component.A Cronbach's alpha higher than .70 was considered to indicate good internal consistency reliability.Confirmation factor analysis was then conducted and construct validity was tested by goodness-of-fit indexes.This study used chi-square values (Note 5), CFI, RMSEA, GFI, and AGFI as goodness-of-fit indexes.
We conducted (1) a pilot study to eliminate extra items and modify the items, and (2) the main study to develop the scale for AID.

Method
Questionnaires were distributed to six seminars in two Japanese universities.The purpose of this questionnaire was explained in the classroom, before distributing it.One hundred and thirteen undergraduates responded to the questionnaire.The number of valid responses was 60 (53.1%).Respondents consisted of 38 men (63.3%) and 22 women (36.6%); 31 students were in the third grade (51.7%), 26 in the fourth grade (43.3%), and three did not specify their grades (4.9%).

Results
(1) Content Validity Assessment Explanatory factor analysis with varimax rotation was conducted to determine content validity.There were nine components with eigenvalues greater than 1.However, the examination of a scree test supported two components; the percentages of variance explained were 24.7% and 20.9% respectively.Finally, ten out of the 13 affective items were left.Eleven out of the 22 cognitive component items were left.
(2) Initial Item Reduction The correlation analysis found eight items of the cognitive dimension and six items of the affective dimension to have a correlation of more than .4with all other variables.Then, explanatory factor analysis with varimax rotation was conducted for these 14 items.This analysis identified two components that were labeled as affective and cognitive respectively; the percentages of variance explained by the affective and cognitive components were 31.5% and 26.3% respectively.No items were eliminated by this explanatory factor analysis.The internal consistency (Cronbach's alpha) was .741for the affective component and .554for the cognitive component.
(3) Confirmatory Factor Analysis This pilot study aimed at eliminating the extra items and modifying the sentences describing the items and therefore did not conduct confirmatory factor analysis.We asked respondents for their feedback regarding the questionnaire and suggestions for improving of the sentences.Consequently, we could make improvements according to the feedback by modifying items that were difficult to understand, before conducting the main study.

Method
The revised questionnaire was distributed during a class in the university.The purpose of this questionnaire was explained in the classroom, before distributing it.One hundred and nineteen undergraduates responded to the questionnaire.The number of valid responses was 37 (31.1%).The respondents consisted of 24 men (64.9%) and 13 women (35.1%); 15 students were in the second grade (40.5%), 10 in the third grade (27.0%), 11 in the fourth grade (29.7%), and 1 in the fifth grade (2.7%).

Results and Discussion
(1) Content Validity Assessment An explanatory factor analysis with varimax rotation was conducted and the Kaiser criterion identified two components.The percentages of variance explained were 43.5% and 35.4% respectively.All affective items showed high factor loadings.Seven of the eight cognitive items showed high factor loadings (see Table 1 for details).Therefore, it may be inferred that the questionnaire had content validity.
(2) Initial Item Reduction The correlation analysis identified six items in the cognitive dimension and six items in the affective dimension that had correlation coefficients of more than .4with all other variables (Table 2).Following this, explanatory factor analysis with varimax rotation was conducted for these 12 items.The results revealed two components that were labeled as affective and cognitive respectively; the percentages of variance explained in Table 1.Factor loadings for content validity assessment of affective and cognitive components were 39.7% and 27.9% respectively (see Table 3 for details).No items were eliminated after this explanatory factor analysis.The internal consistency (Cronbach's alpha) was .883for the affective component and .768for the cognitive component.(3) Confirmatory Factor Analysis The goodness-of-fit indexes of the confirmatory factor analysis indicated that this model fitted the data well (χ 2 (46) = 47.678;CFI = .991;GFI = .904;AGFI = .802;RMSEA = .032).Therefore, the questionnaire has construct validity.The analysis resulted in six affective items and six cognitive items (see Table 4 for details).

Discussion
The present study makes at least two contributions to existing knowledge.Firstly, AID was defined more strictly here than in previous studies, where vaguely defined concepts hindered the discussion of the phenomenon.As AID had not been clearly defined until now, defining it here may lead to and form the basis for future research.
Secondly, this study developed a scale to measure AID.As mentioned before, previous researchers had not studied AID empirically because no such scale had been developed until now.Thus, creating this scale contributes greatly to the development of future empirical studies.
Discussing OID cognitively means discussing the cognitive distance between organizational identity and self-identity and the perceived vector.Previously, this discussion was conducted in terms of two axes: "near or far" and "approaching or leaving."However, bringing the concept of ambivalence into this discussion leads to a new angle-"near and far" and "approaching and leaving."This point was overviewed previously, suggesting the complexity of the concept of identification.The present study could very well provide a lead to understand and acknowledge this complexity.Kreiner and Ashforth (2004) have attempted to measure AID.Although they defined OID cognitively, items in their tool that measured AID consisted of both cognitive and affective dimensions.Thus, there was a disagreement between the scale and the definition used in the study, which affected the scale's content validity.Furthermore, because they did not show how their AID measurement was developed, the construct validity was not also manifested.In contrast, this study considered both dimensions, constructed the items deductively, and tested for content validity.Accordingly, this study is more likely to avoid similar problems of content validity.
As shown from both, the pilot and main studies, the sample size is not sufficient to establish validity (particularly construct validity).The percentage of valid responses suggested that items that looked like "double-barreled" questions might have caused confusion among respondents.However, this problem may be attributed to not the present study but the inherent ambiguity of AID.Considering the previous, subjective approach to AID, these items look like double-barreled questions.Although the objective approach mentioned above (Costarelli & Colloca, 2004) may be useful to solve this problem.Previous studies did not find high intercorrelations between subjective and objective approaches (e.g., Tompson et al., 1995).That is, the objective approach could not be substituted for the subjective approach.Therefore, it is important to acknowledge that we need to develop a valid subjective approach first in order to develop a valid objective approach.
We used undergraduate students as sample.Compared to employees in a company, a university does not work on the undergraduates' socialization to a great extent.Thus, it is likely that they do not pay attention to the university's identity.In addition, Kreiner and Ashforth (2004) argued that when an organization sends inconsistent or contradictory messages to its stakeholders regarding what it stands for, and why and when an organization demands incompatibly from within a give role, AID would occur.These antecedents are all organizational factors.More employees may identify ambivalently with their organization than students may with their university, because the environment requires different behavior from the employees.Therefore, if we use employees instead of undergraduates as the participants, above problems might be resolved.That is, it is likely that employees may have responded to items for AID and consequently we could have gathered a sufficient sample size to establish validity.
Finally, if more future studies focus on AID and its measurement, AID would become a measurable concept, the mechanism of OID might be demonstrated more effectively, and the relationship between creativity and OID would be manifested.

Conclusion
This study focused on defining AID and developing the scale to measure it.Firstly, we examined some concepts of ambivalence, cognitive OID, and affective OID, accordingly defined cognitive AID as "the process in which a member simultaneously and equally perceives and does not perceive himself or herself as part of an organization," and affective AID as "the process in which member simultaneously holds both positive and negative feelings towards organizational identity in equal parts."Secondly, on the basis of above definitions, the scale was developed through pilot and main study.As a result, affective and cognitive scale consists of twelve items (6 affective items and 6 cognitive items).
Many researchers have not empirically tested the hypothesis about AID.The reasons are for the problem of definition and scale as mentioned above.We proposed both, so future researches which tested AID empirically may increase.AID is important concept because it is likely to shed light on organizational phenomena that cannot be demonstrated by identification or disidentification alone.Although it may be difficult for the person to grasp the person's ambivalent state by him/her-self, if this problem will be resolved in the future, this new concept would demonstrate many organizational phenomena.Thus, this study contributes to the foundation for AID researches.
and insulted in being part of this university I am glad and sorry to belong to this university I regret and am satisfied that I belong to this university I am proud of and also complain when I tell people who I study with I feel good and bad when I think about myself as a member of this university I feel happy and sad to be a student in this university I see and am indifferent of myself as belonging to this university I am very interested and uninterested in what others think about this university Belonging to this university is an important and unimportant part of my self-image It is important and insignificant to me that others do not criticize this university Being a member of this university has very little and very much to do with how I feel about myself If this university were criticized, it would influence and not influence how I think about myself

Table 3 .
Factor loadings for initial item reduction