1. Introduction
Women face significant barriers in seeking elected office. This is particularly true for the case of Japan, where, as shown in Figure 1, female legislators make up less than 10% of the politicians in the Lower House due to continuing institutional, structural, and social barriers that prevent equal participation (Kage et al., Reference Kage, Rosenbluth and Tanaka2019).
Figure 1. Democracy and Women in Legislatures, 2022. Democracy data are from the V-Dem dataset released in March 2023 at https://v-dem.net/data/the-v-dem-dataset/, and women in legislature data are from (Inter-Parliamentary Union, 2022). Dots represent countries, and ‘JPN’ shows it is the percent of women in the House of Representative (
$9.9$
%) and Liberal Democracy Index (
$.74$
).
Gendered stereotypes and expectation reward male politicians for discussing certain topics while punishing female candidates for discussing those same topics. It is well known that voters often associate conventional gender roles with particular sets of policy issues, such as national defense, foreign policy, and economy being labelled as ‘male issues’ and education, social welfare and environmental policy being labelled as ‘female issues’.
Although conventional wisdom on how voters’ views are gendered is well established for most policy areas, immigration policy is an exception. Little is known about how these stereotypes function in policy support in the immigration policy area and how voters respond to a more ‘feminine’ or ‘masculine’ framing of immigration. This gap of knowledge is particularly important because immigration is one of the most salient policy issues in many advanced industrialized countries. Whether male or female, electoral candidates are asked about their platform on immigration, and their answers are featured prominently in the media. We therefore ask, how do voters respond when male and female candidates take stances on immigration policies in ways that invoke or challenge conventional gender expectations?
To answer this question, we fielded a survey in Japan that included an experimental vignette where respondents were presented with a hypothetical candidate who is affiliated with a traditionally anti-immigration party but supports expanding immigration. We manipulated the gender of the candidate and the gendered framing of the position and examined their interaction effects on attitudes to the candidate.
As a political issue, immigration has increasingly become more salient. In the past, immigration was largely an issue that lacked salience in much of the advanced industrialized world because as Freeman argues, although publics tend to oppose immigration, the costs of immigration are diffuse, but the benefits were concentrated (Freeman, Reference Freeman1995), so political parties did not see a strategic advantage in focusing on immigration. However, in recent years, as demonstrated by the justification of Brexit with reference to a desire to reduce immigration in the United Kingdom, immigration policy in the United States targeting select groups of immigrants, and the emergence and success of anti-immigration parties in several Western European countries, there has been an increasing politicization of and turn against immigration in many advanced industrialized countries. This has made it increasingly important for political candidates to address and respond to this issue.
The case of Japan presents a great opportunity to explore this question because the country faces demographic pressures stemming from one of the lowest birth rates and one of the longest life expectancies in the world. This crisis has made it increasingly important to include and push historically marginalized people into the workforce, including women and immigrants, despite the pushback both groups have faced (Strausz, Reference Strausz2019; Steel, Reference Steel2019). By including both groups in our analysis, we contribute on immigration policy and gender politics and expand our understanding of how these two groups relate to each other when it comes to people’s attitudes towards them. In addition to our contribution to multiple literatures, our work has important policy implications because it expands our understanding of how female candidates are rewarded or punished for how they communicate their policy positions in a country that has historically struggled to elect more women into political office.
2. Gender bias and immigration
2.1. Gender bias and electoral politics
Female candidates running for office face a myriad of obstacles, including institutional and cultural factors (Eto, Reference Eto2010), bias in the pitch of their voice (Bowern et al., Reference Bowern, Kage, Rosenbluth and Tanaka2023), and socially mandated gender roles that make balancing work and family life difficult (Kage et al., Reference Kage, Rosenbluth and Tanaka2019).Footnote 1 On top of these, barriers are also differing perceptions and expectations that voters have about female and male candidates, namely gender stereotypes. A large literature in the Western context has found that male candidates are perceived to be better at handling political issues coded as ‘male’ (including crime and foreign policy), while women are perceived to be better equipped at addressing issues coded as ‘female’ including education, social security, and disability rights (e.g. Rosenwasser and Seale, Reference Rosenwasser and Seale1988; Sanbonmatsu, Reference Sanbonmatsu2002). American voters are more likely to elect women to city councils and school boards, but more likely to elect men as mayors, a finding consistent with conventional gendered stereotypes about male and female domains of expertise (Anzia and Bernhard, Reference Anzia and Bernhard2022). In addition, voters are more likely to perceive women as more progressive than men (McDermott, Reference McDermott1997) and more likely to evaluate candidates who fit traditional gender stereotypes as being more competent in their respective gender policy areas (Huddy and Terkildsen, Reference Huddy and Terkildsen1993).
In particular, Huddy and Terkildsen conduct a survey experiment that tests whether gender stereotyping in the US context stems from ‘traits’ or ‘beliefs’ that voters associate with genders. An example of the former is that women are viewed as more warm and expressive, and men are viewed as tougher and more aggressive, whereas the latter is that women are thought to be more progressive than men (Huddy and Terkildsen, Reference Huddy and Terkildsen1993, pp. 120–121). Overall, Huddy and Terkildsen find more support for the traits explanation than the beliefs explanation, which suggests that women might be able to overcome gender stereotyping by consciously exhibiting male traits. However, regardless of the traits that their fictional candidates exhibited, ‘(p)articipants in our study were not entirely blind to our fictitious candidate’s gender. The female candidate was seen as more competent on compassion and women’s issues; the male candidate had the edge on military issues’ (Huddy and Terkildsen, Reference Huddy and Terkildsen1993, p. 132).
Similarly, Cassese and Holman examine the vulnerability of male and female candidates to attacks based on both traits and beliefs conventionally associated with their genders. They find that female candidates are particularly vulnerable to trait-based attacks, and especially attacks that focus on female traits. They also find that, while both men and women are vulnerable to belief attacks, ‘female candidates are more susceptible to attacks on a stereotypically female policy issue’ (Cassese and Holman, Reference Cassese and Holman2018, p. 803).Footnote 2
Moreover, gender seems to interact with ideology and partisanship in shaping how voters perceive candidates. In the US context, scholars have found that candidate gender stereotypes differ by ideology of the candidate, with conservative female candidates being more likely to be assessed as competent in traditionally ‘male’ areas compared to their progressive counterparts (Holman et al., Reference Holman, Merolla and Zechmeister2016; Holman et al., Reference Holman, Merolla and Zechmeister2011).
Interestingly, even in a context where women are extensively represented, gender appears to interact with ideology and partisanship in shaping candidate perception. In a study of secondary school students in Norway, a country with extremely high representation of women, Matland (Reference Matland1994, p. 280) finds that when presented with a fictional candidate speech, Labour party supporters are more likely to say that they would vote for that candidate when told the candidate was female while Conservative party supporters are more likely to say that they would vote for that candidate when told that the candidate was male. However, both Labour and Conservative party supporters believe that the male candidate is more competent in policy areas traditionally coded as male, including defense, and less competent in policy areas traditionally coded as female, including women’s rights (Matland, Reference Matland1994, pp. 281 and 284).Footnote 3
Scholars have found similar patterns in Japan as well. In a survey of 3,000 Japanese voters, Endo and Ono (Reference Endo and Ono2021) find remarkable similarities between gendered policy and trait stereotypes held by Japanese voters and American voters. Moreover, in a conjoint experiment comparing voters’ reaction to pairs of hypothetical candidates, Ono and Yamada (Reference Ono and Yamada2020, p. 487) find that ‘female candidates are neither rewarded nor punished for deviating from gender-based expectations in terms of personality traits and ideological positions,’ but that female candidates are ‘rewarded when they show interest and expertise in policy areas that are congruent with a feminine image’ and are punished when they show interest and expertise in incongruent issues. Similarly to US-based studies, Ono and Yamada identify national defense, foreign policy, crime, and the economy as ‘male issues,’ and include education, social welfare, and the environment as ‘female’ issues.
Qualitative research also supports the claim that female politicians in Japan – both from progressive parties and from the conservative Liberal Democratic Party (LDP) – often frame their political activities with reference to both gendered policy and trait stereotypes (LeBlanc, Reference LeBlanc1999).
2.2. The politics of immigration in Japan
Although Japan has a relatively small population of foreign residents, there are major economic and demographic pressures on Japan to greatly increase immigration. Due to among the lowest birth rates and longest life expectancies in the world, Japan is faced with labour shortages that have proven difficult to address in the absence of immigration (Strausz, Reference Strausz2019, pp. 7-9).
These demographic and economic challenges caused by Japan’s aging and shrinking population have pressed increasing numbers of politicians to discuss immigration. Even Shinzo Abe, Japan’s longest serving Prime Minister who was famously opposed to immigration, was persuaded by his Chief Cabinet Secretary, Yoshihide Suga, to advocate a revision to the Immigration Control and Refugee Recognition Act in 2018 which created a new visa category, the ‘Specified Skills Visa’ which would permit Japan to admit foreigners manual labourers in areas including construction, care work, and agriculture, as labourers (rather than as co-ethnics, trainees, or students) for the first time in post-war Japanese history (Harris, Reference Harris2020, p. 316).
Abe regularly justified the policy changes associated with this new visa category as ‘not an immigration policy,’ because, at least in the short term, most holders of these new visas would not be permitted to bring their families and would have to leave after a fixed period of time (Strausz, Reference Strausz2021). Perhaps, partially because of this justification, in the first election after this policy change, the plurality of House of Councillor candidates, 46.8%, agreed that Japan should promote the admission of foreign labour. This was the largest percentage of House of Councillors candidates supporting increasing foreign labour that the Asahi-University of Tokyo survey had registered since it began asking this question in 2010 (Strausz, Reference Strausz and Strausz2025).
Despite this increase, there is some evidence that more politicians are in favour of foreign labour than will publicly admit to being so. During the 2019 House of Councillors election, those politicians who were not up for election were more supportive of foreign labour than those who were even when controlling for gender, age, and partisanship (Strausz, Reference Strausz and Strausz2025). This suggests that at least some politicians continue to fear taking a position on immigration just before an election.
Of course, immigration is not simply an economic issue. People immigrate for a wide variety of reasons, including humanitarian ones. Japan’s refugee policy has come under increasing criticism both inside and outside of Japan for the strictness of the criteria that it uses to determine who gets asylum. Most famously, in 2017, 19,629 people applied for asylum in Japan, but only 20 were granted it (Strausz, Reference Strausz2019, p. 93). Moreover, one of the most reported stories relating to immigration to Japan in recent years has been the story of Wishma Sandamali, a Sri Lankan woman who had come to Japan as a student but quit school because of her inability to pay tuition and ended up overstaying her visa. Ms. Sandamali died in immigration detention in March 2021 after being repeatedly denied provisional release because of medical concerns. She was detained in August of 2020 after calling the police to report an incident of domestic violence (Kishitsu and Itō, Reference Kishitsu and Itō2021).
In sum, both the economic side and humanitarian side of immigration have become politically salient in recent years. Variation in xenophobia among the public is likely to become an important determinant of how politicians position themselves on the issue of immigration.
Despite what is often assumed about Japanese public opinion, given the relatively small proportion of foreign residents in Japan, the Japanese public is not particularly anti-immigration. In a 2018 Pew Survey, 25.57% of Japanese respondents said that Japan should admit more immigrants, which was the third most of the 27 countries studied, behind only the United States and Spain. About 59.18% of Japanese respondents said that Japan should keep the same number of immigrants, which was the largest percentage studied, followed by South Korea and Canada (Pew Research Center, 2018. Analysis by authors).
Kage et al. (Reference Kage, Rosenbluth and Tanaka2022) identify four different sets of attitudes about immigration among the Japanese public: integrators, who appreciate immigration for economic and cultural reasons; diversifiers, who appreciate immigration for cultural but not economic reasons; exploiters, who appreciate immigration for economic but not cultural reasons; and resistors, who are opposed to immigration. They also find that about 60% of the Japanese public supports immigration for some combination of economic and cultural reasons. Moreover, they find that education, job type, age, contact with foreign friends, and right-wing authoritarian predisposition affect whether people support immigration.
Although xenophobia does not appear to be the majority sentiment among the Japanese public, several recent studies conducted in other industrialized democracies do suggest that xenophobes, although the minority, might be relatively more motivated to vote on their prejudices than would non-xenophobes. Phoenix (Reference Phoenix2019) argues that, in the American context, white Americans expect the system to operate in ways that benefit whites (because they are used to things operating in that way), and when it does not, they get angry, while Black Americans do not get as angry when the system works to benefit whites because that is what they have come to expect. In addition to learning from historical examples to lower their expectations, Phoenix (Reference Phoenix2019, p. 246) also argues that Black Americans in the US have learned that, for them, ‘to express anger is to take on the risk of being stigmatized, ridiculed, or to open oneself up to receiving severe pushback’ and that ‘there is a long history of black expressions of dissent—even those that are purely rhetorical or being conducted with strictly non-violent principles—provoking intense surveillance and aggressive responses from state actors.’ This finding is supported by Kustov (Reference Kustov2023), who shows that in the US and the majority of the European countries that he examined, the issue of immigration is more important to the anti-immigrant public than the pro-immigrant public.
Previous studies have not provided evidence on the generalizability of these findings to the Japanese context. These findings suggest that, because those with traditional views on gender and national identity in Japan expect society to act in a way that conforms to their expectations about how society should function, they will get angry and politically active when those expectations are violated. On the other hand, supporters of progressive views on gender and national identity in Japan have no such expectations and are thus less likely to become angry and politically activated when they see politicians and policies that they disagree with. We examine whether this is the case in Japan by analysing heterogeneity among the Japanese public in their attitudes on traditional values.
2.3. Gendering immigration
Migration is a gendered issue. The reason why women immigrate is impacted and shaped by gender inequalities such as gender-based violence and limited access to resources (UN Women, nd). Gendered dynamics impact female immigrants differently and more adversely, as Gomez-Aguinaga et al. (Reference Gomez-Aguinaga, Dominguez and Manzano2021), for example, find in their study on Latina immigrants in the United States during the COVID-19 pandemic. Aside from the differing impacts on immigrants themselves, immigration policy broadly is an arena where gendered structures get reinforced.
Writing about the regulation of marriage immigration in Norway – which in many other contexts, including in Japan or the United States, is heavily regulated as well – Eggebø (Reference Eggebø2010) explains that how a crucial component of regulating income levels of marriage visa applicants is to shift potential economic dependence from the state to the spouse who is marrying an immigrant. However, with this comes the economic dependence of immigrant spouses on their local partners, which tend to be female and male, respectively. This immigration rule hence reinforces traditional gender norms making women dependent on men and putting them in vulnerable positions. Many countries have immigration policies targeted towards ‘high-skilled’ immigrants. While these types of immigrants are viewed positively by many, especially compared to more ‘low skilled immigrants’ (see Hainmueller and Hiscox, Reference Hainmueller and Hiscox2010), policies specifically targeted towards skill-based immigrants produce gendered outcomes. Research on European countries, for example, has found that high-skilled immigrant visas have an overrepresentation of men (Kofman, Reference Kofman2014), showing that this type of policy may reward male advancement in workplace and educational settings that women do not have the same access to. On the other hand, home healthcare workers, which is a visa category that was expanded in Japan, are a female-dominated profession that, as Davison and Peng (Reference Davison and Peng2021) find, is an immigrant category that some Japanese people viewed much more favourably than other types of immigrants. While this does not show that there is a consistent pattern between preferences over male or female immigrants, it does show how immigration policy gets framed matters in attitudes towards them, and how this framing is not gender neutral. The above examples allude to more favourable viewing of immigrants when they fit into gender stereotypical categories, with high-skilled visas, more often taken by men and fitting the gender stereotype of men in managerial and other higher skilled positions, receiving a more positive assessment, and care givers, which are more often visas received by women and a profession often viewed as a ‘female profession,’ receiving a more positive evaluation.
Immigration policy is written in contexts that are shaped by gender hierarchies and norms and is consequently seldom created gender neutrally, even when the language of the policy seemingly is. The facts that immigration policy itself is gendered and that women are rewarded by using gender stereotypical language (Biernat et al., Reference Biernat, Tocci and Williams2012) are important considerations when trying to understand how public attitudes towards immigration are shaped.
2.4. Gender, immigration, and framing
To make sense of how different wording of the same political issue can lead to different assessments of voters, we rely on the concept of framing from the political communication literature (Berinsky and Kinder, Reference Berinsky and Kinder2006; Cacciatore et al., Reference Cacciatore, Scheufele and Iyengar2016). A frame, according to Gamson and Modigiliani (Reference Gamson and Modigiliani1989), is ‘a central organizing idea for making sense of relevant events, suggesting what is at issue.’ One can discuss a set of events (or even facts) using different frames in order to change the implications that the speaker wants the listener to get from hearing about those events. For example, Berinsky and Kinder (Reference Berinsky and Kinder2006) show how the framing of crises impacts whether or not politicians will pay attention, and Iyengar (Reference Iyengar1990, Reference Iyengar1996) shows how issue framing influences to whom people attribute responsibility for poverty or terrorism. In the Japanese context, Horiuchi and Ono (Reference Horiuchi and Ono2023) find that when people have positive interactions with outgroups, they are less susceptible to frames that portray refugees as a threat. Essentially, the way the information is presented affects how people view and what kind of judgments they form about it. In addition, other previous work on Japan and immigration has found that how immigration is discussed, in terms of long- or short-term immigrants and immigrants for specific work sectors, does affect attitudes on whether immigration is supported or not (Davison and Peng, Reference Davison and Peng2021).
How do Japanese voters respond when political candidates frame the same issue in a manner consistent with female or male stereotyped traits? As much previous research suggests, will they punish candidates for framing issues in a way inconsistent with gendered trait expectations and/or reward candidates for frames that are consistent with those expectations?
Immigration is a policy area with major implications for economics, demographics, humanitarianism, culture, and national identity, and it is thus an ideal issue area in which to test the importance of framing. Moreover, LDP politicians have adopted both economic and humanitarian frames to discuss immigration in the past. For example, the 2008 report by the 80-member LDP Diet Members Caucus for the Promotion of Foreign Human Resources used both kinds of frames to call for expanded immigration to Japan (Strausz, Reference Strausz2019, p. 125). Although the LDP is conservative and therefore observers tend to believe that it is open to immigration only for economic reasons, the LDP is also known to have politicians with a wide variety of policy and ideological positions. Among the LDP members in the House of Councillors as of 2024, their positions on immigration range from an explicitly xenophobic one such as Shōji Nishida from Kyoto Prefecture to much more humanitarian concerns about immigration such as Dai Shimamura from Kanagawa Prefecture (Strausz Reference Strausz and Strausz2025).
We argue that using economic frames to justify calls for immigration is consistent with stereotypes about masculinity, while using humanitarian frames is consistent with feminine stereotypes. Because, in Japan and elsewhere, female political candidates are punished by voters for contravening traditional gender expectations, we hypothesize that:
Hypothesis 1a: The public will view female candidates who use humanitarian frames to discuss immigration as more competent than female candidates who use economic frames to discuss immigration.
Hypothesis 1b: The public will view the policies of female candidates who use humanitarian frames to discuss immigration as more effective than female candidates who use economic frames to discuss immigration.
Hypothesis 1c: The public will view female candidates who use economic frames to discuss immigration as less competent than male candidates who use economic frames to discuss immigration.
Hypothesis 1d: The public will view the policies of female candidates who use economic frames to discuss immigration as less effective than male candidates who use economic frames to discuss immigration.
Moreover, we do not expect these effects to be uniform. We hypothesize that people who are xenophobic, people with a primarily ethnic conception of Japanese national identity, and people who are sexist will react more strongly when they see political candidates whose gender does not match the gender expectations attached to the frame that they are using. We expect to see this because we expect that those with traditional views about gender and national identity have learned to expect society to work in their favour and get angry when it does not, while those with progressive views about gender and national identity have no such expectations and are less likely to have these kinds of expectations. Thus, we hypothesize that the effects in Hypothesis 1a, b, c, and d will be bigger for those with traditional views about gender and national identity.
More specifically, we hypothesize:
Hypothesis 2a: the effects in hypothesis 1 (parts a-d) will be larger for people who are xenophobic.
Hypothesis 2b: the effects in hypothesis 1 (parts a-d) will be larger for people with an ethnic, rather than a civic, vision of Japan’s national identity.
Hypothesis 2c: the effects in hypothesis 1 (parts a-d) will be larger for people who are sexist.
Finally, given that women are often viewed as and expected to be compassionate, we hypothesize the following:
Hypothesis 3a: Female candidates who use humanitarian frames to discuss immigration will be viewed as more compassionate than female candidates who use economic frames to discuss immigration
Hypothesis 3b: Female candidates who use humanitarian frames to discuss immigration will be viewed as more compassionate than male candidates who use humanitarian frames to discuss immigration
3. Research Design
To test those hypotheses, we conducted a survey experiment in Japan of 6,033 people where each respondent evaluates a randomly assigned hypothetical electoral candidate. The respondents were recruited through a survey vendor, Cross Marketing Inc., from 28 January to 3 February, 2022. We employed quota sampling by age (19 or above) and gender to match the national census distribution as much as possible. Supplementary Information (SI) A provides an English translation of the questionnaire, and SI B is the original questionnaire in Japanese.
Our
$2 \times 2$
factorial design manipulated both the gender of the hypothetical candidate and the gendered framing of the proposal to expand immigration. The hypothetical candidate shown to a respondent was either male or female and proposed to increase immigration to Japan either for economic or humanitarian reasons. The gender of the candidate and the frame that the candidate uses to discuss immigration are the treatment variables that were randomly assigned in our experiment. We chose these different framing and reasoning because as the literature discussed above has shown, policy areas are gendered.
Although economic policy could, in theory, be discussed in ways consistent with female gender stereotypes, mainstream economic discussion frequently draws on abstract reasoning in a manner that is consistent with male stereotypes. As we discuss below, our economic treatment in particular is consistent with stereotypes about so-called male reasoning. Similarly, while humanitarian policy objectives could be discussed in a way that is consistent with male stereotypes, mainstream discussion of humanitarian policies frequently appeals to compassion and other traits that are consistent with female stereotypes (and that is also true about our humanitarian treatment, which you can find below).
Table 1 shows the wording of each vignette. In the economic treatment, the hypothetical candidate reasoned that immigration should be increased to address labour shortages in Japan because of the decrease in the Japanese population. We interpret this treatment as the framing that is consistent with the male stereotype, since it invokes abstract macroeconomic considerations. In the humanitarian treatment, the hypothetical candidate reasoned that immigration should be increased to help people who have suffered natural disasters, poverty, or war. Contrary to the previous treatment, the humanitarian treatment appeals to the traits of kindness and compassion, and thus it is aligned with female stereotypes.
Table 1. Vignette that respondents were shown. They were either shown the labour treatment with either a male or female politician or they were shown the humanitarian treatment with either a male or female politician
These treatments were block-randomized, where the blocks were party identification (LDP supporters, supporters for the other parties, independents) and the respondent’s gender. The treatments had equal probability within each block. Block randomization reduces the sampling variance of estimated treatment effects when the mean of the outcome variable varies across blocks. Since we measure the respondent’s attitudes towards a hypothetical LDP politician, we expect the outcome depends on the respondent’s party identification.Footnote 4 Moreover, we expect that the respondent’s gender is associated with her evaluation of the candidate, as we present the hypothetical candidate’s gender in the vignette.
As the hypotheses presented in Section 2.4 specify, we are interested in the gender bias against electoral candidates when they propose to expand immigration, which is expected to be unpopular, particularly among those who have a prejudice against women.
For outcome variables, we follow Cassese and Holman (Reference Cassese and Holman2018), p. 795 in measuring candidate evaluation by looking at ‘trait evaluations, policy competence, and vote choice’ (see also Holman et al., Reference Holman, Merolla and Zechmeister2011). Given the similarity of gender stereotypes in Japan and the United States (Endo and Ono, Reference Endo and Ono2021), we expect these measures developed in the US context to be useful in the Japanese context as well.Footnote 5
We focus our measure of trait evaluation on compassion, since compassion is a gendered trait that is clearly invoked by the humanitarian treatment. We asked, ‘do you think that this kind of candidate is compassionate, or not compassionate?’ and we let respondents choose one of seven options ranging from ‘very compassionate’ to ‘not at all compassionate.’
To measure policy competence, we asked ‘Do you think that the policies that this person supports would be good policies, or bad policies,’ and we let respondents choose one of the seven options ranging from ‘very good policies’ to ‘very bad policies.’ This is a particularly effective measure because it asks about the policies that the candidate advocates and not the candidate himself or herself, thus encouraging the respondent to focus on policy rather than personal characteristics.
Finally, to measure vote choice, we asked, ‘Do you think that this kind of politician would be a competent politician or not?’ We let respondents choose one of the seven options ranging from ‘very competent’ to ‘not at all competent.’
The average treatment effects of our interest are estimated by the difference-in-means estimator under block randomization, that is, the weighted average of the difference-in-means estimators across the blocks with the number of observations in each block as a weight. Formally, let
$j$
index the block,
$i$
index the respondent within each block,
${T_{ij}}$
denote the treatment variable, and
${Y_{ij}}$
denote the outcome variable. The average treatment effect is estimated as
$$\hat \tau = \mathop \sum \limits_{j = 1}^J {{{N_j}} \over N}\left\{ {{{\mathop \sum \nolimits_{i = 1}^{{N_j}} {T_{ij}}{Y_{ij}}} \over {\mathop \sum \nolimits_{i = 1}^{{N_j}} {T_{ij}}}} - {{\mathop \sum \nolimits_{i = 1}^{{N_j}} \left( {1 - {T_{ij}}} \right){Y_{ij}}} \over {\mathop \sum \nolimits_{i = 1}^{{N_j}} \left( {1 - {T_{ij}}} \right)}}} \right\}$$
Heterogeneous treatment effects for Hypotheses 2a–c are estimated by the identical estimator within the strata of the moderating variables of interest.
The sampling variance of
$\hat \tau $
is estimated as the sum of the sampling variance of the difference-in-means within each block multiplied by the square of the proportion of the block. We conduct two-sided hypothesis tests with the
$.05$
significance level for our main analysis, and use the adaptive shrinkage method (Stephens, Reference Stephens2017) and the Benjamini-Hochberg method (Benjamini and Hochberg, Reference Benjamini and Hochberg1995) for multiple testing correction as robustness checks. We also compute and show the statistical power of the estimator to examine whether the results are due to underpowered hypothesis tests.
For the hypotheses about heterogeneous treatment effects (Hypotheses 2a, 2b, and 2c), we construct xenophobia, ethnic conception of national identity, and gender stereotype scores using factor analysis and run linear regressions with the interaction terms of the scores and the treatment variables. We measured xenophobia, ethnic conception of national identity, and gender stereotype attitudes before the experimental items.
The gender stereotype score and the xenophobia score for each respondent are constructed by applying the factor analysis to a battery of survey items to measure the respondent’s attitudes in the two areas. For the gender stereotype score, the manifesto survey questions are how strongly the respondent agrees or disagrees with each of the following statements: (1) ‘Men should be prioritized over women when available jobs are scarce.’ (2) ‘A married couple always get into a trouble when the wife earns more than the husband.’ (3) ‘In general, men are better as political leaders than women.’ (4) ‘Higher education is more important for men than for women.’ (5) ‘Married couples should be legally allowed to use separate surnames.’ (6) ‘Even though a couple is legally married, they do not have to have children.’ (7) ‘No matter what, women’s happiness lies in marriage.’ (8) ‘Preschool children will suffer if their mother has a job.’ (9) ‘Husbands should work while wives should look after their home and family.’ (10) ‘No matter what, men’s happiness lies in marriage.’ A higher value of the gender stereotype score indicates greater gender bias. These items are commonly used to measure gender prejudice in Japan in surveys such as the Japanese General Social Surveys.Footnote 6
For the xenophobia score, the manifesto survey items are: (1) ‘Do you think Japanese culture will improve if more foreigners come from abroad? Or do you think it will regress?’ (2) ‘Do you think Japan’s economic situation will improve if more foreigners come from abroad? Or do you think it will deteriorate?’ (3) ‘Do you think Japan’s safety will improve if more foreigners come from abroad? Or do you think it will deteriorate?’ (4) ‘Do you think Japan’s social security system, including medical care, long-term care, and pensions, will improve if the number of foreigners coming from abroad increases? Or do you think it will deteriorate?’ In each survey item, respondents are asked to choose from a five-point ordered scale. A lower score means stronger xenophobic attitudes. We chose these questions because they isolate the economic and cultural reasons that individuals might support or oppose immigration. Those are the dimensions that Kage et al. (Reference Kage, Rosenbluth and Tanaka2022) and others examine when looking at approval of and opposition to immigration.
For the ethnic conception of national identity score, respondents were presented with a table of specific items, all of which they were asked, ‘Different people have different standards for what constitutes a “Japanese person.” What kinds of elements are important to you in deciding whether someone is “a Japanese person?”’ with the order being randomized. The items are: (1) Being born in Japan (2) Having Japanese citizenship (3) Having Japanese ancestors (4) Having Japanese parents (5) Respecting Japanese culture (6) Living according to Japanese customs (7) Having Japanese values (8) Speaking Japanese well (9) Loving the emperor (10) Contributing to Japanese society.
We base this scale on the one that Lindstam et al. (Reference Lindstam, Mader and Schoen2021) used in their study of ethic and civic nationalism in Germany. We kept items such as ‘to have Japanese ancestors,’ dropped items that were less important in the Japanese context including ‘to have a Christian worldview’ and ‘to have democratic convictions,’ and added items specifically relevant to the Japanese context, including ‘loving the Emperor.’
Similarly to the scores above, respondents were asked to respond using a five-point scale, and had the option between important, somewhat important, I can’t say either way, somewhat unimportant, and unimportant. Respondents were also given the option to not answer. Aside from the above mentioned table, the ethnic conception of national identity battery also included two more questions that had a four-point scale: (1) ‘Do you think that a person who has gotten Japanese citizenship as an adult is just as Japanese as someone who was born with Japanese citizenship?’ (2) ‘Do you think that someone who is born as a foreign citizen in Japan but raised in the same way as the Japanese people around them is actually the same as a Japanese person?’ Both questions had the following answers as options: I think there is a difference, I think there is somewhat of a difference, I think they are generally the same, and I think they are the same.
We scored the battery as civic nationalism if respondents had broader, more inclusive understandings of what it means to be Japanese. For example, if respondents answered that having Japanese ancestors is not important to being Japanese, they were considered as having a civic sense of nationalism. On the other hand, if their understanding of Japaneseness was more restrictive such as believing that respecting Japanese culture is important to being Japanese, these respondents were scored as having an ethnic-sense of nationalism.
The study is determined to be exempt from review by the Institutional Review Boards at the University of New Mexico (IRB#: 18921), Texas Christian University (IRB#: 2021-256), and the University of Michigan (Study ID: HUM00209175). The hypotheses and research design are preregistered at the EGAP registry at https://osf.io/gf8pz.
4. Results
As noted in the previous section, our experiment was fielded in Japan from 28 January to 3 February, 2022, with 6,033 respondents. We randomly assigned the respondents to one of the four conditions (female/male candidate with economic/humanitarian justification) within each block of the respondents’ gender and party identification (LDP or not). The covariate balance check (shown in SI C) shows that the p-values of the
$t$
-tests for the mean difference between each pair of the treatment groups are distributed uniformly as expected under no difference. Our outcome measurements are the respondents’ evaluations of the candidates’ compassion and competence, and the effectiveness of the proposed policy (expanding immigration into Japan).
Overall, our empirical analysis does not support the hypotheses that gender stereotypes influence candidate evaluations. On the evaluation of either the candidate’s competence or the policy’s effectiveness (Hypotheses 1a through 1d), no statistically significant differences were found for all comparisons. These null findings are not because heterogeneous treatment effects cancel out each other, but because null effects are estimated for all subgroups defined by Hypotheses 2a through 2c. The respondents’ evaluations of the candidate’s compassion are affected by the framing of the policy as expected in Hypothesis 3a, but not affected by the candidate’s gender contrary to Hypothesis 3b. Our experiment suggests that Japanese voters’ gender bias in the immigration policy context is not as strong as in other policy areas.Footnote 7
Figure 2 shows our results for Hypotheses 1a-d regarding the treatment effects on the candidate’s competence and the effectiveness of the proposed policy. The filled squares and circles in the plot represent estimated treatment effects (
$\hat \tau $
), while the squares and circles with crosses are the point estimates after correction by the adaptive shrinkage method. The vertical bars are the 95% confidence intervals.
Figure 2. Estimated Average Treatment Effects on the Competence and Effectiveness Outcome Variables (Hypotheses 1a-d), Uncorrected (filled symbols) and Corrected (symbols with crosses) for Multiple Hypothesis Testing.
The results shown on the left in Figure 2 suggest that how to frame the policy to increase immigration does not affect the respondents’ evaluations of the candidate’s competence or the policy’s effectiveness. Although the effect of the humanitarian framing on the competence outcome (H1a) is estimated to be marginally significant, both the adaptive shrinkage method and the Benjamini-Hochberg procedure indicate that this estimate is a false positive. Moreover, there is no difference in voter assessment of the immigration policy between the economic justification and the humanitarian justification when those policies are advocated by a female candidate. In fact, additional results shown in SI D suggest that the framing has no effect on both outcomes for a male candidate, too.
Similarly to the policy framing, the candidate’s gender does not have effects on either the competence or effectiveness outcome, when the candidate uses the masculine economic justification for expanding immigration. The results presented on the right of Figure 2 show that both estimates are close to zero and not statistically significant, even without multiple testing correction. Neither Hypotheses 1c nor 1d is supported by our data.
The power analysis shown in Figure 3 suggests that these null findings are not due to a lack of statistical power. For any of the comparisons between treatment groups, the experimental design has sufficient power to detect an effect size of 0.2, which is small in the seven-point scale we used to measure the outcomes.
Figure 3. Power Analysis for the Estimated Average Treatment Effects on the Competence and Effectiveness Outcome Variables Shown in Figure 2. Vertical dashed lines represent the estimated effect sizes.
The null findings on the average treatment effects do not necessarily imply that there are no heterogeneous treatment effects. It is possible that the treatment effects are different for subgroups of the respondents and that these differences cancel each other out to be reduced to the null effect when averaging across the subgroups. However, our analysis of heterogeneous treatment effects does not support this possibility either.
Figure 4 displays estimated treatment effects conditional on the gender stereotype, xenophobia, and ethno-nationalism scores (Hypotheses 2a–c). Each panel from the top to the bottom shows the results in a similar way to Figure 2, except that the two point estimates for each comparison are for the subgroups of the respondents with above and below the median of the respective score. For example, on the left of the top panel under the label ‘Competence,’ the filled square represents the estimated treatment effect of the humanitarian framing relative to the economic justification used by a female candidate on the competence outcome for the respondents with the gender stereotype score above the median, while the open square represents the estimated treatment effect for the respondents with the score below the median.
Figure 4. Estimated Treatment Effects Conditional on the Gender Stereotype, Xenophobia, and Ethno-nationalism Scores. Filled symbols represent the estimated effects for those above the median of the respective score, while open symbols represent the estimated effects for those below the median.
Across all the moderating variables, we do not find statistically significant differences between the subgroups. Respondents with weak gender bias are very similar to those with strong gender bias. This is a striking negative finding because it suggests that even those with stronger than typical gender bias do not reward political candidates for conforming to gender stereotypes or punish candidates for contradicting gender stereotypes.
More xenophobic respondents tend to be negatively affected by the humanitarian framing while evaluating a male candidate lower than a female candidate with the economic justification for increasing immigration, which is in contradiction to Hypothesis 2a. Since these differences are not statistically significant, we conclude that our analysis does not support our hypothesis, but that no evidence for the opposite direction is found, either. People with stronger and weaker ethno-nationalism are both similar to the sample as a whole. Interpreted together, our findings suggest that neither xenophobia nor ethnic nationalism seems to impact the way that the Japanese public evaluates political candidates’ adherence to gender stereotypes when discussing immigration.
Finally, Figure 5 shows the results for Hypotheses 3a and 3b regarding the treatment effects on the candidate’s compassion. The respondents viewed the female candidate who used the humanitarian framing as more compassionate than the female candidate using the economic justification for immigration (Hypothesis 3a). This result is upheld by the adaptive shrinkage method and the Benjamini-Hochberg procedure. However, it is not particularly surprising that voters view those using humanitarian justifications as more compassionate than those using economic justifications, regardless of gender. In fact, the respondents similarly viewed the male candidate using the humanitarian framing as more compassionate than the male candidate using the economic justification, too (see Figure D.1 of SI D). Interestingly, though, and contrary to what the scholarly literature on voter gender stereotypes suggests, the respondents treated the female candidate and the male candidate as having indistinguishable compassion when using the humanitarian justification (on the left of Figure 5). The respondents also treated the female and male candidates using the economic justification as having indistinguishable compassion (also see Figure D.1 of SI D). This means that while the framing of the policy issue affected whether candidates were evaluated as compassionate, the gender of the candidate did not, which contradicts Hypothesis 3b.
Figure 5. Estimated Average Treatment Effects on the Compassion Outcome Variable (Hypotheses 3a-b).
In sum, contrary to what existing literature led us to believe, the candidate’s gender or the framing of immigration policy does not have uniform effects on respondents’ ‘trait evaluations, policy competence, and vote choice’ (Cassese and Holman, Reference Cassese and Holman2018, p. 795). We do not find support for the hypotheses that candidates get rewarded for framing immigration policy in a way that conforms to their respective gender stereotypes. Moreover, we observe minimal to no heterogeneity by the gender bias and xenophobia of the respondents.
5. Concluding remarks
In this paper, we asked the question of how Japanese voters respond when female and male political candidates frame the same issue in a manner consistent with female or male stereotyped traits. We hypothesized that, similar to what past research has found on other policy issues in Japan and other contexts, when female candidates frame the issue of immigration as a more female-coded issue – which we labelled our ‘humanitarian’ frame – they would be rewarded by voters, while male candidates would be rewarded for framing the issue using a ‘labor’ frame. We fielded a survey in Japan in 2022 that included a survey experiment where we manipulated both the gender of a hypothetical LDP politician and the framing of immigration. In all vignettes, the candidate advocated for increasing the number of immigrants either because it was the right thing to do (i.e., the humanitarian frame) or because of the labor shortages in Japan leading to the necessity of increasing immigrants (i.e., the labor frame). Our experiment shows that there is no evidence for gender bias in voters’ evaluations on immigration policies and electoral candidates.
These negative findings contradict the claims of an impressive body of scholarship on voter evaluations of politicians, which suggests that voters reward female candidates for behavior and language that is consistent with gender stereotypes, and punish female candidates for behavior and language that is inconsistent with female candidates. On the other hand, survey experiments about candidate selection typically find that voters express greater voting intentions towards younger, female candidates than to elderly, male candidates (e.g. Eshima and Smith, Reference Eshima and Smith2022; McClean and Ono, Reference McClean and Ono2022; Schwarz and Coppock, Reference Schwarz and Coppock2021; Teele et al., Reference Teele, Kalla and Rosenbluth2018). Our experimental results are at least not in contradiction to this other strand of literature on voting behaviour, suggesting that gender bias in voter evaluations may not be as severe as the literature expects in some policy areas. This insight may point to a direction of future research, which needs to reconcile the discrepancy between evaluations and voting intentions.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1468109924000215
Open access
