How Japan’s COVID-19 vaccination policy shapes trust in governance: a relative deprivation approach

Abstract

This study examines changes in the association between social policy performance and trust in government, focusing on the Coronavirus disease 2019 (COVID-19) vaccination policy implemented by Japan’s central government. Data from the Online Panel Survey of Stratification and Social Psychology (SSPW2021-Panel) were analyzed using two-way fixed effects regression models. The quadratic term of the COVID-19 vaccination rate at the prefecture level had statistically significant effects on the evaluation of the central government’s infection control policies and trust in the central government. This implies that the relative deprivation experienced by unvaccinated individuals weakened trust in the central government in the early stage, and the decline in the number of unvaccinated individuals strengthened trust in the central government in the latter stage. Thus, this paper finds that even if a social policy meets people’s demands, its implementation may temporally damage the government’s reputation through relative deprivation.

1. Introduction

This study examines the association between trust in government and vaccination policy during the Coronavirus disease 2019 (COVID-19) pandemic in Japan. Generalized trust, a component of social capital, significantly affects the performance of democracy and the local economy in a society (Putnam et al. 1994; Putnam 2000). Governments require a high level of generalized trust to effectively implement measures to prevent the spread of a pandemic and maintain society’s overall performance. If the central government’s vaccination policies enhance or reduce generalized trust during the pandemic, society’s overall performance will be affected. To confirm this statement, we need to examine the effects of the COVID-19 vaccination policies implemented by the central government on generalized trust. However, because generalized trust is a multidimensional construct in society and is related to various factors, it is difficult to examine changes in generalized trust directly. Methodologically, this study focuses specifically on trust in central government. This is expected to result in larger effect sizes, thereby increasing statistical power and reducing the risk of failing to detect significant relationships. Additionally, we can assume the following process: the policies implemented by the central government affect citizens’ evaluation of the policies, which affects citizens’ trust in the central government. Finally, changes in trust in the central government influence generalized trust in society. In other words, this study assumes that short-term fluctuations in trust in the central government may lead to changes in generalized trust in the medium to long term, which is beyond the scope of this study’s data.

To overcome this limitation, future research should explore the complex relationships between generalized trust and trust in a specific area, such as trust in the central government. According to the World Values Survey 2017–22 (Wave 7) (https://www.worldvaluessurvey.org/), the level of generalized trust and trust in the Japanese government ranks average globally. Therefore, Japan is suitable for testing the effects of the government’s infection control policies on social trust during the pandemic.

In Japan, due to the delayed approval of the COVID-19 vaccine, the speed of the COVID-19 vaccine rollout in 2021 was extremely low compared to other advanced countries (Kosaka et al. 2021). Japanese residents had to obtain a vaccine voucher issued by each municipality to receive a vaccination. However, the schedule for distributing the vouchers was uncertain. In addition, the speed at which the COVID-19 vaccines were rolled out varied across municipalities because some lacked sufficient medical resources to effectively vaccinate their residents. Furthermore, in 2021, Japan’s central government prepared to host the Olympic and Paralympic Games in Tokyo. Thus, while the central government should have contained the spread of COVID-19 by prompting a vaccine rollout, it risked further stimulating the pandemic and potentially undermining trust in the central government. As a result, residents who wanted to be vaccinated but were unable to were more likely to be dissatisfied with the vaccination policy implemented by the central government.

Fig. 1 shows changes in the number of newly infected individuals in Japan between March 2021 and March 2022, and Fig. 2 traces the changes in the vaccination (first shot) rate in Japan. Figs 1-2 illustrate that while the vaccination coverage was highest at Wave 4, the number of newly infected individuals also peaked during this period. This suggests that the association between the infection control policies and trust in the central government during Wave 4 could be either positive or negative. However, the association between social trust and attitudes toward infection control policies also depends on various social factors, such as socioeconomic status, race, gender, and timing of vaccination (Kawata and Nakabayashi 2021; Liu and Li 2021; Wu 2022; Wu et al. 2022). Therefore, whether this association is positive or negative cannot be determined easily without accounting for these factors. Moreover, changes in the association between social trust and infection control policies remain underexplored in the existing literature.

Based on this, the research question can be formulated as follows: How and why did the infection control policies implemented by Japan’s central government—particularly the delayed rollout of the COVID-19 vaccine—affect trust in the central government during the COVID-19 pandemic? This study, therefore, aims to theoretically identify the social mechanisms through which a social policy can have dual effects on social trust, depending on the social context.

2. Literature review

Several studies have established that trust in the government is crucial in preventing the spread of the COVID-19 pandemic (Stanica et al. 2022; Tan et al. 2022; Lee et al. 2023). Meanwhile, COVID-19 vaccination has been significantly associated with trust in experts and strongly correlated with trust in the government or political parties (Viskupič et al. 2022a). This finding also applies to Japan (Gotanda et al. 2021; Yokoyama and Ikkatai 2022). Therefore, the role of the central government in the rollout of COVID-19 vaccination in Japan cannot be overlooked. Additionally, several studies have argued that various types of social distrust negatively influenced COVID-19 vaccination efforts and contributed to vaccine hesitancy among mistrusting individuals (Niño et al. 2021; Szilagyi et al. 2021; Allen et al. 2022; Choi and Fox 2022; Echánove-Cuevas et al. 2022; Seddig et al. 2022; Van Oost et al. 2022; Viskupič et al. 2022b). That is, while some studies assumed that social distrust, as a stable social factor, generated COVID-19 vaccine hesitancy, it may have been exacerbated during the COVID-19 pandemic (Cowan et al. 2021; Latkin et al. 2021; Blair et al. 2022; Yan et al. 2022; Yuan et al. 2022; N. Liu et al. 2023). Therefore, social distrust may fluctuate depending on the waves of the pandemic. This suggests that social distrust, including distrust in government, is not necessarily a stable social factor for predicting attitudes toward the COVID-19 vaccine. Moreover, the association of social distrust and COVID-19 vaccine hesitancy may not be a unidirectional causal relationship but rather a bidirectional one.

According to other studies, trust in the central government may fluctuate depending on the waves of the COVID-19 pandemic and public evaluations of the central government’s infection control policies (Battiston et al. 2021; Li et al. 2021; Fulkerson et al. 2022). The COVID-19 pandemic has also influenced social trust for various reasons, including economic prospects (De Simone et al. 2022), risk of severe health outcomes (Oude Groeniger et al. 2021), and negative assessments of infection control policies as reported by various media outlets (Rieger and Wang 2022). Additionally, the association between social trust and evaluations of infection control policies is highly complex. This association can be positive or negative depending on social circumstances (Kye and Hwang 2020; Collischon and Patzina 2022). In light of this, social trust does not unidirectionally affect attitudes toward infection control policies, including COVID-19 vaccine hesitancy. Moreover, previous studies have demonstrated that trust in the government is intricately associated with social policies to control the COVID-19 pandemic in various ways (Devine et al. 2021, 2024; Jennings et al. 2021). These associations are often complex and context-dependent. Therefore, the role of social trust during a global crisis cannot be underestimated, and social scientists and policymakers must carefully and accurately analyze its effects.

Previous studies have also highlighted the multidimensional association between infection control policies and social trust. However, they failed to specify how this association varies depending on specific social situations. For instance, the effects of an infection control policy on social trust may be non-uniform and multifaceted due to the heterogeneity within societies. Therefore, this study seeks to contribute to the existing literature by clarifying the heterogeneous effects of infection control policies implemented by Japan’s central government during the COVID-19 pandemic. By examining how the impact of infection control policies on social trust changes based on varying social contexts, this research offers insights that can help policymakers respond more effectively to global crises, such as the COVID-19 pandemic.

3. Theory and hypotheses

Fig. 3 presents a general framework illustrating the association between social policy performance and trust in the government. It suggests that, even when a social policy is perceived as desirable by the public, delays in the distribution of social services may negatively impact trust in the government, resulting in a nonlinear association between policy performance and trust.

To explain changes in trust in the central government during the COVID-19 pandemic in relation to its vaccination policy, this study draws on relative deprivation theory, initially introduced in Stouffer’s The American Soldier (Stouffer et al. 1949) and later refined by Merton (1957). According to relative deprivation theory, individuals do not necessarily experience deprivation simply because they lack a particular status or good. Instead, feelings of deprivation are shaped by comparisons with their reference group members. If many reference group members already possess the desired status or goods, the individual will likely feel a strong sense of deprivation. Moreover, relative privilege can be understood as the reverse of relative deprivation. However, relative privilege and relative deprivation are not perfectly symmetrical. Prospect theory, as proposed by Kahneman and Tversky (1979), suggests that individuals are more sensitive and reactive to deprivation than to privileges.

During the pandemic, residents in Japan required a vaccine voucher issued by their respective municipality to make reservations for COVID-19 vaccination through an online system. These vaccine vouchers were distributed preferentially to medical workers and older adults. Moreover, there were significant differences in the speed of voucher distribution and the performance of the online system between municipalities. Based on this, relative deprivation theory can be applied to understand trust in Japan’s central government during the COVID-19 pandemic as follows: If an individual desired the COVID-19 vaccine but was not vaccinated, they might not necessarily feel dissatisfaction with the central government’s vaccination policies, nor would their trust in the government necessarily decline. However, if most residents within their reference group (e.g., people living in the same prefecture) were already vaccinated, the individual would experience strong dissatisfaction with the policies, reducing trust in the central government. Conversely, if an individual had already been vaccinated, they might not automatically feel satisfaction with the vaccination policies, nor would their trust in the central government necessarily increase. However, if few residents living in the prefecture were vaccinated, the vaccinated individual might feel strong satisfaction with the vaccination policies, thereby enhancing their trust in the central government.

In this study, the relative deprivation in social trust caused by unvaccinated individuals is defined using the following equation:

where $i$ represents an individual, $N$ is the size of the reference group, $n$ is the number of vaccinated individuals in the reference group, $T$ is the level of social trust, and $D$ is the degree of deprivation. It is important to note that $N$ does not equal the size of the entire population, as it excludes people who never intended to be vaccinated. Therefore, $N$ is defined as the size of the group excluding those who did not desire vaccination. Equation (1) demonstrates that deprivation caused by unvaccinated individuals increases as the vaccination rate in the reference group rises. Similarly, relative privilege in social trust caused by vaccinated individuals is defined as:

where $P$ is the degree of privilege. Equation (2) shows that privilege caused by vaccinated individuals weakens depending on the vaccination rate in the reference group. The number of unvaccinated individuals is represented as $N−n$⁠, and the number of vaccinated individuals as $n$⁠. Therefore, the total level of social trust in the reference group $(T)$ can be calculated using the following equation:

From Equation (3), if $Di>Pi$⁠, then function $T$ takes the least value at $n=N2$⁠. Conversely, if $Di<Pi$⁠, then T takes the greatest value at $n=N2$⁠. If $Di=Pi$⁠, then T always takes 0 (or constant).

As formal models of relative deprivation theory, Kosaka’s (1986) model—based on Boudon’s theory—and the relative deprivation index proposed by Yitzhaki (Yitzhaki 1979; Ishida et al. 2014) are well-known. However, since vaccination was exogenously provided through government policies, the Boudon–Kosaka model cannot be applied to this case. Similarly, Yitzhaki’s relative deprivation index, typically calculated using continuous variables such as income, is different from the model, which is based on the discrete variable of vaccination. Therefore, this study employs an alternative model.

Based on the relative deprivation theory, Fig. 4 illustrates how the vaccination rate influences average trust in the central government, showing a curvilinear relationship. By drawing on prospect theory, this study predicts that individuals tend to be more reactive to deprivation than privilege. Suppose the degree of deprivation among the unvaccinated surpasses the degree of privilege among the vaccinated (⁠$Di>Pi$⁠). In that case, the association of average trust in the central government and vaccination rate takes a U-shaped pattern (as shown on the right side of Fig. 4). Initially, average trust in the central government declines as vaccination increases because the deprivation experienced by the unvaccinated intensifies. However, as the number of unvaccinated individuals decreases with rising vaccination rates, average trust in the central government begins to recover after crossing a certain threshold.

From this inference, the following hypothesis can be derived:

 

As mentioned above, $T(n)$ takes the minimum (or maximum) value at $n=N2$⁠. This indicates that the effect of the vaccination rate on trust in the central government changes from negative to positive when the vaccination rate of the reference group reaches approximately 50 per cent. Additionally, the reference group consists of individuals in the region who are eager to receive the COVID-19 vaccine. From this inference, the following hypothesis can be derived:

 

Although this study focuses specifically on the association between the COVID-19 vaccination rate and trust in Japan’s central government during the COVID-19 pandemic, its theoretical framework can be applied to other social policy issues.

4. Data and methods

4.1 Data

To test the proposed hypotheses, I utilized data from the Online Panel Survey of Stratification and Social Psychology in 2021 (SSPW2021-Panel), conducted by the SSP project team (https://ssp.hus.osaka-u.ac.jp/) during the COVID-19 pandemic in Japan. The SSPW2021-Panel consists of four survey waves: March 2021 (Wave 1), July 2021 (Wave 2), November 2021 (Wave 3), and March 2022 (Wave 4). Respondents were selected based on cohorts, gender, and region of residence from online monitors registered with Neo-marketing Inc. (https://neo-m.jp/), a research agency in Japan. The survey targeted individuals aged 25 to 64 years.

Additionally, I used data on the number of newly infected COVID-19 cases per 100 capita at the prefecture level, corresponding to the four survey waves of the SSPW2021-Panel. Similarly, I used data on the number of COVID-19-vaccinated individuals at the prefecture level during the same periods. I collected the data on newly infected individuals for each prefecture from the online site of the Ministry of Health, Labor, and Welfare’s official website (https://covid19.mhlw.go.jp/extensions/public/en/index.html). Data on the number of vaccinated individuals were sourced from Japan’s Digital Agency website (https://www.digital.go.jp)

Data from the SSPW2021-Panel—COVID-19 infection rate at the prefecture level and COVID-19 vaccination rate at the prefecture level—were merged using respondents’ residential information. Consequently, each respondent for each survey wave was assigned two variables corresponding to their residential prefecture: the infection rate (number of newly infected per 100 capita over 1 month) and the vaccination rate (average percentage of vaccinated individuals over one month). Specifically, each respondent had four infection rate values (1st, 2nd, 3rd, and 4th) and four vaccinated rate values (1st, 2nd, 3rd, and 4th) for their residential prefecture. Using these data, the study examined the effects of changes in infection and vaccination rates at the prefecture level on changes in respondents’ evaluation of the infection control policies implemented by the central government and trust in the central government.

4.2 Variables

4.2.1 Dependent variables

Evaluation of infection control policies implemented by the central government and trust in the central government were used as the dependent variables. Respondents of the SSPW2021-Panel were asked to rate the infection control policies with the following question: “How do you rate the COVID-19 infection control policies so far, specifically those implemented by the central government?” Respondents were given four response options: Highly rated (= 4), Somewhat highly rated (= 3), Not highly rated (= 2), and Not at all highly rated (= 1). Similarly, trust in the central government was measured using the question: “How much do you trust the following organizations or institutions: the central government?” The response options were: Strongly trust (= 4), Somewhat trust (= 3), Not very trustable (= 2), No trust at all (= 1). This study treated these 4-point Likert scale variables as continuous variables (Sullivan and Artino 2013).

4.2.2 Independent variable

I used the vaccination rate at the prefecture level as the independent variable in this study. Here, the term vaccinated individuals refers to those who received their first dose of the COVID-19 vaccine during the corresponding period. This variable was employed to examine how and why the performance of the COVID-19 vaccination policies affects individuals’ evaluation of infection control policies and their trust in the central government. Specifically, the study focuses on the effects of the prefecture level vaccination rate on each respondent’s evaluation of the central government’s infection control policies and trust in the central government. It is important to note that all respondents within the same prefecture share the same vaccination rate value, regardless of whether or not they were vaccinated.

4.2.3 Control variable

In this study, I used the COVID-19 infection rate at the prefecture level as the control variable. The infection rate is defined as the number of newly infected individuals per 100 capita for each month corresponding to the survey waves. The levels of the COVID-19 pandemic at the prefecture level may affect changes in individuals’ evaluation of the central government’s infection control policies and changes in trust in the central government. Simultaneously, the levels of the COVID-19 pandemic at the prefecture level may be associated with the speed of the COVID-19 vaccination at the prefecture level, as a high infection rate increases the importance of vaccination policies in the prefecture. Therefore, when estimating the effects of vaccination rates on trust in the central government and evaluations of the central government’s infection control policies, it is necessary to control for the effects of infection rate.

4.3 Analytic strategy

To examine the hypotheses, I employ the two-way fixed effects ordinary least squares (OLS) regression model (Allison 2009). This model can be expressed by the following equation:

where $i$(= 1, …, n) denotes the individuals; $t(=1,…,T)$ denotes the points in time; $yit$ denotes the dependent variables; $x1it$ denotes the vaccination rate of the respondent’s prefecture; $x2$ denotes the infection rate of the respondent’s prefecture; $Zit$ denotes the vector of the time-invariant control variables; $μt$ denotes an intercept that may be different for each wave; $β1$⁠, $β2$⁠, and $β3$ are the coefficients of the independent variables; $γ$ is the vector of coefficients. Equation (4) has two error terms (⁠$αi$ and $εit$⁠). $εit$ is different for each individual at each time point, and $αi$ is different for each individual but same at each time point. If Hypotheses 1 and 2 derived from relative deprivation theory are supported, the quadratic term of the vaccination rate at the prefecture level will be significant.

I calculated the coefficients of the variables in the fixed effects OLS regression model using statistical software R (R Core Team 2018) and plm (Croissant and Millo 2008), which is a package of R. To account for the influence of social contexts at each time point on the dependent variable, time effects were controlled by specifying the option of the effects in plm as “twoways” when estimating the effects of the independent variables.

5. Results

5.1 Descriptive statistics

The descriptive statistics for the variables used in this study (trust in the central government, evaluation of the central government’s infection control policies, number of newly infected per 100 capita per month, and percentage of vaccinated individuals) and demographic characteristics (e.g., age and sex) are summarized in Table 1. The arithmetic mean of trust in the central government is 1.974, which is below 2.5 (=(1[min] + 4[max])/2), indicating that the trust of Japanese people in their central government during the COVID-19 pandemic was relatively low. Similarly, the arithmetic mean of the evaluation of the central government’s infection control policies is 2.116, also below 2.5. This indicates that Japanese residents did not positively evaluate the central government’s infection control policies and that relatively low evaluation of the infection control policies likely contributed to the observed distrust in the central government during the COVID-19 pandemic.

The number of newly infected individuals in the respondents’ prefectures varied across the survey waves. The mean number of newly infected cases was lowest during Wave 3 and highest during Wave 4. Although the average vaccination rates in the residing prefectures also differed across survey waves, the pattern of changes in vaccination rates did not correspond to fluctuations in the number of newly infected cases. The mean vaccination rate was lowest during Wave 1 and highest during Wave 4. Additionally, there were relatively large differences in the number of newly infected cases between prefectures, while differences in vaccination rates between them were relatively small. This suggests that the central government’s vaccination policies were implemented uniformly across the country, regardless of prefectural variation in COVID-19 infection rates.

The mean evaluation of the central government’s infection control policies at Wave 4 is 2.295, which is higher than the means for Waves 1 and 2. Notably, despite the central government’s failure to prevent the spread of COVID-19—as evidenced by the highest mean number of newly infected cases at Wave 4—people rated the central government’s infection control policies more favorably during this period. Similarly, the mean evaluation at Wave 2 is 1.899, lower than that of Wave 1. This occurred despite a sharp increase in the vaccination rate from Wave 1 to Wave 2, suggesting that the promotion of vaccination alone did not significantly influence public evaluation of the central government’s infection control policies. These puzzling trends in public evaluations of the central government’s infection control policies can be effectively explained using relative deprivation theory.

Notably, as respondents were drawn from monitors registered in a research agency, the SSMPW2021-Panel is not nationally representative. Table 1 shows that the panel includes a higher proportion of women than the general Japanese population. Additionally, the SSMPW2021-Panel is unbalanced, as some respondents dropped out between Waves t and t + 1, with younger respondents showing a higher tendency for attrition. To address potential attrition bias, I included all respondents with data from at least two waves in the two-way fixed effects OLS regression models (Müller and Castiglioni 2020). Moreover, this study confirmed the robustness of the analytical results by applying multiple imputation methods (see Supplementary Material for details).

5.2 Two-way fixed effects OLS regression model: infection control polices

The analytical results of the two-way fixed effects OLS regression model predicting the evaluation of the central government’s infection control policies are presented in Table 2. The data is an unbalanced panel since the number of respondents varied across survey waves. Model 1 in Table 2 includes only the vaccination rate at the prefecture level as a predictor of respondents’ evaluations of the central government’s infection control policies. Model 2 incorporates the vaccination rate and its quadratic term to capture potential nonlinear effects. Finally, Model 3 adds the infection rate at the prefecture level as an additional predictor, along with the vaccination rate and its quadratic term.

Based on the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), Model 3 is identified as the best-fitted model among the three. In Model 3, the coefficient for the vaccinated rate is statistically significant and negative, while the coefficient for the quadratic term of the vaccinated rate is statistically significant and positive. This indicates that the effect of the vaccination rate on the evaluation of the central government’s infection control policies is nonlinear. For the infection rate, the coefficient is statistically negative. This suggests that an increasing number of newly infected cases reduces public evaluation of the central government’s infection control policies.

As the quadratic term of vaccination rate is significantly positive, evaluation of the central government’s infection control policies initially declines with an increasing vaccination rate. However, once the vaccination rate surpasses the threshold value, evaluation of the central government’s infection control policies begins to improve with further increases in the vaccination rate. Notably, this nonlinear effect of the vaccination rate on respondents’ evaluations of the central government’s infection control policies remains evident even after controlling for the effects of the infection rate on the evaluation of the central government’s infection control policies. Thus, the nonlinear effect of the vaccination rate is independent of the infection rate.

5.3 Two-way fixed effects OLS regression models: trust in government

The analytical results of the two-way fixed effects regression models predicting the respondent’s trust in the central government are presented in Table 3. As with the models predicting evaluations of the central government’s infection control policies, Model 1 in Table 3 includes only the vaccination rate as a predictor of trust in the central government. Model 2 incorporates the vaccination rate and its quadratic term to account for potential nonlinear effects. Finally, Model 3 adds the infection rate as a predictor alongside the vaccination rate and its quadratic term.

Considering AIC, Model 3 is the best-fitting model among the three. However, based on BIC, Model 2 is considered the best-fitting model. Model 1 cannot be selected as the best-fitting model in either case. These results indicate that the effects of the quadratic term of the vaccination rate on trust in the central government cannot be ignored. However, it remains unclear whether the infection rate significantly affects trust in the central government. In Models 2 and 3, the coefficients for the vaccination rate are statistically significant and negative, while the coefficients for the quadratic term of the vaccination rate are statistically significant and positive. This demonstrates that the effect of the vaccination rate on trust in the central government is nonlinear.

Table 3 illustrates that an increase in the vaccination rate initially decreases trust in the central government. However, once the vaccination rate surpasses the threshold value, trust in the central government begins to increase with further increases in the vaccination rate. This indicates that the impact of vaccination policies on trust in the central government can be either negative or positive, depending on the vaccination rate. Notably, the variable effect of the vaccination rate on trust remains evident even after controlling for the effect of the infection rate on trust in the central government.

Model 4 in Table 3 incorporates the evaluation of infection control policies implemented by Japan’s central government as an additional variable. After controlling for the effects of evaluation, the effects of vaccination and infection rates at the prefecture level on trust in the central government weakened but remained statistically significant. These results suggest that while the relationship between vaccination rates and trust in the central government is mediated mainly by evaluations of infection control policies, a direct effect of vaccination rates on trust in the central government persists.

5.4 Changes in predicted values: infection control polices

Fig. 5 illustrates the changes in predicted values for the evaluation of the central government’s infection control policies as a function of the vaccination rate (Model 3 in Table 2). The figure shows that the predicted values initially decline but later increase as the vaccination rate rises, forming a U-shaped curve. As discussed earlier, the quadratic term of the vaccination rate in the two-way fixed effects regression model has a statistically significant and positive effect on the evaluation of the central government’s infection control policies. This suggests that, in the early stages of vaccination policy, relative deprivation experienced by unvaccinated individuals outweighed the relative privilege experienced by vaccinated individuals, as the unvaccinated represented the majority of the population. However, in the later stages of the vaccination policy, as the unvaccinated became a minority, the relative privilege of vaccinated individuals surpassed the relative deprivation of the unvaccinated. The critical point at which the effect of the vaccination rate on the evaluation of the central government’s infection control policies shifted from positive to negative is approximately 40 per cent. Given that the vaccination rate in Japan eventually saturated at around 80 per cent, this threshold corresponds to half of the potential population demanding COVID-19 vaccination. These findings align with the predictions derived from the relative deprivation theory.

5.5 Changes in predicted values: trust in government

Fig. 6 illustrates how the level of trust in the central government predicted by the two-way fixed effects OLS regression model (Model 3 in Table 3) varies with the vaccination rate. Fig. 6 demonstrates that predicted trust in the central government initially declines as the vaccination rate increases but subsequently rises, forming a U-shaped curve. Then, changes in trust in the central government mirror the U-shaped pattern observed in evaluations of the central government’s infection control policies. This suggests that trust in the central government and evaluations of its infection control policies are interrelated and influenced by its vaccination policy.

From the perspective of relative deprivation theory, trust in the central government weakens as the vaccination rate increases. This occurs because, during the early stages of the vaccination policy, the total effects of relative deprivation experienced by unvaccinated individuals outweigh the total effects of relative privilege experienced by vaccinated individuals, as the unvaccinated constitute the majority of the population. However, in the latter stages of the vaccination policy, as the unvaccinated become a minority, the total effects of relative privilege among vaccinated individuals surpass the total effects of relative deprivation among the unvaccinated. The critical threshold at which the effect of the vaccination rate on trust in the central government shifts from negative to positive is approximately 40 per cent, representing half of the potential population demanding COVID-19 vaccination.

Consequently, trust in the central government and evaluation of the central government’s infection control policies during the COVID-19 pandemic are curvilinearly associated with the COVID-19 vaccination rate. This curvilinear relationship can be rationally explained through the lens of relative deprivation theory. The effect of relative deprivation among unvaccinated individuals on trust in the central government outweighs the effect of relative privilege among vaccinated individuals. However, as the vaccination rate increases and the population of unvaccinated individuals declines, the total effects of relative deprivation among unvaccinated individuals in society diminish, leading to a shift in the association between vaccination rate and trust in the central government.

5.6 Robustness checks

However, the possibility that the significant curvilinear association between the vaccination rate at the prefecture level and the evaluation of the infection control policies implemented by Japan’s central government emerged coincidently due to other time-variant factors cannot be completely ruled out. To address this, I conducted regression models as a robustness check, using trust in mass media and social media as placebo outcomes. Fig. 7 presents the coefficient estimates for each model, including trust in social media, trust in mass media, trust in the central government, and evaluation of infection control policies. The nonlinear effect of the vaccination rate at the prefecture level was not observed for trust in social media. Furthermore, the nonlinear effect of the vaccination rate on trust in mass media was notably weaker than its effect on the evaluation of infection control policies. These findings support the validity of the curvilinear association between the vaccination rate at the prefecture level and evaluations of infection control policies. Moreover, they also suggest that trust in the government changes depending on the evaluation of vaccination policies.

An ordered logistic regression model was also conducted to predict evaluations of the vaccination policies implemented by the central government and trust in the central government (see Supplementary Material for details). The results of these analyses confirmed the validity of the explanation derived from relative deprivation theory.

6. Discussion and conclusions

The results of this study demonstrated that the association between the evaluation of the central government’s infection control policies and the vaccination rate at the prefecture level was contingent on the proportion of unvaccinated individuals. Similarly, the association between trust in the central government and the vaccination rate also depended on the proportion of unvaccinated individuals. Specifically, the evaluation of the central government’s infection control policies and the trust in the central government declined until a critical threshold of vaccination was reached, after which they began to increase. Hypothesis 1 predicted that trust in the central government and evaluation of the central government’s infection control policies would initially weaken with an increasing vaccination rate but eventually recover as the vaccination rate continues to rise. Thus, the findings support Hypothesis 1. Moreover, changes in trust in the central government followed a U-shaped pattern based on the vaccination rate. As outlined earlier with regard to the relative deprivation model, this pattern reflects the degree of deprivation experienced by unvaccinated individuals surpassing the privilege experienced by vaccinated individuals at the early stages of the vaccination policy rollout.

Additionally, the results indicated that evaluations of the central government’s infection control policies and trust in the central government were lowest when the vaccination rate reached approximately 40 per cent. Since the COVID-19 vaccination rate in Japan seems to have saturated at around 80 per cent, 40 per cent corresponds to half the population potentially demanding the COVID-19 vaccine. Hypothesis 2 proposed that when the vaccination rate reaches 50 per cent within the reference group in which individuals are eager to receive COVID-19 vaccination, the effect of the vaccination rate on trust in the central government will shift from negative to positive. Thus, the findings of this study align with Hypothesis 2.

Most residents of Japan were eager to receive the COVID-19 vaccine, making the vaccination policy desirable for the majority of the population. However, the vaccination policy implemented by the central government generated relative deprivation among unvaccinated individuals, temporally lowering the evaluation of the central government’s infection control policies. Furthermore, the analytical results reveal a significant association between the evaluation of the central government’s infection control policies and trust in the central government during the COVID-19 pandemic. Consequently, the performance of the central government’s infection control policies directly influenced trust in the government during this period. Additionally, changes in trust in the central government followed a U-shaped pattern, suggesting that the degree of deprivation experienced by unvaccinated individuals initially outweighed the privilege experienced by vaccinated individuals.

Thus, a form of relative deprivation temporally emerged in Japan during the COVID-19 pandemic. Notably, the emergence of relative deprivation among unvaccinated individuals meant that the COVID-19 vaccination policy, eagerly implemented by the central government, unintentionally and paradoxically affected trust in government. In other words, while the central government’s COVID-19 vaccination policies were significant and necessary, they temporally harmed the central government’s reputation. Although the negative effect of the central government’s vaccination policy on trust in the central government was temporary rather than permanent, policymakers should recognize that even desirable social policies can potentially damage trust. As such, careful consideration is required when implementing such policies. For example, delaying the distribution of the COVID-19 vaccine to mitigate relative deprivation among unvaccinated individuals could have let to a rapid spread of the pandemic. Therefore, despite the temporary reputational cost caused by relative deprivation, the central government in Japan was justified in promoting COVID-19 vaccination to contain the pandemic effectively.

This problem is not confined to the case of the COVID-19 vaccine. The phenomenon where pertinent policies may temporally generate negative societal outcomes can be observed across various fields. Social scientists and policymakers can gain valuable insights from the example of Japan’s vaccination policy during the COVID-19 pandemic. Moreover, this study has highlighted how the effects of vaccination were offset by heterogeneities among residents. Therefore, this finding provides a possible explanation for why trust in the government does not directly correlate with the implementation of its policies.

6.1 Limitations

This study has several limitations. First, it focused on the overall effects of the vaccination rate on the evaluation of the central government’s infection control policies and trust in the central government rather than examining individual-level differences in these effects. Specifically, differences in the impact of the vaccination rate between vaccinated individuals and unvaccinated individuals were not directly analyzed because the exact timing of first-dose vaccinations for each individual was not recorded in the SSPW2021-Panel. Thus, to better specify the relative deprivation experienced by unvaccinated individuals and the relative privilege experienced by vaccinated individuals, future research should directly examine individual-level effects of vaccination in relation to vaccination rates.

Second, this study considered prefectures as the reference group for generating relative deprivation among unvaccinated individuals and relative privilege among vaccinated individuals. However, it is possible that other criteria may be more suitable for examining changes in the effects of vaccination rates. For example, age cohorts could serve as a reference group, as the distribution timing of the COVID-19 vaccine voucher was determined based on age. Nevertheless, the age range of respondents in the SSPW2021-Panel (25–64 years) limited the scope of this study. As a result, differences in the effects of vaccination between individuals who were prioritized in the distribution of vaccine vouchers (those over 65 years old) and those who were not (those under 65 years old) could not be analyzed.

Third, the curvilinear associations between the dependent and independent variables might be related to individual characteristics (time-variant confounders). In the fixed effects OLS regression model, when the associations between the changes in dependent variables and those in vaccination rate are examined, all time-invariant confounders (including unobserved heterogeneity) are controlled. Conversely, the possibility that the time-variant confounders affect dependent variables cannot be denied even when using the two-way fixed effects OLS regression model. If the time-variant confounders affect the dependent variables, the curvilinear association between trust in the central government, the evaluation of the central government’s infection control policies, and the COVID-19 vaccination rate should be carefully examined.

Fourth, the possibility of a reversed causal relationship between the central government’s infection control policies and trust in the central government cannot be entirely ruled out. For example, the ruling party may have prioritized vaccination in regions with a high concentration of supporters. Although this possibility appears to be relatively low, it remains an important consideration for future research.

Fifth, this study was unable to include additional control variables, such as political party support, political ideologies, or political behaviors, as the SSPW2021-Panel did not collect data on these factors. Consequently, the possibility of potential bias in the estimated model due to the exclusion of these variables should also be considered.

6.2 Conclusions

In this study, I examined how and why the COVID-19 vaccination policy implemented by Japan’s central government reduced trust in the government during the early stages and enhanced trust during the later stages of the policy rollout. The findings confirmed that relative deprivation among unvaccinated individuals—and relative privilege among vaccinated individuals—can explain changes in trust in the central government during the COVID-19 pandemic. In the early stages of the vaccination policy, relative deprivation emerged among unvaccinated individuals, who constituted the majority of society at that time. As the number of vaccinated individuals increased, the sense of relative deprivation among the unvaccinated intensive, leading to a decline in trust in the central government. However, after a critical threshold was reached and the unvaccinated shifted from being the majority to the minority, relative deprivation among unvaccinated individuals diminished, while trust in the central government began to recover. This dynamic produced a U-shaped pattern of changes in trust in the central government depending on the COVID-19 vaccination rate. Similarly, relative privilege experienced by vaccinated individuals contributed to changes in the effect of the vaccination rate on trust in the central government. While relative privilege could potentially generate an inverted U-shaped pattern of changes in trust in the central government, the stronger effect of relative deprivation among unvaccinated individuals ensured that a U-shaped pattern was observed.

The COVID-19 vaccination policy implemented by Japan’s central government illustrates that even if a social policy met the demand of large parts of the public, it still might cause harm to the government’s reputation, if only temporarily. Therefore, policymakers should be cautioned to pay attention to the policy implementation process.

Acknowledgments

I would like to thank the anonymous reviewers of the Social Science Japan Journal for their helpful comments and suggestions. I would also like to thank the SSP Project for the permission to use the SSPW2021-Panel survey data.

Conflict of interest statement. There are no conflicts of interest to declare.

Funding

This research is supported by the Grants-in-Aid for Scientific Research of Japan Society for the Promotion of Science (19H00609, 21H00776, 23H00940, 24K05251).

Data availability

The SSP Project (http://ssp.hus.osaka-u.ac.jp/) allowed me to use the SSPW2021-Panel. The data is available upon request to the SSP Project.

Code availability

Contact the corresponding author.

References