![Sample our Area Studies journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5922/TOC-Subject-Sample-2021-Area-Studies.jpg"})
ABSTRACT
Relatively few systematic and categorical studies have been conducted on the variant responses by East Asian states to the rise of China, and even fewer dealt with reasons for such intra-regional variations. This study seeks to fill the void by examining the strategic responses by 15 East Asian states for the period of 2011–2016. This period merits close scrutiny because China became more explicitly assertive. The research proceeds in two phases. First, by way of reading into the expert assessments in academic journals, the strategic responses of the 15 states are categorised into balancing, hedging, or bandwagoning. Second, statistical analyses are conducted to see how the intra-regional variations are related to unit-level factors. Of the seven variables analyzed, three – geographical proximity, democracy, and identity – turn out to be crucial in determining the East Asian states’ strategic responses.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 New Zealand is excluded as it does not view itself as an Asian player nor does it portray itself as an ‘active defense state.’
2 The limited number of observations is certainly an obstacle to statistical analyses (Potter Citation2005), but it is still not uncommon (George and Epstein Citation1992; Chung Citation2009/Citation10). Our statistical analysis can be justified both theoretically and methodologically. First, due to the temporal and geographical purview of our investigation, the current number of observations is the maximum, which is theoretically meaningful. Country-month, instead of current country-year, could be used as a unit of analysis, but a country’s strategic choice seldom changes by the month. Second, the fact that beta estimates were properly calculated with standard errors means that the number of observations was just enough to produce the results. Rainey and McCaskey (Citation2021) recently suggested a new model for the limited observations for binary logits. But we could not use this model since it requires us to collapse our ordinal outcome variable into a dummy variable, which experts do not recommend (Murad et al. Citation2003). Finally, due to the limited number of observations, we were very cautious both in our interpretation of the results and in our robustness checks. We adopted a highly conservative position in reading statistical significance out of our statistical analyses. We discuss only three variables that were consistent in their directions of effect and in their statistical significance throughout all models.
3 Mearsheimer (Citation2001, 139), for instance, uses appeasement and bandwagoning interchangeably.
4 The search was limited to English publications since, unlike English ones, Chinese publications are written almost exclusively by Chinese scholars. The point is that this study is not about Chinese views of how East Asia responded to the rise of China. Google Scholar provides both peer-reviewed and non-peer-reviewed journal articles. We used Google Scholar in order not to exclude non-peer-reviewed articles in advance without the authors’ scrutiny. As it turned out, the number of articles by countries was highly uneven. Moreover, for a few countries, it was difficult to find any pertinent journal articles. For those few countries, we used a supplementary method by looking into some edited volumes (Goh Citation2016; Beeson and Zeng Citation2017; Denoon Citation2017; Ross and Tunsjø Citation2017).
5 Since none characterised Japan as a bandwagoner, in order to illustrate our rationale, we may use the studies on Cambodia. These studies did not hesitate to call Phnom Penh as a ‘client’ (i.e. ‘subordinate’) of Beijing (Ciorciari Citation2013a, Citation2013b).
6 For more details, see Online Appendix: Summaries of Expert Views and Sources Used.
7 National power is measured by the Correlates of War’s Comprehensive Index of National Capabilities (CINC) (Singer et al. Citation1972). Disaggregated individual indices of national power are additionally employed to check the validity of the composite index. We use the SIPRI Military Expenditure database and World Bank database for these disaggregated indices. The results with these indices are nearly identical, and only the result using military expenditure is reported.
8 Balancing is most ‘confrontation[al]’ since the balancer’s message to the aggressor is clear. Bandwagoning could be confrontational toward the hegemon but its immediate effect is conciliatory since it is basically ‘conceding power’ to the aggressor (Mearsheimer Citation2001, 139).
9 The data on the nations’ trade volume come from the World Bank’s World Integrated Trade Solution dataset.
10 We used the Direct Contiguity data (Stinnett et al. Citation2002) and the CIA World Factbook. The results from these indices are identical, and only the result using the length of shared border with China is reported in Model 5.
11 Pertinent information comes from the Formal Alliances data and from the foreign affairs ministry in each country (Gibler Citation2005).
12 Information is obtained from the Intergovernmental Organizations dataset (Pevehouse, Nordstrom, and Warnke Citation2004).
13 The democracy variable is measured by using the Polity IV dataset. V-Dem dataset is also used to gauge the effect of individual characteristics of democracy (Coppedge et al. Citation2018). The results of using these indices are identical, and only the result with the comprehensive democracy score from V-DEM is reported in Model 8.
14 The percentage of overseas Chinese population was measured based on Overseas Community Affairs Council of Taiwan and Poston Jr. and Wong’s research (Citation2016). The number of wars with China was measured by the Correlates of War (Sarkees and Wayman Citation2010). We also considered three other measures but decided not to pursue due to their limitations. First, we considered nationalism. Classical realists stressed the role of nationalism in world politics. The most widely used measure is an expert survey data of the V-Dem project, asking country experts whether the current government promote nationalism. However, no theory predicts that a simple promotion of nationalism without a specific target country changes the strategic calculation against China. Second, we considered the respective image of the US and China, which can also be referred to as ideological affinity or affection. More recently, with the rise of China, survey data on national sentiment or temperature against China is available for some countries for some years. However, this data is not complete in all observation years and not available for at least three countries (Cambodia, Myanmar, Laos). Lastly, we considered the number of Confucius Institutes in each country. However, we decided not to use it since the data mainly reflects China’s initiative.
15 Temporal dependence is generally addressed with lagged dependent variables or binary variables representing the value of each ordinal dependent variable. We did not use these specifications due to the limited number of observations. A test of proportional odds assumption shows that the assumption is violated, which is common (Johnson and Willardson Citation2018). There are alternative models free from the proportional odd assumptions such as generalised ordered logit and multinomial logit, but we did not use these specifications due to the same reason.
16 We conducted sensitivity tests by removing each country, and the results were identical.
17 Among the three relevant variables, democracy and identity are statistically significant in separate bivariate models. The coefficients of the democracy and identity variables in a separate bivariate model are respectively β = -0.319 (SE = 0.145, p = 0.028) and β = -0.031 (SE = 0.017, p = 0.069). The coefficient of the geography variable in a separate bivariate model is not significant (β = -0.826, SE = 1.032, p = 0.423).
18 These alternative variables are significant in separate bivariate models. The coefficient of the democracy and identity variables in a separate bivariate model are respectively β = -8.976 (SE = 3.058, p = 0.003) and β = -1.133 (SE = 0.589, p = 0.055). The coefficient of the geography variable in a separate bivariate model is not significant.
19 A few variables display significance at the 0.1 or 0.05 level. These findings are, however, inconsistent. For example, while the national power variable measured with CINC displays significance in Models 5, 8, and 9, the direction of impact is not consistent. The same conclusion is drawn for the international institution variable in Models 5 and 9, and the alliance variable with the U.S. in Model 5.
20 This switching effect is not present for non-U.S. allies whose probability of hedging remains relatively high regardless of their geographical features.
Additional information
Funding
Notes on contributors
Jae Ho Chung
Jae Ho Chung Professor, Department of Political Science and International Relations, Seoul National University.
Hun Joon Kim
Hun Joon Kim, Professor Department of Political Scinece and International Relations, Korea University.