Abstract
Homelessness is an important social problem in many countries, including the United States. The plight of the homeless is compounded by a high level of stigma associated with the homeless. This study examines the effects of humorous and nonhumorous signs used by the homeless to attract donations. Study 1 shows that nonhumorous signs attracted 10 times as much money as humorous signs. Study 2 shows that subjects felt more comfortable in the presence of homeless not holding a sign and perceived them more positively compared with homeless holding a humorous sign. Positive perceptions of them led to more comfort, which led to more donations. Study 3 shows that subjects perceived homeless not holding a sign more positively compared with homeless holding a nonhumorous sign. These findings suggest that signs make potential donors feel uncomfortable, potentially resulting in diminished donations.
The percentage of people living in the United States who are homeless is difficult to estimate; it changes rapidly and hinges on the method of defining homelessness. Although there may be a relatively small percentage of persons homeless at any one time, the 6.2% lifetime literal homelessness estimate generated by Toro et al. (2007) provides a useful point estimate of the magnitude of this social condition. Notably, this figure is high relative to other developed nations (Toro et al., 2007).
Surveys indicate that perceptions of homeless are negative (Agans et al., 2011). For example, a multidimensional scaling analysis indicates that the homeless are associated with traits such as insignificance, dishonesty, irresponsibility, and a lack of intelligence, as well as a cluster of other negative attributes (Cuddy, Fiske, & Glick, 2008). Interestingly, however, these judgments may arise from faulty premises, as data have shown that certain important perceptions of the homeless are in substantial error. So, for example, the Agans et al. (2011) data indicate that, although the prevalence of factors such as substance abuse and mental illness may be higher among members of the homeless population than among those who are not homeless, those who are not homeless radically overestimate the impact of these factors as causes of homelessness. Specifically, in some cases they overestimate their prevalence in the homeless population by a factor of two or three.
Goffman (1963, p. 3) defined stigma as “an attribute that is deeply discrediting.” People perceived as having negative attributes that distinguish them from most members of society can be said to constitute a socially stigmatized group. Thus, given the negative attributes associated with the homeless, it is reasonable to label them as a stigmatized group. But because stigma arises from social perceptions, and because social perceptions are able to be influenced, the question arises as to how people might be influenced to perceive the homeless more positively.
One framework useful for examining this question is Fiske’s stereotype content model (Cuddy et al., 2008; Fiske, Cuddy, & Glick, 2007). Pivotal to this model are two dimensions of social perception: warmth and competence. Warmth includes characteristics such as being trustworthy, tolerant, and friendly; competence includes characteristics such as being efficient, intelligent, and knowledgeable (Cuddy et al., 2008). Those high on both traits are said to be admired, those high on warmth but low on competence are said to be pitied, and those low on warmth but high on competence are said to be envied. But the low warmth/low competence group, the quadrant into which the homeless are said to fall, is characterized as contemptible or disgusting (Fiske et al., 2007). Moreover, among groups that fall into this quadrant, such as poor Blacks, welfare recipients, and Arabs (Fiske et al., 2007), the homeless are extreme, with mean ratings more than three standard deviations below the mean on both measures (Cuddy et al., 2008). Consistent with these ratings, neuroimaging data indicate that when exposed to stimuli of representative members of groups that fall into the low warmth/low competence space, an increase in activity in the insula, an area of the brain associated with disgust, results (Cuddy et al., 2008).
The stereotype content model implies that the matter of influencing perceptions of the homeless can be reframed fruitfully as one of influencing others to perceive the homeless as warmer and more competent. Warmth is a particularly important dimension to target because people are more sensitive to warmth information. Consequently, warmth judgments are made very rapidly—more rapidly than competence judgments (Fiske et al., 2007). Moreover, a change in warmth without a corresponding change in competence would move perceptions into a region of the warmth/competence space that is characterized by sympathy and pity, a space occupied by groups such as the elderly and the disabled. Groups perceived in this manner tend to receive active help (e.g., assisted living for the elderly), although such help is accompanied frequently by benign neglect (e.g., social isolation).
In contrast, those groups perceived as low warmth/low competence may experience both active hostility and passive neglect (Fiske et al., 2007). Furthermore, to move perceptions from low competence to high competence without changing perceptions of warmth would be to move them into a space characterized by envy. Groups in this space (e.g., the rich) tend not to be isolated, but they may experience overt hostility when circumstances allow it (Fiske et al., 2007). In sum, a move toward both greater warmth, compared with greater competence, would likely result in the homeless receiving more help.
One method that the homeless might employ to influence others to view them as warmer and more competent is to display signs designed to accomplish those ends. Commonly, the homeless display signs designed to solicit monetary donations. The content of these signs can be seen as messages designed to gain the compliance of a target audience, compliance in this case marked by the frequency and size of the donations collected.
Pertinent to such efforts there exists a small, and dated, body of literature that has examined the impact of stigma on compliance. The few studies examining the ability of the stigmatized to gain the compliance of others provide some evidence that the stigmatized are better able to gain the compliance of others than controls. For example, Levitt and Kornhaber (1977) report evidence of a reasonably strong effect of stigma on compliance (r ≈ .36), but the compliance gaining context was such that complying entailed no future interaction with the stigmatized influencing agent. Doob and Ecker (1970) report evidence of an ample effect (r ≈ .30) when compliance implies no future interaction with the stigmatized influencing agent, but no evidence (r ≈ .05) of an effect when compliance would require future interaction. In contrast, Walker, Harriman, and Costello (1980) report no evidence that stigma affected compliance.
The stigmatization inductions investigated to date (eye patches, leg casts, arm in slings) would likely be characterized as high in warmth but low in competence. In contrast, homelessness represents a stigma that may produce more negative and extreme perceptions, and accompanying emotional and cognitive reactions (Phelan, Link, Moore, & Stueve, 1997). The homeless condition is not only visible but may also be perceived as obtrusive and permanent. Consequently, certain features of the results of the Doob and Ecker (1970), Levitt and Kornhaber (1977), and Walker et al. (1980) experiments may not apply to the compliance gaining appeals of the homeless.
Nevertheless, one finding of potential applicability is the challenge of the stigmatized gaining compliance when it requires interacting with the stigmatized party. The failure to find an effect under this condition suggests the possibility that potential targets may be uncomfortable interacting in close physical proximity or for some extended length of time (or both) with a stigmatized person. Signs that convey the sense that it is safe to interact with the stigmatized have the potential to attenuate target discomfort, and hence, to gain compliance in the form of increased frequency or amount of monetary donations. In contrast, signs may convey the opposite message, and, if so, they would be counterproductive for the purpose of obtaining monetary donations.
Observations of the signs used by the homeless to solicit monetary donations indicate that humor is employed frequently. Anecdotal evidence of humor promoting social influence abound. For example, a portion of Henry Kissinger’s diplomatic effectiveness has been attributed to his injection of humor into negotiations (O’Quin & Aronoff, 1981). Additionally, cartoonists such as Jules Feiffer have used humor to attempt to influence political and social views, as have comedians such as Mark Russell and columnists such as Russell Baker. Racehorse Haynes employed humor frequently in the courtroom to promote the defense of his clients. Grady Nutt was noted for employing humor in his sermons as a means of gaining converts. And humor is used frequently in advertisements to promote products. Additionally, however, there is a small, and dated, body of literature indicating that humor is effective for the purpose of persuading, or affecting known mediators of the persuasion process (Gruner, 1967a, 1967b, 1970; O’Quin & Aronoff, 1981; but see Markiewicz, 1974).
On the other hand, humorous signs employed commonly by the homeless may lack some important features associated with warmth, such as helpfulness, sincerity, and reliability (Cuddy et al., 2008). Three studies were designed to examine this issue. The first, an exploratory study drawing on data collected from an online research panel, was designed to assess the relative effectiveness of humorous and nonhumorous signs as a means of obtaining monetary donations.
Study 1
Method
Participants
Data were collected in Michigan, a state ranked 10th in the United States for the number of homeless residents (The National Alliance to End Homelessness, 2015), as part of a larger and more diverse project. Michigan residents ages 18 to 86 years inclusive (Mdn = 51; M = 47.88; SD = 15.21) from an existing online research panel were invited to participate in an online survey. Data pertinent to this study were obtained from 1,341 respondents (males = 48.5%; females = 51.5%). Most respondents were Caucasian (88%) with approximately 9% identifying as Black and 2% as Latino(a). The median annual household income for this group was in the $40,000 to $49,999 range. Approximately 81% of the sample had some college or more education. Finally, 36% identified themselves as Democrats, 23.4% as Republicans, 34.6% as independents, and 4.9% as other.
Design
Each respondent was exposed to six signs said to be employed by the homeless. Three of the signs were designed to be humorous: “Family kidnapped by ninjas, need $4 for karate lessons,” “Why Lie? Need women, beer, and chicken wings,” and “Attention: if you don’t want to spare me some change at least don’t give any to the other bums.” The other three signs avoided humor: “If you can’t help me this time, consider me next time. Thank you,” “Please help me. I’m hungry, unemployed, and homeless. Thank you. God bless you. Luke 6:30,” and “Very embarrassed but I lost my job and need help for my family not to be homeless and have food. Don’t know what else to do.”
Instrumentation
Respondents were instructed to imagine that they had $60 to distribute. They could choose to distribute all of it, none of it, or some of it. If they chose to distribute all or some of it, they could give it to the person associated with one of the signs or give portions to as many of the six signs as they chose. It was made clear that money would not actually be given to the people associated with the signs, and that they could not keep any of the money for themselves.
Results
Inspecting the mean amount of money allocated to each of the signs indicated that there was no evidence of differences among the three humorous signs, and only minor differences among the three nonhumorous signs. Thus, responses to the three humorous signs were summed, and responses to the three nonhumorous signs were summed. A one-way repeated measures analysis of variance performed on these data produced evidence of a substantial difference, F(1, 1,341) = 3,507, p < .001, η2 = .10. The nonhumorous signs garnered, on average, substantially more in contributions than did the humorous signs (M = $49.32 vs. M = $4.82). This difference was by more than a factor of 10.
Discussion
These data could be taken to imply that humor is a counterproductive method for the homeless to employ as a means of soliciting money. This conclusion would, however, be premature because an important control condition is lacking. Namely, there was no condition in which a sign fails to appear.
Additionally, there are other important limitations. For example, respondents were restricted in the amount they could give, and they were given the task of choosing among competing signs rather than making a donation decision is response to only one sign. There was no check on the humor induction to ascertain if the humorous signs were actually perceived to be so and if the nonhumorous signs were perceived to be devoid of humor. There was no picture of a homeless person holding any of these signs, so the experiment lacked a certain amount of both experimental realism and mundane realism. These limitations were addressed in the second experiment.
Study 2
The purpose of the second experiment was to contrast the effectiveness of humorous signs with two control conditions, one in which there is a picture of a homeless person without a sign and a second in which there was no picture and no sign. In this experiment, respondents were unrestricted in the amount that they could give. They were also provided with richer stimuli. In addition to the two control conditions, there were conditions in which respondents were shown pictures of a homeless person holding a sign for the purpose of soliciting funds. Because respondents were assigned to independent groups, they were not given the task of distributing their money across a set of signs, but rather they decided to give (and if so, how much) or not to give in response to the single stimulus person to which they were exposed (or not in the no picture condition). Finally, an induction check and measures designed to probe the mediating process(es) was (were) obtained.
Method
Subjects
A convenience sample of 513 students enrolled in the Department of Communication subject pool at Michigan State University participated in this experiment. The sample was 60.3% female and predominately White (76.2%) with almost equal numbers of freshmen, sophomores, juniors, and seniors. The mean age was 20.21 (SD = 1.96).
Design
Students were assigned randomly to one of five conditions. In the humorous signs conditions some saw a picture of a male actor dressed as a homeless person holding a sign that read “Family kidnapped by ninjas, need $4 for karate lessons” (N = 103), others saw a picture of the same person in the same pose holding a sign that read “Need cash for alcohol research” (N = 102), others saw a picture of the same person in the same pose holding a sign that read “Bet you can’t hit me with a quarter” (N = 103). Those in one of the control conditions saw a picture of the same person in the same pose, but he was not holding a sign (N = 104). Those in a second control condition received instructions, but they did not see a picture (N = 101).
Procedure
Members of the subject pool who agreed to participate in this experiment were given a link to click, and the link took them to the experiment. After completing the consent form, they were initially asked to respond to a set of demographic items.
Next, they received the following instructions.
More than one million people will be homeless in the United States in any given year. Rarely do most homeless people stay homeless for long. The average duration for the homeless is approximately six months. Contrary to what some think, not all homeless people have criminal records. Moreover, not all homeless people suffer from alcoholism, drug addiction, or mental illness. Also many homeless people are employed, at least part time. Following is a picture of a homeless person. After the picture is a set of questions about him.
In the no picture control condition, students did not receive the final two sentences. After these instructions, they were exposed to the stimulus (or lack in one of the control conditions) and asked to respond to a set of several items.
Instrumentation
Four indicators (funny, amusing, entertaining, and humorous) were employed to test the effectiveness of the experimental induction. Seven-point response scales, in which higher numbers indicate perceptions of greater humor, were employed to quantify responses. Responses to these items were gathered for all respondents except those in the no sign condition.
Eight true–false items were employed to measure respondent knowledge of homelessness. Example items included “Most homeless people move from town to town frequently,” “Most homeless people remain homeless for more than five years,” and “Very few homeless have even part time work.” Items were scored so that a higher score indicated more knowledge.
For all respondents, save those in the no picture condition, four dichotomous items (No/Yes) assessed perceptions of the person in the picture. They included asking if the person had a criminal record, if he was addicted to drugs or alcohol, if he had been in a mental hospital, and if he was employed (reflected item). Items were scored so that a higher score indicated a more positive perception of the person in the pictures.
The two items used to assess how comfortable the respondents feel in the presence of the homeless were completed by all respondents. Specifically, the items were “I would feel comfortable having a meal with a homeless person” and “I feel uneasy when I meet homeless people.” Five-point response scales were employed to quantify their answers, and items were scored so that higher scores indicated being more comfortable in the presence of the homeless.
Finally, all respondents were asked if they would be willing to make a contribution to the homeless in the form of making a donation to the “Help the Homeless Foundation.” This variable was scored dichotomously with those making a donation being scored one and those decline being scored zero.
Results
The mean of the four items employed to test the effectiveness of the induction served as the measure of respondents’ perceptions of the humor of the pictures. Scores ranged from 1 to 7. The distribution of responses to these items was symmetrical, albeit platykurtic, with a mean of 4.12 (SD = 1.80). The reliability of the scale was estimated as α = .96. To test the effectiveness of the induction, the mean of responses to the three humorous signs (M = 4.46; SD = 1.80) was contrasted with that obtained in the picture/no sign condition (M = 3.08; SD = 1.32). This difference was statistically significant and ample, t(406) = 7.05, p < .001, d = .70. Notably, one-sample t tests indicate that the picture/no sign mean was less than the midpoint of the scale, and the mean of the humorous sign conditions was higher than the midpoint of the scale.
The distribution of the knowledge scores approximated closely the normal distribution. The number of correct answers was used as the measure of this construct. Scores ranged from 1 to 8, with a mean of 4.57 (57%) and a standard deviation of 1.43. There was no evidence that the experimental induction, including or excluding the no picture/no sign condition, was associated with knowledge scores.
The sum of the perception scores served as the measure of this construct. Scores ranged from 0 to 4, with a mean of 2.87 and a standard deviation of 1.03, a one-sample t test indicating that this mean exceeded the midpoint of the scale. The distribution of the perception scores was skewed negatively, and the reliability of the measure was estimated as α = .50. There was evidence that those in the no sign condition (M = 3.21; SD = 0.88) perceived the stimulus person more positively than did those in the sign conditions (M = 2.76; SD = 1.06), t(407) = −3.88, p < .001, d = −.37.
The mean of respondents’ scores on the two comfort items served as the measure of the construct. These scores ranged from 1 to 5, and approximated closely the normal distribution. A one-sample t test indicated that the mean of 2.89 (SD = 0.89) was lower than the midpoint of the scale. The reliability was estimated as α = .69. There was evidence that those in the no signs conditions (no picture and picture/no sign) reported feeling more comfortable in the presence of the homeless (M = 2.99; SD = 0.88) than did those in the sign conditions (M = 2.82; SD = 0.90), t(508) = −2.18, p = .029, d = −.19.
For the donation item, 13.1% of the respondents pledged to contribute. There was no evidence that the induction affected these donations.
These data were examined further for clues as to the nature of the process that produced these judgments. The causal model presented in Figure 1 depicts the results of these analyses. Of necessity, this model excludes the data from the no picture condition, as one of the variables (the perception of the pictured person) was unable to be measured in that condition.
A causal model of Experiment 2.
This model stipulates that both the experimental induction (contrasting the no sign vs. sign conditions) and the knowledge measure exert direct effects on perceptions of the stimulus person, such that perceptions are more positive in the no sign condition and when knowledge about homelessness is higher. The more positive these perceptions, the more comfortable people feel in being in the presence of the homeless. And the more comfortable people feel in being in the presence of the homeless, the more likely they are to donate in response to the appeal made in this experiment.
Although the parameter estimates for this model are modest, each exceeds zero by more than would be expected from sampling error. Moreover, the fit of the model was excellent, root mean square error = .05. Hence, the data are consistent with this model.
Discussion
As in the initial experiment humorous signs proved unproductive, this time in comparison with a no sign (or no picture) control and without some of the methodological limitations of the first experiment. When the stimulus person held humorous signs he was perceived less favorably and people indicated feeling less comfortable in the presence of the homeless. The intuitively appealing explanation that the impact of these signs was mediated by perceptions of the stimulus person and feelings of comfort in the presence of the homeless was shown to be consistent with these data.
A lingering question remains to be addressed, however. Perhaps this effect is not unique to humorous signs. This question was addressed in a third experiment.
Study 3
The third experiment was an exact replication of the second experiment, save for two features. First, in this experiment, N = 413, and second, the three humorous signs were replaced with the three nonhumorous signs, “Homeless. Alone. Cold. Hungry. Please help,” “Homeless. No family or support. Health problems. Could use some compassion. Really need the help,” and “Homeless. Out of Work. Just need a little help.”
Results
The sample was 61% female and predominately White (73.7%) with similar numbers of freshmen, sophomores, juniors, and seniors. The mean age was 20.23 (SD = 1.78). These figures are very similar to those obtained in the second experiment.
The distribution of responses to the humor scale was skewed positively, and substantially so. Scores ranged from 1 to 5.25 with a mean of 1.39 (SD = 0.08). The reliability of the scale was estimated as α = .94. A one-sample t test indicated that the mean was substantially less than the midpoint of the scale, in fact, it was even less than two, t(244) = −12.05, p < .001. Hence, respondents perceived the signs as devoid of humor. There was no evidence that humor ratings varied across conditions.
The distribution of the knowledge scores approximated closely the normal distribution. Scores ranged from 1 to 8, with a mean of 4.53 (56.6%) and a standard deviation of 1.54, figures very similar to those obtained in the second experiment. Once again, there was no evidence that the experimental induction, including or excluding the no picture/no sign condition, was associated with knowledge scores.
Scores on the perception measure ranged from 0 to 4, with a mean of 3.06 and a standard deviation of 0.94, a one-sample t test indicating that this mean exceeded the midpoint of the scale. The distribution of the perception scores was skewed negatively and leptokurtic, and the reliability of the measure was estimated as α = .48 (standardized item α = .52). There was evidence that those in the no sign condition (M = 3.25; SD = 1.02) perceived the stimulus person more positively than did those in the sign conditions (M = 2.99; SD = 0.91), t(324) = −2.13, p = .034, d = −.23.
Comfort scores ranged from 1 to 5, and approximated closely the normal distribution. A one-sample t test indicated that the mean of 2.99 (SD = 0.85) was within sampling error of the midpoint of the scale. The reliability was estimated as α = .58. There was no evidence that those in the no signs conditions (no picture and picture/no sign) reported feeling more comfortable in the presence of the homeless (M = 3.07; SD = 0.79) than did those in the sign conditions (M = 2.94; SD = 0.89), t(406) = −1.55, p = .123, d = −.15.
For the donation item, 16.3% of the respondents pledged to contribute. There was no evidence that the induction affected these donations.
Although the data were examined for possible mediators, none emerged. And although comfort correlated with donation in much the same manner as in the second experiment, r = .10, 95% confidence interval [.01, .19], the causal model presented in Figure 1 produced unacceptably large errors in this experiment.
Discussion
There were several points of similarity between the second and third experiments. The samples had similar demographic characteristics, the properties of the distribution of the knowledge scores were similar, the percentages donating were similar, and comfort correlated modestly, but positively with donation. On the other hand, there were notable dissimilarities as well. Perception and comfort ratings were slightly more positive, and the causal model that fit the data in the second experiment did not replicate when the signs were not humorous.
Most important was the finding that perceptions of the stimulus person were more positive in the picture/no sign condition than in the sign conditions. The sign induction had a statistically significant effect on the comfort measures in the second experiment, but not in the third experiment. Nevertheless, the d statistic in the second experiment (−.19) is similar to that of the third experiment (−.15). These data indicate that there is a tendency for signs to nudge respondents to be a bit less comfortable in the presence of the homeless.
General Discussion
Fiske’s data (Cuddy et al., 2008; Fiske et al., 2007) indicate that the homeless are perceived to have deeply discrediting characteristics; they are perceived as disgusting and contemptible. The data collected in this set of three experiments indicate that attempts to influence these perceptions in a more positive direction are not facilitated by displaying signs. Indeed, they indicate that signs may be counterproductive in that regard, particularly humorous signs. In the initial experiment, humorous signs produced lower donations than did nonhumorous signs; in the second experiment, a stimulus person holding humorous signs produced less positive reactions than a stimulus person without a sign; and in the third experiment, a stimulus person hold nonhumorous signs produced slightly less positive reactions than the same stimulus person without a sign.
The question arises then as to why so many homeless who are seeking to obtain donations from others employ signs if they are in fact unproductive or counterproductive. Because the homeless are not likely to be conducting controlled experiments to ascertain if signs are more effective than no signs, or which signs are most effective, one possibility is that they are unaware of the effects of their signs on their target audience. But another possibility is that some device is needed to gain the attention of potential targets, and signs are believed, perhaps correctly, to fulfill that function effectively. If so, then the increased probability of gaining attention may offset, or more than offset, the lower probability of donations associated with holding a sign. Put differently, if the expected value of receiving a donation is the product of the probability of gaining a target’s attention and the probability of the target contributing given that one has gained the target’s attention, holding a sign (contrasted with not holding a sign) may produce a substantially larger value for the former which may more than outweigh a lower probability of the latter. In such a case, the ecological validity of the findings presented in these three experiments would be called into question because respondents were directed to attend to the pictures. Future research employing direct observation techniques may be a fruitful method of addressing this possibility.
From a theoretical viewpoint, however, the question may still be raised as to why signs, both humorous and nonhumorous, fail to promote increased feelings of warmth, as might be expected to be reflected in enhanced comfort ratings. One possibility is that the act of calling attention to the homeless can be a distressing experiment for potential targets for donation. Unpleasant affect, such as shame, may arise. Unpleasant cognitions, such as imagining how one might find oneself in the same position, may arise. Negative thought and affect, particularly when coupled with the well-documented desire to believe in a just world (Lerner, 1980), may be sufficient to produce avoidance behavior, and avoidance behavior may be justified after the fact by derogating the homeless.
In addition to the previously mentioned challenge to the ecological validity of these experiments brought about by studying the issue in an artificial setting, there are other shortcomings. For example, student samples, and the limitations accompanying them, were employed in the second and third experiments. Moreover, the reliabilities of some of the measures were not optimal.
Nevertheless, several intriguing findings emerged. Moreover, the outcomes have implications for what might be a fundamental dilemma facing the homeless who are seeking donations on which to help them live. They may find themselves impaled on the horns of either calling attention to their need by holding a sign while lowering their likelihood of obtaining donations, or failing to call attention to their need. Either horn of the dilemma is unsatisfying, and it suggests the importance of humanizing the homeless in the eyes of their fellow citizens in a more effective manner.
Footnotes
Acknowledgements
The authors would like to thank Karen Clarke and Linda Stork at the Institute for Public Policy and Social Research at Michigan State University for their assistance with data collection in Study 1. The authors would also like to thank Arthur Scholbe for his assistance with stimulus materials used in Studies 2 and 3. And, the authors would like to gratefully acknowledge the useful input on measurement and design contributed by Dr. Sandi Smith.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors are grateful to the Urban Communication Foundation and International Communication Association for the James W. Carey Urban Communication Grant, which helped to support this study.