EFFECTIVENESS OF CELEBRITY ENDORSEMENT OF A POLITICAL CANDIDATE AMONG YOUNG VOTERS.
As a presidential candidate in 2008, Barack Obama won the endorsement of the celebrity talk show host, actress, producer, media proprietor, and philanthropist, Oprah Winfrey, and Garthwaite and Moore (2013) found that this endorsement increased the amount of financial campaign contributions, as well as increasing the number of votes Barack Obama received by approximately one million. In fact, Oprah Winfrey's endorsement was found to have increased overall voter turnout. However, according to a Gallup poll of 507 Americans (Carroll & Jones, 2007), her endorsement could have had negative effects, which suggests that celebrity endorsements might have no effect and could even be detrimental to a candidate's chances of winning.
Political campaign managers must carefully consider how an endorsement can help, or undermine, their candidate. Two aspects of celebrity endorsement that the results of many practice-driven surveys have failed to explain is why, and under what conditions, celebrity endorsements of political candidates are effective. By gaining an understanding of the factors affecting voting behaviors, it may be possible to shed light on the conditions, such as party identification and the strength of that identification, that influence the effects of celebrity endorsement. The impact of celebrity endorsement on the political process has not received rigorous scholarly research (Chou, 2014, 2015; Garthwaite & Moore, 2013; Jackson & Darrow, 2005). There is a shortage of studies partly because of the difficulties inherent in conducting experimental studies on the effects of celebrity endorsement. What is more, the results of studies conducted over the past decade concerning the effects of celebrities' endorsement have been inconsistent. The direct or indirect effects of celebrity endorsement are in need of further investigation.
My purpose in the current study was threefold. First, drawing upon reinforcement theory and reactance theory, my aim was to assess the effects of party identification (i.e., with the Democrat Party vs. the Republican Party in the USA) on the evaluation of celebrity endorsement of a political candidate. Second, I examined how, when a voter is processing a celebrity endorsement, he or she is influenced by his or her identification with that celebrity (i.e., weak vs. strong identification). Third, I assessed the varying effects of celebrity endorsement on decided and undecided (i.e., swing) voters, drawing upon the elaboration likelihood model.
Voting Behavior: Key Influences and Theories
People's voting behavior is influenced by four major factors: social identity (i.e., race, gender, and religion), partisan identification, issue positions, and candidate evaluations (Campbell, Gurin, & Miller, 1954). In an examination of the effects of celebrities' endorsements of political candidates, these four factors should be considered as potential moderating variables. In my study I considered partisan identification as being the most important of the factors affecting the political decision-making process.
Partisan identification: Reactance theory versus reinforcement theory. Researchers recognize that how a person votes depends on more than just the candidates and the issues of the election. The most important long-term factor is party identification (Knobloch-Westerwick, 2012). Political parties tend to be quite stable. Partisanship is thus perceived as a long-term component of the political system (Knobloch-Westerwick, 2012). Campbell, Converse, Miller, and Stokes (1960) suggested that party identification influences the major determinants of the individual's vote, such as attitude toward the issues and attitude toward the candidates. Partisanship, thus, serves as a screening function, as voters selectively perceive political information that benefits their preferred party.
To understand how party identification works in the context of celebrity endorsement of a political candidate, reactance theory and reinforcement theory should be elaborated. In reactance theory, it is posited that people resist communication they perceive as being intended to persuade (Raney, Arpan, Pashupati, & Brill, 2003). According to the theory of psychological reactance, it is assumed that people take it as their inherent right to be autonomous agents (Brehm & Brehm, 2013). In reactance theory, it is postulated that persuasive communication poses a potential threat to freedom (Brehm & Brehm, 1981). In this sense, advertising messages are likely to elicit some degree of reactance.
In line with this theory, in the political context a celebrity endorsement of a candidate would produce reactance in the supporters of the candidate for the opposing party, who are more likely to interpret the celebrity's message as having persuasive intent than are supporters of the candidate for their own political party. In other words, supporters of a candidate for the opposing party are likely to evaluate the celebrity negatively, and are very unlikely to be moved to vote for the endorsed candidate. On the other hand, in reinforcement theory it is postulated that people look for, and store, information that provides cognitive support for their preexisting attitudes and beliefs (Staddon, 2013). Therefore, when a celebrity endorses a political candidate whom the voter supports, this will serve as reinforcement to that voter's preexisting beliefs about, and/or attitude toward, a candidate. The above discussion led to the following hypothesis:
Hypothesis 1: After being exposed to celebrity endorsement of a candidate standing for the political party with which they identify, voters will tend to hold a more favorable attitude toward (a) the celebrity endorser and (b) the endorsed political candidate than will voters who identify with the opposing party.
Identification with a celebrity endorser: Social identification theory. Kelman (1961) proposed three processes of social influence: compliance, identification, and internalization. Individuals could adopt an attitude or behavior from another person because they associate it with a satisfying self-defining relationship with that person (Kelman, 1961). According to Bandura's social cognitive theory (1986), a person's identification with the model has impacts on the likelihood of that person performing a behavior. That is, when an individual perceives him- or herself as similar to the model, he or she is more likely to perform whatever behavior that person models.
Brown and Basil (1995) also viewed identification as a critical factor underlying the effectiveness of celebrity endorsements. They cited U.S. basketball star player Magic Johnson's announcement of testing positive for HIV, and found that young adults who had strongly identified with him were more likely to be personally concerned about their health and more likely to report intended changes in their sexual behavior than were those who did not identify with him (Brown & Basil, 1995). In sum, these authors suggested that the more strongly that people identified with a celebrity, the more likely they were to credit that celebrity with positive attributes. Identification with celebrities encourages individuals to change aspects of their own attitudes, beliefs, and behaviors in order to emulate those of the celebrities more closely (Basil 1996; Brown & Basil 1995; Brown, Basil, & Bocarnea, 2003). Based on this evidence, I proposed the following hypothesis:
Hypothesis 2: People with a higher level of identification with a celebrity endorser will have more favorable attitudes toward (a) the celebrity endorser and (b) the endorsed political candidate than will people with a lower level of identification.
Decided versus undecided voters. The effectiveness of persuasion depends on whether the likelihood of elaboration of the communication situation is high or low (Petty, Cacioppo, & Schumann, 1983). Petty et al. (1983, p. 137) contended that when the likelihood of elaboration is high, the central route to persuasion is effective, which means that when involvement is high, people are more likely to make "the cognitive effort required to evaluate the true merits of an issue or product." When the likelihood of elaboration is low, that is, when there is a low probability of a message- or issue-relevant thought occurring, the peripheral route is more effective than the central route. Under a low involvement condition, people are not motivated to evaluate the true merits of an issue or product.
Petty and Cacioppo (1981) found that rational appeals, in which arguments or issues relevant to the facts (central cue) are employed, are effective when the likelihood of elaboration is high. Emotional appeals, in which the physical attractiveness of celebrity endorsers or affective messages (peripheral cue) are used, are effective when the likelihood of elaboration is low. With regard to the political arena, celebrity endorsement as a peripheral cue in a persuasion tactic ought to affect the undecided (swing) voters most. In processing a given message, decided voters are liable to have high likelihood of elaboration and undecided voters are liable to have low likelihood of elaboration. Hence, a celebrity endorsement is less likely to produce an attitude change in the latter group (the swing voters). Based on the central and peripheral routes to persuasion proposed in the elaboration likelihood model, the following hypothesis was proposed:
Hypothesis 3: After being exposed to a celebrity's endorsement of a political candidate, undecided voters will be more likely than are decided voters to have favorable attitudes toward (a) the celebrity endorser and (b) the political candidate.
Study Design, Participants, and Procedure
To test the proposed hypotheses, a 2 x 2 x 2 between-subjects factorial design was employed. The three factors were voters' party identification (Democrat or Republican), level of identification with a celebrity endorser (low vs. high), and status of voting intention (decided vs. undecided). Party identification and status of voting intention were measured variables whereas level of identification with a celebrity was measured and then divided, by median split, into low versus high groups.
Via the Internet, I recruited 256 students enrolled in a lower level and upper level of a journalism course at a major university in the upper Midwest in the USA to take part in the study. I deemed this group--a convenience sample--appropriate because of a significant connection between the lives of young people and celebrity culture. Chaffee and Frank Kanihan (1997) underscored the importance of the mass media in politically socializing young people. As Turner (2004) suggested, there is a significant connection between the lives of young people and celebrity culture. For these reasons, for my study I selected young adults as a group for the purpose of analyzing their responses to the endorsements and opinions of celebrities. The importance of young adults' participation in elections has grown and few doubt that, in the future, young adults will play a more pivotal role in the political process (Austin, Van de Vord, Pinkleton, & Epstein, 2008).
The students participated for extra credit in their course. They were directed to a main survey website on which the study was explained. After reading the explanation on the main website and clicking the "continue" button, they were directed to the experimental condition.
Because using actual celebrities in a study assures a sense of realism in examining the effects under consideration (Kamins, 1990; Money, Shimp, & Sakano, 2006; Walker, Langmeyer, & Langmeyer, 1992), in this study I utilized an actual celebrity, Denzel Washington, an American actor, filmmaker, director, and producer. For the pretest I selected 10 celebrities who had publicly endorsed Barack Obama as a presidential candidate. Dimensions of likeability, trustworthiness, familiarity, and identification were measured for these 10 selected celebrities. If respondents viewed a celebrity differently in any dimension and by a significant margin, according to gender, those people were excluded as potential celebrity endorsers. I selected Denzel Washington as the celebrity for the main study because, regardless of the respondents' gender, there were no significant differences in scores for him for familiarity, likeability, trustworthiness, and identification.
The respondents were first asked to give their general feelings about the celebrity endorser (likeability and credibility). The celebrity's likeability was measured with a three-item, 7-point semantic differential scale (Tripp, Jensen, & Carlson, 1994) where the most negative items were coded as 1 and the most positive items as 7: "very likeable-very unlikeable," "very unpleasant-very pleasant," and "very agreeable-very disagreeable" (a = .87). The celebrity's credibility was measured on a seven-item, 7-point semantic differential scale where negative items were coded as 1 and the most positive items as 7 (McCroskey, 1966): "dishonest-honest," "sincereinsincere," "trustworthy-untrustworthy," "biased-not biased," "crediblenot credible," "believable-not believable," and "disreputable-reputable" ([alpha] = .89).
Then participants were asked to state which presidential candidate they identified with more and how they felt about that candidate. They were also asked whether or not they had decided which candidate they would be voting for. This was to gauge how many decided and undecided voters there were and how likely they were to vote for Barack Obama, as he was the candidate being endorsed by the selected celebrity in this study. After responding to these general questions, the respondents were exposed to the stimulus.
Party identification was measured by the response to the question: "Generally speaking, do you usually think of yourself as a Republican, a Democrat, an independent, or something else (please specify)?" The students' strength of party identification was then measured with the response to the question: "If you think of yourself as a Republican or a Democrat, how strong is your party identification?" This was measured on a 7-point scale where 1 = very weak and 7 = very strong. For those who had identified themselves as an independent, a follow-up question was posed, "Do you think of yourself as closer to the Republican or Democrat party?"
To measure celebrity identification on a scale of 1 to 7, where 7 = strongly agree and 1 = strongly disagree (Lammie, 2007), the students were asked to respond to five items. The statements were "I like Denzel Washington," "I can easily relate to Denzel Washington," "I think of Denzel Washington as a good friend," "I have no doubt Denzel Washington and I would work well together," and "Denzel Washington is a personal role model." In this study, the reliability was .82.
Attitude toward a celebrity endorser and an endorsed political candidate was measured using five 7-point semantic differential items anchored by the most positive items scored as 7 and the most negative items scored as 1: "not believable-very believable," "not attractive-very attractive," "not competent-very competent," "not persuasive-very persuasive," and "not likeable-very likeable" (Williams & Qualls, 1989). In this study the respondents' attitude toward a celebrity endorser and the endorsed political candidate were measured twice, firstly before and then after the reading of Denzel Washington's endorsement of Barack Obama. The reliability was .90.
Using the Statistical Package for the Social Sciences (SPSS) version 22.0, all reliability of the measures, independent sample t tests, and a series of two-way analyses of variance (ANOVAs) were performed.
Hypotheses 1a and 1b: Partisan Identification
In the first hypotheses, I predicted that celebrity endorsement of the candidate standing for voters' identified political party would result in those voters having a more favorable attitude toward the celebrity endorser and the endorsed candidate than it would for voters who identified with the opposing party. As can be seen in Table 1, results indicated that the Wilks' lambda for party identification was significant. Because party identification was a significant factor, further analyses were conducted to examine its effects on two dependent variables. Univariate analyses indicated that party identification affected both voters' attitudes toward the celebrity endorser and voters' attitude toward the endorsed political candidate.
When I examined the mean differences between voters identifying with the Democrat Party and voters identifying with the Republican Party, the results suggested that party identification influenced attitudes toward the celebrity endorser, Democrat Party identification M = 5.45 versus Republican Party identification M = 4.86, and also influenced attitude toward the endorsed political candidate, Democratic Party identification M = 5.63 versus Republican Party identification M = 3.83. Therefore, Hypothesis 1a and Hypothesis 1b were supported. The findings suggest that, after being exposed to a celebrity endorsement of the candidate standing for a voter's identified political party, that voter evaluates the celebrity endorser as well as the endorsed candidate more positively than do voters aligned with the opposing party.
Hypotheses 2a and 2b: Identification with a Celebrity Endorser
I posited that people with a high level of identification with a celebrity endorser would evaluate the endorsed political candidate more positively than would people with a low level of identification. As can be seen in Table 1, the results indicate the Wilks' lambda for identification with a celebrity endorser was significant. Because identification with a celebrity endorser was a significant factor, its effects on the two dependent variables were further analyzed. Significant differences were found on attitude toward the celebrity endorser between people with high identification, M = 5.53, versus people with low identification, M = 4.78; F = 16.1, p < .001, and attitude toward the endorsed candidate, people with high identification, M = 5.02, versus people with low identification, M = 4.43; F = 9.84, p < .01. The findings suggest that people with high identification tend to evaluate both the celebrity endorser and the endorsed political candidate more positively than do people with low identification. Thus, Hypothesis 2a and Hypothesis 2b were supported.
Hypotheses 3a and 3b: Decided versus Undecided Voters
I posited that undecided voters, after being exposed to a celebrity's endorsement of a political candidate, would tend to hold more favorable attitudes toward the celebrity endorser and the endorsed political candidate than would decided voters. As the results in Table 1 show, the Wilks' lambda for voters' decision status was not significant. It is plausible to assume that individuals' voting intention did not influence either their evaluation of the celebrity endorser or that of the endorsed political candidate. Hence, Hypothesis 3a and Hypothesis 3b were not supported.
However, as shown in Figure 1, the results of univariate ANOVA showed that there were significant two-way interactions between party identification group (Democrat vs. Republican Party) and voters' decision status (decided vs. undecided; F = 18.85, p < .001) on attitude toward the endorsed candidate. Those who were undecided but who identified with the Republican Party had a tendency to hold a more favorable attitude toward the endorsed political candidate than did those who identified with the Republican Party and who had also decided. However, those who were decided and who identified with the Democrat Party had a tendency to hold a more favorable attitude toward the endorsed political candidate than did those who identified with the Democrat Party but were still undecided.
The focus in this study was to examine how celebrity endorsement of a political candidate influences voters' attitude. My findings in the study support the idea that the reactance and reinforcement theories are valid in explaining voters' information processing and the political decision-making process. In short, celebrity endorsement could work in two ways: serving as reinforcement to voters' preexisting beliefs and/or attitude toward a candidate, and serving as reactance to the candidate standing for the opposing party to the one with which that voter identified. In addition, I examined the effects of voters' identification with a celebrity endorser of a political candidate. The findings suggest that people with a higher level of identification with a celebrity endorser evaluate the endorsed political candidate more positively and are more likely to vote for the endorsed political candidate than are people with a low level of identification. As shown in previous studies (Basil, 1996; Brown & Basil, 1995; Brown et al., 2003), identification with a celebrity endorser is known to be a good predictor of voters' evaluation of the endorsed political candidate as well as their voting intention.
In my study I tested the elaboration likelihood model, predicting that decided voters are very likely to process the given message with a high level of elaboration (i.e., high motivation and ability) whereas undecided voters are likely to process with less elaboration (i.e., low motivation or low ability), resulting in little attitude change among decided voters and great attitude change among decided voters. The findings did not support my hypothesis but do suggest, however, that celebrity endorsement of a political candidate can enhance the evaluation of the candidate among undecided voters who are aligned with the political party of the endorsed candidate.
Practical implications can be drawn from this study. Celebrity endorsement of a political candidate is deemed to be an effective strategy, especially for swing voters who are seen as likely to lean toward the endorsed candidate's party. Based on the findings of the study, I would advise political campaign managers to employ celebrity endorsement to attract undecided voters to support their party and candidates. Furthermore, I found that people with a higher level of identification with a celebrity endorser evaluated the endorsed political candidate more positively than did people with a low level of identification. Thus, when practitioners are employing celebrity endorsers in political campaigns, they should consider various celebrity endorsers who can appeal specifically to male and female, young and old voters, respectively.
Limitations in this study are, first, that I used a real celebrity endorser, which could result in bias among respondents in processing the celebrity endorsement message. It may be desirable to use a fictitious celebrity endorser in order to avoid consumers' having a preexisting bias toward or against a celebrity endorser. Second, in this study I limited responses to college students in the USA. University students' perceptions of celebrity endorsement of a political candidate may differ from that of the general population. For future research, broadening samples demographically by including various age groups could be useful in investigating the effects of political celebrity endorsement. To make the study results more representative and generalizable, it is essential to use the general population in the future.
It would be worthwhile for future researchers to examine the impact of celebrity endorsement of a political candidate in countries other than the USA, such as Asian countries where the use of celebrity endorsement has been even more prevalent than in the USA. A cross-cultural study on celebrity endorsement of a political candidate would broaden understanding of the impact of celebrity endorsement. In addition, future research may also include an examination of the third person effect in celebrity endorsement of a political candidate. According to the third person effect, individuals view media messages as wielding greater persuasive influence on people other than themselves (Phillips Davison, 1983). I believe that it would be interesting to delve into whether or not the third person effect occurs in the context of celebrity endorsement of a political candidate.
Austin, E. W., Van de Vord, R., Pinkleton, B. E., & Epstein, E. (2008). Celebrity endorsements and their potential to motivate young voters. Mass Communication and Society, 11, 420-436. https://doi.org/c9b9r5
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall.
Basil, M. D. (1996). Identification as a mediator of celebrity effects. Journal of Broadcasting & Electronic Media, 40, 478-495. https://doi.org/cqghpj
Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance: A theory of freedom and control. New York, NY: Academic Press. Brown, W. J., & Basil, M. D. (1995). Media celebrities and public health: Responses to "Magic"
Johnson's HIV disclosure and its impact on AIDS risk and high-risk behaviors. Health Communication, 7, 345-370. https://doi.org/dg83qt
Brown, W. J., Basil, M. D., & Bocarnea, M. C. (2003). Social influence of an international celebrity: Responses to the death of Princess Diana. Journal of Communication, 53, 587-605. https://doi.org/dw3556
Campbell, A., Converse, P., Miller, W., & Stokes, D. (1960). The American voter. New York, NY: John Wiley & Sons.
Campbell, A., Gurin, G., & Miller, W. (1954). The voter decides. Oxford, UK: Row, Peterson.
Carroll, C., & Jones, M. (2007). Most voters not moved by Oprah's endorsement of Obama. Retrieved from https://www.gallup.com/poll/102301/most-voters-moved-oprahs-endorsement-obama. aspx
Chaffee, S. H., & Frank Kanihan, S. (1997). Learning about politics from the mass media. Political Communication, 14, 421-430. https://doi.org/fj9tmh
Chou, H.-Y. (2014). Effects of endorser types in political endorsement advertising. International Journal of Advertising, 33, 391-414.
Chou, H.-Y. (2015). Celebrity political endorsement effects: A perspective on the social distance of political parties. International Journal of Communication, 9, 523-546.
Garthwaite, C., & Moore, T. J. (2013). Can celebrity endorsements affect political outcomes? Evidence from the 2008 U.S. Democratic presidential primary. Journal of Law, Economics, & Organization, 29, 355-384. https://doi.org/brgr
Jackson, D. J., & Darrow, T. I. A. (2005). The influence of celebrity endorsements on young adults' political opinions. The International Journal of Press/Politics, 10, 80-98. https://doi.org/bdf4rh
Kamins, M. A. (1990). An investigation into the "match-up" hypothesis of celebrity advertising: When beauty may be only skin deep. Journal of Advertising, 19, 4-13. https://doi.org/bkn6
Kelman, H. C. (1961). Process of opinion change. Public Opinion Quarterly, 25, 57-78. https://doi.org/fj7mgz
Knobloch-Westerwick, S. (2012). Selective exposure and reinforcement of attitudes and partisanship before a presidential election. Journal of Communication, 62, 628-642. https://doi.org/brgs
Lammie, K. E. (2007). The (not so) "powerless elite" celebrity endorsements of political candidates (Doctoral dissertation). Retrieved from ProQuest. (Accession No. AAI3260938.)
McCroskey, J. C. (1966). Scales for the measurement of ethos. Speech Monographs, 33, 65-72. https://doi.org/ccq
Money, R. B., Shimp, T. A., & Sakano, T. (2006). Celebrity endorsements in Japan and the United States: Is negative information all that harmful? Journal of Advertising Research, 46, 113-123. https://doi.org/bxscfz
Petty, R. E., & Cacioppo, J. T. (1981). Attitudes and persuasion: Classic and contemporary approaches. Dubuque, IA: W. C. Brown.
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement, Journal of Consumer Research, 10, 135-146. https://doi.org/g72
Phillips Davison, W. (1983). The third-person effect in communication. Public Opinion Quarterly, 47, 1-15. https://doi.org/dn7pqk
Raney, A. A., Arpan, L. M., Pashupati, K., & Brill, D. A. (2003). At the movies, on the Web: An investigation of the effects of entertaining and interactive Web content on site and brand evaluations. Journal of Interactive Marketing, 17, 38-53. https://doi.org/b43jh2
Staddon, J. E. R. (Ed.). (2013). Limits to action: The allocation of individual behavior. New York, NY: Academic Press.
Tripp, C., Jensen, T. D., & Carlson, L. (1994). The effects of multiple product endorsements by celebrities on consumers' attitudes and intentions. Journal of Consumer Research, 20, 535-547. https://doi.org/cmkjnc
Turner, G. (2004). Understanding celebrity. London, UK: Sage.
Walker, M., Langmeyer, L., & Langmeyer, D. (1992). Celebrity endorsers: Do you get what you pay for? Journal of Consumer Marketing, 9, 69-76. https://doi.org/fphnpv
Williams, J. D., & Qualls, W. J. (1989). Middle-class black consumers and intensity of ethnic identification. Psychology & Marketing, 6, 263-286. https://doi.org/d6h7z2
Wood, N. T., & Herbst, K. C. (2007). Political star power and political parties: Does celebrity endorsement win first-time votes? Journal of Political Marketing, 6, 141-158. https://doi.org/d7wcpk
Nam-Hyun Um, School of Advertising and Public Relations, Hongik University.
Correspondence concerning this article should be addressed to Nam-Hyun Um, School of Advertising and Public Relations, Hongik University, 2639 Sejong-ro, Jochiwon-eup, Sejong, 339-701, Republic of Korea. Email: firstname.lastname@example.org
Caption: Figure 1. Interaction between party identification and voting decision status.
Table 1. Multivariate Analysis of Variance Results for Effects of Party and Celebrity Identification on Voter Decision Status Effects Wilks' df F P lambda Main effects Party identification (A) .71 (2, 245) 50.36 < .001 Celebrity identification (B) .94 (2, 245) 8.53 < .001 Voting decision status (C) .99 (2, 245) 1.58 > .05 A x B .97 (2, 245) 2.53 > .05 A x C .87 (2, 245) 18.95 < .001 B x C .99 (2, 245) .48 > .05 A x B x C 1.00 (2, 245) .02 > .05
|Printer friendly Cite/link Email Feedback|
|Publication:||Social Behavior and Personality: An International Journal|
|Date:||Oct 1, 2017|
|Previous Article:||PSYCHOLOGICAL CONTRACT BREACH, ORGANIZATIONAL DISIDENTIFICATION, AND EMPLOYEES' UNETHICAL BEHAVIOR: ORGANIZATIONAL ETHICAL CLIMATE AS MODERATOR.|
|Next Article:||HOW HIGH-COMMITMENT WORK SYSTEMS ENHANCE EMPLOYEE CREATIVITY: A MEDIATED MODERATION MODEL.|