Harnessing the Identifiability Effect to Enhance Public Support for Redistribution

Abstract

This article proposes a new method of increasing public support for redistribution through tax-and-transfer law by harnessing the psychological phenomenon known as the identifiability effect. Enhancing such support is crucial, as implementing or sustaining redistributive measures is difficult in the face of unsympathetic public opinion. The article reports on the findings of three original, pre-registered experiments—conducted in Israel, England, and the United States—that substantiate the theoretical argument. The findings reveal that minimal and noninvasive identification of a single recipient can enhance people’s support for redistributive taxes—including direct income transfers—that would benefit the entire group of welfare beneficiaries. These results are relevant for the literatures on redistribution and the identifiability effect, as well as for policymakers.

1. Introduction

A fundamental feature of tax-and-transfer law, which may negatively impact public support for this primary method of redistribution, is that givers and receivers are far removed from and entirely anonymous to each other. A wealthy person is not obliged to send a check to a specific welfare recipient, or make a wire transfer to a designated beneficiary. Instead, the money passes through several intermediaries—such as the giver’s employer (who withholds the tax from the employee’s salary), the state, and various administrative agencies—before it eventually reaches its unknown recipient. In these circumstances, it is not surprising if the transferors feel little empathy or responsibility toward the unidentified beneficiaries, or lack confidence that they can effectively help them. This is regrettable, as direct income transfers are widely regarded as the most efficient and effective method of redistributing wealth (Kaplow & Shavell, 1994; Polinsky, 2019, pp. 155–164). Yet, such transfers encounter the greatest public resistance (Gilens, 1999, pp. 1–9, 60–79), and it has proven difficult to sustain welfare programs over time in the face of unsympathetic public opinion (Kluegel & Miyano, 2011, p. 81). Consequently, insufficient redistribution occurs and widespread inequality persists and even worsens (Hartwich & Becker, 2023, p. 168; Kuziemko et al., 2015, p. 1478; Page & Jacobs, 2009, pp. 4–9).

Direct income transfers do not enjoy widespread support because they are often perceived as undermining incentives to work and as aiding the “undeserving poor.” Furthermore, people’s beliefs about poverty and the deservingness of welfare-recipients are often based on deeply entrenched stereotypes, biases and misconceptions (Gilens, 1996, 1999; Piff et al., 2020), and thus, even in the face of relevant information, their opinions on redistribution are unlikely to change (Kuziemko et al., 2015; Van Oorschot & Roosma, 2017, p. 10). Nevertheless, if the state wishes to reduce inequality and ensure every person an adequate level of well-being, it must boost public support for redistributive measures. ¹ Unfortunately, various methods suggested in the literature to promote tax compliance and support—like aligning the source of the tax and its expected use, relying on local (rather than national) taxes, or allowing taxpayers to allocate some of their tax bill to goals that they value—are mostly inapplicable in the context of wealth redistribution.

This article explores a new and plausibly cost-effective way to increase support for redistributive taxes. To date, neither scholars nor policymakers have examined whether the psychological phenomenon labelled the identifiability effect may be harnessed to enhance support for redistribution. The identifiability effect is the tendency of people to react more strongly to identified individuals than to unidentified ones. The main line of research has focused on prosocial behavior and uncontroversial goals, such as voluntary monetary donations. For example, it was found that victims of natural disasters or individuals who are gravely ill attract more generous contributions from the public when they are minimally identified by name and/or photo, rather than anonymous (Kogut & Ritov, 2005b; Small et al., 2007).

To examine the existence of an identifiability effect in the more contentious context of coercive transfers through redistributive taxes, we conducted three original, preregistered vignette experiments—the first two in Israel and the third in England and the United States. We focused on means-tested welfare benefits designed to alleviate poverty and provide safety-net support, such as income assurance allowances. ² Our findings indicate that minimal, noninvasive identification of a single recipient increases support for redistributive measures that would benefit the entire group of anonymous recipients. Furthermore, such identification enhanced support not only for a relatively popular welfare measure like funding daycares (to enable poor parents to work), but also for one that encounters more resistance—direct income transfers to the poor. Importantly, identifiability had a broad positive impact beyond participants’ demographics and self-interest (in terms of the prospect of receiving the benefit themselves). Interestingly, identification increased individuals’ support for the tax despite not affecting their beliefs about the worthiness of the state’s redistributive goals or the causes of poverty. In contrast, two other support-enhancing techniques we tested—a reminder that people’s current financial condition may deteriorate, and framing the situation as a loss rather than a gain—did not influence participants’ support for the tax. Our experimental results contribute to the literatures on redistribution and the identifiability effect, as well as to policymaking. Among other things, the findings demonstrate that the scope of the identifiability effect is wider than previously believed. It can extend beyond the private sphere, to include interactions with the state. Furthermore, they may encourage policymakers to consider experimenting with the identifiability effect when promoting and implementing redistribution (and other) programs.

Two clarifications about the goals and scope of this article are in order. Like many studies of public attitudes toward governmental policies, our study focuses on individuals’ civic attitudes and the ways these attitudes may be influenced by governmental entities. As prior research has shown, individuals’ attitudes toward governmental policies may diverge from their day-to-day behavior: people who support state redistribution may refrain from donating to charities, and those who favor strict environmental policies may nonetheless litter public spaces. Although such divergence may be perceived as hypocrisy, it may instead reflect a rational recognition that meaningful redistributive (or environmental) outcomes require coordinated, collective state action, whereas uncoordinated, individual contributions are unlikely to have significant effects (Lewinsohn-Zamir, 1998). Because we examine participants’ civic attitudes rather than their personal behavior, we did not include monetary or other incentives in our experiments.

In the same spirit, this article does not examine voter decision-making. Voting choices in general elections are shaped by a wide range of deliberative and non-deliberative factors (such as partisan identity), and the redistributive policies of elected candidates may have a limited effect on individual voting behavior. At the same time, previous studies show that elected policymakers (and, indirectly, unelected ones as well) are influenced by public opinion (Page & Shapiro, 1983; Burstein, 2003; Shapiro, 2011; Gilens, 1999, p. 213; Katz & Zamir, 2021, pp. 422–423), and employ various techniques to shape it (see generally Zamir & Teichman, 2018, pp. 408–409). Our focus is on such attitudes and influence, not on electoral behavior.

The Article is structured as follows. Part 2. reviews previous studies on redistribution and the identifiability effect. This review supports our hypothesis that factors such as identification may increase support for redistributive taxes. Part 3. presents our experiments and their findings. Part 4. discusses the results, highlights their contribution to the literature, and explains their practical significance. Additionally, it addresses concerns and limitations of the experiments, and suggests avenues for future research.

2. Theoretical Background

Tax-and-transfer law—including progressive taxation, negative income taxes, cash assistance to needy families, unemployment compensation, and disability benefits—is the primary method employed by the state to redistribute wealth. There is an ongoing debate between scholars who argue that tax-and-transfer law should be the sole method of redistribution (Kaplow & Shavell, 1994; Weisbach, 2003, 2023; Cooter & Ulen, 2012, pp. 7–8, 106–08; Polinsky, 2019, pp. 155–164), and scholars who contend that other types of legal rules—such as rules of private law—should also be used to attain redistributive goals (Fennell & McAdams, 2016; Kronman, 1980; Lewinsohn-Zamir, 2006; Liscow, 2022). The latter position does not maintain that these other rules should replace tax-and-transfer mechanisms, but, rather, that they complement them. ³ Thus, there is general agreement that tax-and-transfer law should play a major role in wealth redistribution (for classic theories of the optimal progressive income taxation see Mirrlees, 1971; Saez, 2001).

This is easier said than done. People’s beliefs about poverty and their attitudes towards welfare programs are complex. Poverty can be attributed to personal factors, such as the impoverished person’s lack of skills, laziness, or loose morals; or it can be attributed to social factors such as discrimination, lack of equal opportunity, or failure of the economic system. The first type of explanation implies that inequality is fair, a product of people’s own behavior and responsibilities, whereas the second implies that this phenomenon is unfair and should be corrected (Katz, 2013, pp. 3, 6–7; Piff et al., 2020). Although in the United States, for example, personal explanations of poverty are more common than in other western, free-market societies, Americans still attribute poverty to both personal and social causes (Kluegel & Smith, 1986, pp. 23–24, 76–88; Kluegel et al., 2011, pp. 189, 203–206; Piff et al., 2020, pp. 497, 502).

The co-existence of these conflicting beliefs affects people’s attitudes toward different types of redistribution. Support for a redistributive program may depend, among other things, ⁴ on whether it is framed in a way that makes either the social or the personal explanations for poverty salient. Direct income transfers to the poor typically receive the least public support, because they are often perceived as undermining individuals’ incentive to work and better their own position, thus aiding the “undeserving poor.” In contrast, the public widely supports governmental assistance that would provide equal opportunities for achievement and enable the “working poor” to help themselves—including job creation, job training programs and daycare funding (Gilens, 1999, pp. 1–9, 60–79; Heclo, 1986, pp. 326–332; Kluegel & Smith, 1986, pp. 152–157, 163–165; Page & Jacobs, 2009, pp. 22–23, 56–63, 72–73). ⁵ For similar reasons, there is greater support for welfare transfers that are tied to specific necessities—such as housing and healthcare—than to no-strings-attached transfers of money (Burtless, 1986, p. 24; Liscow & Pershing, 2022). Put differently, people’s opposition to redistribution may be based not on a principled objection to helping the poor, ⁶ but rather on the concern that recipients are taking advantage of the system.

Interestingly, although countries differ in the degree of public support for redistribution (Alesina et al., 2018; Mengel & Weidenholzer, 2022, pp. 3, 7–8), they nonetheless share important similarities. For instance, people across societies appear to care about welfare recipients’ own efforts to improve their circumstances and believe that assistance should be provided to those who make an effort to help themselves. Aarøe and Bang Petersen (2014) show that when presented with explicit cues portraying recipients as either lazy or industrious but unlucky, Americans and Scandinavians express similar views about whether such individuals should—or should not—receive welfare. Thus, cross-national differences in support for redistribution stem not only from factors such as cultural values related to individualism, levels of ethnic homogeneity, meritocratic beliefs, or ideological positions on redistribution per se, but also from individuals’ beliefs about whether most welfare recipients are, in actuality, idle or diligent (Bang Petersen, 2012; Heuer & Zimmerman, 2020; Van Oorschot, 2000; Van Oorschot & Roosma, 2017). ⁷ These beliefs may be based on stereotypes, misconceptions, biases, and racial prejudice (Gilens, 1996, 1999). A prominent example is the image of the “Welfare Queen”, who allegedly prefers to cash-in welfare checks than to earn a living (Monnat, 2010, pp. 643–644). This offensive depiction of Black single mothers is gradually being replaced with that of immigrants and young Black males (Katz, 2013, pp. 8–9; Van Oorschot & Roosma, 2017, pp. 11–13). ⁸ Other biases that undermine support for redistribution include the tendency to view the status quo as fair (Jost, Banaji, & Nosek, 2004), to overestimate social mobility (Shariff, Wiwad, & Aknin, 2016; Page & Jacobs, 2009, pp. 11–12, 51–53), and to favor personal over social explanations for poverty (Hunt & Bullock, 2016). These biases serve, among other things, to satisfy individuals’ need to believe in a just world (Piff et al., 2020, pp. 496, 501).

To be sure, opposition to redistribution is based not only on concerns about the deservingness of welfare recipients. Redistributive measures may negatively affect taxpayers’ incentives to work, save, and earn income, thereby reducing the tax base and generating deadweight losses. The efficiency costs of redistribution must therefore be taken into account when pursuing the goals of greater equality and social welfare (Diamond & Saez, 2011; Kaplow & Shavell, 1994; Tuomala, 1990). This article does not engage with the debate over the equity-efficiency tradeoff. Instead, it focuses on people’s perceptions of welfare beneficiaries and examines ways to increase support for the redistributive programs that the state chooses to adopt.

The impact of individuals’ perceptions of self-effort and fault on their inclination to help others is buttressed by the psychological literature on charitable giving. As numerous studies have shown, people are more empathic toward—and hence more willing to aid—those who are not viewed as responsible for their plight and those who attempt to help themselves (Erlandsson, 2021, p. 1120; Fong, 2007; Jin et al., 2021; Lee, Winterich, & Ross, 2014). It has also been found that victims of sudden emergencies and naturally-caused disasters (such as a tsunami) elicit more donations than victims of ongoing emergencies and human-caused disasters (such as a war) (Slovic, 2007; Zagefka et al., 2011). The former types of events are perceived as involving less personal blame and fewer opportunities for self-help efforts. From this perspective, poverty is mostly an ongoing and humanly-caused condition, that is likely to be, at least to some degree, attributed to poor people’s choices and behavior.

Note, that individuals may also vary in their preferences for redistribution. Women, for example, tend to be more charitable than men (Winterich et al., 2009, p. 201), and show more support for redistributive policies (Blekesaune & Quadagno, 2003, pp. 417, 420, 425; Mengel & Weidenholzer, 2022, p. 8). In a similar vein, people holding liberal or left-wing political views are more supportive of redistribution than people holding conservative or right-wing views (Alesina et al., 2018, pp. 543–545, 549–552; Bang Petersen, 2012, p. 1; Hartwich & Becker, 2023). The effect of income appears to be mixed (Karadja et al., 2017, pp. 201–202). Some studies found greater support among individuals with lower income. This may be due to the likelihood that they themselves would benefit from welfare, or because they endorse social-democratic values and feel more compassion towards welfare recipients. ⁹ In contrast, other studies observed that the association between economic status and attitude towards welfare is rather modest, as many poor people oppose welfare and many wealthy ones support it (Gilens, 1996, p. 594; Kuziemko et al., 2014, pp. 107–108, 133–145; Mengel & Weidenholzer, 2022, pp. 4, 7).

Importantly for our purposes, it was found that individuals’ basic opinions about redistribution are unlikely to change, even in the face of information about inequality and poverty (Katz, 2013, p. 2; Van Oorschot & Roosma, 2017, p. 10). Thus, for instance, Kuziemko et al. (2015) showed that, while information can significantly increase people’s concern about inequality and poverty, it has only a very small effect on their ensuing support for poverty-alleviation policies. ¹⁰ This is troubling, because the extent of public support affects both the methods and the scope of redistribution (Burtless, 1986, p. 48; Gilens, 1999, p. 213; Heclo, 1986, p. 313). As Gilens (1999, p. 205) states, suspicion towards welfare recipients “limits the easiest and by many accounts most effective antipoverty measure: giving money to those who lack it.” However, researchers have observed that even more popular forms of redistribution do not receive adequate funding. Consequently, the current level of redistribution in society is much lower than it should be (Gilens, 1999, p. 207; Liscow, 2022, p. 511; McCaffery, 2021).

If giving people accurate—and more—information does not suffice, are there additional means that those who favor redistribution can employ? The general literature on taxation has suggested various ways to enhance individuals’ support for, and compliance with, taxes (for a theoretical discussion of this issue see Rivlin, 1989). ¹¹ These methods, however, are unlikely to solve the problem of insufficient redistribution. One suggested method is to earmark tax revenues for specific expenditures. For instance, it was found that earmarking revenues of fuel tax and carbon tax for environmental policies, increases public support for the tax (see, respectively, Saelen & Kallbekken, 2011; Gevrek & Uyduranoglu, 2015). But as Hemel and Porter (2022, pp. 472–480) demonstrated experimentally, the increased support for earmarked taxes does not stem from earmarking per se, but rather requires a close alignment between the source of the tax and the expected use of the revenues (as when taxes on tobacco are used for cancer research, or motor fuel taxes are employed to build highways). The alignment requirement severely restricts the application of this popularity-enhancing technique for redistributive purposes, since the well-off bearing the tax burden are neither the cause of the poor’s plight (in the way that pollutants contribute to the need for environmental measures), nor stand to benefit themselves from the intended use of the revenues.

Another support-increasing method is to rely more on local—rather than national—taxes, because taxpayers are more likely to identify with their local community, and the public goods provided at the lower level of government would be more salient to them than those provided at the higher level of government (Listokin & Schizer, 2013, pp. 200–202). This technique, too, is largely inapplicable in the redistribution context. A main reason is that due to the mobility of taxpayers, redistribution is primarily undertaken at the national—rather than the local government—level (Shobe, 2016, pp. 333, 361–377; Galle, 2010, pp. 531–537). Yet another method suggested is to allow taxpayers to allocate some of their tax bill to goals that they value. However, even proponents of this measure admit that it would be mostly symbolic (say, 1% of a person’s tax bill), and limited to goals enjoying broad public support, ¹² like health care and education (Listokin & Schizer, 2013, pp. 204–215). Quite obviously, unpopular redistributive measures like direct income transfers to the poor, would not qualify as such a goal.

In light of the above, those who favor redistribution should seek new, cost-effective ways to increase public support for redistributive taxes, and particularly for those that, although unpopular, are highly efficient. Without enough support, it would be extremely difficult to implement welfare programs or sustain them over time. To date, scholars and policymakers have not examined whether the psychological phenomenon known as the identifiability effect, can be harnessed to achieve this goal. Our article attempts to fill this gap.

The identifiability effect (hereafter, IE) is the tendency of people to react more strongly to identified individuals than to unidentified ones. The main line of psychological research on the IE examined pro-social behavior, such as monetary donations. Studies have found that people are significantly more willing to help another person when the individual in need—e.g., a cancer patient, a tsunami victim, or a starving child—is identified rather than unidentified (Jenni & Loewenstein, 1997; Kogut & Ritov, 2005a, 2005b; Ritov & Kogut, 2011; Slovic et al., 2013). Importantly, identifiability can have this effect even when it is minimal and conveys no meaningful information about the recipient (Burnham, 2003 (identification by photograph); Charness & Gneezy, 2008 (identification by name); Small & Loewenstein, 2003; Haran & Ritov, 2014 (identification by number)). Knowing a person’s name, for example, does not make them more familiar or deserving of help than an unnamed person (Kogut & Ritov, 2011, p. 134)—and yet merely knowing such a detail is enough to induce people to allocate more money to that person, even at the contributors’ own expense (Charness & Gneezy, 2008, p. 32).

Some studies explored the effect of identifiability beyond the paradigmatic case of voluntary monetary donations. Thus, scholars have documented identifiability’s impact on willingness to modify school regulations in favor of students (Kogut, 2009) and giving less biased advice in conflict-of-interest situations (Sah & Loewenstein, 2012). In addition, a line of research has examined the effect of identification on reactions to blameworthy behavior, with mixed results. Kogut (2011) showed that when an individual is seen as responsible for his own plight—a person who contracted AIDS as a result of drug use—identification (by name and photo) leads to a decrease in willingness to help him finance expensive medication. Similarly, Small and Loewenstein (2005) observed that identification by a number alone can lead to an increased monetary penalty for non-cooperators in a social-dilemma laboratory game. In contrast, Lewinsohn-Zamir et al. (2017) found that identification by first name benefitted wrongdoers in legal settings. The study compared the judgments individuals made as policymakers formulating rules for unidentified people, with people’s judgments as decision-makers concerning identified individuals. The data revealed that participants were more lenient toward wrongdoers in their capacity as decision-makers than when acting as policymakers. Similarly, Barak-Corren and Lewinsohn-Zamir (2019) showed that, in sexual harassment scenarios, while victims do not benefit from identification by their first name, offenders do. In sum, depending on the specific context, identifiability causes people to react either more generously or more punitively towards others.

In the context of allocating scarce resources—such as university admissions or ventilator assignments—a few studies examined the effect of identifying the party who loses a resource, rather than the party who gains it. As the recipient’s gain comes at another person’s expense, identifying the latter draws attention to the loss caused by allocation decisions. In such cases, identifiability may result in reluctance to support an overall beneficial allocation (Ritov & Garcia, 2022, 2023; Ritov & Zamir, 2014).

Psychological studies have further shown that, at least under certain circumstances, the IE occurs only with regard to a single identified person and does not extend to a group of identified people. Furthermore, assistance to a single identified recipient may exceed that given to a group of identified recipients, even when the amount needed to save one victim or several victims are kept constant (Kogut & Ritov, 2005a, 2005b; Markowitz et al., 2013; Slovic et al., 2013). This phenomenon has been labeled the singularity effect. Since contributing more money to a single victim than to a group of victims may be regarded as inefficient, scholars have attempted to debias this effect, with disappointing results. For example, it was found that presenting an identified single victim (such as a starving African child) jointly with statistical information about the scope of the problem (the extent of starvation in Africa), only reduced people’s willingness to donate to the single identified victim, bringing it down to the lower level of donations for the numerous victims (Small et al., 2007; For similar results in other charity-giving contexts, see Erlandsson et al., 2016; Moche et al., 2022).

Additional factors that affect the incidence and scope of the IE are perceptions of group belonging and the existence or absence of intergroup conflict (Kogut & Ritov, 2007; Ritov & Kogut, 2011). Among other things, it was found that in scenarios perceived as involving intergroup conflict, identifiability increased generosity toward an individual from the opposing group. Thus, when eliciting donations for someone from a rival group—such as a member of another political faction or a fan of a competing soccer team—identifying the beneficiary increased people’s willingness to contribute (Ritov & Kogut, 2011).

Turning to the IE’s underlying mechanism, it appears that the chief source of the effect is the stronger emotional reactions elicited by an identified individual (Erlandsson et al., 2015, p. 10; Slovic, 2007, pp. 88–90). In the context of helping behavior, empathic emotions—such as sympathy, compassion, and distress at the plight of another—depend on adopting the other person’s perspective and imagining how he or she feels. Such empathy is more likely to occur when an individual is identified rather than anonymous. Similarly, when blameworthy behavior is involved, people may find it easier to feel anger and attribute responsibility to an identified person than abstract individuals (Kogut & Ritov, 2005a; Lee & Feeley, 2018; Markowitz et al., 2013). Researchers have also found that people process individuals and groups differently. A single person—unlike a group—is regarded as a psychologically coherent unit and therefore stronger impressions are formed about them (Slovic et al., 2013, p. 132). People more readily and confidently make attributions and judgments regarding a single individual, than about a group (Hamilton & Sherman, 1996; Susskind et al., 1999), so emotions are more likely to play a decisive role when the identifying details concern an individual than a group (Kogut & Ritov, 2011, pp. 134–135). Additional, complementary explanations for the IE emphasize people’s tendency to feel more responsible and effective when they decide to help a particular identified individual than an anonymous person or a group (Cryder & Loewenstein, 2012; Cryder et al., 2013). For a meta-analysis of the IE literature, see Lee and Feeley (2016).

Based on this literature, it is possible that the IE may also manifest in the context of redistribution. Although this effect has mostly been observed with respect to a single identified person rather than a group of identified individuals, it is conceivable that identifying one person within an anonymous group would produce an IE. If so, then by minimally identifying a future beneficiary of a redistributive tax, the state can potentially increase public support for a planned welfare measure. Plausibly, such identification would enhance empathic emotions and feelings of responsibility towards the potential recipients of the redistributive taxes, who are typically both remote from and anonymous vis-à-vis the taxpayers.

In addition to the IE literature, the well-studied phenomenon of psychological distance also buttresses our conjecture that identification may result in increased support for redistributive policies. Psychological distance is a subjective experience that something is either close to, or far away from, the self. Scholars have shown that people’s judgments and decisions regarding themselves, the here and now, and things or events that are close to them, are different from those they make about others, or about objects and occurrences perceived as remote (Trope & Liberman, 2003). One dimension of psychological distance that may be relevant in the context of redistribution is social distance, ¹³ since the taxpayers and the beneficiaries of the tax revenues ordinarily belong to different socio-economic classes. Another dimension, that was examined in relation to the IE, is that of spatial distance. Kogut et al. (2018) found that an increase in participants’ geographical distance from the victim (a resident of a city in an earthquake zone or a sick child), decreased their willingness to donate money. However, identifying the victim (by name and/or picture) attenuated the detrimental effect of perceived geographical remoteness. Plausibly, the IE could also have a positive impact on people’s support for redistributive taxes.

Testing the IE in the domain of tax-and-transfer law can advance both the legal literature on redistribution and the psychological literature on this phenomenon. First, existing studies of the IE focus on voluntary giving, such as donations to various causes. To the best of our knowledge, we are the first to examine coercive giving in the form of redistributive taxes. This new and important context relates to interactions with the state and judgments about its actions, rather than opinions or behavior in the private sphere.

Second, studies finding a favorable effect for the identified party have largely addressed uncontroversial or “neutral” topics and goals, such as charitable giving to cancer patients (Kogut & Ritov, 2005b); aid to refugees (Erlandsson et al., 2016), post-mortem organ donation (Harel & Kogut, 2021); financial advice (Sah & Loewenstein, 2012), decisions by educators (Kogut, 2009), and conservation of endangered species (Markowitz et al., 2013). ¹⁴ Redistribution of wealth, however, is a more contentious issue and some of its forms—particularly direct income transfers to the poor—encounter public resistance. If, nevertheless, the IE can increase support for redistributive taxes, it will assist in attaining a major social goal.

Third, in studies of the IE, participants were typically able to help a particular person—that is, the beneficiary was the identified party itself, and not just a representative of recipients in general. ¹⁵ Clearly, in the context of redistribution, it would hardly be credible for the state to claim that money raised from a certain taxpayer is earmarked for a particular recipient’s needs. Furthermore, as explained above, scholars have found that giving to a single individual decreases when the individual is embedded in a group of persons in need (Small et al., 2007). In the present study, we test the opposite effect: When assistance is to be provided to the whole (statistical) group—such as poor people in general—would presenting a single member of the group increase willingness to help the group, even if such help will not be targeted at the specific identified individual? An affirmative answer to this question would imply that the IE is wider in scope than currently thought, and that its usefulness extends beyond the private sphere—namely as a tool that policymakers should consider employing in the public sphere.

3. Experimental Findings

We conducted three studies to identify factors that influence people’s support for redistribution through tax and transfer. In particular, we tested whether the IE can increase such support and, if so, what accounts for this impact.

3.1. Study 1

The first study examined two factors that may enhance people’s support for redistributive taxes. The first was identifying a single beneficiary of the proposed redistributive transfer from within the group of potential recipients, in an attempt to harness the IE in favor of the group. The identifying information we used was minimal: first name, age, family situation, and city of residence. The second factor was a note on future uncertainty, namely, that individuals’ current economic situation may change for the worse. This reminder may cause people to consider that they too could need similar assistance in the future (thereby making the redistributive tax more self-relevant), or to view welfare recipients as not fundamentally different from themselves, potentially attenuating perceived social distance between givers and receivers (or both). The two factors were tested with respect to two types of redistributive measures: A tax collected to fund direct income transfers to the poor (“income assurance allowances”), and a tax to help fund the payment of daycare for the children of poor people, to enable them to go out to work.

Participants were informed that in order to reduce inequality of wealth and ensure individuals a dignified existence, the state is considering increasing the income tax for all taxpayers, with the additional revenues collected to be used for one of the two redistributive measures. They were asked to indicate the extent to which they supported the tax increase. ¹⁶ After recording their level of support, participants were asked additional questions, including about the emotions they felt towards beneficiaries of the tax, the extent to which the tax’s purpose is worthy, and their beliefs about the causes of poverty. Based on the literature review in Part 2., our first hypothesis was that both identifiability and future uncertainty would increase people’s support for the redistributive tax, in comparison to the control condition, in which neither factor is mentioned. Plausibly, identifying a recipient would arouse more positive emotions toward the group as a whole, particularly if the beneficiary depicted is not distinctly to blame for their situation. And noting that an individual’s present economic situation can deteriorate in the future, could also cause people to view welfare recipients in a more positive light. ¹⁷ Our second hypothesis was that financing daycare payments would meet with greater support than financing direct income transfers to the poor. The former type of assistance is clearly meant to remove barriers to employment, thereby enabling the poor to help themselves, while the latter is often perceived as undermining incentives to work, and consequently encounters more public resistance. The study was preregistered. ¹⁸

3.1.1. Participants

A total of 602 people were recruited by the Midgam survey company to participate in an online study. ¹⁹ The participants constituted a representative sample of the Israeli adult population in terms of gender and age. One hundred and twenty-nine participants (21%) did not answer the attention check correctly. The following analyses includes the remaining 473 respondents.

3.1.2. Experimental Design

The study included two types of scenarios: a tax to fund income assurance allowances (“Income”) and a tax to fund daycare payments (“Daycare”). Each scenario had three versions: control, identified recipient, and future uncertainty. Thus, the study contained six between-subject conditions (2 types of redistribution x 3 influencing factors). Participants were randomly assigned to one of the conditions and asked to indicate the extent to which they support the tax increase, on a scale ranging from 1 (strong opposition) to 7 (strong support). ²⁰ To illustrate, the three versions of the “Income Assurance Allowance” scenario read as follows:

Inequality in wealth in our society is immense. Some people have very low income, or no income at all. Among its various goals, the state aims to reduce income disparities and ensure a dignified existence for its residents. One way to achieve this goal is by providing the poorest individuals with a monthly payment (“income assurance allowance”). Assume that the current funding allocated for income assurance allowances is insufficient to achieve this goal, and therefore the state is considering increasing the income tax by 1% for all taxpayers. The additional revenue collected from this tax will be used to fund income assurance allowances.

Control: The people who will receive the income assurance allowance will consequently be able to maintain a dignified existence.

Identified Recipient: One of those eligible to receive an income assurance allowance is David. David, who lives in Ashdod, ²¹ is a 34-year-old widower raising his two young children alone. David and other people who will receive the income assurance allowance will consequently be able to maintain a dignified existence.

Future Uncertainty: Life experience shows that even people who are currently financially stable may, at some point in their lives, encounter severe economic difficulties and need an income assurance allowance. The people who will receive the income assurance allowance will consequently be able to maintain a dignified existence.

To what extent do you support increasing the income tax to fund income assurance allowances? Please indicate your answer on a scale from 1 to 7, where 1 signifies that you very strongly oppose increasing the tax rate, and 7 signifies that you very strongly support increasing the tax rate.

Likewise, the “Daycare Funding” scenario asked participants to indicate the extent of their support for an income tax increase that would provide poor people with a monthly payment for daycares for their children, to enable them to work. Similarly, it had three versions: control, identified recipient, and future uncertainty.

After indicating the degree of their support for the tax, participants were asked to rate the extent to which they felt anger, pity, and responsibility toward people receiving an income assurance allowance (or daycare funding), on a scale from 1 (not at all) to 7 (very much). Following that, they were asked whether, in their opinion, funding such benefits is a worthy goal for the state, on a scale from 1 (not worthy at all) to 7 (worthy very much). Participants also estimated the likelihood that they or a family member would ever need this benefit in the future, on a scale from 1 (no chance at all) to 7 (very high chance). In addition, we asked them whether they, or a family member, currently receive the benefit, or have received it in the past (a yes/no question). Finally, participants were presented with a list of potential causes of poverty, in a random order. ²² Some of the items attributed poverty to dispositional factors, such as the behavior of the poor themselves (e.g., “poor people do not save, they spend money foolishly”; “low morals and alcoholism”; “poor people do not try hard enough”). Other items attributed poverty to situational factors, that originate in the state or society (e.g., “inadequate government policy”; “exploitation by the rich”; “the overall economic situation is unfavorable”). Participants were asked, with respect to each item, to what extent they believe it to be an important cause of poverty, on a scale from 1 (not important at all) to 7 (very important). ²³

3.1.3. Results

Identifiability increased support for redistributive taxes. ANOVA of tax support by condition and scenario yielded a significant effect of condition (F(2,467) = 4.253, p = .015, Partial Eta Squared = .018). ²⁴ Neither scenario (the type of redistributive measure), nor the interaction, were significant (p > .7 for both). Post-hoc contrasts treatment conditions to the control one, yielded a significant difference between the “identified recipient” and the control conditions (mean difference = .60, p = .004, CI[1.02, .19]), but no significant difference between the “future uncertainty” condition and the control (mean difference = .19, p = .352, CI[.61, −.22]). In light of these results, the following analyses focus on the effect of identifiability, comparing only the condition of identifiable recipient with the control condition. Mean support for the redistributive tax by scenario and condition is presented in Figure 1.

Figure 1.

Mean support for the tax by scenario and condition, in study 1

Excluding 40 respondents (8%) who wrote that they (or their relatives) benefit from such welfare transfers did not significantly change the results: ANOVA of tax support by condition and scenario yielded a significant effect of condition (F(1,270) = 4.679, p = .031, Partial Eta Squared = .017), and no other significant effects. Including participants’ self-estimated future need for these transfers as a covariate did not change the results either (F(1,309) = 8.583, p = .004, Partial Eta Squared = .027, for the effect of identifiability). Testing for the effects in each scenario separately, we find that identifiability was marginally significant in the Daycare scenario (F(1,164) = 3.220, p = .075, Partial Eta Squared = .019), and significant in the Income scenario (F(1,146) = 5.381, p = .022, Partial Eta Squared = .036).

A question related to support for the tax is that of support for the goal of the tax (e.g., “to what extent is funding income assurance allowances a worthy goal for the state?”). ANOVA of the response to this question yielded a higher support for funding daycare payments than funding income assurance allowances (F(1,310) = 4.948, p = .027, Partial Eta Squared = .016), but no other significant effects.

Participants’ answers to the emotion questions shed light on the source of the favorable IE. As expected, support for the tax was negatively correlated with anger (r = −.260, p < .001), and positively correlated with pity (r = .166, p < .001) and responsibility (r = .445, p < .001). Pity varied with condition. ANOVA of pity ratings by scenario and identifiability yielded a significant effect of condition (F(1,310) = 5.328, p = .022, Partial Eta Squared = .017), whereby pity was higher in the identified condition than in the control one (M = 4.07, SD = 1.797 and M = 4.52, SD = 1.789, for the control and the identified conditions, respectively). No other effects approached a significant level. Analyses of anger by identifiability and scenario yielded only a significant main effect of scenario (F(1,310) = 5.328, p = .022, Partial Eta Squared = .017), whereby anger was higher in the Income scenario than in the Daycare scenario (M = 2.43, SD = 1.548 and M = 2.07, SD = 1.508 for the Income and the Daycare scenarios, respectively). Analysis of responsibility ratings by the same predictors did not yield any significant effect.

As pity was significantly affected by condition, we tested whether the effect of condition on support for the tax is due to its effect on feelings of pity. Including pity as a covariate in the ANOVA of support by condition and scenario yielded a significant effect of pity (F(1,309) = 6.719, p = .010, Partial Eta squared = .021), as well as a significant effect of condition (F(1,309) = 6.614, p = .011, Partial Eta Squared = .021). Thus, the positive effect of identifiability on support for the tax cannot be fully accounted for by increased pity toward the recipients.

The Poverty Attribution questionnaire included 4 rating scales related to dispositional attribution, and 5 rating scales related to situational attribution (see Appendix). We computed separate mean agreement with dispositional statements (Cronbach Alpha = .772) and with situational statements (Cronbach Alpha = .709). The difference between the situational mean and the dispositional mean reflects the extent to which poverty is attributed to situational factors relative to dispositional factors, with higher values reflecting greater weight to the situational factors. Predicting support for the tax from the above difference between situational and dispositional poverty attribution, identifiability, and the interaction between the two factors yields only a significant effect of identifiability (B = .750, t = 3.311, p = .001). No other effects reached a significant level.

We examined the effects of five demographic attributes: age, gender, income, education, and political attitudes (liberal/conservative) across the two scenarios. None of the five attributes had a significant effect on support for the redistributive taxes. More importantly for the present research, the effect of mentioning an identified recipient did not interact with any of the demographic attributes. Income level (coded as a 5-point scale) marginally affected support for the taxes (greater support among higher-income respondents). The effect of identifiability, however, remained significant even when controlling for income level. ²⁵

In sum, Study 1 identified a factor that may enhance public support for redistribution, namely, identifiability: Minimal identification of a single recipient increased people’s support for redistributive taxes that would benefit the entire group of potential recipients. Furthermore, in terms of enhanced support, there was no statistically significant difference between a relatively popular measure (daycare funding), and one that encounters more resistance (income assurance allowances). In this respect, our hypothesis—based on the welfare literature—that there would be a difference between the two types of redistribution, was not corroborated. This result, however, is encouraging, because direct income transfers to the poor, although unpopular, are considered to be the most effective method to redistribute income.

3.2. Study 2

Study 2 aimed to probe deeper into the IE found in Study 1. Specifically, we examined whether people’s support for redistributive taxes could be further enhanced by employing a loss—rather than a gain—frame. According to prospect theory, individuals typically perceive outcomes as gains and losses, rather than final states of welfare, and losses loom larger than gains of similar value (Tversky & Kahneman, 1979, 1992). Consequently, the framing of a situation as either a gain or a loss may affect people’s reaction to it (Tversky & Kahneman, 1981; Zamir & Teichman, 2018, pp. 42–48). In the context of voluntary donations, for instance, it was found that a loss message (“your action can prevent a death”) is more likely to increase monetary contributions than a gain message (“your action can save a life”) (Bhati & Hansen, 2020, p. 6; Chou & Murnighan, 2013). Plausibly, a similar framing effect may manifest itself in the context of redistribution.

Since Study 1 found no statistically significant difference in the impact of identifiability on redistributive measures that vary in their public support, Study 2 focused on the less popular measure, namely, direct income transfers to the poor. It tested two factors that may influence people’s support for a tax to fund income assurance allowances: Identification of a recipient; and framing the consequences for the group of beneficiaries as either a gain or a loss. The basic structure of the questionnaires was similar to that of Study 1 (see Appendix). Participants first indicated the extent of their support for the tax, and then answered the additional questions. However, we added the emotion of “empathy” to the emotions of anger and pity employed in Study 1; and we phrased the responsibility question somewhat differently (“Please rate the extent to which you felt […] a sense of responsibility toward people receiving income assurance allowances”). Our first hypothesis was that identifiability of a single member of the recipient group would increase people’s support for the redistributive tax, compared to situations where the entire group remains anonymous. We further hypothesized that there would be greater willingness to support the tax when the situation is framed as a loss rather than when it is framed as a gain, and that framing will have a greater impact in the identified conditions. The study was preregistered. ²⁶

3.2.1. Participants

A total of 403 individuals were recruited by the Midgam survey company to participate in an online study. ²⁷ Participants in Study 1 were excluded from Study 2. The participants were a representative sample of the Israeli adult population in terms of their gender and age. Forty-one participants (10%) failed the attention check and were therefore removed from our sample. The analyses include the remaining 362 respondents.

3.2.2. Experimental Design

The study employed one scenario: a tax increase to fund income assurance allowances. There were two types of framing (loss/gain) and two types of identification (no mention/mention of an identified recipient). We used the same form of minimal identification for the identified recipient: First name, age, family situation, and city of residence. Thus, the study included four between-subject conditions (2 frames x 2 identifications), and participants were randomly assigned to one of them. The basic scenario read as follows:

Inequality in wealth in our society is immense. Some people have very low income, or no income at all. Among its various goals, the state aims to reduce income disparities and ensure a dignified existence for its residents. One way to achieve this goal is by providing the poorest individuals with a monthly payment (“income assurance allowance”). Assume that the current funding allocated for income assurance allowances is insufficient to achieve this goal, and therefore the state is considering increasing the income tax by 1% for all taxpayers. The additional revenue collected from this tax will be used to fund income assurance allowances.

Unidentified Conditions: If income assurance allowances are [Gain: provided / Loss: not provided] to poor people, they will consequently [Gain: enjoy a reasonable existence with dignity / Loss: suffer from a distressful existence without dignity].

Identified Recipient Conditions: One of those eligible to receive an income assurance allowance is David. David, who lives in Ashdod, is a 34-year-old widower raising his two young children alone. If income assurance allowances are [Gain: provided / Loss: not provided] to David and other similarly situated poor people, they will consequently [Gain: enjoy a reasonable existence with dignity / Loss: suffer from a distressful existence without dignity].

To what extent do you support increasing the income tax to fund income assurance allowances? Please indicate your answer on a scale from 1 to 7, where 1 signifies that you very strongly oppose increasing the tax rate, and 7 signifies that you very strongly support increasing the tax rate.

As in Study 1, after indicating the degree of their support for the redistributive tax on a 7-point Likert scale, participants were asked the additional questions about their emotions, the worthiness of the tax’s goal, the likelihood that they or their family will need the benefit (or already receive it), and their beliefs about the causes of poverty.

3.2.3. Results

The findings reveal that whereas identification increased support for the redistributive tax, the gain/loss framing did not affect support. ANOVA of tax support by identifiability and framing yielded a significant effect of identifiability (F(1,358) = 10.064, p = .002, Partial Eta Squared = .027). ²⁸ Neither the type of frame, nor the interaction, were significant (p = .999, and p = .220, respectively). ²⁹ Mean support for the redistributive tax by identification and framing is presented in Figure 2.

Figure 2.

Mean support for the tax by identifiability and framing, in study 2

As in Study 1, excluding the 42 participants (11%) who wrote that they (or their relatives) benefit from the welfare transfer, did not significantly change the results: ANOVA of tax support by identifiability and framing yielded a significant effect of identifiability (F(1,315) = 11.182, p < .001, Partial Eta Squared = .034), and no other significant effects. Including participants’ self-estimated future need for the transfer as a covariate did not change the results either, yielding again a significant effect of identifiability (F(1,357) = 9.466, p = .002, Partial Eta Squared = .026). Self-estimated future need had only a marginally significant effect on support for the tax (F(1,357) = 3.105, p = .079, Partial Eta Squared = .009). Overall support for the tax was correlated with support for the goal of the tax (r = .452, p < .001). However, the extent of support for the goal was not significantly affected by either identification or framing (p > .4 for each factor).

Participants’ answers to the emotion questions shed light on the source of the favorable IE. As expected, support for the tax was negatively correlated with anger (r = −.236, p < .001), and positively correlated with empathy (r = .350, p < .001) and responsibility (r = .519, p < .001). The correlation between pity and support for the tax was not significant (r = .086, p = .101), nor did it vary by condition (p > .2 for all factors in an ANOVA of pity by framing and identifiability). Empathy did not vary with condition. ANOVA of empathy ratings by framing and identifiability yielded no significant effects (p > .136 for all factors). Analyses of anger by identifiability and framing yielded only a marginally significant main effect of identifiability (F(1,358) = 3.779, p = .053, Partial Eta Squared = .010), whereby anger was higher in the unidentifiable conditions than in the identifiable ones (M = 2.39, SD = 1.518 and M = 2.09, SD = 1.383, respectively). Thus, identifying a certain member of the group somewhat reduced anger towards the group of welfare recipients. Analysis of responsibility ratings by the same predictors yielded a significant effect of identifiability (F(1,358) = 5.239, p = .023, Partial Eta Squared = .014), whereby responsibility was higher in the identifiable conditions than in the unidentifiable ones (M = 4.15, SD = 1.722 and M = 3.74, SD = 1.658, respectively). No other effects yielded a significant result.

As responsibility was significantly affected by identifiability, we tested whether the effect of identifiability on support for the tax is wholly due to its effect on feelings of responsibility. Including responsibility as a covariate in the ANOVA of support by identifiability and framing yielded a significant effect of responsibility (F(1,357) = 124.360 p < .001, Partial Eta Squared = .258), as well as a significant effect of identifiability (F(1,357) = 5.348, p = .021, Partial Eta Squared = .015). Thus, the positive effect of identifiability on support for the tax cannot be fully accounted for by increased sense of responsibility toward the recipients.

We analyzed the answers to the Poverty Attribution questionnaire as in Study 1. Predicting support for the tax from the difference between the means of situational and dispositional poverty attribution, identifiability, and the interaction between the two factors yields significant effects of identifiability (B = .697, t = 3.108, p = .002) and poverty attribution (B = .314, t = 4.023, p < .001), with no significant interaction between the two factors (p > .2). Thus, while both factors contribute to predicting tax support, they do so independently from one another.

Finally, we examined the effect of age, gender, income, education, and political attitudes (liberal/conservative), and found that none of them influenced the impact of identifiability on support for the redistributive tax. Testing for interactions with identifiability did not yield any significant effect. The only effect we found was a main effect of gender: Women tended to support the tax more than men (M = 3.61, SD = 1.861 and M = 3.18, SD = 1.806 for women and men, respectively, F(1,224) = 11.403, p < .001, Partial Eta Squared = .048, in ANOVA predicting tax support from gender, identifiability, and the interaction of the two factors). However, gender did not interact with the other factors. Thus, beyond the effect of gender, identifiability increased support for the tax.

In summary, Study 2 replicated the main results of Study 1 with respect to the IE’s ability to enhance support for redistributive taxes. Furthermore, Study 2 identified a new factor that accounts for this favorable influence: It appears that minimal identification of a single future beneficiary can heighten people’s sense of responsibility toward the group of welfare recipients.

3.3. Study 3

Study 3 aimed to address two potential concerns. The first relates to the identified person in the vignettes—a widower named David—who may be viewed as an atypical welfare recipient and as someone particularly likely to evoke sympathy. Accordingly, Study 3 employed a more “ordinary” type of beneficiary: A single mother identified by her first name. Such a person may be perceived as more responsible for her circumstances, as single parenthood can result from being unmarried or divorced—conditions that are often socially perceived as involving greater personal agency than widowhood. The second concern was that Studies 1 and 2 were conducted using a representative sample of the Israeli adult population. As discussed in the literature review, levels of public support for redistribution vary across societies. Therefore, Study 3 tested the IE in two additional countries, England (Study 3.a) and the United States (Study 3.b). ³⁰ Once again, we focused on the relatively unpopular welfare measure of direct income transfers to the poor (using the term “Universal Credit” in England and “Income Support” in the United States). We hypothesized that identifiability of one member of the recipient group would increase support for redistribution, compared to the control condition in which the entire group is anonymous. ³¹ The experiments were preregistered. ³²

3.3.1. Study 3.a

3.3.1.1. Participants

A total of 302 individuals were recruited via Prolific to participate in an online study. The participants were a representative sample of the British adult population in terms of their gender, age, and political affiliation. Nine participants (3%) failed the attention check and were therefore removed from the sample. The analyses include the remaining 293 respondents.

3.3.1.2. Experimental Design

The structure of the questionnaire was similar to that used in Studies 1 and 2 (see Appendix). Study 3.a used a single scenario involving a tax increase to fund Universal Credit payments. Employing a between-subjects design, the experiment included two conditions: an identified recipient condition and a control condition (with no mention of an identified recipient). After reading the basic scenario describing the government’s goal of reducing poverty and its consideration of raising taxes to fund Universal Credit payments, participants in the identified condition read the following text:

One of those eligible to receive Universal Credit payments is Emma. Emma is a 34-year-old single parent raising her two young children on her own. Despite working, she struggles to make ends meet. Emma and other people who will receive Universal Credit payments will consequently be able to maintain a dignified existence.

Participants first reported their support for the tax increase and then rated their feelings of empathy, pity, anger, and responsibility toward individuals receiving Universal Credit payments. Subsequently, participants in the identified condition were presented with an additional question assessing whether they perceived the identified person as a typical recipient of Universal Credit payments, on a scale from 1 (not typical at all) to 7 (very typical).

3.3.1.3. Results

As predicted, the findings reveal that identification increased support for the redistributive tax (M = 4.31, SD = 1.981 and M = 3.45, SD = 1.936, for the identified and control conditions, respectively; t(291) = 3.771, p < .001, CI[.412, 1.313]). Furthermore, the individual presented in the identified condition was judged by participants to be typical: the mean typicality rating was = 5.21, with a median of 5.

Excluding the 100 participants (34.1%) who wrote that they (or their relatives) have benefited from the welfare transfer did not significantly change the results: support for the tax in the identified condition was still significantly higher than in the control (M = 4.06, SD = 1.979 and M = 3.17, SD = 1.941, for the identified and control conditions, respectively; t(191) = 3.151, p = .002, CI[.333, 1.448]). Including participants’ self-estimated future need for the transfer as a covariate did not change the results either. As could be reasonably expected, self-estimated future need significantly increased support for the tax (F(1,290) = 13.902, p < .001, Partial Eta Squared = .046). Yet even controlling for future need, the effect of identifiability remained significant (F(1,290) = 14.341, p > .001, Partial Eta Squared = .047).

Examining participants’ answers to the emotion questions, we find, as expected, that support for the tax was negatively correlated with anger (r = −.530, p < .001), and positively correlated with empathy (r = .708, p < .001) and pity (r = .379, p = .011). Responsibility was also highly correlated with support for the tax (r = .740, p < .001). Identifiability significantly increased empathy towards the recipients of the payments (M = 4.82, SD = 1.708 and M = 4.11, SD = 1.703, for the identified and control conditions, respectively; t(291) = 3.539, p < .001, CI[.313, 1.098]). It similarly increased pity (M = 3.59, SD = 1.740 and M = 3.12, SD = 1.564, for the identified and control conditions, respectively; t(291) = 2.431, p = .016, CI[.089, .850]). Anger was marginally decreased by identifiability (M = 2.50, SD = 1.364 and M = 2.83, SD = 1.564, for the identified and control conditions, respectively; t(291) = 1.813, p = .094, CI[.089, .850]). Finally, the sense of responsibility marginally increased by identifiability (M = 3.76, SD = 1.947 and M = 3.37, SD = 1.780, for the identified and control conditions, respectively; t(291) = 1.813, p = .094, CI[-.034, .823]).

Does the effect of identifiability on empathy, pity, anger, and the responsibility felt for the recipients account for the observed effect on support for the tax? In order to answer this question, we included all four ratings as covariates in the ANOVA of support by identifiability. As expected, empathy, anger, and responsibility significantly contributed to prediction of tax support (F(1,287) = 15.155 p < .001, Partial Eta Squared = .050 for empathy; F(1,287) = 12.957, p < .001, Partial Eta Squared = .043 for anger; F(1,287) = 88.241, p < .001, Partial Eta Squared = .235 for responsibility), while pity did not have a significant contribution F(1,287) = .088, p = .766, Partial Eta Squared<.001). In addition, even when including these four ratings, identifiability remained significant (F(1,287) = 7.317, p = .007, Partial Eta Squared = .025). Thus, we find that the positive effect of identifiability on support for the tax cannot be fully accounted for by increased empathy and sense of responsibility, or by reduced anger, toward the recipients.

Finally, we examined the effect of age, gender, income, education and political attitudes (left wing/right wing) by testing, separately for each of these measures, a model predicting support for the tax from identifiability, the demographic measure, and the interaction of the two factors. Two demographic factors yielded significant main effects: political views and education level. Support for the tax increased, albeit marginally, with education level (B = .628, t = 1.958, p = .051) and increased significantly with left wing orientation (B = −.040, t = 6.490, p < .001). However, none of the demographic measures influenced the impact of identifiability on support for the redistributive tax. Testing for interactions with identifiability did not yield any significant effects.

3.3.2. Study 3.b

3.3.2.1. Participants

A total of 302 individuals were recruited via Prolific to participate in an online study. The participants were a representative sample of the American adult population in terms of their gender, age, and political affiliation. Thirteen participants (4%) failed the attention check and were therefore removed from the sample. The analyses include the remaining 289 respondents.

3.3.2.2. Experimental Design

The structure of the questionnaire was identical to that used in Study 3.a, with two minor adjustments for the American context: The term “Income Support” was used for the direct income transfer (instead of “Universal Credit”), and the identified woman’s name was “Jessica” (instead of “Emma”).

3.3.2.3. Results

As in the previous studies, the findings reveal that identification increased support for the redistributive tax (M = 4.78, SD = 2.081 and M = 4.22, SD = 2.130, for the identified and control conditions, respectively; t(278) = 2.247, p = .025, CI[.069, 1.045]). Furthermore, as in Study 3.a, the individual presented in the identified condition was judged by participants to be a typical welfare recipient: mean typicality judgment was = 5.22 on a scale of 1 to 7, with a median of 5.

Excluding the 50 participants (17.3%) who wrote that they (or their relatives) benefit from the welfare transfer, did not significantly change the results: support for the tax in the identified condition was still significantly higher than in the control (M = 4.73, SD = 2.121 and M = 4.15, SD = 2.163, for the identified and control conditions, respectively; t(237) = 2.075, p = .039, CI[.029, 1.123]). Including participants’ self-estimated future need for the transfer as a covariate did not change the results either. Self-estimated future need significantly increased support for the tax (F(1,286) = 37.257, p < .001, Partial Eta Squared = .115). Yet even controlling for future need, the effect of identifiability remained significant (F(1,286) = 4.904, p = .028, Partial Eta Squared = .017).

Examining participants’ answers to the emotion questions, we find, as expected, that support for the tax was negatively correlated with anger (r = −.415, p < .001), and positively correlated with empathy (r = .632, p < .001) and pity (r = .191, p = .001). Responsibility was also highly correlated with support for the tax (r = .673, p < .001). Identifiability significantly increased empathy for the welfare recipients (M = 5.34, SD = 1.553 and M = 4.82, SD = 1.809, for the identified and control conditions, respectively; t(287) = 2.599, p = .01, CI[.125, .908]). Although descriptively the pattern of responses to the other emotion questions was similar to the ones observed in Study 3.a, identifiability did not significantly affect pity (M = 3.62, SD = 1.898 and M = 3.61, SD = 1.792, for the identified and control conditions, respectively; t(287) = .049, p = .961, CI[.-.417, .438]), anger (M = 1.95, SD = 1.480 and M = 2.222, SD = 1.635, for the identified and control conditions, respectively; t(287) = 1.476, p = .141, CI[.-.633, .090]), or responsibility (M = 4.14, SD = 2.034 and M = 3.85, SD = 1.947, for the identified and control conditions, respectively; t(287) = 1.241, p = .216, CI[-.170, .751]).

In order to test whether the effect of identifiability on empathy, pity, anger, and the responsibility felt for the recipients, accounts for the observed effect on support for the tax, we included all four ratings as covariates in the ANOVA of support by identifiability. Each of the covariates significantly contributed to the prediction of tax support: Support for the tax increased with empathy (F(1,283) = 25.112, p < .001, Partial Eta Squared = .082), pity (F(1,283) = 4.293, p = .039, Partial Eta Squared = .015), and responsibility (F(1,283) = 7.500, p = .007, Partial Eta Squared = .026), and decreased with anger (F(1,283) = 7.500, p = .007, Partial Eta Squared = .026). When these four ratings are included, identifiability no longer has a significant effect (F(1,287) = 1.027, p = .312, Partial Eta Squared = .004). Thus, we find that in this case, the positive effect of identifiability on support for the tax can be fully accounted for by increased empathy, increased pity, a heightened sense of responsibility, and reduced anger toward the recipients.

Finally, we examined the effect of age, gender, income, education and political attitudes (liberal/conservative) by testing, separately for each of these measures, a model predicting support for the tax from identifiability, the demographic measure, and the interaction of the two factors. The only factor that yielded a significant main effect was political views. Support for the tax increased significantly with liberal orientation (B = −.029, t = 5.784, p < .001). Again, none of the demographic measures influenced the impact of identifiability on support for the redistributive tax. Testing for interactions with identifiability did not yield any significant effects.

In summary, Study 3 replicated the main findings of Studies 1 and 2 regarding the IE’s ability to enhance support for redistributive taxes, using a more typical identified beneficiary and samples from two additional countries. The next part discusses the results of all three studies, addresses key concerns and limitations, and outlines directions for future research.

4. Discussion: Findings, Limitations and Future Research

The three studies demonstrated how minimal identification of a welfare beneficiary can increase people’s support for redistributive taxes. In all studies, identification of a potential recipient had a favorable impact on the entire, anonymous group of recipients. Identifiability increased support for the welfare measures, despite the fact that it did not affect respondents’ beliefs about the worthiness of the state’s redistributive goal and the causes of poverty (dispositional vs. situational factors). The observed increase in support for the tax due to the IE was partially accounted for by the more positive emotions aroused with respect to the group when one of its members is identified: More pity (in Studies 1 and 3.a), a greater sense of responsibility toward welfare recipients (in Study 2), and more empathy (in Studies 3.a and 3.b). ³³ This is in line with the psychological literature, which largely attributes the IE to the stronger emotional reactions elicited by an identified person. Furthermore, identifiability increased support for the redistributive tax regardless of participants’ demographics (age, gender, income, education and political attitudes) and self-interest (actual or expected enjoyment of the benefit themselves). The fact that the IE increases support for redistribution across age, gender, income, education, political attitudes, and self-estimated future need makes it a particularly attractive persuasive tool. ³⁴ While the IE appears to be a robust phenomenon in the contexts we examined, and was manifest in all three studies, the other two factors we tested had no impact. Neither a reminder that individuals’ current financial state may deteriorate (in Study 1), nor framing the situation as a loss rather than a gain (in Study 2), had any effect on participants’ support for the redistributive taxes. But, of course, it is difficult to draw conclusions from a null result. Possibly, the manipulation of these factors was too subtle.

Although examining the link between identifiability, distinct emotions, and support for the tax does not provide completely consistent replications across the studies, similar patterns are revealed in all of them. Across conditions, anger decreases support for the tax, while empathy, pity, and responsibility increase that support. However, more interesting in the context of the present study, presenting an individual needy recipient affects pity, empathy, and responsibility, while hardly affecting anger. This pattern suggests, more generally, that the IE is associated with more positive and fewer negative emotions toward recipients, while the specific emotions may vary across studies.

These findings are in line with the longstanding distinction in the psychology of emotion that separates discrete emotions from broader dimensions of affect, most commonly described in terms of positive versus negative valence. Discrete-emotion approaches emphasize qualitatively distinct emotional states—including, among others, anger, pity, and empathy—that differ in their eliciting conditions, subjective experience, and action tendencies. By contrast, dimensional models of affect focus on evaluative states, typically organized around valence (positive or negative) and arousal (Ekman, 1992; Lerner et al., 2015; Mauss & Robinson, 2009). Measures of positive and negative affect often predict outcomes such as life satisfaction or general approach–avoidance tendencies. In the current study, we examined the effects of distinct emotions. However, the pattern of findings described above could suggest that the identifiable individual effect may be better accounted for by affective change related to the identifiability of a potential recipient. Future research is needed to obtain further insight into the underlying emotional mechanism.

From a theoretical perspective, our findings contribute to the IE literature, by showing that this phenomenon is wider in scope than previously thought. ³⁵ First, the IE is not limited to voluntary giving, but also extends to coercive giving by way of taxation. Second, it may enhance support not only for uncontroversial goals (such as donating to the gravely ill or victims of natural disasters), but also for more contentious goals, like redistribution of wealth in society. Third, an IE can arise not only when people are able to aid the specific, identified person. It appears that when assistance is to be provided for an entire, anonymous group—such as poor people in general—presenting a single member of the group can increase willingness to help the group, even if such help will not be targeted at the specific identified individual.

Besides their theoretical contributions, our findings may also have practical significance. The fact that the identified person does not need to be the one directly helped, but can instead serve as an illustrative example of a broader group of needy individuals, expands the potential usefulness of the IE as a tool for enhancing public support for social policies. It would be difficult for the state to claim, for example, that money raised from a particular taxpayer is earmarked for a specific individual’s needs. Fortunately, the findings suggest that such a direct connection between giving and receiving may not be necessary to evoke a favorable IE. Furthermore, Study 1 found no difference between a relatively popular redistributive measure (daycare funding to enable poor parents to work), and one that encounters more public resistance (income assurance allowances): With respect to both measures, minimal identification of a recipient increased people’s support for the tax. ³⁶ This too may have practical significance, because direct income transfers, despite being less popular, are commonly considered by scholars as the most efficient method to redistribute income. ³⁷ Therefore, policymakers of various types may wish to consider ways to reduce opposition to this form of redistribution. ³⁸

Politicians seeking to rally support for increased redistribution can, for example, employ the IE in their speeches, press releases, or other forms of public communication. In contrast to President Ronald Reagan, who popularized the negative image of a stylized, generic “Welfare Queen” (Baekgaard et al., 2023, p. 594), politicians may instead minimally identify a real, prospective welfare recipient who is deserving of assistance. In other contexts, politicians have occasionally used real individuals’ full names and personal stories to garner support for their initiatives. For example, to secure passage of the Lilly Ledbetter Fair Pay Act, president Barack Obama cited the true case of a woman who sued her employer for pay discrimination. ³⁹ Our findings demonstrate the efficacy of minimal identification and suggest that public support may be enhanced even without fully revealing the identity or life story of specific individuals.

Governmental agencies may also employ the IE to strengthen support for redistributive measures. In addition to their primary responsibility of administrating and implementing programs, such agencies sometimes seek to persuade the public about the programs’ merits in order to increase public acceptance and compliance. A well-known example is the Disney animated short film The New Spirit (1942), ⁴⁰ which starred Donald Duck and was produced specifically for the United States Department of the Treasury and the Internal Revenue Service (IRS). The film was part of a broad public information campaign promoting the expansion of the income tax and highlighted the critical role of Americans’ tax compliance in funding the weapons needed to win the war against the Axis Powers (Jones, 1988, pp. 699–719). Viewed by millions, the film influenced Americans’ willingness to pay their taxes promptly (Jones, 1988, pp. 716–717).

Even in non-emergency contexts, agencies sometimes produce short films, video clips and public announcements that are posted on agency websites, shared on platforms like YouTube, or broadcast on television. These communications often highlight the benefits people derive from agencies’ activities. ⁴¹ An example is a short video featuring an IRS commissioner, explaining the importance of paying taxes by referencing the public services funded by tax revenues, including Social Security and Medicare. ⁴² Similarly, the Israeli National Insurance Institute, which administers several types of welfare payments, has occasionally produced short educative and promotional films. These include depictions of a (fictional) employer who failed to pay social security contributions for her domestic worker and is therefore required to cover the worker’s losses after she is injured in a car accident; ⁴³ as well as a person recounting his real-life-story of rehabilitation following a severe workplace injury, made possible by the Institute’s assistance. ⁴⁴ The Israel Tax Authority has sought to encourage tax compliance and the reporting of tax evasion (through the so-called “Justice Line” initiative), not only by posting information on its website but also through short films and public service announcements. These materials visually and vividly explain—using actors—how tax evasion leads to the “disappearance” of medicines from hospitals and computers from schools, ⁴⁵ or imposes a heavier burden on those who do pay their taxes. ⁴⁶

Governmental agencies thus engage in persuasion campaigns in order to increase public acceptance and compliance. In this context, we suggest that agencies also consider employing the milder IE, particularly in the context of redistribution. Evoking an IE requires only minimal identification of ordinary people who are neither fictional nor celebrity figures.

Finally, the IE may be valuable in journalistic reporting aimed at building public support for redistribution. The possibility of minimal identification is important, because the complete anonymity of welfare recipients can detrimentally affect news reporting on poverty and redistribution. Generally speaking, news reporting tends to be personified, focusing on individuals rather than abstract social issues, based on the assumption that people are more interested in, and better able to identify with, other people (Harcup & O’neill, 2001, pp. 269–274; Ostfeld & Mutz, 2014, pp. 53–54). Moreover, news stories are often framed episodically rather than thematically. Episodic framing conveys information through the presentation of a specific event or case, whereas thematic framing presents information in a more general manner and situates an issue within a broader context (Ostfeld & Mutz, 2014, pp. 54–57). Thus, to be considered “newsworthy” and thereby inform public opinion about redistribution, reporting cannot rely solely on statistical data or expert commentary on poverty. Instead, it should also highlight the case of a particular potential beneficiary. The IE makes this possible.

There may be additional advantages to employing the IE in the sphere of redistributive taxes. First, our studies succeeded in evoking an IE with minimal information about the identified individual—merely first name, age, and family situation. Contrary to other identifiability experiments and charitable appeals, we did not use a photo, ⁴⁷ family name, or other information that invades a person’s privacy or might cause shame and stigmatization. ⁴⁸ Perhaps similar, limited identification would suffice to arouse an IE in real-life. Second, employing the IE on behalf of the poor is plausibly a cost-effective measure. For example, the cost of producing a very short film, video clip or public service announcement, or of adding a few words or sentences to a government website, a press release, a billboard, or notices on broadcast and communication platforms, is relatively low compared to other methods proposed in the literature for increasing support for taxation. ⁴⁹ While the effects uncovered in our studies were generally small, if they truly represent the population’s preferences, these small effects would produce meaningful shifts in public support for the policies we studied. Third, the fact that identifiability can increase individuals’ sense of responsibility toward welfare beneficiaries (in Study 2) may prove important. Since the group of taxpayers is extremely large, this factor is liable to diminish the wealthier classes’ perceived responsibility vis-à-vis the recipients. Indeed, with respect to donations and volunteer work, Garcia et al. (2002) showed that simply thinking about a group (without actually being in a group of potential helpers), caused people to feel less responsible and reduced their willingness to help others. The IE may attenuate such diffusion of responsibility, thereby enhancing taxpayers’ acceptance of redistributive measures. ⁵⁰ Evoking more empathic emotions towards the poor is likely essential, because the welfare literature shows that people’s basic opinions about redistribution are unlikely to change, even in the face of information about inequality and poverty. ⁵¹ In our experiments as well, identifiability did not affect respondents’ opinions about the causes of poverty or the worthiness of the state’s redistributive goals. Rather, enhanced support for the taxes in the identified conditions was partially accounted for by the more positive emotions toward welfare recipients. Finally, the results suggest that the IE can have a favorable impact beyond individuals’ demographic characteristics or self-interest. Since people differ in their basic preferences for redistribution, ⁵² the potential for the IE to have a broad positive influence supports testing its use in real-life.

Various concerns and limitations of our findings need to be addressed. Arguably, using a psychological phenomenon to increase people’s support for redistribution is manipulative. To the extent that the IE is driven by respondents’ emotional reactions to an identified person, it may be based on their intuitive and automatic System-1 thinking, rather than on their deliberative and conscious System-2 thinking. ⁵³ In response, we note that it is far from clear that judgments and decisions made under the latter mode of thinking are necessarily superior to those made under the former mode (Zamir & Teichman, 2018, p. 23). It is possible that the interests of the anonymous group of beneficiaries are unduly discounted, and that emotions stirred by minimal identification of a single recipient correct this bias and lead to more appropriate decisions (Lewinsohn-Zamir et al., 2017, pp. 534–536; see also Rachlinski, 2005, pp. 549–550). ⁵⁴ This is indeed likely in the context of redistribution. As explained in the literature review (Part 2.), scholars acknowledge that the current level of redistribution in society is too low, and that providing people with additional data about inequality and poverty is unlikely to change their beliefs, which are often based on stereotypes, biases, and misconceptions about the deservingness of the poor. For example, contrary to common belief, many individuals with full-time jobs still live below the poverty line (Ehrenreich, 2001; Herbst, 2013, p. 134; Krumer-Nevo, 2020, p. 84). In general, people tend to downplay the many personally uncontrollable causes of poverty (Piff et al., 2020, pp. 496–497). Insofar as a higher level of redistribution is deemed appropriate, it is legitimate to explore new ways to increase taxpayers’ support for redistributive measures, including by fostering positive feelings toward welfare recipients. Note that the identifiability technique does not involve deception, as the selected beneficiary can be a real person, who is eligible to receive the redistributive transfer. Our experiments also demonstrate that there is no need to use shocking or disturbing images—such as a photograph of a starving, crying child. An IE can emerge from minimal factual information about a potential beneficiary, without visuals or detailed data about their personal characteristics and history.

A related concern is that the identified person in the vignettes was relatively deserving, rather than a “challenging” type of beneficiary. The vignettes did not depict someone likely to arouse antipathy in the respondents, but they also did not present an ideal example of deservingness, such as an elderly or disabled person clearly unable to work (see Meuleman et al., 2020, p. 3). In Studies 1 and 2, participants received no information about the work efforts of the 34-year-old widower, and could not assume that he was blameless for his situation. Studies 3.a and 3.b similarly found an IE with a more typical and less sympathetic identified beneficiary—a single mother. Participants did not know whether she had chosen not to marry or was divorced, and whether she works full time or is less diligent. Note, that the aim of our article is to examine whether the IE manifests in the context of redistribution, thereby enhancing public support for this social goal. We do not seek to identify the boundaries of a positive IE. Nonetheless, we acknowledge that the impact of identifiability may not be universal but could depend on the characteristics of the identified individual. In other words, it remains unclear whether more controversial beneficiaries would evoke a positive IE.

Another possible concern is that our studies were conducted with representative samples of the Israeli, British and American adult population. As discussed in the literature review, countries differ in their citizen’s support for redistribution. Although Israel is highly polarized along dimensions such as nationality, ethnicity, religion, political ideology, and class (Harel et al., 2020, p. 54; Yaish, 2001, pp. 410, 414, 434), its citizens may nonetheless share a relatively strong sense of social solidarity. The replication of our findings in England and the United States suggests that the IE is not unique to the Israeli context, while leaving open the possibility that identifiability may operate differently in other national settings. Note, that as people everywhere seem to care about welfare recipients’ deservingness (Aarøe & Bang Petersen, 2014), to the extent that beliefs on this issue are based on stereotypes, biases, or misconceptions, the IE may attenuate their negative impact on support for redistribution.

A different concern is the potential gap between individuals’ responses to questionnaires and their behavior in the real world. Although this limitation—common to vignette studies—cannot be ruled out, it should not be overstated. Our experiments focused on attitudes towards redistributive taxes (rather than self-reported behavior), which makes them less susceptible to this concern than studies that ask participants how they would act in real-world situations. This approach is consistent with public opinion surveys, which routinely ask individuals to express their attitudes, opinions and preferences on a wide range of issues without requiring them to incur any financial (or other) cost. Such responses are taken into account by policymakers (alongside other relevant data), as indicators of public opinion and levels of public support (Burstein, 2003; Page & Shapiro, 1983; Shapiro, 2011; Soroka & Wlezien, 2005). In a similar vein, we believe that our findings provide useful evidence on how public support for redistributive measures can be enhanced in practice. Furthermore, because minimal identification of a welfare beneficiary is a relatively inexpensive intervention that does not harm the identified individual, there is little to lose from testing its effects in real-world settings.

There are several avenues for further research. The primary aim of our studies was to introduce a new, relatively simple method that could increase support for redistributive programs. It would be worthwhile to explore the duration of a favorable IE in real-life. Would the state’s “identifiability intervention” have only a short-term impact on people’s support for redistributive transfers? And if repeated interventions are necessary, would their effectiveness diminish over time as people become accustomed to them? Note, that even if recurrent governmental action is required, it may still prove cost-effective compared to alternative methods. Future research should also examine the IE in relation to other types of redistributive measures (e.g., in-kind rather than monetary transfers), and explore different types of identified recipients (e.g., a single man without children or an immigrant). Another promising avenue would be to compare the impact of identifiability across additional countries with varying cultures and welfare regimes.

5. Conclusion

It is widely acknowledged that welfare programs cannot be implemented or sustained over time in the face of unsympathetic public opinion. As a result, insufficient redistribution occurs, and inequality not only persists but also increases. To date, various proposals aimed at boosting public support have not been sufficiently successful. This article identifies a relatively simple measure that has not been previously tested in the context of redistribution. Our findings indicate that the identifiability effect may be harnessed to increase support for redistributive taxes, thereby contributing to the social goals of reducing income disparities and ensuring a dignified existence for the poor.

In his book The Needs of Strangers (1984, pp. 9–18), Ignatieff observes that our responsibilities toward the poor are mediated through numerous governmental agents, which isolates us from one another and “ensures that those with resources and those in need remain strangers” (id., p. 17). Consequently, the welfare state may provide people with what they need in order to survive, but not with what they need to flourish—such as solidarity and respectful treatment from others. This seems especially true of redistributive taxes. Revenues collected from any taxpayer pass through several third parties and only reach a recipient long afterwards; the taxpayer may live far from the beneficiary; and the parties involved in the coerced transfer belong to different social classes. It is, therefore, unsurprising that redistributive taxes are often perceived as a form of charity—other people’s money being transferred to possibly undeserving beneficiaries. Identifiability, however, evokes empathic emotions and strengthens our sense of responsibility toward welfare recipients. Simply put, it makes them less of strangers to us. We encourage policymakers to experiment with the IE for this reason as well.