Race and Gender-Based Perceptions of Older Adults: Will the Youth Lead the Way?

Abstract

Background:

Older individuals encounter the greatest racial/gender biases. It is unknown whether younger generations, who often lead culture shifts, have racial and gender biases against older populations.

Methods:

Using Amazon’s Mechanical Turk’s crowdsourcing, we identified how an individual’s race and gender are associated with perceptions of individuals aged mid 60s. Participants were asked to rate photograph appearances on Likert Scale (1-10). Interactions between participant and photograph race and gender were assessed with mixed effects models. Delta represents rating differences (positive value higher rating for Whites or women, negative value higher rating for African-Americans or men).

Results:

Among 1563 participants (mean 35years± 12) both non-Hispanic White (WP) and all Other race/ethnicity (OP) participants perceived African-American photos as more trustworthy [Delta WP −0.60(95%CI-0.83,−0.37); Delta OP −0.51(−0.74,−0.28), interaction p=0.06], more attractive [Delta non-Hispanic White participants −0.63(−0.97,−0.29); Delta Other race/ethnicity participants −0.40(−0.74,−0.28), interaction p<0.001], healthier [Delta WP −0.31(−0.53, −0.08); Delta OP −0.24(−0.45, −0.03), interaction p=1.00] and less threatening than White photos [Delta WP 0.79(0.36,1.22); Delta OP 0.60(0.17,1.03), interaction p<0.001]. Compared to OP, WP perceived African-American photos more favorably for intelligence (interaction p<0.001). Both genders perceived photos of women as more trustworthy [Delta Women(WmP) 0.50(0.27,0.73); Delta Men Participants(MnP) 0.31(0.08,0.54); interaction p<0.001] and men as more threatening [Delta WmP −0.84(−1.27, −0.41), Delta MnP −0.77(−1.20, −0.34), interaction p=0.93]. Compared to MnP, WmP perceived photos of women as happier and more attractive than men (interaction p<0.001). Compared to WmP, MnP perceived men as healthier than women (interaction p<0.001).

Conclusions:

Among a young generation, older African-Americans were perceived more favorably than Whites. Gender perceptions followed gender norms. This suggests a decline in implicit bias against older minorities, but gender biases persist. Future work should investigate whether similar patterns are observed in healthcare.

INTRODUCTION

Racial and gender bias contribute to healthcare disparities [1] [2] [3]. Bias negatively influences healthcare professionals’ clinical decision-making [1] [4]. Often racial minorities and women face the greatest healthcare biases compared to Whites and men [5]. The combination of race and gender bias compounds among older patient populations who also deal with ageism [6]. The older population may be less inclined to seek out healthcare due to physician implicit age bias [7].

Historically, younger generations have led changes in culture that promote diversity [8]. With increased awareness of racial and gender bias, younger generations have rejected marginalization and promoted equality through various movements (i.e. Black Lives Matter, Women’s March, campus anti-sexual assault) that confront racial and gender disparities [9]. The younger population’s perspective will shape the future and determine the projection of current bias. However, the younger population’s perception of diverse racial older individuals is unknown.

It is important to examine how younger populations view older populations since the global population is aging and diversifying rapidly [7]. Younger health care workers will soon be responsible for the care of older populations [10]. A previous systematic review analyzed the extent to which implicit racial/ ethnic bias existed among healthcare professionals and how it influenced healthcare outcomes. The results of the review suggested that implicit bias against Blacks, Hispanics/ Latino and dark-skinned individuals is present among most healthcare providers with a positive attitude towards Whites and negative attitudes towards people of color [11]. Implicit bias in healthcare can be manifested through poor communications and ratings of care by African American patients [12] as well as worsened clinical outcomes that contribute to healthcare disparities [13].

Further studies have shown that healthcare professionals exhibit the same levels of implicit bias as the wider population [14]. Therefore, we examined race and gender bias of older African-American and White individuals using an international crowdsourcing marketplace composed of a young population, Amazon’s Mechanical Turk (MTurk). We hypothesized that participants’ race and gender would have no significant influence on perception of older African-American and White individuals with and without a disease, heart failure.

METHODS

Photo Development

Although there are other face databases, such as the Chicago Face Database and AgeGuess, many of these databases had a limited number of photos of Non-White older individuals. Therefore, we created our own older diverse photo database. The study team took photos of 132 individuals in a metropolitan city in Arizona. Inclusion criteria included being 60+ years of age and African-American or White race. Similar to other face databases, we provided matching apparel. Models wore gray T-shirts, removed glasses, jewelry, hair accessories, and hats. Photos were taken against White cloth background with Nikon DX camera using portrait capture. Models were directed to provide neutral facial expressions resembling a passport photo. Adobe Photoshop was used to correct for differences in photo brightness and size. However, we did not resize core facial features to make all facial features equivalent like the Chicago Face Database [15]. Written consent was obtained for study participation as a model.

Photo Selection

The study team individually narrowed the 132 photos into smaller categories of similar appearance in: age of mid 60s, facial expression, hairstyle, and facial features. Then the team convened to review the photos as a group, narrowing photographs to a total of 16 photos varying by race and gender. Four photos were chosen for each category: African-American men, African-American women, White men, and White women (Figure 1). Each photo was duplicated with a description stating that the patient has heart failure and another version that had no statements on medical health for a total of 32 photos. The number of photos used in the study was reduced down to 32 in order to ensure that the photos were as similar in appearance, i.e. facial features/expression, as possible. By selecting comparison photos as illustrated in Figure 1 that have similar facial structure, hair, age, and facial expression, we were able to understand the perceptions based upon race and gender, which was the purpose of this study. Including all photos would have diluted any potential effect and resulted in findings that could be biased by hair, facial expression etc.

Figure 1. Survey Photographs.

Photographs of older African-American and White men and women.

Participant Survey

Participants were randomized to 16 of the 32 photos using Qualtrics survey software. Similar to the Chicago Face Database study, each photo included a series of questions where the participant estimated the individual’s age, gender (male or female), and race (White, African-American, American Indian or Alaska Native, Asian, Native Hawaiian or other Pacific Islander), and rated the individual using Likert scale (1-10, strongly disagree to strongly agree) as healthy, attractive, intelligent, trustworthy, and facial expression (sad, neutral, happy, and threatening). Each survey ended with participant demographic questions (age, gender, race, and ethnicity). Survey questions were tested among laypersons and healthcare professionals for comprehension prior to use in this study. Unlike other studies, our study specifically examined only older individuals. These attributes were chosen because previous studies have shown that physical attributes contribute to implicit bias by healthcare providers [3].

Participant Selection

Surveys were placed in Qualtrics survey database and administered through Amazon’s crowdsourcing marketplace Mechanical Turk (MTurk). MTurk is generally composed of a young, college educated White population [16]. By using Amazon’s Mechanical Turk marketplace, we could obtain a broader range of opinions and perspectives by allowing international participation. Inclusion criteria included adults 18 years of age or older with intent of reaching at least 50% U.S. participants. Exclusion criteria included refusal to consent for participation (n=7), inattention to survey [gender reported incorrectly by participant (n=87), completion time less than the 10^th percentile or greater than 90^th percentile of the median time (n=400), age of the photo reported as less than 20 years or greater than 120 years (n=188)], missing participant demographics (n=36). The first survey was available to international and U.S. participants. It was available for 28 days and closed when it reached 1000 responses. The second survey was only available to U.S. based MTurk workers in order to obtain enough U.S. responses. At the time of the study, MTurk did not have a means to limit the proportion of U.S. and international participants in a single survey. Participants received $0.07 for completion of the survey. This study was approved by the University of Arizona Institutional Review Board.

Statistical Analysis

The comparison of non-Hispanic Whites to all other races/ethnicities and men to women scores was performed using mixed models that included random effects for both participants and photos to account for correlations among responses by the same participant and correlation across responses to the same photo. Race and ethnic groups other than non-Hispanic White were grouped together since the majority of participants were White, and no other group was large enough to make statistically powered comparisons in the model. Most characteristics were categorical and modeled as indicator variables with a reference category. Continuous predictors were centered at their means. The model included age, photo race and sex, heart failure (presence/absence), participant race/ethnicity, participant age and sex, location (US versus international), and interactions between photo race and participant race as well as interactions between photo sex and participant sex (Supplementary Table 1). Estimated marginal mean differences (Delta) with standard errors were computed for White and African-American photos or women and men photos (positive value higher for Whites or women, negative value higher for African-Americans or men). These estimations may vary for different levels of other factors in the model. However, differences between Deltas that define interaction effects in figures do not depend on the levels of the other factors in the models. All statistical analyses were performed using R Version 1.0.136., and all tests of statistical significance were performed at an alpha level of 0.05.

RESULTS

Participant Demographics

Among the 1563 eligible participants, 72% were from the US and 41% were non-Hispanic White women (Table 1). A majority of the participants were non-Hispanic White (63%) and Other race/ethnicity participants included Non-Hispanic American Indian (1%), Native Hawaiian or other Pacific Islander (<1%), Non-Hispanic Asian (13%) and Non-Hispanic African American (8%). The average age of the women participants was 36.3 years old and men participants was 34.7 years old.

Table 1.

Participant Demographics

	Women N=956 (%)	Men N=607 (%)
Age, mean ± SD	36.3 ± 12.0	34.7 ± 11.6
Non-Hispanic White (%)	635 (66.4)	343 (56.5)
Other Race/Ethnicity
Non-Hispanic African American (%)	100 (10.5)	32 (5.3)
Non-Hispanic American Indian or Alaskan Native (%)	6 (0.6)	17 (2.8)
Non-Hispanic Asian (%)	94 (9.8)	102 (16.8)
Native Hawaiian and Other Pacific Islander (%)	4 (0.4)	2 (0.3)
Hispanic (%)	117 (12.2)	111 (18.3)
International (%)	239 (25.0)	206 (33.9)

SD indicates standard deviation.

Summary of Unadjusted Results by Race

All Other race/ethnicity participants rated African-American photos as more attractive, trustworthy, and neutral appearing when compared to the White photos (Table 2). Non-Hispanic White participants rated African-American photos as more attractive, intelligent and trustworthy when compared to White photos. All participants rated White photos as more threatening when compared to African-American photos. There were no statistically significant differences for other independent ratings of White and African-African photos.

Table 2.

Other Race/Ethnicity and White Participant Ratings Of Photos By Race & Women and Men Participant Ratings Of Photos By Gender

All Other Race/Ethnicity Participant Ratings				Non-Hispanic White Participant Ratings
	AA Photos Mean ± SD	White Photos Mean ± SD	P value	AA Photos Mean ± SD	White Photos Mean ± SD	P value
Age, years	64 ± 6	65 ± 6	0.12	64 ± 5	65 ± 6	0.16
Healthy	5.80 ± 2	5.51 ± 2	0.26	5.22 ± 3	4.91 ± 2	0.47
Attractive	5.71 ± 2	5.29 ± 2	0.01	5.58 ± 2	4.94 ± 2	<0.001
Intelligent	6.37 ± 2	6.19 ± 2	0.08	6.46 ± 2	6.08 ± 2	<0.001
Trustworthy	6.28 ± 2	5.78 ± 2	<0.001	6.41 ± 2	5.82 ± 2	<0.001
Neutral	6.24 ± 2	6.01 ± 2	0.04	6.56 ± 2	6.28 ± 2	0.07
Happy	5.03 ± 2	4.58 ± 2	0.08	4.25 ± 2	3.68 ± 2	0.07
Sad	4.32 ± 2	4.61 ± 2	0.06	3.83 ± 2	4.1 ± 2	0.18
Threatening	4 ± 3	4.55 ± 3	0.01	2.97 ± 2	3.75 ± 3	0.01
Women Participant Ratings				Men Participant Ratings
	Women Photos Mean ± SD	Men Photos Mean ± SD	P value	Women Photos Mean ± SD	Men Photos Mean ± SD	P value
Age, years	65 ± 5	63 ± 6	<0.001	65 ± 5	63 ± 6	<0.001
Healthy	5.18 ± 2	5.22 ± 2	0.89	5.13 ± 2	5.44 ± 2	0.36
Attractive	5.68 ± 2	5.31 ± 2	0.11	5.02 ± 2	5.12 ± 2	0.62
Intelligent	6.49 ± 2	6.32 ± 2	0.19	6.05 ± 2	6.08 ± 2	0.04
Trustworthy	6.43 ± 2	5.94 ± 2	<0.001	6.09 ± 2	5.79 ± 2	0.04
Neutral	6.42 ± 2	6.53 ± 2	0.61	6.09 ± 2	6.13 ± 2	0.85
Happy	4.46 ± 2	3.67 ± 2	0.01	4.63 ± 2	4.24 ± 2	0.08
Sad	3.86 ± 2	3.96 ± 2	0.61	4.45 ± 2	4.41 ± 2	0.78
Threatening	2.97 ± 2	3.79 ± 3	<0.001	3.61 ± 2	4.39 ± 3	<0.001

AA indicates African-American; SD, standard deviation.

Primary Outcomes by Race

In the fully adjusted analyses, both non-Hispanic White and Other race/ethnicity participants perceived African-American photos as more trustworthy [Delta non-Hispanic White participants −0.60 (95%CI-0.83, −0.37); Delta Other race/ethnicity participants −0.51 (−0.74, −0.28); interaction between photo race and participant race p=0.06], more attractive [Delta non-Hispanic White participants −0.63 (−0.97,−0.29); Delta Other race/ethnicity participants −0.40 (−0.74,−0.05), interaction p<0.001], modestly more healthy [Delta non-Hispanic White participants −0.31(−0.53, −0.08); Delta Other race/ethnicity participants −0.24 (−0.45, −0.03), interaction p=1.00], and less threatening than White photos [Delta non-Hispanic White participants 0.79 (0.36, 1.22); Delta Other race/ethnicity participants 0.60 (0.17, 1.03), interaction p<0.001; Figure 2]. Non-Hispanic White participants perceived African-American photos more favorably for intelligence compared to Other race/ethnicity participants who perceived no racial differences (interaction p<0.001). Ratings for happiness, sadness, and neutral did not vary significantly between Non-Hispanic White and Other race/ethnicity participants.

Figure 2. Differences between Ratings of White and African-American photographs according to Participant Race.

Blue line indicates White participants; green line, Other race/ethnicity participants; horizontal line, 95% confidence interval; p value indicate interactions between photo race and participant race.

Summary of Unadjusted Results by Gender

Women participants rated photos of women as more trustworthy and happier when compared to photos of men (Table 2). Men participants rated photos of men as modestly more intelligent when compared to photos of women. Men participants rated photos of women as more trustworthy. Both men and women participants rated photos of men as more threatening. There were no statistically significant differences for other independent ratings of women and men photos.

Primary Outcomes by Gender

In the fully adjusted analyses, both genders perceived photos of women as more trustworthy [Delta women participants 0.50 (0.27,0.73); Delta men Participants 0.31 (0.08,0.54); interaction between photo gender and participant gender p<0.001] and men as more threatening [Delta women participants −0.84 (−1.27, −0.41), Delta men participants −0.77 (−1.20, −0.34), interaction p=0.93; Figure 3]. Compared to men participants, women participants perceived women as happier and more attractive than men (interaction p<0.001). Compared to women participants, men participants perceived men as healthier than women (interaction p<0.001). Although women and men participants perceived no significant differences between women and men photos for other factors, significant interactions by participant gender were observed with scores favoring the participants’ gender for intelligence and sadness. Heart failure has no statistically significant effect on the perceptions of photos with the exception of health which was rated 2 points lower for photos with stated heart failure (Supplementary Table).

Figure 3. Differences between Ratings of Men and Women photographs according to Participant Gender.

Blue line indicates men participants; red line, women participants; horizontal line, 95% confidence interval; p value indicates interactions between photo gender and participant gender.

DISCUSSION

To our knowledge, this is the first study to examine racial and gender perceptions of photographs of older African-Americans and Whites in their seventh decade of life using an online interface. Although other studies have investigated racial bias and found similar results [17], the utilization of an online international survey market created a unique opportunity to assess perception bias amongst various populations. Using an international survey market of a young population, older African-Americans were perceived more favorably than Whites; however, perceptions based on gender followed many gender norms. In fully adjusted analyses, White participants and Other race/ethnicity participants viewed African-American photos as more trustworthy, attractive, modestly healthier, and less threatening than White photos. Only White participants perceived African-American photos as more intelligent. Other races/ethnicities perceived no differences between races for all other ratings.

Both men and women participants perceived women photos as more trustworthy and men photos as more threatening. Participants favored their own gender for other factors with the exception of the neutral facial expression which suggest gender bias persists among the young population. Participants’ bias towards their own gender may be due to decreased awareness of their own biases [18]. Recent studies have shown gender bias continues to persist despite increasing female representation in the work force due to lack of recognition of workplace discrimination [19]. Gender bias may be reduced by promoting interventions that increase awareness and gender diversification [18] [20].

Multiple studies have analyzed the evolution of racial attitudes using surveys. The “Obama effect” study utilized the American National Election Studies (ANES), an academically run national survey, to evaluate White voters’ perception of African-Americans in regards to work ethic and intelligence. It demonstated that White voters’ perception of African-Americans was becoming more positive over time [21] but Schmidt & Nosek (2010) showed there was very little evidence of systematic change in explicit and implicit racial attitudes during President Obama campaign and early presidency [22]. A recent Pew Survey revealed that majority of races and ethnicities in the U.S. believe that race relations are getting worse, and the majority of Whites believe that racial discrimination is not an issue in the U.S. Non-Whites believed that racial discrimination remains an issue in the U.S [23].

Conversely, outside of the U.S., surveys have revealed that the young British population has been more accepting of racial diversity when compared to the older population [24]. Some survey studies have shown a shift in racial attitudes that favor minority groups, particularly among younger populations. Perceptions of race and gender are driven by longterm exposure to social messages [25].

Our findings suggest that a third subdivision of stereotypes within these groups is age. Ghavami and Peplau (2013) found different stereotypes within various racial and ethnic groups by gender [25]. We should also consider that the interactions of White participants with older African Americans in person or through media portrayals may underpin our findings. African Americans in their 60s lived at least part of their lives in the U.S. under Jim Crow laws which were not abolished until 1968. They were of necessity socialized to interact with Whites differently than current generations of African Americans, sometimes exhibiting deferential behaviors which might reinforce positive assessments of older African Americans. The importance of age in these positive assessments of African Americans by White participants is underscored by a series of smaller studies which also recruited participants from Amazon’s MTurk who were similar in age to those in our study and showed them facial pictures of young White and African American men [26]. They found considerable racial bias with consistent assumptions that young African American men were significantly larger, more capable of harm, and more deserving of police force in an altercation without a weapon than non- Hispanic White men of identical weight and height. This underscores the need to transform the way African Americans are perceived throughout the lifecourse.

Our face photo database allowed us to focus on older populations and how they are perceived in society. An online makertplace permited access to a young diverse international population. Similar to previous studies, our findings revealed a reduction in racial bias amongst a young population but no major changes in gender bias [27]. This study’s use of photographs to evaluate perception bias creates an opportunity for further investigation of visual perception and the impact on clincal decision making.

In the times of a growing aging population with significant chronic diseases and a younger workforce, the next steps of investigation should explore whether bias of the older patient population exists amongst younger healthcare professionals. Additional studies are needed to determine whether younger healthcare professionals’ biases are consistent with the results of our study and how it impacts their clinical decision-making.

Limitations

The background of the photos used in the survey varied in shades of gray because of the lighting available, this may have impacted the results. As an online survey, environmental factors such as screen resolution, room lighting, or the presence of other people could not be controlled for and may have affected participant responses. However, one study showed that data obtained in-person versus through social media or MTurk do not vary significantly [26]. Although duplicate responses were not allowed through the MTurk website, we cannot confirm duplicate participation in the U.S only version of the study. Exclusion of automated tasks software was attempted but could not be guaranteed. Another potential limitation of this study is that a majority of MTurk participants self-identified as Non-Hispanic White, which may skew responses towards the White participant category by underrepresenting other racial and ethnic groups.

CONCLUSIONS

This international study investigates the perception of older adults greater than 60 years of age based on gender and race. White participants viewed African-Americans more favorably than non-White participants. While many aspects of race relations have worsened, our results show promise that perceptions based on race of older individuals may be changing, particularly among the younger adult population. However, perceptions based upon gender followed many gender stereotypes. Further investigation is needed to examine how these biases are shaped through healthcare.