Introduction
Discrimination against older workers in the workplace
Policies in developed countries have promoted labour force participation among older adults since the 1990s, while policies on early retirement had been recommended before then (Philipson, Reference Philipson2012). Two factors contributed to these changes. One factor was the necessity to mitigate the increases in the number of retirees that otherwise negatively affected the sustainability of public pension systems (Pilipiec et al., Reference Pilipiec, Groot and Pavlova2021). To reduce expenditure in developed countries, redesigns of pension systems have included increasing the pension eligibility age and reducing benefits (Henkens and van Solinge, Reference Henkens, van Solinge, Ferraro and Carr2021). Another factor was the requirement to treat the shortage of young adults in the labour force due to declining fertility rate (Organisation for Economic Co-operation and Development (OECD), 2020). Whatever the case may be, older adults have been required to remain in the workforce to sustain their standard of living.
However, older adults face economic and psychosocial barriers when they remain and participate in the labour force. The system of seniority wages, where wages increase over time, has proved to be an economic barrier (Henkens and van Solinge, Reference Henkens, van Solinge, Ferraro and Carr2021). This system assumes that during the first phase of their career, employees’ earnings tend to be lower, and during the last phase of their career, their earnings tend to be proportionally higher than their actual productivity. The prospect of increased wages through this system operates as an incentive to stay in the enterprise. The abolition of this system could contribute to increased employment of older people owing to the resultant labour cost savings. Older adults also face psychosocial barriers, such as ageism. The OECD (2006) reports that employers’ negative stereotypes about the abilities and productivity of older workers are critical barriers to their employment prospects.
‘Ageism’ refers to stereotypes, prejudice and discrimination against a particular age group, especially elderly people (Iversen et al., Reference Iversen, Larsen and Solem2009). Studies on such stereotypes reported that employers and managers of enterprises had negative views of older workers. These included low productivity, creativity, flexibility, and willingness to be trained and learn new technology (Gringart et al., Reference Gringart, Helmes and Speelman2005; Van Dalen et al., Reference Van Dalen, Henkens and Schippers2009, Reference Van Dalen, Henkens and Schippers2010). Conversely, previous studies have observed that employers and managers value older employees’ social skills, reliability and commitment to the organisation (Van Dalen et al., Reference Van Dalen, Henkens and Schippers2009, Reference Van Dalen, Henkens and Schippers2010). Regarding discrimination, studies elucidated that the hiring and retention of older adults, compared to young adults, were a lower priority or possibility for employers (Van Dalen et al., Reference Van Dalen, Henkens and Schippers2009; Karpinska et al., Reference Karpinska, Henkens and Schippers2011). In addition, studies generally reported that these negative stereotypes were related to discrimination against older workers (Taylor and Walker, Reference Taylor and Walker1998; Gringart et al., Reference Gringart, Helmes and Speelman2005; Lu et al., Reference Lu, Kao and Hsieh2011; Karpinska et al., Reference Karpinska, Henkens and Schippers2013; Fasbender and Wang, Reference Fasbender and Wang2017).
Limitations in studies on discrimination in the workplace and the novelty of this study
There are some limitations in previous studies regarding ageism in the workplace. First, study participants were limited to employers, managers, students and older workers (Harris et al., Reference Harris, Krygsman, Waschenko and Rudman2018), resulting in few studies in which young workers participated (Patel et al., Reference Patel, Tinker and Corna2018). As there has been a rise in the proportion and number of older adults in the workforce due to ageing populations, it is becoming increasingly important for workers from different generations to be able to work together effectively (Patel et al., Reference Patel, Tinker and Corna2018). Despite this, as some previous studies have shown, young adults demonstrate more pronounced ageism than other age groups, while other studies report the contrary (Donizzetti, Reference Donizzetti2019; Marques et al., Reference Marques, Mariano, Mendonqa, De Tavernier, Hess, Naegele, Peixeiro and Martins2020). In addition, studies have reported that age-based stereotypes contribute to workers’ negative job attitude and their decisions to retire, especially among older workers (von Hippel et al., Reference von Hippel, Kalokerinos and Henry2013, Reference von Hippel, Kalokerinos, Haanterä and Zacher2019). Thus, in order to create environments that are conducive to the wellbeing of workers, especially older workers, it is important to gain insight into ageism among young workers as a distinct party in the workplace. However, this ageism remains unclear.
Second, although there have been many studies that focused on each individual concept, such as the stereotypes and discrimination which constitute ageism, few exist on the structures concerning the concepts and psychosocial predictors of ageism in the workplace. Moreover, studies apply only one model, the Intergroup Contact Theory, to explore the psychosocial predictors of ageism. Even in studies on psychosocial predictors of ageism outside the workplace, few utilise multiple models to explore them. In addition, few studies examine the paths of influence of psychosocial predictors on discrimination. As studies indicate significant relationships between stereotypes and discrimination, among the concepts of ageism, based on the Attitude–Behaviour Model (Lu et al., Reference Lu, Kao and Hsieh2011), it is possible that the influence of psychosocial factors on discrimination constitute direct and indirect paths through stereotypes.
Third, many studies on workplace ageism have been conducted in Western countries with few investigating Eastern countries (Harris et al., Reference Harris, Krygsman, Waschenko and Rudman2018). Thus, it is unclear whether findings in the Western population can be generalised to Eastern cultures that tend to be more respectful of older people (Wilińska et al., Reference Wilińska, de Hontheim, Anbäcken, Ayalon and Tesch-Römer2018).
This study explores the psychosocial predictors of young workers’ discrimination against older workers in Japan by using multiple analytical models. The novel contributions of this study include (a) focusing on discrimination against older workers in young workers, (b) using the multiple types of the analytical model, (c) examining paths of influence of psychosocial predictors on discrimination directly or indirectly through stereotypes, and (d) conducting in a country other than Western countries.
Models of psychosocial factors of discrimination and the hypotheses of this study
We formulated the hypotheses based on four analytical models: Intergroup Contact Theory (ICT), Knowledge–Attitude–Behaviour Model (KABM), Terror Management Theory (TMT) and Frustration–Aggression Theory (FAT). Regarding ICT, Allport (Reference Allport1954) provided a theory by which the positive effects of intergroup contact occurred only in situations marked by four key conditions. These were equal group status within the situation, common goals, intergroup co-operation, and the support of authorities, law or custom (Allport, Reference Allport1954). This theory indicated that mere contact between groups was not sufficient for the reduction of negative prejudice or stereotypes. While a few studies examined the relationships between contact with older adults and their discrimination in the workplace, their results were inconsistent. In a study by Henkens (Reference Henkens2005), the frequency of contact with older workers significantly predicted managers’ attitudes towards their early retirement. Conversely, Lu (Reference Lu2010) and Lu et al. (Reference Lu, Kao and Hsieh2011) reported that the quality of contact with older workers did not relate significantly to either managers’ intention to hire older adults or co-workers’ intention to work with them.
Chung and Park (Reference Chung and Park2019) applied KABM to explore the predictors of discrimination of older adults. According to this model, reasonable attitudes were products of reasoned knowledge, and these attitudes could be expressed explicitly, leading to specific behaviours towards an indicated target. In empirical studies that targeted workplace discrimination, stereotypes measured as attitude indicators had a significant influence on discrimination (Taylor and Walker, Reference Taylor and Walker1998; Gringart et al., Reference Gringart, Helmes and Speelman2005; Henkens, Reference Henkens2005; Lu, Reference Lu2010; Lu et al., Reference Lu, Kao and Hsieh2011; Karpinska et al., Reference Karpinska, Henkens and Schippers2013; Fasbender et al., Reference Fasbender and Wang2017). However, few studies have examined the validity of the causal model of knowledge–attitude–behaviour in the workplace. Only one study, using the general population as survey participants, examined the validity of this model and obtained the results that validated it (Chung and Park, Reference Chung and Park2019).
Death, according to TMT, represents a potent threat to the human psyche, and people defend against it by clinging to cultural systems of belief and striving to maintain a sense of self-esteem (Martens et al., Reference Martens, Goldenberg and Greenberg2005). According to this theory, as older adults arouse contemplation of mortality, physical distancing becomes the direct way to assuage this sentiment. In addition, psychological distancing also serves to minimise this perceived threat. There have been no studies, however, based on this theory, that have examined the relationship between death/ageing anxiety and the workplace discrimination of older adults. Nevertheless, some studies with participants outside the workplace provided results which supported this model (Allan and Johnson, Reference Allan and Johnson2008; Boswell, Reference Boswell2012; Allan et al., Reference Allan, Johnson and Emerson2014; Barnett and Adams, Reference Barnett and Adams2018; Donizzetti, Reference Donizzetti2019; Cooney et al., Reference Cooney, Minahan and Siedlecki2021).
Finally, we also investigated the FAT model, according to which, frustrations can give rise to aggressive inclinations because they are aversive and presumably produce an instigation to aggression only to the extent that they are unpleasant to people affected (Berkowitz, Reference Berkowitz1989). Although aggression is frequently directed towards the agent perceived to have provoked frustration, sometimes other features of the situation elicit restraint. When any of these constraining factors are present, direct aggression is often controlled and, furthermore, it is alleged to be redirected towards or displaced on to less-powerful or more-available targets (Marcus-Newhall et al., Reference Marcus-Newhall, Pedersen, Carlson and Miller2000). Palmore (Reference Palmore, Palmore, Branch and Harris2005) provided this theory as an individual source of ageism. According to this theory, young workers who feel dissatisfied at work have a greater likelihood of experiencing negative emotions, which are likely to promote discrimination against their older counterparts, instead of the agents perceived to have provoked job dissatisfaction, as a possible response to these negative emotions. Previous studies have neither examined the validity of this theory in the workplace nor in other settings.
Allan and Johnson (Reference Allan and Johnson2008), Barnett and Adams (Reference Barnett and Adams2018), Boswell (Reference Boswell2012) and Cooney et al. (Reference Cooney, Minahan and Siedlecki2021) have conducted studies that employed three models to explore the psychosocial predictors of ageism. However, some of these studies did not explicitly explain their theoretical model. Notably, the aforementioned four studies included predictors related to ICT, KABM and TMT. While the results in the first three studies supported all the models, the results in the last study supported only KABM and TMT.
Our analytical framework is shown in Figure 1. Based on ICT, KABM, TMT and FAT, our framework included each possible predictor of interactions with older workers, knowledge about older adults, fear of ageing and job dissatisfaction. Each predictor could have direct and indirect influences on workplace ageism through positive and negative dimensions of stereotypes. Therefore, we hypothesise the following:
• Hypothesis 1: Based on ICT, interactions with older workers have an influence on discrimination against older workers directly and through stereotypes about older workers.
• Hypothesis 2: Based on KABM, knowledge of older workers has an influence on discrimination against older workers directly and through stereotypes about older workers.
• Hypothesis 3: Based on TMT, fears of getting older workers have an influence on discrimination against older workers directly and through stereotypes about older workers.
• Hypothesis 4: Based on FAT, job dissatisfaction has an influence on discrimination against older workers directly and through stereotypes about older workers.
Figure 1. The analytic framework.
Selection of Japan as the study setting
We selected Japan as the study setting. Its population has aged more rapidly than in most developed countries (OECD, 2018). Due largely to the increased demographic pressures in Japan, an extension of older peoples’ working lives attempts to mitigate the decline in the workforce and the rising health and pension expenditure (Williamson and Higo, Reference Williamson and Higo2009). While older workers, especially males, continued to constitute a significant demographic in the labour market compared to other developed countries (OECD, 2021), the government intervened to secure employment for employees until the age of 65. The pillar of this employment policy was the increase in the mandatory retirement age. Through the Act on the Stabilization of Employment of Older Persons enacted in 1986, employers were obliged to make efforts to ensure that the mandatory retirement age did not fall below 60. Through the revised act in 1994, employers were prohibited from adopting a mandatory retirement age less than 60 years. Through the revised act of 2004, employers were obliged to adopt one of three measures to secure employment for employees until the age of 65 years: (a) extend the mandatory retirement age to 65 years; (b) re-hire older workers who wished to continue working after 60 years, which is the mandatory retirement age; or (c) abolish the mandatory retirement system. The 2020 Status of Employment of Elderly Persons report stated that almost all enterprises with 31 or more full-time workers provided employment security until the age of 65. Approximately 80 per cent of them re-hired their older workers (Ministry of Health, Labour and Welfare, 2021). Owing to these policy trends, there are more opportunities for intergenerational co-operation than ever before, but it also gives rise to the possibility of intergenerational conflict. However, there have rarely been any Japanese policies which combat negative attitudes towards older workers, thereby contributing to manage intergenerational conflict and creating an inclusive age-diverse culture in the workplace (OECD, 2018). Compared to Western nations, it is a popular belief that East Asian culture prescribes a notable tradition of respecting elderly people. The Chinese, the Japanese and the Koreans have shared this tradition for many generations (Sung, Reference Sung2001), although some empirical studies do not support this perspective (Lin and Bryant, Reference Lin and Bryant2009). Under the employment policy for older adults coupled with the tradition that values respect for older adults, young workers may have predictors of discrimination that differ from their counterparts in Western nations. Accordingly, this study in Japan could examine a generalisation of psychosocial predictors of discrimination hypothesised on the basis of theoretical frameworks of Western culture.
Methods
Participants
We used cross-sectional data collected through a web survey in 2020. The participants comprised young workers based on the following four criteria: (a) 25–39 years of age, (b) employees, (c) male, and (d) living in cities of the Tokyo metropolitan area, including Saitama, Chiba, Tokyo and Kanagawa prefectures. The reasons why we selected participants based on these criteria were as follows.
Age
Previous studies have not established a precise age range for ‘young workers’. However, an ‘older worker’ has typically been operationalised in the literature as a worker between the ages of 55 and 65 years, while a ‘young worker’ has ranged between the ages of 24 and 34 years (Finkelstein et al., Reference Finkelstein, Burke and Raju1995, Reference Finkelstein, Ryan and King2013). We made the decision to distinguish the age range for young workers clearly from that of older workers as follows. None of the reviewed research considered workers aged 40 years and below to be ‘older workers’ and none aged 55 years and above as ‘young workers’ (Gringart et al., Reference Gringart, Helmes and Speelman2013). Consequently, we defined the age range of young workers as 25–39 years old.
Employees
There were no difficulties in clarifying the actual situation of discrimination towards older workers among workers aged 25–39 years even if we limited possible participants to employees, because the proportion of employees among workers aged between 25 and 39 years was 88.1 per cent (Statistics Bureau of Japan, 2017).
Male
Young male workers are likely to display higher ageism in the workplace than young female workers. Males are found to have more work-centrality than females (Mannheim, Reference Mannheim1993; Harpaz and Fu, Reference Harpaz and Fu1997), and work centrality among young workers is related to ageism in the workplace (Ospina, Reference Ospina2015). In fact, according to a systematic review by Marques et al. (Reference Marques, Mariano, Mendonqa, De Tavernier, Hess, Naegele, Peixeiro and Martins2020), out of 67 studies which explored determinants as sex, 23 indicate significant higher ageism among males towards older adults and only three studies reported significantly higher ageism among females. In terms of younger adults, Smith et al. (Reference Smith, Bergeron, Cowart, Ahn, Towne, Ory, Menn and Chaney2017) and Luo et al. (Reference Luo, Zhou, Jin, Newman and Liang2013) found that younger males displayed stronger ageism towards older adults than younger females. In addition, as described formerly, as we used web survey data, there are concerns regarding sampling bias in the data. We used the data from a representative sample of males aged 25–39 years in 2003 (Harada et al., Reference Harada, Sugisawa, Sugihara, Yamada and Shibata2004). Using the data from the 2003 survey, we can examine the levels of sampling bias in the data used in this study.
Living in cities of the Tokyo metropolitan area
A substantial concentration of male employees aged between 25 and 39 years lived in these areas of Japan (Statistics Bureau of Japan, 2018).
Procedures
This study utilised a voluntary web survey. We asked a web survey company (Macromill Inc.) to collect the complete data of 1,500 participants and conducted the web survey methodically as follows. First, based on the sampling criteria described before, the web survey company selected possible participants (N = 37,936) from a survey panel (1,295,027 as of 1 June 2020). They sent co-operative survey requests via email to these participants. Only individuals on the panel who consented to participate in this survey could respond to the questionnaire. The survey ran for three days in November 2020, until we obtained 1,500 participants. Successfully conducting the online survey without sampling bias raised concerns related to the possibility that participants would not fully engage their efforts to answer the questionnaires truthfully (Heerwegh and Loosveldt, Reference Heerwegh and Loosveldt2008). In response to this, we required the participants to take an oath to answer truthfully (TO) and added an instructional manipulation check (IMC), to filter out unmotivated or dishonest participants (Masuda et al., Reference Masuda, Sakagami and Morii2019). If they did not take the oath, we could prevent them from further participation. We added two kinds of instructional manipulation in the IMC. First, we added two instructions to select a specific choice in matrix-style questions. Second, we gave response directions in relatively longer-form questions. When participants did not read questions well, they tended to provide answers that differed from the instructions. We excluded participants that gave one or more wrong answers. As a result, we utilised data from 874 participants for further analyses in this study.
Measures
Discrimination
To evaluate young workers’ prejudices, we employed a scale developed by Chiu et al. (Reference Chiu, Chan, Snape and Redman2001). In the question text, we defined ‘older workers’ as those aged 60 and above. This age range would define ‘older workers’ in the other scales employed in this survey. This measure included five items. An example of an included item was, ‘It is a better investment to train younger workers rather than older workers.’ We included one reversed item in this scale. All items are shown in the Appendix. There were five choices (scores to quantify) with which to respond, from ‘strongly agree’ (+2) to ‘cannot say either’ (0) to ‘strongly disagree’ (−2). Chiu et al. (Reference Chiu, Chan, Snape and Redman2001) employed each item separately as a dependent variable because this scale had low reliability as a five-item scale (alpha-coefficient = 0.4). Moreover, principal component analyses extracted the first component that accounted for only 37 per cent of the total variance. However, according to our data, model fit statistics for a one-factor model by confirmatory factor analysis (CFA) were 0.070 for the root mean square error of approximation (RMSEA), 0.978 for the comparative fit index (CFI) and 0.944 for the Tucker–Lewis index (TLI). These statistics indicated moderate acceptance of criteria regarding goodness of fit with an alpha-coefficient of 0.696. Accordingly, we created a one-dimension scale composed of five items. The result for each participant was the average of their total scores. This meant that the participants with a score greater than zero had a stronger tendency towards discrimination than those in a more neutral position.
Stereotypes
We based the scales that evaluated participants’ views on older workers on those that Loretto et al. (Reference Loretto, Duncan and White2000) adapted from Lyon and Pollard (Reference Lyon and Pollard1997). This measure included 13 items, which are shown in the Appendix. There were three choices (scores to quantify) with which to respond, from ‘agree’ (+1) to ‘cannot say either’ (0) to ‘disagree’ (−1). While previous studies have not provided factor structures for this scale, according to a review article by Harris et al. (Reference Harris, Krygsman, Waschenko and Rudman2018), positive and negative dimensions largely comprised stereotypes towards older workers. The CFA examined the model validity of the two dimensions. According to CFA for these data, model fit statistics for the two-factor model were 0.065 for RMSEA, 0.946 for CFI and 0.925 for TLI. These figures indicated moderate acceptance of criteria regarding goodness of fit. The alpha-coefficients for the positive and negative dimensions were 0.760 and 0.724, respectively. We calculated the score for each participant using the same method as that of the discrimination scale. This meant that the participants with a score greater than zero had a stronger tendency towards the positive or negative dimension of stereotypes than those in a neutral position.
Interactions with older workers
We characterised contact with older workers as supportive or uncomfortable, in addition to avoidance (accommodation, non-accommodation and avoidance for each name in the scale by McCann and Keaton) to evaluate interactions with others by referring to a scale created by McCann and Keaton (Reference McCann and Keaton2013). It appears that influences of contacts with older adults on positive/negative stereotypes of older adults differ by either positive or negative contacts (Cadieux et al., Reference Cadieux, Chasteen and Packer2019). We have added the dimension of avoidance, as higher avoidance could also be associated with lower expressed satisfaction to intergenerational contacts (McCann et al., Reference McCann, Dailey, Giles and Ota2005). These dimensions have been assessed based on five choices, ranging from ‘strongly agree’ (+2) to ‘cannot say either’ (0) to ‘strongly disagree’ (−2). We employed CFA to examine configural invariance. As McCann and Keaton (Reference McCann and Keaton2013) provided information on the factors that the items were strongly associated with, we determined the configural invariance of both our scale and the original scale (Gregorich, Reference Gregorich2006). The CFA provided model fit statistics for the three-factor model that were 0.033 for RMSEA, 0.994 for CFI and 0.988 for TLI. These figures indicated an acceptance of criteria regarding goodness of fit. The alpha-coefficients for supportive contact, uncomfortable contact and avoidance were 0.829, 0.906 and 0.871, respectively. We calculated the interaction score for each participant using the same method as that of the discrimination scale. This meant that participants with a score greater than zero had a stronger tendency towards a given dimension regarding interactions with older workers than those in a neutral position.
Knowledge about older adults and ageing
We utilised Palmore's (Reference Palmore1980) Facts on Aging Quiz. The original scale comprised 25 items. We deleted the final item because it inquired about a future forecast, 2010 to be specific, at the time of its formulation in the 1970s. We altered questions about ageing and the admission rate among older adults for the purpose of our study. Choices included ‘yes’, ‘no’ and ‘do not know’. We calculated the score for each participant using the number of incorrect answers and ‘do not know’, out of the total number of questions (i.e. 24).
Fear of being an older worker
This study used fear of being an older worker instead of death anxiety, focusing on discrimination against older workers which were not directly associated with death. In addition, studies have stated that ageing anxiety predicts death anxiety (Benton et al., Reference Benton, Christopher and Walter2007) and that fear of ageing was a significant factor influencing discrimination against older adults; however, death anxiety, which was a primary predictor from TMT, was not significant in the study by Chonody et al. (Reference Chonody, Webb, Ranzijn and Bryan2014). Sargent-Cox et al. (Reference Sargent-Cox, Rippon and Burns2014) developed the scale regarding anxiety about ageing which comprised four dimensions. We employed only one dimension, in this scale, namely ‘fear of loss’. This construct is referred to as ‘fear of ageing’ hereafter. We revised the wording of each item to indicate fear of loss when one becomes an older worker, as opposed to fear of loss due to ageing. Three of the five items in this dimension are as follows: ‘I fear that when I become an old worker, I will not be able to have casual conversations with my colleagues’ and ‘The older I get, the more I worry about my diminishing productivity’. All items are shown in the Appendix. The dimension was assessed based on five choices, ranging from ‘strongly disagree’ (−2) to ‘cannot say either’ (0) to ‘strongly agree’ (+2). As we utilised only one dimension comprising five items, we examined the configural invariance of both our scale and the original scale. The CFA provided model fit statistics for a one-factor model as follows: 0.095 for RMSEA, 0.975 for CFI and 0.938 for TLI. These figures indicated moderate acceptance of criteria regarding goodness of fit with an alpha-coefficient of 0.803. We calculated the score for each participant using the same method as that of the discrimination scale. This meant that participants with a score greater than zero had a stronger tendency to fear loss than those in a neutral position.
Job dissatisfaction
We created this scale for this study, because existing ones included relatively numerous items. Existing scales for job satisfaction include dimensions that comprise salary, promotions, colleagues, supervision and working conditions (Macdonald and MacIntyre, Reference Macdonald and MacIntyre1997; Özpehlivan and Acar, Reference Özpehlivan and Acar2016). This study employs a five-item scale to evaluate dissatisfaction that distributed one item to each dimension: salary, promotions, colleagues, supervision and working conditions. These were assessed based on five choices, ranging from ‘very satisfied’ (−2) to ‘cannot say either’ (0) to ‘very dissatisfied’ (+2). The CFA provided model fit statistics for a one-factor model: 0.088 for RMSEA, 0.976 for CFI and 0.941 for TLI. These figures indicated moderate acceptance of criteria regarding goodness of fit with an alpha-coefficient of 0.754. We calculated the job dissatisfaction score for each participant using the same method as the discrimination scale. This meant that participants with a score greater than zero had a stronger tendency towards dissatisfaction than those in a neutral position.
Control variables
Control variables included age, education, annual income, marital status, job status, occupational categories, the number of full-time employees in the enterprise, and the proportion of employees aged 60 years and above. In terms of education, participants were asked to indicate their highest academic qualification from the following list: ‘junior high school graduate’, ‘high school graduate’, ‘vocational school graduate’, ‘junior college graduate’, ‘university graduate’ or ‘master's or doctoral-level graduate’. The numbers 9, 12, 14, 14, 16 and 18 were assigned to each category to quantify the responses. These numbers reflect the number of years in education according to the Japanese educational system. Annual income including tax was measured by using 11 categories, which ranged from ‘less than one million yen (9,100 dollars)’ to ‘more than 10 million yen (91,000 dollars)’. The mid-point of each category was used for quantification. For example, the categories ‘1–2 million yen’ and ‘2–3 million yen’ were assigned 1.5 and 2.5 million yen, respectively. Marital status was categorised as having or not having a spouse. Job status was categorised as full-time and part-time contract or temporary. Occupational categories were as follows: clerk, sales, services, skills, professionals and others. The number of full-time employees was grouped into five categories: 0–29, 30–299, 300–999, and 1,000 and over. The proportion of employees aged 60 years and above was grouped into four categories: no one, less than 10 per cent, 10 per cent and over, and unknown.
Measurement invariance
There were great concerns about sampling bias of the web survey used in this study (Couper and Miller, Reference Couper and Miller2008). As mentioned under ‘Participants’, we use data from a survey conducted in 2003 for males aged 25–39 years who lived in the same areas; mail distribution and in-person visits were employed for administering and collecting self-report questionnaires (Harada et al., Reference Harada, Sugisawa, Sugihara, Yamada and Shibata2004). The potential participants in the 2003 survey were obtained by stratified random sampling using the residential registry, and the response rate was 40 per cent. We evaluated the measurement invariance of three scales (discrimination, stereotypes and job dissatisfaction), as these were used both in this study and in the 2003 survey.
The multi-group CFA examined the metric invariance model that required corresponding factor loadings to be equal across two groups. The multi-group CFA determined the metric invariance model of all three scales. In fact, the change of each fit index, namely CFI and RMSEA, in the metric invariance model, in comparison to the configural invariance model, was 0.01 and 0.007, respectively, for discrimination, 0.004 and 0.000, respectively, for stereotypes, and 0.002 and 0.015, respectively, for dissatisfaction. A change less than 0.01 in CFI, accompanying a change less than 0.015 in RMSEA, would indicate invariance (Chen, Reference Chen2007). Confirmation of measurement invariances of the three scales between the two models indicated the low sampling bias in this study.
Statistical analysis
We utilised Mplus version 8.1 (Muthén and Muthén, Reference Muthén and Muthén1998–2017) for data analyses. We employed multiple mediation analysis, which Preacher and Hayes (Reference Preacher and Hayes2008) proposed, to estimate direct and specific indirect influences from possible psychosocial predictors of discrimination against older adults. We standardised all variables without nominal values in the model to compare the direct and indirect influences. We utilised bootstrapping to estimate the total direct and specific indirect effects of these mediators. Using the null hypothesis, we determined point estimates and 95 per cent confidence intervals. With regard to missing values, 25.4 per cent of the participants responded, ‘do not know’ to the question about the participation rate of employees aged 60 years and above. We included them in our analyses, creating ‘do not know’ as a new category. Scales regarding interactions with older workers included ‘no chance of interaction with older adults’ as one of the choices. At least 22.9 per cent of the participants selected this choice, and we assigned them a missing value for this scale. For the question concerning educational attainment, 0.2 per cent of the participants chose ‘others’. Therefore, we treated ‘others’ as a missing value. In our analysis, we employed a full information maximum likelihood approach to handling missing cases in these scales (Muthén and Muthén, Reference Muthén and Muthén1998–2017).
We evaluated our overall model fit using RMSEA and CFI. For good model fit, RMSEA values should be below 0.05 (Browne and Cudeck, Reference Browne and Cudeck1992) and CFI values should be above 0.90 (Bentler, Reference Bentler1990).
Ethical considerations
The study complied with the guidelines of the Declaration of Helsinki. The research ethics board at J. F. Oberlin University approved all procedures therein. We sent a mail of invitation that clearly explained the aims and content of the study to each potential participant. Only potential participants who consented to participate answered the questionnaires. We adopted data collection procedures and storage management that ensured confidentiality. Participation was voluntary, and we guaranteed strict confidentiality.
Results
Characteristics of participants
Table 1 shows the distribution of discrimination and stereotypes of older workers and the possible psychosocial predictive and control variables. The average discrimination score was 0.63, which indicated that half of participants selected ‘cannot say either’ and ‘strongly agree’ on average. In terms of stereotypes, although both positive and negative dimensions were near neutral, the negative dimension was slightly more dominant than the positive. Among possible predictors, participants experienced fewer negative interactions. The Facts on Aging Quiz yielded an incorrect answer rate of 56.9 per cent.
Table 1. Distribution of discrimination of older workers, independent variables and control variables of study participants
Notes: N = 874. 1.Figures were calculated before standardisation. 2.Values were obtained only from participants who had no missing values for each variable. The proportion of missing cases in education and interaction was 0.2 and 22.9 per cent, respectively. There were no missing cases in other variables outside these variables. 3.One unit: one million yen (roughly US $10,000). SD: standard deviation.
Influences of possible psychosocial predictors on discrimination
Table 2 shows the influences that possible psychosocial predictors can have on discrimination.
Table 2. Possible psychosocial predictors of direct and indirect influences on discrimination against older workers
Notes: All variables were standardised. Fitness index: root mean square error of approximation = 0.040, comparative fit index = 0.936. Ref.: reference category.
Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.
The overall model fit was 0.040 for RMSEA and 0.936 for CFI, which indicate a moderate fit. Hypothesis 1 based on ICT was supported for two out of the three indicators on interactions with older workers. Both higher supportive interactions and higher uncomfortable interactions were significantly and directly related to lower and higher discrimination, respectively. Additionally, higher supportive interactions were significantly and indirectly related to lower discrimination through both dimensions of stereotypes (via higher positive and lower negative dimensions). Conversely, higher uncomfortable interactions were significantly and indirectly related to higher discrimination through higher positive and lower negative dimensions of stereotypes. Hypothesis 4 based on FAT was supported. Higher job dissatisfaction was indirectly related to higher discrimination through a lower positive dimension of stereotypes. Interestingly, contrary to Hypothesis 3, which is based on TMT, a greater fear of being an older worker was significantly related to lower discrimination through a higher positive dimension of stereotypes. Hypothesis 2 based on KABM was not supported by the results. More knowledge about older adults and ageing had no significant direct or indirect influences on discrimination through stereotypes regarding older workers.
Discussion
We can compare the results regarding stereotypes and young workers’ discrimination against older workers with the results from studies in other countries. Evidently, there are some limitations due to the differences in participants’ characteristics and survey periods. Chiu et al. (Reference Chiu, Chan, Snape and Redman2001) compared stereotypes and discrimination of older workers among part-time students with experiences of paid work in the United Kingdom (UK) and Hong Kong. The scales utilised were similar to some employed in this study. Although the study did not provide the measure of discrimination in both countries, we estimated them from their means, standard deviations and correlation of related variables. Discrimination against older workers was slightly weaker than it was against young workers among both countries’ students, while young workers in Japan displayed more discrimination against older workers than towards young workers. In addition, they showed that both positive and negative dimensions of stereotypes for older workers (referred to as ‘effectiveness’ and ‘adaptability’) were almost the same in the UK. However, the levels of the negative dimension exceeded those of the positive dimension in Hong Kong. In another UK student survey (Loretto et al., Reference Loretto, Duncan and White2000), the positive dimension of stereotypes exceeded those of the negative, contrary to the results in this study. While there are 20-year time lags between the surveys by Chiu et al. (Reference Chiu, Chan, Snape and Redman2001) and Loretto et al. (Reference Loretto, Duncan and White2000), this study suggests that young workers in Japan discriminate more severely against older workers than young people in the UK.
This study explores direct influences of psychosocial predictors on discrimination against older workers and indirect influences through stereotypes in Japanese young male workers. We developed four hypotheses based on the ICT, KABM, TMT and FAT models. Hypothesis 1 based on ICT indicated that mere quantity of interactions between groups was not sufficient for the reduction of negative prejudice or stereotypes (Allport, Reference Allport1954). However, as few studies examined differences in the relationships between interactions with older adults and discrimination due to differences in quality of interactions, we used three scales to measure interactions with older workers evaluating different qualities (supportive, uncomfortable and avoidance). The results of this study supported Hypothesis 1 for two scales of interactions. Higher supportive interactions with older workers were significantly related to lower discrimination both directly and indirectly through higher positive and lower negative stereotypes. Additionally, higher uncomfortable interactions with older workers significantly contributed to higher discrimination directly and indirectly through lower positive and higher negative stereotypes. These results also suggest that influences of positive and negative interactions on discrimination mediate changes in both negative and positive stereotypes. In this study, avoidance neither directly nor indirectly influenced discrimination. As Allport (Reference Allport1954) highlighted, the quality of contacts with older adults, rather than frequency, is more likely to increase the positive dimension of stereotypes regarding older adults (Marques et al., Reference Marques, Mariano, Mendonqa, De Tavernier, Hess, Naegele, Peixeiro and Martins2020). As avoidance seems to be a characteristic from the quantitative aspect of interpersonal interactions rather than the qualitative, it is possible that the influences of the avoidance scale could be weaker than those of supportive and uncomfortable interactions.
Hypothesis 4 based on FAT was supported by the results in this study. Higher job dissatisfaction influenced higher discrimination against older workers through lower positive stereotypes of older workers, and not through higher negative stereotypes. These indirect influences indicate the mechanism of influences. In cases where aggression cannot be directed at the source of frustration, the person might direct their aggression at another social group who are a weaker target. According to Hinton's point of view, the scapegoated group targeted through the redirected aggression are seen in negative and stereotypical terms (Hinton, Reference Hinton2000); it is possible that discrimination against older workers appears through the lower positive stereotypes regarding older adults due to work dissatisfaction.
Hypothesis 2 based on KABM was not supported by the results: more knowledge of older adults did not have an influence on lower discrimination against older workers both directly and indirectly through the stereotypes about older workers. However, we analysed the model on knowledge of older adults after excluding three scales – interactions with older workers, job dissatisfactions and fear of being an older worker – related to analytical models other than KABM. Our analysis revealed that participants with higher knowledge of older adults had lower discrimination through higher positive stereotypes regarding older workers (not presented in the results). Although previous studies obtained mixed results regarding whether knowledge of older adults has a negative influence on ageism (Allan and Johnson, Reference Allan and Johnson2008), there are a few studies which controlled influences of factors related to other models of origin of ageism. In addition, Allan and Johnson (Reference Allan and Johnson2008) found that although knowledge of older adults affects age discrimination, its influence was not direct and was indirectly mediated through the effect on anxiety. Our results indicate that knowledge of older adults influences discrimination against older workers indirectly through predictors related to other models as mediators.
This study does not support Hypothesis 3 based on TMT despite the support of most previous studies that examined psychosocial predictors. We obtained results that were contrary to this theory. Higher fear of being an older worker was indirectly and significantly related to low discrimination through a higher positive dimension of stereotypes. When we excluded the positive dimension of stereotypes from the model, the higher fear of being an older worker contributed significantly to higher discrimination through the higher negative dimension of stereotypes (not presented in the results). That is, adding positive stereotypes caused results that differed from previous studies. Vail et al. (Reference Vail, Juhl, Amdt, Vess, Routledge and Rutjens2012) recommended a closer look at the conceptual foundations of TMT. This would be in consideration of some of the more positive, or optimal, trajectories that terror management efforts can foster. For the most part, deleterious consequences frequently featured in the TMT literature have overshadowed these trajectories. Most previous studies examine the direct influences that variables related to TMT have on discrimination or ignore the positive trajectories of the fear of loss through ageing on reducing discrimination. This study suggests that fear of loss may help reduce discrimination through positive stereotypes.
This study has practical implications for reducing discrimination against older workers by young male workers. There have been few studies which applied the frustration–aggression model to examine empirically the origins of discrimination of older adults. We indicate that job dissatisfaction may lead to not only lower positive stereotypes but also discrimination for older workers. An organisation needs to focus on the attenuation of young male workers’ dissatisfactions related to their job to reduce their discrimination of older adults. While there have been few studies on young workers, a lot of studies show that boosting the quality of interactions with older adults contributes to reduced negative stereotypes (Lytle and Levy, Reference Lytle and Levy2019). This study suggests that not only boosting the quality of interactions but also reducing bad interactions with older workers contribute to lower discrimination. A company needs to focus on promotion of collaborative works and reduction of intergenerational conflicts between young and older workers.
Furthermore, there are concerns regarding the result that higher positive stereotypes are mediators between the quality of interactions with older workers and discrimination. It is suggested that positive self-stereotypes are salutary to older adults’ health and wellbeing (Levy et al., Reference Levy, Slade, May and Caracciolo2006, Reference Levy, Slade, Murphy and Gill2012; Söllner et al., Reference Söllner, Dürnberger, Keller and Florack2022), and they serve as a way for older adults to be treated negatively (Mulders, Reference Mulders2020). Despite the ‘positive’ nature of positive stereotypes, they are not without fault and pose their own set of challenges. First, positive stereotypes may provoke negative emotion through depersonalisation (Siy and Cheyan, Reference Siy and Cheyan2013). Positive stereotypes signal the stereotyper's judgement of the targeted group. It represents depersonalisation based solely on their group membership rather than on their individual traits and attributes, even if they are recognised as having positive characteristics. Such disregard for one's individuality constitutes a threat, especially for people who define the self as unique and distinct. The sense that one is being reduced to one's group membership leads to the derogation of those who instigate the threat and provoke negative emotions such as anger. Second, positive stereotyping is a resource that helps apply pressure on the targeted group to comply with the expectancy context related to this stereotype. Positive stereotype is more prescriptive than negative stereotypes which tend to be more descriptive (Czopp et al., Reference Czopp, Kay and Cheryan2015). Prescriptive stereotypes create an expectancy context that is more likely to encourage and reinforce stereotype-consistent behaviours than the descriptive properties of negative stereotypes. Under the expectancy context, people may feel great anxiety. Third, complimentary relationships between negative and positive stereotypes reinforce awareness of being discriminated against among the targeted group (Kahalon et al., Reference Kahalon, Shnabel and Becker2018). In the case of age discrimination, positive stereotypes regarding older workers, as opposed to negative stereotypes, put older workers on a metaphorical pedestal by highlighting their patient, reliable and mature attributes. Despite its seemingly positive tone, by the interdependency and complementarity of both stereotypes of older workers, the perspective of anti-age discrimination subtly implies that older workers need younger workers’ help by characterising younger workers as being designed naturally for priority positions because older workers lack flexibility and productivity (Kahalon et al., Reference Kahalon, Shnabel and Becker2018). As few studies explore positive stereotypes regarding older workers and their negative influence on older workers, we need to explore these issues further.
Some limitations of this study may bias the results, and thus, we provide some recommendations to reduce these biases. First, our data did not include female participants. A study indicated that although the Facts on Aging Quiz was negatively associated with ageism, this effect was moderated by sex: males who knew less about the ageing process reported more negative ageist behaviour (Stahl and Metzger, Reference Stahl and Metzger2013). Compared with men, women also experience greater anxiety about their own ageing (Cummings et al., Reference Cummings, Kropf and De Weaver2000; Abramson and Silverstein, Reference Abramson and Silverstein2006), perhaps stemming from the construction of ageing as a more negative experience for women. Consequently, young female workers more strongly exhibit discrimination. By including young female workers as participants, it is essential to examine whether the results of this study are observed across both sexes.
Second, the study participants may not fully represent the population. We selected them through a voluntary web survey, which involves selection bias because possible participants are limited to individuals who register with web research companies. This survey panel is also likely to have higher information and communication technology skills and higher socio-economic status (Bethlehem, Reference Bethlehem2010). Using the 2015 census data, we determined that the educational attainment of our participants was higher than that of male employees in the general population. In addition, the proportion of our participants who were professionals and clerks was higher and the proportion of our participants who worked in manual labour was lower than in the general population of male employees. However, a study reported that a voluntary web survey was beneficial for explanatory research (Fabo and Kahanec, Reference Fabo and Kahanec2018). In addition, comparisons to determine representativeness of the sample tended to support the measurement invariance of some scales in our data. Accordingly, it is possible that the selection bias displayed by the voluntary web survey has only a weak influence on the results in this study.
The third bias concerns influence by social desirability. Caution is warranted given that social desirability bias may have lowered rates of reported discriminatory behaviours (Harris et al., Reference Harris, Krygsman, Waschenko and Rudman2018). As a web survey can provide higher confidentiality to participants than interview survey, social desirability bias may not necessarily be of concern.
Fourth, panel data may be used as it is possible to confirm causal relationships between psychosocial predictors and discrimination with more certainty than by using cross-sectional data. While this study indicated that more positive contacts with older workers reduced discrimination, the results can be interpreted as people who had already not discriminated against older workers were also more likely to engage in high-quality communications with older workers.
Despite these limitations, this study has overcome some of the shortcomings of previous studies. There have been few studies on psychosocial predictors of workplace discrimination against older adults based on multi-theoretical models. This study suggests that three theories regarding ageism outside the workplace can be applied to elucidate young workers’ discrimination against older workers in Japan.
Financial support
This work was supported by the J. F. Oberlin University Research Promotion Fund (grant number 20-74).
Ethical standards
The study complied with the guidelines of the Declaration of Helsinki. The research ethics board at J. F. Oberlin University approved all procedures therein.
Conflict of interest
The author declares no conflicts of interest.
Appendix: Questionnaires
(1) Discrimination against older workers:
• It is a better investment to train younger workers rather than older workers.
• Older employees should ‘step aside’ (take a less-demanding job) to give younger employees advancement opportunities.
• Younger employees should be given priority to stay if a company contracts its business.
• Given a choice, I would not work with an older worker on a daily basis.
• If two workers had similar skills, I'd pick the older worker to work with me (reversible).
(2) Stereotypes for older workers:
Older workers
• are better at team work.
• are better at interpersonal skills.
• are more patient.
• are more conscientious.
• are more reliable.
• are more committed.
• are more mature.
• have lower expectations.
• are less flexible.
• are less productive.
• are prone to higher absenteeism.
• are resistant to change.
• are more difficult to train.
(3) Interactions with older workers:
(a) Supportive interactions: Older workers
• are supportive.
• are helpful.
• give useful advice.
(b) Uncomfortable interactions: Older workers
• act superior to me.
• talk as if they know more than me.
• speak as if they were better than me.
(c) Avoidance: Older workers
• I remain silent if my opinion conflicted with older workers.
• I restrain myself from arguing with older workers.
• I hold back my opinions.
(4) Knowledge of older adults:
• The majority of old people are senile (i.e. defective memory, disoriented or demented).
• All five senses tend to decline in old age.
• Most old people have no interest in, or capacity for, sexual relations.
• Lung vital capacity tends to decline in old age.
• The majority of old people feel miserable most of the time.
• Physical strength tends to decline in old age.
• At least one-tenth of the aged are living in long-stay institutions (i.e. nursing homes, mental hospitals, homes for the aged, etc.).
• Aged drivers have fewer accidents per driver than drivers under the age of 65 years.
• Most older workers cannot work as effectively as younger workers.
• About 80 per cent of the aged are healthy enough to carry out their normal activities.
• Most old people are set in their ways and unable to change.
• Old people usually take longer to learn something new.
• It is almost impossible for most old people to learn something new.
• The reaction time of most old people tends to be slower than that of younger people.
• In general, most old people are pretty much alike.
• The majority of old people report that they are seldom bored.
• The majority of old people are socially isolated and lonely.
• Older workers have fewer accidents than younger workers.
• Over 30 per cent of the Japan population are now aged 65 years or above.
• Most medical practitioners tend to give low priority to the aged.
• The majority of older people have incomes below the poverty level (as defined by the Federal Government).
• The majority of old people are working or would like to have some kind of work to do (including housework and volunteer work).
• Older people tend to become more religious as they age.
• The majority of old people report that they are seldom irritated or angry.
(5) Fears of getting older:
• I fear that when I become an old worker, I will not be able to have casual conversations with my colleagues.
• The older I get, the more I worry about my diminishing productivity.
• I get nervous when I think about someone else making decisions for my job.
• I worry that my colleagues will ignore me when I become an older worker.
• I am afraid that there will be no meaning in my work when I am an older worker.
