When the Japanese government surrendered to the United States at the end of World War II (1945), it left behind a lasting legacy in East Asia. During the War, as the Imperial Army advanced first through China, then through Southeast Asia, and finally into British India, it committed large-scale violence. In one of the more infamous episodes, the Nanjing Massacre, the Japanese took control of the key Chinese city. Over the next six weeks, they looted, raped, and murdered anywhere between 50,000 and 300,000 (Fogel Reference Fogel2000; Honda and Sandness Reference Honda and Sandness2015). In another episode, the Sook Ching Massacre, the Japanese rounded up the ethnic Chinese in Singapore and massacred them to purge the island of enemy sympathizers. Over 70,000 Chinese were summarily executed in the two-week period (Blackburn Reference Blackburn2000; Lee Reference Lee1998). By no means were the ethnic Chinese always the target of the violence. In some cases, they were made complicit in these atrocities. As the war dragged on, the Japanese addressed manpower shortages by conscripting over 200,000 able-bodied men into the army from Taiwan (Kleeman Reference Kleeman2003; Lamley Reference Lamley and Rubinstein1999)—a Japanese colony captured in the First Sino-Japanese War (1895).
Conventional wisdom would suggest strong negative attitudes stemming from a legacy of historical hatred (e.g., Huntington Reference Huntington1996; Tir and Singh Reference Tir and Singh2015). In 2005, there were a series of anti-Japanese protests throughout East Asia—from the Koreas (Jager Reference Jager2005) to Taiwan (Valencia Reference Valencia2006; Weiss Reference Weiss2014), from Mainland China (Cody Reference Cody2005; Watts Reference Watts2005) to the Philippines (Conde Reference Conde2005). In 2012, a series of demonstrations erupted across the Greater China—inclusive of Hong Kong, Macau, and Taiwan—over the disputed Diayou/Senkaku islands (Bradsher, Fackler, and Jacobs Reference Bradsher, Fackler and Jacobs2012). And in 2015, Mainland China commemorated the seventieth anniversary of the Japanese defeat by hosting a 12,000-strong military parade (BBC News 2015).
These anti-Japanese sentiments are widespread. When asked about Japan's influence in their country, over 60 percent of respondents in Mainland China said either “rather bad” or outright “bad” (see Figure 1). While the same figures are much lower in Singapore and Taiwan, it bears mentioning that one out of every four respondents in Singapore and one out of every three in Taiwan remain ambivalent. But at the same time, we also see nontrivial—if not outright sizable—populations of people who see Japan in a positive light. These attitudes have been matched by booming tourismFootnote 1 and student exchangeFootnote 2 numbers between Japan and each of the three political units. In this article we ask: What explains public attitudes in Mainland China, Singapore, and Taiwan towards a former aggressor state—i.e., Japan?
Attitudes towards Japan
We contend that proficiency in a third-party's language—a lingua franca (see Ostler Reference Ostler2006)—is one channel for tempering preexisting negative attitudes. Languages are more than just vehicles of communication; they transmit the values and norms of a distinct culture (Sapir Reference Sapir1929, 207–214). And so when an individual is proficient in a lingua franca—especially one unencumbered by the history of the two states, i.e., English in East Asia (see Liu Reference Liu2015; Ricento Reference Ricento2015)—they have a different channel for understanding how the world works (Whorf Reference Whorf1940). In this alternative framework, individuals in victim states can consume positive narratives put forth by other parties—e.g., the United States and other western democracies—about the former aggressor state.
In this article, we examine public attitudes towards Japan in the Chinese-speaking world—i.e., Mainland China, Singapore, and Taiwan. We focus on this set of cases for three reasons. First, Japan's military history in the twentieth century renders it empirically an important case. It was ambitious in geographic scope: the Japanese Imperial Army stretched from Japan to India—with ambitions beyond the Indian Ocean to lands as far as Madagascar (Rigge Reference Rigge1980; Thomas Reference Thomas1996). It was also noteworthy for its bloodiness. Aside from battlefield deaths and civilian atrocities committed, it was also the target of the two nuclear bombs.
Second, Mainland China, Singapore, and Taiwan constitute the full population of political units in the Chinese-speaking world. This allows for the systematic testing of cultural clash theories between distinct “civilizations” (Huntington Reference Huntington1996). On the one hand, in all three political units, ethnic Hans are the majority—91.5 percent in Mainland China, 74.1 percent in Singapore, and 95 percent in Taiwan—and Mandarin Chinese is an official language of government services, public education, and mass media. On the other hand, however, the nature of the Japanese occupation varied substantially across the three political units (more below).
The third reason for case selection has to do with language. In the context of Japan and the Chinese-speaking world, English serves as a lingua franca. It is not the mother tongue—the ancestral language—for any sizable subset of the population. Yet it is a compulsory subject in schools. Given the high enrollment rates in all three political units, a priori we have little reason to believe selection effect is serious concern. We do, however, test for this later in the article. We find that while learning opportunities may vary at the household and school levels, the results remain robust when we control for individual and regional characteristics. This in turn allows us to examine the effects of English proficiency on attitudes towards the Japanese.
Of course, being proficient in the Japanese language is another useful—if not obvious—avenue for people to consume alternative narratives about Japan. There is certainly a non-trivial population of Japanese-speakers in all three political units. Moreover, there is a growing trend in the number of new people studying Japanese each year. In fact, Mainland China and Taiwan rank second and fifth in the world for the number of people learning the language (Japan Foundation Reference Foundation2007). Yet, there are two empirical concerns with focusing on Japanese proficiency. The first is one of relativity. Compared to English, Japanese is still a minor foreign language for most of the Chinese-speaking world. This is the case both in terms of the number of people learning language and the positions in the formal education (Japan Foundation Reference Foundation2007; Wei and Su Reference Wei and Su2012). In fact, a few years ago Japanese dropped to second behind Korean as the second foreign language in the Taiwanese education curriculum. The second empirical concern has to do with selection effect. Individuals who are predisposed to liking Japan are probably more likely to study the language—and vice versa. And since the Japanese language is not a mandatory language of study in any part of the Chinese-speaking world, exposure to and proficiency in Japanese is not randomly distributed. Given these concerns, in this article we focus on the effects of English proficiency—although the theoretical implications are certainly not confined to the one language when generalizing beyond the Chinese-speaking world.
This research proceeds as follows. We begin by discussing the Japanese military occupation in the Chinese-speaking world, demonstrating why this historical legacy remains easy political fodder and how the perceived aggression was far from uniform. We then shift to a theoretical discussion, offering a language-based explanation for public attitudes towards Japan. We posit that when individuals know a foreign language—specifically, a lingua franca—they can access an alternative set of narratives about the Japanese that is on average more positive than the ones put forth domestically and in their mother tongues. The third section introduces the research design, followed by a discussion of the empirical results. We show the attitudinal effects of English are robust even when we consider whether some individuals are predisposed to being cosmopolitan; whether some individuals have more opportunities to learn English; and whether the linguistic effects are symptomatic of American soft power. The article concludes by highlighting the policy implications and the generalizability beyond the Chinese-speaking world.
THE JAPANESE IN THE CHINESE-SPEAKING WORLD
In this section we briefly discuss the Japanese military occupation in the Chinese-speaking world. We do so for two reasons. First, negative attitudes do not appear randomly without cause. Instead, they are often shaped by cunning political elites looking to mobilize the masses (Snyder Reference Snyder2000). But to achieve this objective, there must be some issue on which the masses can direct their focus. What the Japanese did during World War II—and what they have allegedly not done since (e.g., sincerely apologize)—are easy political fodder. If our theoretical argument rests on monolinguals being able to access only the narratives in the Chinese language, it is imperative that we establish why biased, anti-Japanese narratives exist. Second, while the Japanese military campaign was noted for its general aggression—inspiring both awe and fear among its subjects (Lee Reference Lee1998)—it is important to recognize this aggression was far from uniform. There was substantial variation in how the Japanese treated the ethnic Han Chinese—both spatially and temporally. In fact, at one point, in the Taiwan case, the Chinese were considered adequately “Japanese” (more below). This discussion will set the foundation for our model specification, i.e., multilevel modeling with three levels and our choices of control variables.
The rise of the Great Japanese Empire dates to 1868. In its first major war, Japan defeated the Chinese Qing government (First Sino-Japanese War 1894–1885). The victory gave Japan its first foreign colony: Taiwan. The subsequent Japanese rule in Taiwan went through three stages: from colonization (1895–1915) to assimilation (1915–1937; Doka) to Japanization (1937–1945; Kominka). While the Taiwanese were initially considered as biologically inferior—if not outright inhuman—they were later seen as good Japanese subjects with the privilege of dying for the emperor (Ching Reference Ching2001).
With Germany's defeat in World War I, Japan was given German-controlled territories–courtesy of alliances with the British (1902 Anglo-Japanese Alliance): Qingdao (China) and Samoa (Pacific). In 1931, Japan invaded Manchuria, thereby setting in motion the Second Sino-Japanese War (1937–1945). During the war, Japan and Mainland China would engage in open conflict. In the first years, the Japanese captured Shanghai, Nanjing, and Wuhan. The fighting was particularly brutal in Nanjing. For six weeks, the Japanese Imperial Army inflicted large-scale atrocities on the Chinese population. As many as 300,000 Chinese were killed in the massacre (Fogel Reference Fogel2000; Honda and Sandness Reference Honda and Sandness2015)—an event that remains contentious in contemporary China–Japan relations. As the fighting moved inland, the Japanese were stretched too thin: The resulting stalemate in Changsha, Guanxi, and Shaanxi marked a turning tide. With the American involvement in the Pacific Front, Japan was unable to force a Chinese surrender. By October 1945, 10 to 25 million Chinese civilians—along with another four million military personnel from the two sides—died from the war (Paine Reference Paine2012).
In 1942, the Japanese Imperial Army marched into Southeast Asia. As they advanced, people in the Malayan Peninsula—the colonial British, the local Malays, and the immigrant Chinese and Indians—all fled south to Singapore. Despite being an island, Singapore's water source came from across the peninsula. Once the Japanese blew up the causeway, panic ensued in Singapore, and the British immediately surrendered—an event that Churchill would call the “worst disaster” in British military history (Churchill Reference Churchill1959/2002, 518). Upon taking Singapore in 1942, the Japanese purged those deemed anti-Japanese—and in this case, this meant the ethnic Chinese. Using an established network of informers, as many as 70,000 Chinese were arrested and killed. In one particular incident, the Sook Ching Massacre, the Japanese rounded up fifty trucks full of Chinese men. The six thousand men were taken out to a beach where they were tied together and ordered to march out to sea. And as they marched, they were gunned down and then bayoneted (Lee Reference Lee1998). Japanese presence in Singapore formally ended in Singapore on September 12, 1945.
Since World War II, Japan has taken steps to atone for its past transgressions both regionally in East Asia and globally. But whether these acts of penance are enough for a Han Chinese to forgive and forget depends on whether they have a channel for learning about these conciliatory behaviors. In this article, we argue this opportunity to learn manifests when individuals are proficient in a foreign language, specifically, a lingua franca. In the next section, we elucidate the mechanism linking foreign language proficiency and public attitudes towards former aggressor states.
ENGLISH PROFICIENCY AND ATTITUDES TOWARDS THE JAPANESE
Languages are vehicles of communication. They allow individuals to convey a message. For example, consider that you are a server at a restaurant. A patron says to you, “My cousin and I are thirsty. We would like two glasses of ice water.” While what the patron says may seem straightforward, at the same time, the two sentences also convey linguistic markers that are unique to the culture. If the scenario had taken place in the Chinese language, the patron would not have been able to simply say “cousin.” “Wo he biaoge” would not be correct—unless the cousin was older and from the patron's mother's side of the family. If the male cousin were younger—but still from the mother's side, he would be a biaodi; but if he were from the patron's father's side, a tangdi. In short, phrases such as “Wo he biaoge” and “Wo he biadi” depict two different meanings. If the exchange in the restaurant had taken place in Spanish, a patron would say primo if the cousin were male; otherwise, they would have to say prima. If the scenario had taken place in Indonesian, the patron could not say “Kita ingin” for “We want.” If they did, it would suggest that the two glasses of water are to be shared between the cousin, the patron, and the server. Kita is an inclusive third-person plural pronoun. In this situation, the proper pronoun would be the exclusive third-person plural pronoun: kami. And finally, if the restaurant were situated somewhere in northern Finland, Sweden, or Norway where Saami is spoken, the request for ice water better differentiate between áhttán (ice on salt water), čoddi (coating of ice formed by frozen rain on stones or trees), rovda (weak ice that cannot bear the weight of reindeers), sáisa (packed ice pressed up against the shore), and spildi (a layer of thin ice on top of the water)—amongst many more (Magga Reference Magga2006).
Each of these four languages is embedded with a linguistic feature that reflects something larger about the culture. In Chinese, the family and relations of each individual to the others is important. In Spanish, gender demarcates boundaries—and not only with animate objects but inanimate as well. In Indonesian, the default is on the group as inclusive. But the exclusive “we” is used by speakers when there is need to create distance from those in the out-group, e.g., when the Indonesian nationalists were proclaiming their independence against the colonial Dutch. And in Saami, the physical environment manifests in the vocabulary.
Whether they are about family, gender, group boundaries, or ice permutations, new words are created, and their addition to everyday parlance in any language can be inherently political. In Korean, for example, the word “capitalist” came to mean something negative per Marxist-Leninist principles in the north. Likewise, words such as “workers’ party,” “people's army,” and “people's liberation war” have been codified (Kaplan and Baldauf Reference Kaplan and Baldauf2003, 41). This differentiation is certainly not exogenous. When governments—or a society at large—have a narrative to tell, words can be created to perpetuate the facts. These facts can be either positive or negative—for example, the origins of an ethnic group, their associations to some sacred land since time immemorial, the contributions to world civilization, or the aggression of a neighboring state. And as people speak the language and use these words, the narrative inherently goes hand-in-hand with the culture.
When the issue at hand is about the state's relationship with that of a former aggressor state, the likelihood of the narrative being negative is far from trivial. The victim state may find it difficult to forgive and forget, especially if it perceives the repentance lacking. This hostility can be reinforced by government rhetoric—from classroom materials to media broadcasts. When people can only speak one language, their mother tongue, they are restricted in the narratives they can access. The only narrative that is available to them is the one espoused by the government in their mother tongue. Under such conditions, reconciliation between the two states can be hard to come by.
People need not be monolingual, however. They can learn to speak second or third languages, even if not fluently (Liu and Pizzi Reference Liu and Pizzi2018). When individuals are proficient in a foreign language, they have linguistic access to an alternative set of narratives—which may or may not reinforce those perpetuated by the mother tongue. But when they are incongruent, this forces the individual to confront—if not reconcile—the disagreement. For example, the historical aggressor state may believe it has atoned for its past wrongdoings. By adjudicating between the two conflicting narratives and de-emphasizing the one espoused in the mother tongue, it is possible for individuals in victim states to see the former aggressor state in a non-negative light.
Of course, not all languages are equal in their association with cosmopolitanism. Knowing Zulu will most likely have no effect on how an Albanian sees Serbia. Likewise, a Yemeni who is fluent in Basque is probably no more cosmopolitan and accepting of Saudi transgressions than their monolingual co-nationals. Instead, languages that are more likely to have a cosmopolitan effect are lingua francas. Lingua francas are third-party languages; they are used as vehicles for inter-ethnic communication (Crystal Reference Crystal2003). Here it is important to note that no natural language (even English) is a lingua franca in all contexts—for example the status of English in Canada (Medeiros Reference Medeiros2017).
When individuals are proficient in a lingua franca, this language of inter-ethnic communication provides them with access to an alternative set of narratives. Whether an individual agrees with these accounts is not the point here. Instead, what matters is that being proficient in a lingua franca exposes individuals to different perspectives and allows them to consume positive narratives of the former aggressor state. Note here that we do not assume the alternative narratives are always positive. We do contend, however, that the alternative narratives are on average more positive than the ones being put forth in the mother tongue. There are two reasons why. The first is that the narrative in the mother tongue is inevitably biased. It is possible that there are positive narratives about the former aggressor state. But we contend that we are more likely to see ones about the barbaric brutality. Of all narratives about the former aggressor state, the one emanating from the victim state and in the victim state language will be the most negative. The second reason is that the lingua franca narrative is a smattering collection of perspectives. It includes the non-biased perspectives of third parties. It also includes the biased—but in the positive direction—of those from the former aggressor state. This can include the “we have atoned for our past sins” perspective. And so, in aggregate, the lingua franca narrative will be more positive. And when individuals are proficient in the lingua franca, they become exposed to these positive narratives.
In East Asia, English is the lingua franca (Ricento Reference Ricento2015). Unlike Europe, the United States, or Australia, English is not associated with a politically dominant group. And unlike Africa or South Asia, it is not (generally) associated as the colonial language and that of an upper-class. For sure, we see some inkling of this in the former British Malaya and to some degree in the Philippines, but the trend in general is quite different. Instead, English is seen as a global lingua franca. It is not just a foreign language; instead, it is a skill of great economic importance. Proficiency in English is akin to computer literacy and mathematical analysis—all necessary for future job prospects (Ricento Reference Ricento2015). In most—if not all—of the East Asian countries, English is a compulsory subject in the school curriculum (Kaplan and Baldauf Reference Kaplan and Baldauf2003; Kim et al. Reference Kim, Liu, Tuxhorn, Brown and Leblang2015). Here, we should also note that no East Asian country has a sizable population that speaks an indigenous Indo-European language as its mother tongue. This is important because language learning is supposedly more difficult when the distance between two languages is high (Laitin and Ramachandran Reference Laitin and Ramachandran2016). Thus, what we have here is the same linguistics challenges for all parties—further rendering English as a true lingua franca.
When individuals in the Chinese-speaking world knows only Chinese, they are restricted to the set of narratives put forth in that language. Often, the narratives are negative: the Japanese will not relinquish the disputed island (Jin and Zhao Reference Jin and Zhao2012). The Japanese have not apologized for the “comfort women”—the abduction of women to serve as prostitutes to Japan soldiers (Chen Reference Chen2018). The Japanese have not really apologized for their war time aggression as evident by their silence on the Nanjing Massacre (Zhang Reference Zhang2018). As a result, for these individuals their attitude of the Japanese is accordingly negative (Zhao Reference Zhao2017). However, when an individual in the Chinese-speaking world is proficient in English, they are exposed to a set of narratives about Japan that are incongruent with that in the Chinese language. There are narratives put forth by the western countries about the importance of Japan in the global order. Politically, it is a democracy. Economically, it is a key player in the international market. And culturally, it is a society that has transformed from one of “military cults into … a peace loving” one (Cai Reference Cai2008). There are also narratives put forth by the Japanese themselves that talk about what Japan is doing to move on from its past transgressions. In 2015, Prime Minister Abe reiterated his predecessors’ official apologies for the damage his country inflicted on others (New York Times 2015). These narratives—both from the west and from Japan—are in stark contrast to those in Chinese. While the Chinese-speaker may not agree with all of these perspectives, their exposure to this positive narrative allows them to question the negative ones put forth in the Chinese language. And when they do, their attitudes towards the Japanese are more likely to be less negative—if not outright more positive—than their monolingual counterparts.
If familiarity with a lingua franca can expose individuals to new narratives, the following hypothesis summarizes this discussion:
Hypothesis: When individuals are proficient in a lingua franca, attitudes towards a former aggressor state are more likely to be positive.
RESEARCH DESIGN
To understand sentiments towards the Japanese in the Chinese-speaking world, we use survey data from the AsiaBarometer. Note that the AsiaBarometer is not the same as the Asian Barometer Survey. The former is headquartered in Japan (University of Niigata Prefecture, Tokyo Satellite Office),Footnote 3 the latter in Taiwan (Academia Sinica and National Taiwan University).Footnote 4 The choice to use the AsiaBarometer and the particular wave—2006—was motivated strictly by data availability for our key independent variable of interest: English proficiency.
ATTITUDES TOWARDS JAPAN
The outcome of interest is attitudes towards Japan. Respondents were asked, “Do you think [Japan] has a good influence or bad influence on [Chinese-speaking political unit]?” After removing the “don't know” responses, all remaining responses range on a five-point scale from 0 (good influence) to 4 (bad influence). The median value is a 3 for Mainland China and a 1 for Singapore and Taiwan. We plotted the sample distribution for each political unit in Figure 1 at the beginning of the article.
ENGLISH PROFICIENCY
The AsiaBarometer asked respondents to self-evaluate their English language skills, specifically: “How well do you speak English?” Possible answers included “not at all” (1), “very little” (2), “I can speak it well enough to get by in daily life” (3), and “I can speak English fluently” (4). There was also a “don't know” option, but for the purposes of this article, we drop these responses. When we look at the distribution of responses (Figure 2), it is not a surprise that English proficiency levels are the highest in Singapore—incidentally a former British colony. In contrast, in Mainland China and Taiwan, more than half of the respondents answered that they did not speak English with any proficiency.
English Proficiency Levels
While proficiency is relatively easy to measure—whether through a standardized test or via self-evaluation as in this case—the concept itself can be more challenging. If the argument is that proficiency exposes individuals to these alternative narratives, the question becomes at what proficiency level does this exposure happen? On the one hand, it can be just a little. It is a matter of simply getting the proverbial foot in the door. Understanding basic vocabulary and grammar can set the necessary foundation in place. Whatever is ambiguous can be clarified with translation software; and whatever remains unclear can be inferred from context or ignored. On the other hand, the proficiency required may be at the highest level. Only when an individual is fluent in English can they pick up the nuance in the argument. Here, we sidestep this debate in part by conceptualizing proficiency both as a stock and as a level. Knowing a little English can expose an individual to an alternative narrative. But knowing it well opens the individual to more narratives. Given this discussion, we make two predictions. One is about proficiency as a stock; the other is about proficiency as a level:
Prediction 1.1 (Stock): When individuals are proficient in the English language, their attitudes towards Japan are more likely to be positive.
Prediction 1.2 (Level): As an individual's proficiency in the English language increases, their attitudes towards Japan is also more likely to increase positively.
DEMOGRAPHIC CONTROLS
Inevitably, individual characteristics can affect how an individual evaluates the Japanese. Age, for example, can be important. The general consensus in the literature is that older respondents are more hostile of outsiders (Wilkes, Guppy, and Farris Reference Wilkes, Guppy and Farris2008). And in the context of this article, older respondents are likely still to have memories of the Japanese atrocities. Conversely, younger respondents may have a different perception of the Japanese, from the 1980s economic boom to the proliferation of Hello Kitty and other cultural icons. It is possible that these different associations are driving older respondents to be more negative towards the Japanese but not the younger ones.
We consider the respondent's gender as well. Here, the literature has been split on the effects of gender on tolerance. On the one hand, there are those who contend women are less tolerant than their male counterparts (Golebiowska Reference Golebiowska1999). On the other hand, there are those who argue women are less hostile (McLaren Reference McLaren2003). And then there are those who have found no gender effect whatsoever (Hayes and Dowds Reference Hayes and Dowds2006).
We also include a measure for educational attainment. Specifically, we look at whether the respondent has completed secondary education. If education is a vehicle for people to learn about diversity and to be exposed to alternative narratives, respondents who have not completed secondary schooling are more likely to be hostile to those in the out-group—the Japanese (Bobo and Licari Reference Bobo and Licari1989). And finally, we control for income, since wealth can foster tolerance (Lipset Reference Lipset1959). Respondents were asked to place their household gross income along a ten-point scale.
REGIONAL VARIANCE
We also control for the variance in attitudes across the three political units and within each of the three units. There are multiple factors that may affect English proficiency, whether socioeconomic differences or cultural preferences. Moreover, attitudes towards Japan may differ across regions. The Japanese occupation also varied across the three political units, as we discussed above. In addition to the Japanese legacy, the three political units also vary in their (historic) receptiveness to the English language. Prior to 1972, the US was heavily involved in Taiwan—and remains largely so in the present-day. Moreover, in the 1970s and 1980s, Taiwanese constituted one of the largest Asian student populations in the US (Harvie and Lien Reference Harvie and Lien2017; Lien and Harvie Reference Lien and Harvie2018). And in Singapore, as a British colony, the English language legacy is strong. In 1979, a change in the national education curriculum would make English the de facto first language for the state (Liu Reference Liu2015). For Mainland China, these historic factors are absent. And while Mainland China is now the number one sending state of international students abroad (Leung Reference Leung2017), this development is more contemporary.
The attitudinal differences can also exist within each political unit. Japan's occupation was far from uniform. In Taiwan, the military proceeded in three stages: the first stage in the north, the second stage in the west, and the final stage in the south. The surrender of Tainan effectively marked the end of the island's resistance efforts. Likewise, in Mainland China, the Japanese started in the northeast (Manchuria). From there it moved down the coast to Shanghai and then eventually inland to Wuhan. Even in Singapore, there was variance in how the Japanese treated the locals. It is possible these experiences across the different stages of occupation manifest in the present. In sum, given this discussion, we estimate our models with fixed effects at both the cross-unit and within-unit levels.
EMPIRICAL EVIDENCE
We estimate our models using a three-level model (individual-subnational unit-political unit). Since our dependent variable (Attitudes towards Japan) is a five-point construct, we estimate with ordered logit. As a reminder, higher values in the dependent variable correspond to more negative sentiments. We begin with a baseline model. In Model 1 we operationalize English Proficiency as an ordinal variable. The lowest value (1) indicates no command of the language whatsoever and the highest value (4) suggests fluency. As we see in the first row of top panel in Figure 3, the effect of English proficiency is statistically significant and in the expected direction. The dots are the point estimates and the whiskers are the 95 percent confidence intervals (note that for clarity of presentation, we have omitted the controls; see the online appendix for the full numerical results). In particular, increasing proficiency levels translate into less negative attitudes towards Japan on average (β = −0.15). This corroborates Prediction 1.2.
Results of Multilevel Analyses: English Proficiency's Effect on Public Attitudes toward Japanese (Observation: 3803)
At the political unit level, we allow the effects (slope and intercept) of English Proficiency to vary across these political units. In doing so, we can account for differences in saliency: specifically, how this saliency affects attitudinal distribution in Mainland China, Singapore, and Taiwan (see Figure 1). Note that a post-estimated likelihood ratio test shows a χ2(df = 2) of 4.4 with a p-value of 0.11. This suggests that that the slope variance does not significantly improve the goodness of fit of the model. Put differently, allowing the slope to vary across political units does not make a difference. As such for parsimony purposes, we only report the results from the intercept-varying model.
The findings from Model 1 assume English Proficiency is ordered in some meaningful way. While this may be the case qualitatively, mathematically this requires a shift from no proficiency to some to have the same attitudinal effect as a shift from some proficiency to sufficient. In Model 2 we relax this assumption. We consider English Proficiency as a categorical variable. Specifically, we rerun the baseline model, but this time we disaggregate our concerned variable into four dummies: English: None (1), English: Little (2), English: Sufficient (3), and English: Fluent (4). Here, we set English: Fluent (4) as the reference category. With this model specification, we can examine whether the effects from Model 1 are consistent across each proficiency level. The results (the lower three rows of the top panel in Figure 3) suggest that indeed they are: An individual with limited English proficiency—even at the most basic level—is statistically more likely to express anti-Japanese attitudes than their English fluent counterparts. And while the coefficients for each dummy decrease as proficiency increases, it is noteworthy that they are statistically no different. This corroborates Prediction 1.1.
In Model 3 we relax another assumption: the values of the dependent variable are meaningfully ordered and mathematically equivalent. Instead, we operationalize each sentiment level (“good,” “rather good,” “neutral,” “rather bad,” and “bad”) as distinct attitudes towards Japan. Our reference category is the “neutral” attitude. Hence, a negative coefficient for “good” suggests a decreasing likelihood of the respondent answering in the affirmative vis-à-vis “neutral.” Conversely, a positive coefficient for “bad” indicates an increasing likelihood of the respondent answering negatively vis-à-vis the reference category. As shown in the bottom panel of Figure 3, the effects of English proficiency on each attitude level are substantive, statistically significant, and robust. Those who cannot speak English fluently are less likely to see Japan as “rather good” or “good.” Instead, they more inclined to have “rather bad” or “bad” sentiments.
SPURIOUS EFFECTS: PREDISPOSITION TO A GLOBAL VIEW
The argument has been that the English language exposes an individual to alternative narratives that emanate from abroad. And these narratives are likely to contradict the negative narratives about the former aggressor state extant in the victim state. While we have focused on English, it possible that, with technological developments, people can still access the outside world despite the linguistic barriers. For example, there are countless sources—both online and offline—that can translate contents of foreign media into Chinese. And more immediately, the rapid progress of machine translations enables people to browse effectively any foreign website in their own mother tongues. People can also access foreign narratives through their family and friends who live abroad; through foreign television programs, movies, and music videos with Chinese subtitles; or through their jobs with multinational organizations. These different sources can all increase both the possibility of people being exposed to alternative, positive narratives of Japan and people's general interest in learning English. If this is the case, then the attitudinal effects we found in the previous section might be a spurious correlation.
Furthermore, it is possible that there are people naturally predisposed to having a broader worldview. Even after controlling for age, gender, and socioeconomics, some people may be more inclined to think in global terms. And so their attitude towards Japan may have less to do with Japan per se; instead, it may reflect their general attitudes about their political unit vis-à-vis other countries. For instance, a xenophobe with extreme national pride will probably answer “bad influence”—whether the political unit in question is Japan, the US, or Maveth. Similarly, someone who believes in open borders and globalization will probably answer “good influence” regardless of the political unit being asked. In such case, globally inclined individuals may naturally hold more positive attitudes towards Japan than their xenophobic counterparts. And in the meantime, the former may be more willing to learn English than the latter. Again, the findings in the previous section could be spurious in this case.
Here, we take the above concerns into account in two ways. First, we measure people's exposure to the outside world—independent of language. In the AsiaBarameter Survey 2006, respondents are given a battery of statements about their foreign connections and are asked whether any of the individual statements apply to them. The statements are follows:
1. A member of my family or a relative lives in another country.
2. I have traveled abroad at least three times in the past three years on holiday or for business purposes.
3. I have friends from other country [sic] who are in [the Chinese-speaking political unit].
4. I often watch foreign-produced programs on TV.
5. I often communicate with people in other countries via the Internet or email.
6. My job involves with organizations or people in other countries.
Using these questions, we construct a variable called Foreign Culture Exposure and add it to our previously run Model 3 (Model 4). The variable—ranging from 0 to 6—is a count of the different types of exposure to foreign culture exposure and experiences.
Second, we consider the respondent's attitude towards other third-party countries as a proxy of their global view. The third-party country needs to have little salience in everyday Chinese-speaking life. But at the same time, it cannot be completely random or wholly unknown—e.g., Kiribati, Moldova, or Togo. The distribution needs to be normal, with “neither good nor bad influence” as a convincing modal response—all the while keeping the “don't know” numbers to a minimum. We settled on two countries that met these conditions. The first was India, where the distribution of attitudes across the three political units is normal: 54.2 percent said “neither good nor bad”—while 14.2 percent had no response. The second country was Turkey: 64.8 percent of the respondents (of those with an explicit opinion) had an indifferent response, 8.5 percent said (rather) good influence, and 12.6 percent indicated (rather) bad influence.
Despite these normal distributions, we are aware that neither country individually is a perfect proxy for the world view of the Chinese-speaking world. After all, India is a salient regional power in Asia; and Turkey is a Muslim country (note that two of the three political units in our analysis have a sizable Muslim minority). As such, it is difficult to hold a truly neutral opinion of either country. We also do not expect individuals in the three political units to view these two countries in the same light as they do Japan—considering the economic power and cultural influence of the latter in Asia. Nevertheless, we believe the attitudes towards the two countries combined can give us some insight into how people in the Chinese-speaking world see the world at large rather than rooting from specific state characteristics.
The results after incorporating the above variables tell an interesting story (see Figure 4). At first glance, the coefficient for exposure to foreign culture is signed in the correct direction: negatively. This is consistent with what we would expect. A broader view is correlated with lower levels of Japanese hostility. However, it is important to note that the coefficient is not statistically significant in any of the model specifications. Even if we disaggregate the exposure indicators into a set of binary measures the results do not change (results not presented here but available upon request). And the coefficients for attitudes towards India and Turkey are both signed in the correct direction. Individuals who are generally positive towards other countries are more likely to hold the same attitudes towards Japan.
Robustness Check: Predisposition to a Global View
Yet despite these additional controls, the coefficient for our concerned variable English Proficiency remains significant, robust, and in the predicted direction: respondents who are proficient in the English language are less likely to see Japan as having a negative influence on their political unit. So, when we put these results together, regardless of people's experience and view of the world at large, it seems English language proficiency has some independent effect on anti-Japanese attitudes.
SELECTION EFFECTS: OPPORTUNITIES FOR ENGLISH PROFICIENCY
While the above section considered the possibility that the English and Japanese attitude correlation is of a pseudo relationship (answer: no), here we look at whether there is a selection effect. Specifically, while English proficiency leads to less hostility to Japan, it is possible that English proficiency is not randomly distributed. So, what ends up accounting for the anti-Japanese attitudes is not language per se, but rather the factors explaining English proficiency in the first place. While not perfect, we seek to address these selection problems using a two-prong approach. We first compare the mean proficiency scores to assess whether the opportunity to learn English is largely uniform. We then conduct a coarsened extract matching, which allows us to construct a semi-experiment with observed data. The results from both tests suggest selection effects are of minimal concern.
For sure, the opportunities to learn English are not uniformly distributed. Age can matter. In Mainland China, for example, we expect the youths today to have more opportunity to learn English—and with greater ease—compared to their elders who suffered through the Cultural Revolution. In Singapore, we expect those who went to school after 1979—when the education system was reformed to make English the first language of the curriculum—to have more opportunities to be proficient in English than their parents, who may have in fact gone to a Chinese language school. Income can also matter. We expect people coming from wealthy families to be able to afford English language tutors or private school education. While these differences are important, at the individual level they cannot explain the average effects that we observe across all three political units.
Alternatively, there could be some spatial, regional effect. We would expect people in the capital and economic centers to have more opportunity to learn English. They are more likely to encounter foreigners—tourists, business executives, missionaries, or English language teachers. This contact can translate into higher levels of English proficiency. While there may be more opportunities in certain areas, it is important to recognize that there are few areas where there is no opportunity to learn English. In all three political units (Mainland China, Singapore, and Taiwan), English is a compulsory subject in schools—and this is not a recent phenomenon. In Singapore and Taiwan, English has always been a compulsory subject since independence (Liu Reference Liu2015; Ostwald, Ong, and Gueorguiev Reference Ostwald, Ong and Gueorguiev2019). Even in Mainland China, where the education system was severely disrupted by the Cultural Revolution, the education ministry introduced English as a compulsory subject at the middle school level in 1976 and starting in the fourth grade in 1978 (Cowen and McLean Reference Cowen and McLean1983; Kaplan and Baldauf Reference Kaplan and Baldauf2003). Since enrollment numbers are nearly universal, from a theoretical standpoint, the opportunity to learn English is uniformly distributed.
To empirically demonstrate this, we compare the average English proficiency score for each AsiaBarometer region in the three political units. We plot the mean scores in Figure 5. Recall, the maximum value of “4” is equivalent to “I can speak English fluently.” Mainland China is divided into three regions: central, east, and west. If there is a regional difference, we would expect to see highest proficiency levels in the east—home to the big cities including Beijing, Shanghai, and Guangzhou. Conversely, we would see the lowest levels in the west where distance is a barrier to economic activities. While the pattern of bar heights is consistent with these statements, the averages are not statistically different.
English Proficiency across Regions (4 = Fluent)
Similarly, in Singapore, despite being an island city-state the size of Washington, DC, there is also notable variation. Central Singapore is the primary hub for the multinational banks and major commercial businesses. In contrast, eastern Singapore includes some of the poorer neighborhoods and reclaimed land. And again, we see the patterns are generally consistent—but they are not statistically differentiable. Finally, in Taiwan, we would expect the north—home to the capital and some of the largest high tech companies—to have the lowest English proficiency. Yet again, the averages across the three Taiwanese regions (north, south, and west) are statistically non-differentiable.Footnote 5
We also conduct a coarsened extract matching to test selection effects more generally (Iacus, King, and Porro Reference Iacus, King and Porro2012). Matching allows us to control for the impact of potential biases from model specification and data selection on the results by constructing a semi-experiment with observed data. To construct the treatment variable, we convert the English proficiency variable into a binary one. The treatment group is whether the respondent knows at least a little English; conversely, the control group includes those who “don't know English at all.” We split the proficiency at this point for two reasons. The first is theoretical: We know from model 2 (see Figure 3) that proficiency—at any level—has a significant effect on attitudes. The second reason is empirical: Cutting it at this point ensures we have the greatest balance between the control (43 percent) and treatment (57 percent) groups.Footnote 6 The results show that, with this semi-experimental design that can effectively diminish model dependency, there is a significant difference between the two groups (Iacus, King, and Porro Reference Iacus, King and Porro2012). Those who have some familiarity of English are on average -0.2 less likely to be hostile towards the Japanese with a variance of 0.1 (see the numeric tables in the online appendix).
PARSING OUT THE MECHANISM: THE ENGLISH LANGUAGE VERSUS AMERICAN HEGEMONY
In this article, we contend that English proficiency can dampen anti-Japanese attitudes. This is because English—as a lingua franca—can bridge people of diverse backgrounds together. It equips speakers with a global view. However, there is an alternative explanation for the purported relationship between language and attitudes. While English is a lingua franca in the region, it is not just a lingua franca. It is also the language of the one of the strongest global hegemons: the United States. American soft power has been relatively unmatched since World War II (Nye Reference Nye1990)—and if anything, has strengthened with the information revolution (Nye Reference Nye2002).
Thus, it is possible that the positive effects of English are not about the language itself, but rather about a hegemonic country. With soft power comes the “power of persuasion through ideas, culture, and policies” (Nye Reference Nye2002, 60). These ideas—in the American case—can include those associated with western liberalism—from freedom of speech, assembly, and press to tolerance of cultural diversity. They can also include policy positions vis-à-vis other countries—e.g., Japan. Since World War II, American rhetoric on Japan has been largely positive. From political cooperation to military interests, from economic competition to cultural exchanges, Japan is regularly portrayed in a non-negative—if not outright amicable—narrative (Armacost Reference Armacost1996; Buckley Reference Buckley1995).
We employ two strategies to assess whether people who are—and can be—receptive to the American message are more positive towards Japan. First, we look for a mediation effect—i.e., all the effects of English proficiency on Japanese attitudes are due to the individual's approval of the United States. To this end, we look at whether respondents hold positive attitudes towards the United States. This question—like that for Japan, India, and Turkey—is on a five-point scale. We examine this variable's effects through a mediation model. In the model, an individual's English proficiency can have either a direct or indirect effect through their attitudes towards the United States. If the mediation effect is the main mechanism for the influence of English proficiency, we predict the following:
-
Prediction 2.1 (Mediation): English proficiency affect people's attitudes towards Japan primarily through their attitudes towards the United States rather than directly.
We conduct the estimation using a specific type of structural equation model. We adjust for education, income, age, gender, attitudes towards India and Turkey, exposure to foreign culture, and spatial variances at the regional and political unit levels—as in the previous analyses. Furthermore, we bootstrapped with 10,000 repetitions to diminish the potential biases of the variance estimation (Yuan, Hayashi, and Yanagihara Reference Yuan, Hayashi and Yanagihara2007).Footnote 7
We see the results in Figure 6. The mediation effect of English proficiency is represented by the paths a and b—via the attitudes towards the United States. The direct effect is represented by the path c. The total effect (ab + c) is statistically significant at the 0.01 level indicating that an individual's English proficiency alters their attitudes towards Japan—as found in the previous sections. The mediation effect (ab), however, is not significant. Specifically, liking the United States does reduce people's negative attitudes towards Japan. Yet, the contribution of people's English proficiency to their affection of the United States—while positive—is inconclusive. At the same time, the direct effect (c) of English proficiency is significant and reduces the negative attitudes towards Japan as our theory expects.
The Mediation Effects of Attitudes towards the US
According to the result above, American hegemony does not serve as the underlying primary reason for the English influence. But, does it make the influence more pronounced? While people do not hold a positive attitude towards Japan just because of American narratives, there is the possibility that they are more likely to be positive if they are committed the to the values and norms that the Americans hold. To test this moderation mechanism, we interact the variable of attitudes towards the United States with the respondent's English proficiency. If the effect of English proficiency is stronger among those who evaluate the United States favorably, we would have reason to believe that there is some American influence at play. In other words, we predict the following:
-
Prediction 2.2 (Moderation): The effects of English are even more pronounced when people are favorable towards the United States.
The moderation effect is estimated with a conditional effect mode with the interaction between English proficiency and the attitude towards the United States. We include the same battery of controls from the previous models. Note that with interaction models, we cannot simply interpret the coefficients of either the constituent or the interaction terms. Doing so risks drawing insufficient and potentially misleading conclusions (e.g, Aiken, West, and Reno Reference Aiken, West and Reno1991; Clark, Gilligan, and Golder Reference Clark, Gilligan and Golder2006; Hainmueller, Mummolo, and Xu Reference Hainmueller, Mummolo and Xu2019). We thus present the results with a more comprehensive and reliable way: marginal effects.
In Figure 7, the line is the marginal effects of English proficiency along with attitudes towards the US. The whiskers are the simulated 95 percent confidence interval. The plot shows that the effect of English proficiency does not increase when people are favorable to the United States. Instead, the effect seems to be in the negative direction! More importantly, we can only consider the conditional effects significant when the intervals do not saliently overlap. The “CI (Max-Min)” provides precise evidence here. It is the difference between the conditioned effect (the effect of English proficiency) at the minimum and maximum values of the conditioning variable (whether the United States has a positive influence). If the interval contains zero—as is the case here—the conditional effect is considered statistically non-significant. As such—and in conjunction with the findings of the mediation mechanism—we can say with confidence that the effects of English proficiency on anti-Japanese attitudes are driven by linguistically related effects and less so by American hegemony.Footnote 8
The Moderation Effects of Attitudes towards the US
DISCUSSION: GENERALIZABILITY BEYOND THE CHINESE-SPEAKING WORLD
This article examined attitudes towards the Japanese within the Chinese-speaking world. Specifically, we argued when individuals can speak English—a lingua franca in East Asia—they have an alternative channel through which they are able to see Japan in a less hostile perspective. Proficiency in a third-party language exposes individuals to a set of narratives that are inevitably more positive than the one being espoused in the Chinese language. Put differently, knowing English allows them to learn different narratives that are not available to their monolingual counterparts. And as a result, these individuals are more likely—all else being equal—to view the Japanese in a positive light. We subject these results to alternative explanations, including whether some people are simply more global in their outlook; whether some individuals have better resources to learn English; and whether the linguistic effects are being driven by American hegemony. We find the attitudinal effects of English proficiency remain significant.
The results are not just robust; they are normatively encouraging. The effects of English proficiency are as substantively important as other factors commonly believed to be relevant for public opinion—i.e., education and household income. In fact, at the individual level, learning English can negate some of the effects of having been dealt the proverbial bad cards by nature. This matters for at least two reasons.
First, it shows how languages can be an instrument for diffusing ethnic tensions. In the ethnic politics literature, language is often considered a marker of a distinct identity. Individuals understand their environs through this lens. Yet, we contend and show that when individuals are proficient in a second or third language (i.e., ones unencumbered by their own culture), they can see the world through alternative perspectives. The narratives about the “other”—whether it is another ethnic group or another state—will be framed differently. In the East Asian context, we are encouraged by the evidence showing that using English can breed tolerance—even for those who are associated with a former aggressor state. And when there is tolerance, social stability—both domestically and regionally—is much more likely.
The second reason English matters in East Asia has to do with attracting investments. Investors go where they are welcomed. Where anti-Japanese sentiments are widespread, this can deter Japanese investments—whether it is of the economic or cultural persuasion. If a government is looking to diversify and attract more investors from abroad, it cannot afford to alienate those from Japan. Instead, for the Chinese-speaking governments, it is important that they have a new generation that is broadly global in its outlook and specifically pro-Japanese (or at the very least, non-anti-Japanese) in its attitudes. And when they do, political stability—both domestically and regionally—is also much more likely.
CONFLICTS OF INTEREST
Yue Hu and Amy H. Liu declare none.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/jea.2019.41.
