Our sample of 70 independent states is drawn from the International Monetary Fund’s (IMF) ranking of countries by estimated nominal gross domestic product (GDP). [
60] We limit our sample to the upper one-third of the GDP list, so as only to include those states whose economies are likely large enough to carry some international weight. These are states that the US and China, one would presume, care the most about; their support, or their lack of support, can plausibly make a significant difference in the growing great-power rivalry. Our panel thus encompasses 70 sovereign states that are, albeit to varying degrees, caught in the middle of the intensifying tug-of-war between the United States and China. A majority of the states included are, again to varying degrees, friends and allies (i.e., clients) of the US, though there are also several states in the sample that take a more neutral position. The panel also includes some countries that lean toward Beijing, most notably Iran and Russia [
61‐
63].
The Dependent Variable
Using information from international news media, we coded the dependent variable –
Y_Huawei – on a scale from 1 to 4, where 4 denotes a position of full rejection of Huawei (in accordance with US, but not Chinese, preferences), and 1 denotes a stance that is fully acceptive (in accordance with Chinese, but not US, preferences). Table
1 describes the content of each value on the scale.
Table 1
Categories of the dependent variable
4 | Full ban: The state’s government has issued a full ban of Huawei’s equipment in their 5G network |
3 | Considerable government restrictions/operators are rejective: The state’s government has imposed laws that pose substantial barriers for Huawei; has signed 5G security agreements with the US; and/or some of the state’s major telecom operators have given 5G contracts to Nokia, Ericsson, and/or local actors, thus circumventing Huawei |
2 | Open approach/future use: To some degree, the state’s government is ignoring the US’s warnings, stating that it will not meddle in operators’ decisions on 5G; and/or the state uses Huawei in various technological areas, for example in 5G trials. The 5G network may still be in its early phases, but most evidence is pointing toward potential full use of Huawei in the future |
1 | Full use: The state’s government is using Huawei without any apparent restrictions, and their 5G networks have launched or are close to launching |
Table
2 shows the value given to each of the sample’s 70 countries on the dependent variable, while the Appendix presents a brief account and justification of the coding decision for every state in the sample. Two important notes are in order. First, there are, to be sure, fluid borders between some of these categories. A few states, such as Japan and Australia, placed full bans on Huawei early on [
64,
65]. Others, such as Russia [
66] and Indonesia [
67], have already signed major deals signaling full acceptance of the company. Cases such as these ease the coding. A handful of countries fall in-between multiple categories, such as Mexico [
68], Greece [
69], and Ireland [
70]. Such cases involve challenges that are also related to a second caveat, namely that the policies of some individual states on this issue are in flux. For example, responses to Huawei investments by European states have been mixed even if they seem, over time, to have become more rejective. There have also been some signs lately of a push-back against Huawei from some Asian countries. Others, like Brazil and the United Arab Emirates, are, at the time our coding concluded, in the midst of particularly intense pressure from the US to drop Huawei as a telecom partner, and the outcomes are unknown [
71,
72].
Table 2
Countries’ values on the dependent variable (N = 70)
Argentina | Algeria | Belgium | Australia |
Bangladesh | Angola | Bulgaria | Japan |
Ethiopia | Austria | Canada | Poland |
Hungary | Brazil | Czech Republic | Romania |
Indonesia | Chile | Denmark | Sweden |
Kazakhstan | Colombia | Ecuador | United Kingdom |
Kenya | Cuba | Finland | |
Kuwait | Dominican Republic | France | |
Morocco | Egypt | Germany | |
Oman | Ghana | Greece | |
Philippines | Guatemala | India | |
Qatar | Iran | Israel | |
Russia | Iraq | Italy | |
Saudi Arabia | Ireland | Luxembourg | |
South Africa | Mexico | Malaysia | |
Switzerland | Myanmar | Netherlands | |
Thailand | Nigeria | New Zealand | |
Unit. Arab Emirates | Pakistan | Norway | |
| Peru | Portugal | |
| South Korea | Singapore | |
| Sri Lanka | Slovakia | |
| Turkey | Spain | |
| Ukraine | Vietnam | |
18 | 23 | 23 | 6 |
(25.71%) | (32.86%) | (32.86%) | (8.57%) |
The use of prominent and readily available news media for the coding of the dependent variable was a necessary strategy considering our objective to include a large sample of states. When selecting the sources on the basis of which coding decisions were made, we pursued a fairly simple method. Using
Google News as our point of departure, we applied three search terms – Huawei, 5G, [state X] – for each of the 70 states in our sample for the relevant time frame (which was primarily the year up to October 15, 2021, although in a few cases where information was deemed insufficient, we also needed to rely on news articles – and sometimes also official reports – from before or after this time period). We then made a pre-selection of articles based on the perceived relevance of the headlines and the trustworthiness of the sources. Following an initial reading of this material, we selected a more limited final sample of sources (typically consisting of three to four per country, some of which are referenced in the
Appendix).
Granted, such a strategy does come with some selection-bias and reliability concerns.To mitigate these, we made two additional steps in the coding process. First, we scrutinized recent scholarship in order to check for any potential mismatches that would require further inspection. In the literature, coverage of the policies of secondary states is generally good with respect to Europe [
73‐
76] and South Korea, Japan, Australia, and New Zealand [
77‐
79]. Moreover, Gregory Moore’s study, which was recently published in the
Journal of Chinese Political Science, largely corroborates our coding also for (a few) states situated outside of these regions (such as Kenya and Thailand). [
75] Overall, this comparison with other, recent works, many of which draw on official government statements, diplomatic white papers, and the like, worked to increase our confidence in our own coding.
Second, in the present analysis we only separate between four categories of responses to Huawei. Identifying countries belonging to the two “extreme” categories is a relatively uncomplicated task. The “middle” categories are a bit more challenging; but judgments proved easier, also when compared with assessments made by the aforementioned studies, when restricting these to just two (a point helpfully made by one of the anonymous reviewers). However, we also constructed a more nuanced six-category dependent variable, where we split each of the two middle categories into two separate ones, to check for the robustness of our models. Results were substantially the same, though coefficients of the six-category version were generally significant at a somewhat higher level of confidence (and naturally so since this increases the variation in the dependent variable).
One last point to consider is that our coding was conducted – and concluded – as of October 15, 2021. This means that our data constitute a snapshot of the policies in existence at that date. The implied attendant caveat, of course, is that some of these policies, and the circumstances surrounding them, may have changed recently – or they may change in the near or far future. To cite but two such examples. Turkey has now arguably moved from “2” to “1” on our dependent variable (see Table
1), as Turkish telecom provider Türk Telecom in March 2022 signed a memorandum of understanding with Huawei to develop the country’s 5G network [
80]. At the other end of the scale, In May 2022. Canada moved to ban the Chinese company altogether from its 5G network, thus placing the country in the category “4.” [
81]
Independent Variables
The independent variables can be separated into three clusters, each of which corresponds to one of the theoretical models reviewed earlier. First, we operationalize and measure three of Stephen Walt’s four components of threat perceptions. (The fourth – “perceived aggressive intentions” – is virtually impossible to operationalize.) First, considering that the relational aspect of power is notoriously difficult to measure [
82], we employ size of the economy as a proxy of a country’s aggregate power. We therefore include nominal GDP, drawing on estimated numbers for 2021 from the International Monetary Fund [
83]. This variable is highly skewed, so we use the logarithmically transformed version of it in all models (
Nominal GDP). Second, the military dimension of (offensive) power is also of import. We therefore use the measure of military expenditure for 2020 estimated by the Stockholm International Peace Research Institute (
Military expend.). [
84] Third, for the “geographical proximity” dimension of Walt’s balance-of-threat-theory, we use a measure of the distance from a state’s capital to Beijing (
Distance). This we extracted from the
CShapes dataset [
85]. On the basis of this distance, we introduce a fourth variable as well, a dummy that categorizes a country as belonging to China’s “neighborhood” if its capital is separated from Beijing by no more than 6,000 km (
Neighbor).
For the second cluster of independent variables, we include three essential indicators of states’ patron-client relationship with the United States. The first variable –
Alliance – is a dichotomy that is coded 1 for any country that is either a member of NATO, a major non-NATO ally of the United States [
86], or a strategic partner of Washington [
87]. Second, we include a logged measure of the number of forward-deployed active-duty US military personnel (
ustroops), as of June 2021, with data from the US Defense Manpower Data Center [
88] (
US troops). Third, we use an estimation of a country’s arms imports to calculate the percentage of a state’s total arms purchases accounted for by sales from the United States (
Arms imports). [
89] To account for any “outlier” years with unusually large/small arms imports for specific countries, we use the average for the period 2015–2020.
The third cluster of independent variables is associated with Hirschman’s theory on trade relations and foreign-policy convergence. We retrieved the numbers for imports from China (
Import China), imports from the US
(Import US), exports to China (
Export China), exports to the US (
Export US), and total trade for both (
Trade China and
Trade US) from the World Bank’s
World Integrated Trade Solution (WITS). We use the latest data available, which for most countries is 2019 [
90]. All numbers are calculated as percentages of the states’ total imports, exports, and trade, respectively.