"Ageism at work: What happens to older workers who benefit from preferential treatment?"

In order to increase the activity rate of older workers, the Organisation for Economic Cooperation and Development (OECD) recommends that national governments implement policies promoting the employment of this category of workers. However, policies that favour minority groups have been shown to produce detrimental effects such as devaluing members of these groups. In two studies, we examined whether age-related preferential treatment reinforces ageist attitudes in the workplace. A first study revealed that policies favouring 50 years old workers increased negative perceptions toward them. In a second experimental study, results indicated that, compared to a merit-based treatment, a preferential treatment increased negative perceptions, emotions, and behaviours toward an old target. As a set, our findings shed new light on ageism at work and on the role of context. In order to increase the activity rate of older workers, the Organisation for Eco- nomic Cooperation and Development (OECD) recommends that national governments implement policies promoting the employment of this category of workers. However, policies that favour minority groups have been shown to produce detrimental effects such as devaluing members of these groups. In two studies, we examined whether age-related preferential treatment reinforces ageist attitudes in the workplace. A first study revealed that policies favouring 50 years old workers increased negative perceptions toward them. In a second experimental study, results indicated that, compared to a merit-based treatment, a preferential treatment increased negative perceptions, emotions, and behaviours toward an old target. As a set, our findings shed new light on ageism at work and on the role of context.

In order to increase the activity rate of older workers, the Organisation for Economic Cooperation and Development (OECD) recommends that national governments implement policies promoting the employment of this category of workers. However, policies that favour minority groups have been shown to produce detrimental effects such as devaluing members of these groups. In two studies, we examined whether age-related preferential treatment reinforces ageist attitudes in the workplace. A first study revealed that policies favouring 50 years old workers increased negative perceptions toward them. In a second experimental study, results indicated that, compared to a merit-based treatment, a preferential treatment increased negative perceptions, emotions, and behaviours toward an old target. As a set, our findings shed new light on ageism at work and on the role of context.
In order to increase the activity rate of workers aged 50 and more, OECD governments have decided to implement public policies that promote the employment of this category of workers (e.g., by reducing wage costs for older workers) and that encourage them to stay longer in the labour force (e.g., by improving their working conditions). Clearly, older workers have become a target for employment policies. Unfortunately, research has repeatedly shown that policies that favour members of specific groups (e.g., woman or foreigners) may trigger an unanticipated set of reactions such as decreasing beneficiaries' perceived competence and reinforcing the salience of their minority status (Heilman & Welle, 2006). A similar phenomenon may be induced by employment policies that favour older workers. In other terms, we argue that employment policies that treat preferentially some workers because they are older are likely to reinforce ageism, i.e., the stigmatisation and discrimination of older workers as a group (Butler, 1969). The purpose of the present efforts is to analyse the effects of preferential treatment based on age on younger workers' attitudes towards older workers. ysis by Gordon and Arvey (2004) confirmed the existence of negative attitudes toward older people compared with younger individuals. For example, Armstrong-Stassen and Templer (2005) revealed that fewer than ten per cent of the surveyed organisations provided training for older workers. Taylor and Walker (2003) showed that some managers have a positive experience with a senior workforce but when it comes to recruiting new employees, younger people are their first choice. In Belgium, a worker who is older than 50 is five times less likely to be engaged than a younger worker (Eurostat, 2007).
One important factor that may affect people's attitudes toward older workers is the organisational context (Finkelstein & Farrell, 2007). To the best of our knowledge, studies on ageism in the workplace have never investigated the influence of organisational policies favouring older workers on attitudes toward them. This is all the more surprising that this issue turns to be central with respect to recent policies which favour employment of older workers as a target group and that encourage them to stay longer in the labour force (e.g., the Belgian Agreement for Intergenerational Solidarity, 2005 [1] ). Indeed, research has shown that preferential treatment may well have detrimental effects on perceptions of the beneficiaries whose perceived competence is questioned (e.g., Gilbert & Stead, 1999;Heilman, Lucas, & Block, 1992).

Preferential treatment
In the larger context of affirmative action programs (Crosby, Iyer, Clayton, & Downing, 2003), diversity-management initiatives include efforts to systematically recruit, promote, and retain a heterogeneous array of employees throughout the ranks of the organisation (Ivancevich & Gilbert, 2000). One way to implement diversity is to treat preferentially members of some target groups. Unfortunately, deleterious effects of preferential treatment on target's perceived competence have been repeatedly shown both when the association between the target and an affirmative action program was made explicit and when this link was only implicit (e.g., Heilman, McCullough, & Gilbert, 1996;Heilman & Welle, 2006;Little, Murry, & Wimbush, 1998;Moore, 1999;Nacoste, 1990). As a case in point, Heilman and colleagues (1992) presented a bogus female hire as being associated or not with an affirmative action program. Results showed that a female target associated with preferential treatment 1. In 2005, in Belgium, the government, employers' organisations, and employees' associations formed an agreement (entitled "Le Contrat de Solidarité entre Générations") aimed at retaining older workers in the workforce. This agreement consists in favouring the employment of older workers and encouraging them to stay longer in the labour force (e.g., by improving the financial benefit of staying in the workforce and reducing working hours). The complete French text of this agreement can be downloaded from http://premier.fgov.be/fr/051011_contrat_solidarite_generations. was perceived as being less competent than a female target not associated with preferential treatment. In another study, Heilman and Welle (2006) showed that a general diversity rationale for how a team was assembled (i.e., in order to ensure that the organisation's demographic diversity was represented) was sufficient to decrease the perceived competence of an Afro-American target. Current employment issues are concerned with increasing older individuals' participation to the workforce. Some of these policies implement preferential treatment (e.g., labour costs reduction) for older workers in order to encourage employers to hire them. Building upon previous work about race and gender-related preferential treatment, we assume that age-related employment policies should have the unintended effect of reinforcing ageist attitudes in the workplace.
In two studies, we examined the influence of policies favouring older workers on ageist attitudes. First, we wanted to test the relevance of the preferential treatment hypothesis for age-related policies. Therefore, we examined attitudes associated with a real age-related policy. Secondly, we wanted to test this relationship experimentally by varying the context of social judgment. To this end, in a second experiment, we manipulated the type of a hiring context (preferential treatment versus merit, Heilman et al., 1996;Heilman & Welle, 2006) and analysed its effects on ageism.

Study 1
According to the literature on preferential treatment (e.g., Gilbert & Stead, 1999;Heilman & Welle, 2006) and according to the stereotypical beliefs identified by Warr and Pennington (1993), we predicted that younger workers would see older workers as less effective and less adaptable in their job than themselves when they perceived that their organisation implemented particular policies that favour older workers. In other words, we expected that when younger workers perceived preferential treatment for older workers, they would report more bias against them in terms of work effectiveness and adaptiveness.

Participants and procedure
Data were collected in a sample of 187 French-speaking workers of a Belgian hospital aged less than 50 years old (i.e., employees who don't belong to the group of older workers as defined by the OECD, 2005). The average age was 37.08 years (SD = 7.63). Most of participants were females (86%) and worked full-time (54.5%). Some 63% of respondents had completed high school, 16% held a university degree and 21% had not finished high school. The majority of participants were nurses (54%), 25% paramedical nurses and 21% administrative workers.
Participants completed a paper open-ended questionnaire which was accompanied by a covering letter signed by the researchers and indicating that the purpose of the study was to examine "people's attitudes towards age diversity and intergenerational relationships at work". An envelope was provided so that completed questionnaires could be returned sealed to the researchers. Questionnaires were anonymous and included the personal and organisational variables first, followed in order by predictor variable (preferential treatment) and criterion variable (stereotypes) [2] .

Measures
At the beginning of the questionnaire, participants were informed that the group of older workers included workers of 50 years old and more.
Preferential treatment. The measure of preferential treatment was adapted from the concept of "preferential treatment" as depicted by Heilman and colleagues (1996;Heilman & Welle, 2006) and corresponded to effective practice of preferential treatment in the context. Participants were asked to indicate to what extend they think that in their organisation "older workers have the right to a specific diminution of their working hours [3] ", and "specific policies for older workers are implemented". Because these two statements were positively and significantly correlated (r = .38, p < .001), we computed an average preferential treatment score.
Stereotype bias. The measure of stereotypes on older workers as a group was adapted from Warr andPennington (1993 in Redman &Snape, 2002). Work effectiveness was measured with six items for older workers (α = .87) and six items for younger workers (α = .95). Participants used a 11-point scale ranging from 1 (= 0%) to 11 (= 100%) to indicate how many older workers and how many younger workers are conscientious, reliable, work hard, are effective in their job, are loyal to the organisation, and have interpersonal skills.
Adaptability was measured by six items for older workers (α = .89) and six items for younger workers (α = .90). Using the same 11-point scale, participants rated the extent to which they thought that older workers and younger workers are able to grasp new ideas, adapt to change, accept the 2. These questions were included in a broader questionnaire about variables unrelated to the issue of age-related preferential treatment. 3. The right to a specific reduction of the working hours concerns workers aged 50 years or more (Collective convention of work N° 77 Bis, December, 19, 2001). A weak version of this right (i.e., smaller reduction of working hours) may apply to workers of 45 years old (see Collective convention of work, October, 26, 2006). In the context of our study, according to the HR Manager, this particular policy mainly concerned workers over 50.
introduction of new technology, learn quickly, are interested in being trained, and are receptive to direction. The bias toward older workers was computed by subtracting participants' mean older workers ratings (outgroup) from the mean of younger workers ratings (in group) in both the work effectiveness and the adaptability dimension, such that higher (or positive) scores indicated that participants (all younger workers) evaluated older workers' group less favourably than their own age group.
Personal and organisational variables. Age, gender, education, workweek, and professional status were each measured using single items. Except for age which was a continuous variable (chronological age in years), all variables were multiple-choice questions.

Results
Means, standard deviations and Pearson's coefficients are presented in Table  1. Hierarchical regressions were computed to test a two-step model for each dependent variable (work effectiveness bias and adaptability bias, see Table  2) [4] . We entered personal and organisational variables on the first step to control for their effects. On the second step, preferential treatment was entered in the regression.
Firstly, results indicated that only participants' age (all p > .10 for other control variables) had an impact on both work effectiveness (β = -.316, p < .001) and adaptability (β = -.212, p < .01) bias. In other terms, the more younger workers were coming nearer to 50 years old, the less they were prejudiced toward older workers. Secondly, the relationship between preferential treatment and the bias regarding older workers' effectiveness was significant and positive (β = .155, p < .05) whereas the relationship with the bias regarding older workers' adaptability was not significant (β = .006, ns). Therefore, according to our hypothesis, the more younger workers perceived their organisation as treating preferentially older coworkers, the more they viewed older workers in general as less effective in their job than themselves. But, contrary to our expectation results revealed that preferential treatment did not reinforce the bias regarding adaptability.

Discussion
Supporting our hypothesis, the more younger workers perceived their organisation as treating preferentially older coworkers, the more they viewed older workers in general as less effective in their job than younger workers. However, the present results also revealed that preferential treatment did not reinforce the bias regarding adaptiveness. It thus appears that preferential treatment had deleterious effects only on the dimension on which older workers were perceived more positively than younger workers.
Interestingly, the relationship between preferential treatment and work effectiveness bias was influenced by participants' age. Specifically, the increase in work effectiveness bias linked to preferential treatment was observed for younger workers low in age but nor for younger workers coming High Age nearer to 50 years old. This non-significant relationship between preferential treatment and bias among "older-young" workers might be explained by the fact that some of them (i.e., workers aged 45 and more) could be already beneficiaries of a weak version of the preferential treatment (see Footnote 3). Moreover, age groups being permeable (e.g., Chasteen, 2005), workers coming nearer to 50 years old are likely to know that they will become potential beneficiaries of age-related preferential treatment. In other terms, the awareness of being beneficiaries in the near future could have buffered the negative impact of preferential treatment.
To sum up, Study 1 showed that preferential treatment in a real context can be related to more bias toward older workers. However, because the design was correlational, we cannot draw any firm conclusions in terms of causality. We therefore conducted a second study in order to experimentally test the effects of preferential treatment on ageism at work. Moreover, Study 2 allowed us to make sure that none of the participants were beneficiaries of the manipulated preferential treatment.

Study 2
Building on the literature on intergroup relations (for a review, see Yzerbyt & Demoulin, 2010), we decided to rely on a tripartite view of intergroup attitudes inasmuch as it constitutes a potentially useful framework to deal with the complexity of ageist attitudes (e.g., Finkelstein & Farrell, 2007). The tripartite view holds that intergroup bias comprises three components: a cognitive component (i.e., beliefs about members of a group), an affective component (i.e., feelings with respect to these members) and a behavioural component (i.e., acts and behavioural intentions toward them). Like any other intergroup bias, ageist attitudes can be thought as a constellation of these three factors. As it turns out, hardly any research measured ageism with a proper consideration of all three aspects. In particular, although intergroup emotions have been shown to predict attitudes toward outgroups as well or even better than does cognition (Nelson, 2009;Yzerbyt & Demoulin, 2010), the literature about ageism reveals that the affective component is seldom conceptualised and, indeed, rarely measured (Finkelstein & Farrell, 2007).
As an offspring of the stereotype content model (Fiske, Cuddy, Xu, & Glick, 2002;Fiske, Xu, & Cuddy, 1999), the bias map model (Cuddy, Glick, & Fiske, 2007) gives some insight by considering the tripartite view of intergroup bias. Indeed, this model predicts emotions and behaviours toward members of a group as a function of people's perceptions of this group. The purpose of our second study was to analyse the effects of preferential treatment on age bias by developing a tripartite approach of ageism including affective and behavioural components besides perceptions. With this goal in mind, we extended the analysis of participants' perceptions by including warmth stereotypes in addition to competence stereotypes. According to the bias map model (Cuddy et al., 2007), each behavioural pattern is predicted by a particular combination of perceived competence and warmth and one of four affective states (admiration, contempt, pity, and envy) induced by these perceptions . For example, groups that are perceived as high on both warmth and competence (e.g., the ingroup) would elicit admiration feelings and (passive and active) facilitation behaviours (i.e., explicit and implicit aims to benefit a group) while groups that are judged as low on both warmth and competence (e.g., welfare recipients) would elicit contempt feelings and (active and passive) harm behaviours (i.e., explicit and implicit intends to hurt a group and its interest). Because perceptions of competence and warmth are positively related when it comes to judging a single target (i.e., halo effect, Judd, Hawkins, Kashima, &Yzerbyt, 2005;Kelley, 1950;Rosenberg, Nelson, & Vivekananthan, 1968;Thorndike, 1920), preferential treatment should decrease the perception of competence and warmth of the target. According to the bias map model (Cuddy et al., 2007), this perceptual pattern should then reinforce contempt and active harm toward the target.
In the second study, based on Heilman and colleagues' work (e.g., Heilman et al., 1996;Heilman & Welle, 2006), we investigated experimentally how a preferential treatment may influence ageism at work by manipulating the context of a personnel selection policy that focused either on preferential treatment or on personal merit of an "older worker". Specifically, we tested two hypotheses: Hypothesis 1: In line with the stereotype content model  and the halo effect (e.g., Judd et al., 2005), we predicted that an older worker would be seen as less competent and less warm in the presence of a preferential treatment based on age ("preferential treatment condition") than in a context which highlights the individual's merit ("merit condition").
Hypothesis 2: Building on the bias map model (Cuddy et al., 2007), we also hypothesised that the target would elicit more contempt and harm behaviours in the preferential treatment condition than in the merit condition.

Participants
We relied on a between-subject design with hiring context (preferential treatment versus merit) as the manipulated variable. Fifty-eight French-speaking Belgian workers (32 females) aged from 20 to 58 (M = 33.21; SD = 10.69) took part to the study. Most participants were white-collar workers (80%) and worked in the medical (47%) or commercial (32%) sectors (the 21% others worked in sectors like law, journalism or research).

Procedure
Data were obtained via a snowball sampling method using e-mail address contact.
Participants completed an open-ended questionnaire online with the software DORIS (UCL/PSP, version 1.5). They were informed that a multinational company had recently hired a large number of new employees and wished to know how the new employees would be perceived by their colleagues. Subjects were told that their participation was needed as a pre-test of the questionnaire for the company. This instruction was intended to reduce participant's social desirability concerns.
All participants read first information about a target allegedly randomly chosen among the new employees. The target was a 60 years old man. Additional information was given which briefly depicted the target on a series of personal (e.g., his family status, hobbies) and work-related (e.g., education, work history) variables.
Next, participants were randomly assigned to a preferential (n = 32) or merit condition (n = 26) and read a short text that provided them with information about the organisational context where applicant had been hired. In the preferential treatment condition, the text presented on the screen informed participants that the target "was hired in accordance with an employment policy that favoured workers over the age of 50". In the merit condition, the text mentioned that the target "was hired in accordance with an employment policy that favoured the most competent workers". This short text served as the preferential treatment/merit manipulation adapted from Heilman and colleagues' work (1996; Heilman & Welle, 2006) [5] . Participants were then instructed to complete the stereotypes scale, the emotions scale, and the behavioural tendencies scale.

Materials
Stereotypes. Stereotypes were assessed by means of a French adaptation of Fiske and colleagues' (2002) scale. Participants were asked to use a 7-point scale ranging from 1 (= totally disagree) to 7 (= totally agree) in order to indicate their agreement with four traits depicting the target's competence ("How competent, confident, intelligent, and independent will the hired person look 5. Based on Heilman and colleagues' work (e.g., Heilman et al., 1996; see Study 1 of Heilman & Welle, 2006), we did not include a control condition. In particular, Heilman and Welle (2006) included a control condition in two studies and found that a female target (Study 2) and a black man target (Study 3) were perceived as less competent in the preferential condition compared to the control and merit conditions with no difference between the control condition and the merit condition. In fact, the merit condition could be considered as a more realistic control condition, as a traditional recruitment policy is to hire a worker on the basis of his/her high competence.
like in the eyes of their colleagues", α = .75) and with four traits depicting the target's warmth ("How tolerant, warm, good-natured, and sincere will the hired person look like in the eyes of their colleagues", α = .67).
Emotions. Contempt was measured by means of three items adapted from Cuddy and colleagues (2007). Using a 7-point scale ranging from 1 (= totally disagree) to 7 (= totally agree), participants rated how much the target would elicit contempt (contempt, disgust, and repulsion, α = .77).
Behavioural tendencies. Finally, harm behavioural tendencies were measured by means of four items adapted from Cuddy and colleagues (2007). Using the same 7-point scale, participants rated the extent they thought that the target would elicit harm behaviours (denigrate, harass, jeer, and exclude him, α = .94).
Control variables. Gender, age, activity sector (commercial, healthcare and other), and occupational position (blue-collar workers, white-collar workers, executive) were also assessed.
Manipulation check. A final question checked whether participants had paid attention to the manipulation of the type of treatment (preferential versus merit). Specifically, the sentence "This person has been hired under a policy that favoured […]" could be answered with one of the three possible responses: (a) workers over 50 years old; (b) the most competent workers; and (c) I don't know.

Manipulation check and preliminary analyses
A total of 81 workers participated to the study but only 62 workers completed the entire questionnaire. The data of four participants who selected "I don't know" as an answer to the manipulation check (three participants in the preferential condition and one in the merit condition) were dropped from the analysis, leaving 58 participants in the sample.
Each dependent measure was analysed by means of a series of one-way ANOVAs using gender, the activity sector, or the occupational position of participants as independent variables. None of these variables was shown to significantly impact the criterion variables. Correlations between age and each dependent measure revealed no significant relationships [6] . In light of this, no control variable was taken into account in the following analyses. Correlations among variables are presented in Table 3. 6. We also computed regressions analyses to test the interaction between age and preferential treatment according to results of Study 1. However, the interaction was not significant for all dependant variables (all. p > .10) indicating that contrary to Study 1, age didn't seem moderate the effects of preferential treatment.

Main analyses
To test our hypotheses, we conducted separate one-way ANOVAs on the two dimensions of stereotypes, the contempt emotions, and the harm behavioural tendencies using treatment (preferential treatment vs. merit) as our independent variable. Means ratings of stereotypes, emotions and behaviours by conditions are presented in Table 4.
Regarding stereotypes, the target (i.e., an older worker) was perceived to be less competent (M = 4.85, SD = 0.87) than warm (M = 5.09, SD = 0.75, t = 2.09, p < .05). Both dimensions were positively correlated, r = .40, p < .001. Our first hypothesis held that the target would be seen as less competent and less warm in the context of a preferential treatment based on age than when the context highlighted individual merit. The experimental manipulation was significant for competence, F(1, 56) = 4.43, p < .05, indicating that the target was perceived as significantly less competent (M = 4.63, SD = 0.85) in the preferential condition than in the merit condition (M = 5.11, SD = 0.84). The difference failed to be significant for warmth, F(1, 56) = 2.49, p > .10, although the difference was in the expected direction (M = 4.96, SD = 0.66 vs. M = 5.24, SD = 0.83 for preferential treatment vs. merit).
In line with the bias map model (Cuddy et al., 2007), our second prediction was that the target would elicit more contempt and more harm behaviours in the preferential condition than in the merit condition. The experimental manipulation was marginally significant for contempt, F(1, 56) = 3.79, p < .10, indicating that the target tended to elicit more contempt in the preferential condition (M = 1.51, SD = 1.01) than in the merit condition (M = 1.11, SD = 0.36). Concerning harm behaviours, the effect of the experimental manipulation was significant, F(1, 56) = 4.78, p < .05, indicating that in the preferential condition, the target elicited more harm behaviours (M = 1.40, SD = 0.77) than in the merit condition (M = 1.06, SD = .20). Our second hypothesis was thus marginally supported for contempt and fully supported for harm behaviours. Note. N = 32 for the preferential condition and 26 for the merit condition. All ratings were provided on 7-point scales (the higher the mean the more favourable the rating) In sum, we found causal evidence that preferential treatment increased negative attitudes toward older workers. Moreover, in this second study, we showed that negative effects of preferential treatment were far from being limited to perceptions but that they also increased negative emotions and behavioural tendencies toward older workers.

General discussion
The purpose of this study was to investigate whether preferential treatment related to age triggers ageism in the workplace. Findings revealed that agerelated diversity goals in employment practices result in social devaluation of older workers when the organisational setting treats preferentially older workers or casts a doubt on individuals' merit with the rationale of age.
In line with previous research which has shown that women and racial minorities who are thought to be beneficiaries from preferential treatment were later affected with a stigma of incompetence (Heilman et al., 1996;Heilman & Welle, 2006), we showed in two studies that preferential treatment toward older workers induces a perception of lower competence.
In Study 1, we examined perceptions of competence on the basis of Warr and Pennington's (1993) two-factor model. Results revealed that younger workers who perceived that older workers were treated preferentially by their organisation were more suspicious about the effectiveness of these workers. Unexpectedly, results of Study 1 revealed that the adaptiveness dimension was preserved from preferential treatment deleterious effects. In other terms, preferential treatment seems to alter the characteristics on which older workers are usually positively rather than negatively judged (Chiu et al., 2001;Redman & Snape, 2002;Warr & Pennington, 1993).
Design of Study 1 was correlational precluding to draw any conclusions in terms of causality. Therefore, a second study was conducted replicating and extending findings of Study 1 in several ways. First, the deleterious impact of age-related preferential treatment was confirmed experimentally: when an older worker was associated with an age-related preferential treatment, he was perceived as less competent than when he was not associated with this kind of treatment. Secondly, building upon studies on social perceptions which have shown that social perceptions are two-dimensional Fiske et al., 1999), we not only investigated perceptions of competence but also perceptions of warmth. This complements work on preferential treatment which focused mainly on target's competence (e.g., Heilman & Welle, 2006). Interestingly, although the two dimensions were positively correlated, warmth perceptions proved insensitive to the experimental manipulation. This finding is reminiscent of a study by Heilman and colleagues (1992) who showed that the interpersonal characterisation of a female target was unaffected by affirmative action. In other words, the deleterious effect of preferential treatment doesn't seem to affect all domains of social perception.
Thirdly, Study 2 investigated the emotional and behavioural consequences of perceptions on the basis of Cuddy and colleagues' (2007) bias map model. Harmful consequences of preferential treatment on attitudes towards beneficiaries appeared to extend to both emotions and behavioural intentions. Regarding emotions, results indicated that when the target was associated with a preferential treatment, he elicited more contempt. These findings complement previous research which has shown that preferential treatment induced negative emotions as feelings of unfairness and guilt among non-beneficiaries (Boeckmann & Feather, 2007). Similarly, other studies demonstrated that preferential treatment increased negative emotions because of feelings of injustice related to preferential treatment (Nacoste, 1990;Richard & Kirby, 1999). However, whereas these studies examined the emotions linked to the procedure itself, the present study shows that preferential treatment triggers negative emotions also toward beneficiaries themselves.
As far as behaviours are concerned, Study 2 evidenced potential negative effects of diversity goals on behavioural tendencies: Preferential treatment was likely to increase harmful behaviours such as denigrating or excluding older workers. These data confirm, in line with several studies (e.g., Heilman, Simon, & Repper, 1987;Little et al., 1998), that affirmative action can contribute to reinforce the discrimination of targets. For example, Heilman, Block, and Stathatos (1997) showed that lower wages were proposed to a female applicant who was presented as benefiting from an affirmative action in hiring process than to a control female target. Similarly, the strong pressure towards retirement that many older workers experience (McCann & Giles, 2002) may well reflect an increase in harmful behaviours due to perceptions of lower competence and feelings of contempt.
Eventually, these studies underline the role of age. As a matter of fact, Study 1 showed that the increase in work effectiveness bias linked to preferential treatment was observed for younger workers low in age but not for younger workers coming nearer to 50 years old. Of course, as mentioned before, some participants aged 45 and more in Study 1 could themselves be beneficiaries of a weak version of preferential treatment. Alternatively, because of age groups permeability (e.g., Chasteen, 2005), the awareness of being beneficiaries in the near future could have buffered the negative impact of preferential treatment in Study 1. Turning to Study 2 where we made sure that participants could not be beneficiaries, age did not moderate the impact of preferential treatment. Together, the results of Study 1 and Study 2 converge in showing that the deleterious effects of age-related preferential treatment emerge only among non-beneficiaries. Future studies should thus examine whether the relative salience of group permeability and/or common fate (i.e., being a future beneficiary of preferential treatment related to age) actually plays a role as a buffer of the preferential treatment effects.

Limitations and perspectives
Future work on these issues should take into account possible limitations of the present studies. First, although the tripartite view based on the bias map model seems to be a promising route to investigate ageism in the workplace, measures of stereotypes should be adapted for the population of middle-aged individuals. In fact, Study 2 revealed that older workers were perceived as competent whereas Study 1 revealed that older workers were perceived as more competent on the work effectiveness dimension than on the adaptiveness dimension. In particular, distinguishing between traits related to crystallised intelligence (e.g., the worker's experience) and fluid intelligence (e.g., the worker's information processing speed) could better predict which traits are positively or negatively rated depending on aging viewed as either a source of improvement (e.g., experience due to tenure) or a cause of decline (e.g., slower information processing due to senescence) at work. Moreover, Study 1 showed that preferential treatment toward older workers only affects perceptions of their work effectiveness. Therefore, future research should investigate stereotypes, emotions, and behavioural tendencies according to the tripartite view but should also take into consideration the distinction between work effectiveness and adaptiveness perceptions.
Second, the present set of studies did not examine the various processes through which ageism is reinforced by preferential treatment. Several studies showed that affirmative action violates principles of both distributive and procedural justice among non-beneficiaries (Boeckmann & Feather, 2007;see Crosby & Franco, 2003). Therefore, perceptions of unfairness linked to age-related preferential treatment could explain its consequences on ageism at work. In this perspective, some studies showed that when an affirmative action is perceived as unfair, it increases negative attitudes toward this type of treatment among non-beneficiaries (Beaton & Tougas, 2001;Cropanzano, Slaughter, & Bachioche, 2005). However, these studies investigated effects of preferential treatment and fairness on attitudes toward this type of treatment but not on intergroup attitudes. Moreover, these studies considered preferential treatment toward woman and foreigners but not toward older workers. Future research should therefore investigate whether perceptions of unfairness can explain how preferential treatment increases ageism at work.
A final limitation concerns the generalizability of our results. Study 1 was conducted in a hospital because of the effectiveness of the preferential treatment practices toward older workers. However, it would be interesting to analyse preferential treatment in other organisational contexts where similar practices exist (e.g., reduced working hours). Moreover, it would be interesting to investigate the relationships between preferential treatment and ageism toward older people outside the workplace. For example, future research could study whether preferential treatment like price reductions for seniors (e.g., in public transport or museum) can increase prejudice toward them. In the same way, it could be interesting to make varying the age of the target of Study 2 to support the results with a "young-older" worker (e.g., a 55 years old man). In sum, future research is needed to explore if detrimental effects of age-related preferential treatment may generalise to other contexts and to other populations of older individuals.

Practical implications
Our research points to the need to promote valuing differences linked to generations at work rather than simply activating self-categorisation as an "older worker" for justifying preferential treatment devices. Indeed, several studies showed that a typical multiculturalism perspective proposes that group differences and memberships should not only be acknowledged and considered, as it is the case when a policy of preferential treatment is adopted, but also celebrated (Richeson & Nussbaum, 2003;Yinger, 1994). Such a perspective emphasises the benefits of a diverse workforce and explicitly recognises employee differences as a source of strength (Plaut, Thomas, & Goren, 2009;Stevens, Plaut, & Sanchez-Burks, 2008). This approach is similar to the "integration and learning" perspective proposed by Ely and Thomas (2001) which consists in encouraging all employees to value and express themselves as members of their racial identity groups. Studies provide growing evidence that intergroup bias reduction and positive diversity climate are promoted by a multiculturalism perspective (Huo, Smith, Tyler, & Lind, 1996;Richeson & Nussbaum, 2003;Verkuyten, 2005;Wolsko, Park, & Judd, 2006). Therefore, we may expect the detrimental effects of preferential treatment to be diminished for underrepresented groups, like older workers in organisation endorsing a multiculturalism diversity perspective.

Conclusion
The present data provide additional evidence that affirmative action policies may entail some harmful consequences on about the way their beneficiaries (in this case older workers) are being perceived. As a matter of fact, older workers were more likely to be associated with a stigma of incompetence and to be the targets of contempt and less cooperative behaviours. Our efforts extend traditional research on age bias at work which has focused on stereotypes and discrimination without taking into consideration the affective com-ponent of ageism. To conclude, these findings should serve as a warning. Indeed, although achieving age diversity in work may be both a sensible and desirable objective, the method used to reach this goal can also trigger the unfortunate outcome of increasing ageism at work.