The perception of the poor: Capturing stereotype content with different measures

Abstract

Even though Sweden stands out in many ways, with for example a well-developed welfare system, there are some indications that Swedish egalitarianism does not include tolerance for people with low income. The present research concerns the content of the Swedish stereotype of poor people, and examines whether the poor are associated with just as negative traits as they are in other countries. Three different measurement methods were employed to investigate the contents and strength of the stereotype. In Study 1, participants freely associated on the perceived characteristics of Swedish citizens who receive welfare benefits. They also provided ratings of this group on traits related to warmth and competence. Study 2 employed both trait ratings and implicit measures (Single Category-IATs) in order to estimate the strength of the associations between the poor and warmth/competence, respectively. Across the different measures, the poor were associated with low competence and low–medium warmth. This corroborates the findings from previous research related to the Stereotype Content Model in other countries.

Keywords:

• stereotype
• poor
• Stereotype Content Model
• Single-category implicit association test
• Sweden

Introduction

Stereotypes are beliefs about the attributes that characterize a group of people (Ashmore & Del Boca, 1981). The core of the stereotype is its content, i.e. the set of characteristics that are depicted by it. Stereotype content can be studied in a systematic fashion. In fact, the study of stereotype content is a classic topic of social psychological research, dating back to some of the very first studies on stereotyping (e.g. Katz & Braly, 1933), but it was set aside when the field turned its attention to how stereotypes affect judgments and decisions. However, as the content of stereotypes partly determines how people will feel, think, and behave toward those who belong to the stereotyped group (Cuddy et al., 2007), it is essential to be aware of the content of prevailing stereotypes in society. And because the content of stereotypes may differ geographically and shift over time, they need to be examined empirically rather than inferred or assumed.

It is therefore fortunate that stereotype content research has revitalized recently, partially thanks to the arrival of the Stereotype Content Model (SCM; Fiske, Cuddy, & Glick, 2007; Fiske, Cuddy, Glick, & Xu, 2002). In the model, stereotypes are arranged into the two dimensions of warmth and competence. The warmth dimension concerns the human desire to determine others’ intentions: Is this person sincere, friendly, and trustworthy? The competence dimension concerns the human desire to determine others’ capability of pursuing their intensions: Is this person skillful, creative, and confident? The warmth dimension and the competence dimension are generally independent of one another. For example, a social group can be either generally disliked (low competence–low warmth); respected but not liked (high competence–low warmth); or liked but not respected (low competence–high warmth; see e.g. Cuddy, Fiske, & Glick, 2008; Cuddy et al., 2009; Fiske et al., 2002). Groups may also be perceived as “high warm” (HW) and “high competent” (HC), or even as “medium warm,” (MW) and “medium competent” (MC). When a group is perceived as MW/MC, it does not stand out in either the warmth–cold dimension or in the competent–incompetent dimension, as compared to other groups in society, but could rather be described as somewhere between the extremes.

The SCM states that groups which enjoy high status in society are normally seen as having high power and competence (Fiske et al., 2002). Conversely, groups that have low status, such as the poor, are usually seen as having low in power and competence. Furthermore, groups that are perceived to constitute some kind of threat, e.g. for common societal resources, are generally perceived as cold. Accordingly, to the extent that poor people in Sweden are perceived to be a burden on public expenses, it follows from SCM that they should be perceived as cold.

The present research approaches the Swedish stereotype of the poor from different methodological angles, exploring it in terms of a free response format task, explicit trait ratings, and a categorization task designed to capture implicit associations. In what follows, we will describe previous related research. Even though we will relate the stereotype content of the poor in a Swedish context to the same stereotype in other cultures, it should be noted that the aim of our research is not comparative as such, but rather to explicate the contents of the stereotype of the poor in a Swedish context.

Stereotypes of the poor

Studies across a range of countries indicate that the core content of the stereotype of the poor is similar. One widespread negative belief about poor people is that they are irresponsible, not having internalized a culture of savings (Bertrand, Mullainathan, & Shafir, 2004; Mullainathan & Shafir, 2009). Another negative belief about the poor is that they are dishonest. For example, one common notion of welfare recipients is that they spend their money on drinking and drugs (Bullock, 1999; Underlid, 2005). Similar examples are extensive and widespread: The stereotype of the poor includes unfavorable individual properties, such as having low intelligence, being lazy, and not being interested in self-improvement (see e.g. Bullock, 1995; Cozzarelli, Wilkinson, & Tagler, 2001; Kreidl, 2000; Lott, 2002; Woods, Kurtz-Costes, & Rowley, 2005). Poor individuals are often held to have themselves to blame (Bolitho, Carr, & Fletcher, 2007; Kluegel & Smith, 1986). As a consequence, welfare recipients sense that society sees them as criminals, “milking the system.” (Underlid, 2005)

Although the content of the stereotype of poor people as outlined in earlier research primarily originates from North American studies (as in Bullock, 1999; Cozzarelli et al., 2001; Gilmore & Harris, 2008; Henry, Reyna, & Weiner, 2004; Kluegel & Smith, 1986; Woods et al., 2005), researchers have identified similar stereotype content across post-communist countries, such as Hungary and Russia, where the acceptance of inequality and poverty seems to be lower than in the United States (Kreidl, 2000).

The societal perceptions of the poor, as identified in earlier research, can clearly be arranged according to the two dimensions of the SCM: Traits belonging to the warmth dimension include dishonest and predatory, whereas traits belonging to the competence dimension include irresponsible, unintelligent, and lazy.¹ As all of these traits have negative valence, it follows that the attitude towards the poor is negative. The magnitude of the negative attitude toward poor people appears to depend on whether the stereotype implies that they are responsible or not for their situation (see e.g. Lepianka, van Oorschot, & Gelissen, 2009 for a discussion). People often believe that there are several different and complex determinants of economic disadvantage (Kluegel & Smith, 1986). Many nevertheless tend to think that internal causes, such as being irresponsible or lacking willpower, are more important than external causes, such as having a low wage because of injustice in society; Bolitho et al., 2007; Weiner, Osborne, & Rudolph, 2011).

Attributions for poverty

The current study fills a gap, in that it analyzes the content of the stereotype of poor individuals in a Swedish context. The Scandinavian countries (Sweden, Norway, Denmark, Finland) stand out in many ways, with for example a well-developed and large-scale welfare system (Blanc-Noel, 2013), with relatively generous welfare benefits, contributing to keeping the differences in Socio-Economic Status (SES) low (Barth, Moene, & Willumsen, 2014; Fochesato & Bowles, 2015). In addition to this, Sweden is among the most tolerant countries in the world (World Values Survey, 2015). Swedes put high value on both cooperation and caring for the weak, and see poverty as the most pressing global issue (Lindberg & Svensson, 2011).

Previous research about the poor in Sweden has been conducted within the fields of social work or sociology, and has concerned social assistance in relation to poverty, or individual risk factors for poverty (see e.g. Gustafsson, 2013; Kangas & Palme, 2000; Marttila, Johansson, Whitehead, & Burstrom, 2012). It has not concerned the stereotype of the poor. The most closely related research from Sweden has instead analyzed the attribution for poverty (Kallio & Niemelä, 2014; Lepianka, Gelissen, & van Oorschot, 2010; Shirazi & Biel, 2005; van Oorschot & Halman, 2000). According to these studies, Swedes seem to perceive the causes of poverty as relatively external, and are not blaming the poor for being poor to the same extent as in other countries. In contrast to these findings, there are some indications that the Swedish strivings for equality do not necessarily include high tolerance for people with low income. For example, attitudes emphasizing the personal responsibility of individuals who have slipped through the social safety net have been reported to be strong in Sweden (Fitzpatrick & Stephens, 2014). This is interesting, not least from a political perspective. Proposals to benefit those in poverty are likely to meet resistance if the tax-payers and/or policy-makers do not consider poor people worthy of aid.

Supporting the picture that the poor are derogated in egalitarian welfare states too, relevant research from Norway, one of Sweden’s cultural neighbors, indicates that the stereotype of poor people is similar to the stereotype identified in North America (Bye, Herrebrøden, Hjetland, Røyset, & Westby, 2014; Underlid, 2005). However, when comparing how poor and welfare recipients are depicted in the news-media, they were found to be more negatively depicted in Britain as compared to Sweden and Denmark (Larsen & Dejgaard, 2013).

To summarize, earlier findings indicate that the content of the stereotype of poor people may be similar across countries, but that there are differences especially concerning attribution for poverty. Moreover, it has been suggested that comparison across countries is problematic, due to cultural differences (Nasser, Abouchedid, & Khashan, 2002). Shirazi and Biel’s (2005) comparison of 14 different countries regarding attributions for poverty further underscores the need for country-specific research reflecting social and political contexts. As no research has explicated the content of the stereotype of poor people in the Swedish context, there is reason to systematically investigate the content of the stereotype in Sweden specifically. The results may interest researchers and decision-makers alike. The perception of poor individuals is of great importance when it comes to societal decisions concerning, for example, social benefits. It is thus useful to identify the stereotype content of the poor, and create awareness of its nature. To the extent that the stereotype of the poor in a Swedish context is similar to that of other countries, it can be expected to be characterized by low competence and low warmth, which is also what is suggested by the SCM (Fiske et al., 2007).

Explicit and implicit measures of stereotypes

Most research on stereotype content relies on explicit measures, e.g. surveys and questionnaires. Even if these methods have been fruitful, their reliability and validity can be questioned because they, for example, are vulnerable to social desirability concerns (Fazio & Olson, 2003). In research on the SCM, the main strategy for counteracting such influence has been to ask respondents to make their ratings according to how the target group is viewed in society, instead of asking about their own personal views, i.e. asking about the societal stereotype rather than the personal stereotype. The method appears to have been rather successful, but does not guarantee that social desirability concerns are handled. Therefore, implicit measures of warmth and competence have been developed (Carlsson & Björklund, 2010; Ebert, Steffens, & Kroth, 2014; Rohmer & Louvet, 2012) and found to be useful complements for analyzing and validating stereotypes. In the current study, both explicit and implicit measures will be used, to triangulate the measurement of the stereotype content and thereby increase the validity of our results.

Aim of study

The primary aim of the present research is to identify the content of the stereotype of the poor in a Swedish context. We will also compare and discuss the different measurement methods employed in the study, as they represent the methods that have been used to analyze stereotype content in earlier research (although seldom in the same study).

Study 1 – Free associations

In the first study, we wanted to outline the content of the stereotype of the poor in a Swedish context. Participants performed a free association task, and also made explicit ratings of warmth and competence according to the SCM (Fiske et al., 2002). In the free association task, we were specifically interested in whether participants would spontaneously report characteristics related to warmth and competence.

Methods

Participants

One hundred and two individuals (self-identified as 68 women, 33 men, 1 non-binary) between 20 and 81 years (M_age = 38, SD = 12) participated in the first study. They were recruited by notices on Internet forums, where they were asked to follow a link to the questionnaire. The questionnaire was open for all. One hundred and seven individuals opened the questionnaire; accordingly, five of them did not finish it. For confidentiality reasons, the system did not save any demographic data when participation was cancelled. However, because the questionnaire was in Swedish, and the participants responded in Swedish, we assume that the participants were Swedish.

Questionnaire and procedure

Participants completed a questionnaire online consisting of two tasks. Because the concept “poor” may evoke different connotations for different persons (see e.g. Lepianka et al., 2009), we specified the concept of “poor” as “Swedish citizens receiving welfare benefits.”

In the first task, the participants were asked to associate on the societal stereotype of “Swedish citizens receiving welfare benefits,” and write down as many words they could think of in an open-ended question. In the other task, they were asked to rate “Swedish citizens receiving welfare benefits” on warmth and competence (cf. Fiske et al., 2002) using 5-points scales. The warmth scale consisted of 4 items (tolerant, warm, good natured, sincere), and the competence scale consisted of 5 items (competent, confident, independent, competitive, intelligent). In both task 1 and task 2, the instructions emphasized that the participants should respond according to the societal stereotype rather than according to their own.

Results

Free association on Swedish citizens receiving welfare benefits

In total, the participants produced 578 words (M = 5.7 words / participant) when associating on “Swedish citizens receiving welfare benefits.” We (the three authors) sorted the 578 words into categories, based on theme of content (cf. thematic analysis; Povee & Roberts, 2015): The themes were continually reworked by us to ensure that each category had sufficient content. We first sorted the words, where those relating to the same specific theme were added together. For example, words categorized as “lazy” consisted of that specific word, and synonyms such as “listless”; the category “uneducated” consisted of that specific word, and synonyms such as “ill-educated”; the category “drug addict” consisted of that specific word and other similar words, such as “alcoholic” and “junkie.” For validating these themes, we applied a within-methods triangulation, where we met to discuss the categorization, and to confirm agreement of themes developed (cf. Jonsen & Jehn, 2009). We aimed at being as open-minded as possible during this process, applying an inductive approach with no preunderstanding of what themes that “should” occur.

After the sorting process, 20 categories had ≧10 words. Figure 1 shows the frequency distribution for these categories. Among them, 9 categories could be related to characteristics of the external situation indicating for example illness (physical or mental) and family situation (single parent, many kids), whereas the remaining 11 categories could be related to personal attributes. As expected from earlier research on stereotype content, all 11 could be related to the SCM dimensions warmth (dishonest, predatory) or competence (lazy, uneducated, addiction, and smoker as indications of irresponsible, unintelligent, passive, incapable, work-shy, uneconomical). When relating the categories to the dimensions of warmth and competence (i.e. to SCM; cf. Fiske et al., 2002), their content imply low values. In other words, the free association task indicates that the stereotype content of the poor constitutes “low competence, low warmth” LW/LC.

Figure 1. The frequency distribution of the 20 categories with ≧ 10 words from the free association on the concept “Swedish citizens receiving welfare benefits.” Categories related to warmth/competence have solid black bars.

Ratings of warmth and competence

There were no significant differences between women and men in the ratings of warmth t(99) = .26, p = .80, d = .0528 or competence t(99) = .82, p = .90, d = .0440. Hence, all data were collapsed across gender. When calculating the mean value of the warmth scale (α = .73) and the competence scale (α = .79), it was shown that our target group “Swedish citizens receiving welfare benefits” received relatively low ratings on warmth and competence (M_warmth = 2.54, SD = .62; M_competence = 1.75, SD = .59) according to the interpretation of the results made by Fiske et al. (2002). Ratings were significantly lower on competence than warmth, t(101) = 13.17, p < .001, d = 1.242. The ratings of warmth and competence were significantly correlated, r(100) = .55, p =< .001.

Discussion

When Swedish participants freely associated on how Swedish citizens receiving welfare benefits are perceived in society, they generated many of the personal traits familiar from earlier research on stereotypes toward poor people, such as lazy, uneducated, unintelligent, and being addicted to alcohol or drugs. All but two of these personal traits (9 of 11) were related to the competence dimension of the SCM (Fiske et al., 2002), and the remaining two were related to warmth. The results of the free association task suggest that the competence dimension in the stereotype of individuals receiving welfare benefits is more important than the warmth dimension. Hence, low competence appears to be a signifying characteristic of the stereotype of the poor. Also in task 2, where participants were asked to provide ratings on scales reflecting warmth and competence with regard to how Swedish citizens receiving welfare benefits are perceived in society, the levels were relatively low (i.e. low competence, low–medium warmth [LC-L/MW]). The competence ratings were significantly lower than the warmth ratings. It is worth noting that the stereotype content of the poor in a Swedish context, identified by free association and explicit ratings of warmth and competence, appears quite similar to the stereotype content of the poor in other national contexts.

Study 2 – Single Category Association Test

In the second study, we wanted to further analyze the warmth dimension and the competence dimension in the stereotype of citizens receiving welfare benefits in a Swedish context. As opposed to Study 1, where only self-report measures were used, Study 2 employed both self-report measures and implicit measures. Implicit measures are often designed to capture the strength of association between categories. This is highly relevant for stereotype research, as stereotypes concern the associations between a social group and specific attributes (such as warmth and competence).

Methods

Questionnaire and procedure

Study 2 was divided into two parts (2a and 2b), which had two tasks each. For both parts, the participants completed the first task on a laptop provided by the experimenter, and the second task on a paper-and-pencil questionnaire. The first task was a Single Category Implicit Association Test (SC-IAT; Karpinski & Steinman, 2006), a sorting task that was developed from the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998). The SC-IAT has an important advantage to the standard IAT. It allows for testing the associations between a target concept (such as “the poor”) and attribute concepts without having a complementary target category (such as “the rich”). This is very useful for the present purposes, as we were interested in the content of the implicit stereotype of the poor, regardless of how it is related to the stereotype of e.g. the rich (i.e. in the absolute strengths of associations rather than the relative). Using two separate SC-IATs allowed us to estimate both the automatic association between “the poor” and “warmth,” and the automatic association between “the poor” and “competence.” A result where participants primarily automatically associate the poor more strongly with incompetence would corroborate the results from Study 1.

In Study 2a and 2b alike, the target category “the poor” contained (Swedish synonyms of) the words poor, impoverished, broke, and indigent. Study 2a concerned the warmth dimension. The attribute category “warmth” contained the words tolerant, warm, nice, and sincere, whereas the attribute category “coldness” contained the words frosty, impolite, dishonest, and fraudulent. Study 2b concerned the competence dimension. The attribute category “competence” contained the words confident, effective, competent, and intelligent, whereas the attribute category “incompetence” contained the words unable, blundering, unproductive, and dull.

Both SC-IATs consisted of two parts, where the participants were to sort the words in the category “poor” together with either “warmth/competence” (version 1; positive valence) or “coldness/incompetence” (version 2; negative valence). They were randomly assigned to which version they began with (version 1 or version 2). Both tasks started with a short practice round of 20 trials, each task consisted of 60 trials. The keys “d” and “k” were mapped to categories represented by labels that appeared at the top left and right side of the screen, respectively. Within each version (1 or 2), participants were randomly assigned to one of two key combinations: In version a, the association “poor-warmth” or “poor-competence” could for example be linked to either the key “d” or the key “k.” The measure of interest is the difference between version 1 and 2, and is based on the idea that when the response mapping fits the respondent’s mental representation of the association between the categories (e.g. “poor people are incompetent”), the response will be made more quickly and accurately than when then the response mapping misfits. If there is no association between the target category and the attribute category in the respondent’s mental representation, there will be no facilitation/impeding in any of the response mappings, and accordingly no difference between version 1 and 2.

Participants were instructed to perform the task as fast as they could. Using “d” and “k,” they rapidly classified stimuli appearing in the middle of the screen to the correct category. When sorting a word incorrectly, they received feedback in the form of a red X appearing on the screen. For each trial, the response time was recorded from the time the stimulus was presented until the participant made the response. If no response had been given within 1200 ms the red X appeared and the word was recorded as incorrectly sorted.

When having finished the SC-IAT, participants performed the second task. This was the same as the second task in Study 1, where they were asked to rate “Swedish citizens receiving welfare benefits” on warmth and competence (Fiske et al., 2002) using 5-points scales. The instructions emphasized that the participants should respond on the basis of how the poor are viewed by Swedish society, not by themselves.

Participants, Study 2a

One hundred undergraduate students (55 women, 44 men, 1 being non-binary) between 18 and 30 years (M_age = 22, SD = 2.5) participated in Study 2a. They were recruited by two experimenters at a Swedish university campus, and received lottery tickets for participating.

Participants, Study 2b

Eighty-three undergraduate students (49 women, 33 men) between 20 and 46 years (M_age = 33, SD = 14) participated in Study 2b. They were recruited by an experimenter at a Swedish university campus, and received lottery tickets for participating.

Results, Study 2a

Single Category Association Test

Fifty-two participants were randomly assigned to begin with version 1, sorting “poor” with “warmth,” and the remaining 48 participants to begin with version 2, sorting “poor” with “cold.” There were no significant effects of version order, and the data were collapsed across versions. There were no significant differences between women and men, and the data were collapsed across gender. Following the procedure described in Karpinski and Steinman (2006), a reliability analysis on the SC-IAT measures from Study 2a revealed a reasonable level of internal consistency (α_correct = .60; α_RT = .68).

Figure 2 shows the results from the SC-IAT. As can be seen in the figure, the participants had a larger proportion correct responses when associating “poor” with “cold” compared to when associating “poor” with “warm,” and made these responses faster. Consequently, lower proportions of correct responses were obtained for conditions producing longer RTs, and these conditions were present when the response mapping associated “poor” with “warmth.” The differences between the poor-warm and poor-cold associations were significant, both concerning the proportion of correct responses, t(99) = 8.501, p < .001, and concerning response time, t(99) = 11.84, p < .001. Using the pooled standard deviation (for poor-warm and poor-cold associations) as the effect size unit, the effect size for the implicit association was d = 1.01, suggesting a large effect.

Figure 2. Results from the Single Category Association Test. Primary axis shows correct responses in % (black bars), secondary axis shows response time in ms (patterned bars). X-axis indicates if the category “poor” was associated with “warm” or “cold.”

Ratings of warmth and competence

There were no significant differences between women and men in the ratings of warmth or competence. Hence, all the data were collapsed across gender. When calculating the mean value of the warmth scale (α = .85) and the competence scale (α = .82), it revealed that Swedish citizens receiving welfare benefits received relatively low ratings on both scales (LC-L/MW; M_competence = 2.22, SD = .64; M_warmth = 3.09, SD = .76), with significantly lower ratings on competence than warmth, t(99) = 12.15, p < .001, d = 1.24. The ratings of warmth and competence were significantly correlated, r(98) = .51, p =< .001.

Relation between the implicit and explicit task

There were no significant correlations between the results from the SC-IAT and the explicit warmth ratings (r_RT = −.002; p = .99; r_Correct = −.003; p = .97).

Results, Study 2b

Single Category Association Test

Forty-one participants were randomly assigned to begin with version 1, sorting “poor” with “competence” and the remaining 42 to begin with version 2, sorting “poor” with “incompetence.” There were no significant effects of version order, and the data were collapsed across versions. There were no significant differences between women and men, and the data were collapsed across gender. A reliability analysis on the SC-IAT measures from Study 2b revealed a reasonable level of internal consistency (α_correct = .68; α_RT = .72).

Figure 3 shows the results from the SC-IAT. As can be seen in the figure, the participants had a larger proportion of correct responses when associating “poor” with “incompetence” compared to when associating “poor” with “competence,” and they made these responses faster. Similar to the results in Study 2a, lower proportions of correct responses were obtained for conditions producing longer RTs, and these conditions were present when the response mapping associated “poor” with “competence.” The differences between the poor-competence and poor-incompetence associations were significant, both concerning the proportion of correct responses, t(164) = 5.83, p < .001, and concerning response time, t(164) = 4.06, p < .001.

Figure 3. Results from the Single Category Association Test. Primary axis shows correct responses in % (black bars), secondary axis shows response time in ms (patterned bars). X-axis indicates if the category “poor” was associated with “competence” or “incompetence.”

Using the pooled standard deviation (for poor-competence and poor-incompetence associations) as the effect size unit, the effect size for the implicit association was d = .62, which is smaller than in Study 2a but still relatively substantial.

Ratings of warmth and competence

There were no significant differences between women and men in the ratings of warmth or competence. Hence, all the data were collapsed across gender.

When calculating the mean value of the warmth scale (α = .91) and the competence scale (α = .91), it was revealed that Swedish citizens receiving welfare benefits received relatively low ratings on both scales (LC-L/MW; M_competence = 2.37, SD = .81; M_warmth = 3.06, SD = .80) although with significantly lower ratings on competence than warmth, t(81) = 11.56, p < .001, d = .857. The ratings of warmth and competence were significantly correlated, r(80) = .78, p =< .001.

Relation between the implicit and explicit task

There were no significant correlations between the results from the SC-IAT and the explicit competence ratings (r_RT = .129; p = .25; r_Correct = −.101; p = .37).

Discussion

When participants were asked to rate Swedish citizens receiving welfare benefits on SCM-scales reflecting warmth and competence in the explicit task, the results showed moderately low ratings on both warmth and competence, where the competence ratings were significantly lower than the warmth ratings. The results from the SC-IATs too indicate both low warmth and low competence. As shown in Figures 2 and 3, it is obvious that participants implicitly associated the category “poor” with coldness rather than warmth; as well as with incompetence rather than competence. In sum, the results from the implicit and explicit tasks in Study 2 suggest the stereotype content of the poor to be LC-L/MW.

General discussion

The present research concerned a classic social psychological issue, stereotype content. More precisely, it aimed to identify the content of the stereotype of poor people in a Swedish context, with its unique egalitarian society model. In contrast to previous research, we did this by means of both traditional (explicit) and modern (implicit) measures. Methodologically, it is likely the most systematic study of the contents of the stereotype of poor people to date. The results supported the application of the SCM on the Swedish stereotype of the poor, and fit well with earlier findings regarding this stereotype in other countries (cf. Asbrock, 2010; Bye et al., 2014; Cuddy et al., 2008; Durante et al., 2013): The content of the stereotype of poor people consists of relatively low ratings on both competence and warmth (i.e. LC-LW or LC-L/MW), always with a higher rating on warmth than competence. Implicit associations confirmed the significance of both low warmth and low competence.

Warmth and competence across studies

Free associations to social categories are a valuable source in stereotype content research, as they unobtrusively provide information indicating the availability of certain traits in the stereotype of the relevant group. Interestingly, in the results from the free association task in Study 1, almost all categories describing personal traits could be related to low levels of the competence dimension. Only two categories could be related to “warmth,” also this with negative valence. Hence, the identified semantic content of the stereotype was similar to earlier findings in other countries, where “poor” has been associated with “LC–LW” or “LC-L/MW.” In line with earlier research, spontaneous associations of poor people did not evoke strong associations to the warmth dimension.

The content of the Swedish stereotype of the poor appears similar to findings in other countries, despite its egalitarianism and relatively low difference in SES. To validate this assumption, we have plotted the explicit warmth/competence ratings from Study 1, Study 2a, and Study 2b together with the SCM mapping from other countries in Figure 4. The competence ratings in Figure 4 are within the “low” range, whereas the warmth ratings range from “low” (Hong Kong; Cuddy et al., 2009) to “medium.” (Study 2a and Study 2b) Hence, the content of the poor stereotype varies between LC–LW and LC–MW, and could be summarized as LC-L/MW. One tendency is striking in Figure 4: The competence dimension is constantly lower than the warmth dimension, not only in the present studies, but overall. This indicates that even though the specific ratings of warmth and competence demonstrate a variation between studies, the stereotype contents nevertheless represent medium-to-low values on warmth, with consistently lower values on competence.

Figure 4. Explicit ratings of the poor on the two SCM-dimensions, from nine different studies (a–i).

The results from the SC-IAT in Study 2a and Study 2b showed that implicit associations concerning the poor were significantly more negative than positive, as participants more easily associated “poor” with words describing “low warmth” and “low competence” rather than “high warmth” and “high competence”. The results can be taken to suggest that even though the explicit associations to the poor primarily concern low competence, the implicit associations still show that the stereotype also includes low associations to the warmth dimension.

Methods discussion

The non-existing correlation between the implicit and explicit tasks are in line with those of Akrami and Ekehammar (2005) and Devine (1989), who describe how individuals may hold one explicit, deliberate attitude toward a social group and at the same time automatically activate implicit stereotypes of that group. Here, it is worth noting that implicit and explicit tasks measure different things, distinct but related (as described in Nosek, 2007; Nosek & Smyth, 2007): Explicit tasks measure a more controllable response made with awareness, whereas implicit tasks measure a more automatic response. Our suggestion to future research is that even though explicit and implicit tasks both could be used for identifying a stereotype, researchers should ask themselves whether the interesting object of analysis is the controllable response or the automatic response.

Limitations

One general concern regarding Study 2 is the difference between the concepts “poor” and “citizens receiving welfare benefits,” as these concepts could activate different associations (see e.g. Henry et al., 2004). The concept “poor” could, for example, activate associations to a single parent; an EU migrant begging on the street; as well as an unemployed Swedish citizen receiving welfare benefits (see Lepianka et al., 2009 for a discussion), and these distinct associations could result in different responses or feelings (Henry et al., 2004). Because of this, it could be problematic not to specify what is meant by the concept “poor.” (see e.g. Lepianka et al., 2009) In SC-IAT it is, however, crucial that the categorizations can be made quickly, hence sentences or long and complicated words are avoided. Even though our definition of poor was clearly stated before the SC-IAT too, in the instructions on the screen, this might not have been sufficient to give the category “poor,” the exact same meaning in the SC-IAT as in the explicit rating task, which may in turn have affected the measures.

The same comment could also be made concerning Study 1, where we clearly specified what we meant by poor, and hence were more precise than much of the earlier research, analyzing the broader concept of “poor.” The difference between our specific definition of poor and the broader definition of poor could, to some extent, render comparisons between studies more difficult.

Moreover, just as in previous stereotype content studies, the role of stimulus valence was not scrutinized in the present research. It is possible that some of the results that we obtained were influenced by the evaluative tone of the stimuli, blurring the distinction between stereotype and attitude. In the implicit association tests, participants may have sorted the words according to valence instead of content. In the rating tasks, participants may have been affected by the evaluative tone of the traits. Future research may address this issue by systematically varying the valence of the stimulus words, and then relate valence to response times in the IATs, and to endorsement levels in the trait ratings.

Another concern regarding Study 2 is that it had a student sample, with young and well-educated individuals, meaning that it might be difficult to generalize the results to a broader population. We note, however, that the results from Study 1 were similar to those of Study 2, where participants were sampled from a wider population, although not nationally representative. Future research may benefit from using representative samples. Also, it is possible that different groups in society attribute somewhat different qualities to poor people. Identifying such differences was not an aim of the present study, but could be worthwhile pursuing in the future. Despite this limitation, we believe our findings are important, not least because the stereotype content identified in a Swedish context resembles earlier findings from other cultures.

Finally, it is worth noting that earlier studies referred to in Figure 4 did not have the stereotype of the poor as their unique focus, but rather constructed SCM maps with ratings of several different groups.

Concluding remarks

According to the SCM, societal groups that are perceived as LC-LW can be expected to receive extremely negative responses from members of other groups, including contempt and passive as well as active harmful behavior (Cuddy et al., 2008). So based on the stereotype of their group, the poor are more likely to face negative than supportive behavior. This is relevant information for policymakers. Up-to-date knowledge regarding the perception of the poor can be useful in decisions regarding whether to introduce new policies and, once implemented, to evaluate their effects. Furthermore, negative stereotypes are not only harmful in intergroup relations, where the poor may be disregarded or discriminated by individuals belonging to other groups in society. They can also have detrimental effects through self-categorization (Fell & Hewstone, 2015). Thinking of oneself as a poor person not only implies feeling incompetent, but may also influence one’s predicted as well as actual performance, which may affect life-choices in both children and adults. That poor people are affected not only by how they are treated, but also by the contents of the stereotype itself, is another piece of useful information to policy-makers. Similar arguments have recently been put forward with regard to gender stereotypes (e.g. in the PISA-report; OECD, 2015).

In the present research we set out to identify the content of the Swedish stereotype of the poor. We used multiple methods, not only the explicit ones that psychologists traditionally use, but at the same time treat somewhat skeptically. All of the methods provided results suggesting that the content of the stereotype of the poor is LC-L/MW. They all appear to capture the same basic content, which suggests that researchers may feel free to use either of them when investigating stereotypes about the poor, but explicit measures if the controllable stereotype is of primary interest, and implicit measures if the implicit stereotype is of primary interest.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the Swedish Research Council [grant number 2014-203].

Acknowledgments

The authors thank Emil Rytterstedt, Eva Högström and Johanna Albinsson for data collection. We are also grateful to Øyvind Jørgensen for assistance with the SC-IAT programming.

Notes

1. There is however an ongoing discussion in social psychology regarding which attributes that mirror which dimension, see e.g. Abele, Uchronski, Suitner, and Wojciszke (2008).