3.1.1 Benchmark regression model.

In order to verify H1 and explore whether the digital governance level can affect the degree of political stability. The benchmark regression model in this study is set as follows:

(1)

Although the individual fixed effects model effectively addresses a major source of omitted variable bias by eliminating individual characteristics that do not change over time, such as a country’s historical culture and geographical environment, it cannot automatically guarantee that the error terms are homoscedasticity. The heteroscedasticity in the sample data of this study may be manifested as different variances of error terms in different countries (individuals). If heteroscedasticity is ignored, although the coefficient estimator remains unbiased and consistent, its standard error will be biased, which may lead to inaccurate confidence intervals. Therefore, we adopted a multi-level processing strategy: 1. Perform logarithmic transformation on the variables to alleviate skewed distribution and extreme values; 2. Using Cluster-Robust Standard Errors to deal with any form of intra-group correlation, that is, allowing for any form of correlation and heteroscedasticity among error terms within individuals (different years in the same country). Therefore, to more accurately measure the relative changes between variables and mitigate potential heteroscedasticity, logarithmic transformation was applied in this study. In addition, logarithmic calculation was not carried out for negative values. In model (1), ln PSit is the explained variable, indicating the degree of political stability of the country i in the year t; DGit represents the digital governance level of the country i in the year t; is the individual fixed effect respectively; ε it is the random disturbance term; β₀ represents the intercept term of the model; β₁ represents the estimation coefficient of the independent variable; Controlit represents a series of control variables, which include educational level (el), population size (ln ps) and the rule of law (ln rl).

3.1.2 Intermediate effect model.

To verify H2 and examine whether the anti-corruption efforts mediates the effect of the digital government on political stability, model (1) is combined with the stepwise regression method. The intermediate effect models in this study are set as follows:

(2)

(3)

In model (2), ln ACit indicated the anti-corruption efforts of the country i in the year t; α1 represents the estimated coefficient of the digital governance level. In model (3), λ1 represents the estimated coefficient of the digital governance level; λ2 represents the estimated coefficient of anti-corruption. According to the analysis steps of the stepwise regression method, the first step is to regress the benchmark model (1). If β1 is significant, the digital governance level significantly affects the degree of political stability. The second step is to regress model (2), and the intermediate variable is the anti-corruption efforts. If α1 is significant, then an increase in the digital governance level can enhance or reduce the anti-corruption efforts. In the third step, intermediate variables are introduced to the regression model (3). If λ1 is significant, the digital governance level can influence the degree of political stability in a country by influencing the strength of anti-corruption.

3.2.1. Explained variable.

This study focuses on the variable of political stability degree (ln PS). This study adopts Worldwide Governance Indicators (WGI), which was developed by the World Bank and is used to measure the comprehensive governance quality of various countries. The explained variable PS is derived from the “Political Stability and Absence of Violence/ Terrorism” indicator in WGI. It is worth mentioning that WGI is not only widely used in academic research, policy-making, and international comparison, but also provides an important reference for scholars to study governance issues, government policy-making, and international organizations to assess national development. However, the way WGI is measured is still being questioned. For example, indicators are too complex, there is a lack of time and space comparability, and there are hidden biases [41]. Undeniably, despite its limitations, WGI remains one of the authoritative indicator systems for measuring the state of global governance.

The development of WGI traces its origins to 1989 when the World Bank’s report “Sub-Saharan Africa: From Crisis To Sustainable Growth” was published, the term “Crisis of governance” was first mentioned to describe the challenges facing the region [42]. Subsequently, the World Bank developed WGI based on its definition of governance: “ ‘ the traditions and institutions by which authority in a country is exercised. ‘ “ [43] The WGI incorporates existing data sources from more than 30 think tanks, international organizations, non-governmental organizations, and private companies around the world, and WGI aggregates these indicators into six governance dimensions covering more than 200 economies. The six dimensions of governance include: Voice and accountability (VA), political stability and absence of violence/ terrorism (PS), government effectiveness (GE), regulatory quality (RQ), rule of law (RL), and control of corruption (CC). According to Kaufmann and Kraay [44], the measurement standard of indicator PS is “capturing perceptions and views of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including politically motivated violence and terrorism.” More precisely, The numerical calculation steps of each WGI’s indicator are as follows: (1) Fully absorb the information of each data source – complete the matching of the six indicators with other independent data sources; (2) Collate and standardize each source data, and control the value range between 0 and 1; (3) Adjust the source data using Unobserved components model (UCM) to ensure that the composite index value of each governance indicator is between −2.5 and 2.5. The closer the value is to 2.5, the higher the quality of the governance indicator is. On the contrary, it indicates that the quality of the governance indicator is lower.

3.2.2. Core explanatory variable.

The principal explanatory variable examined in this study is the digital governance level (DG). Government website construction, as the foundation and key form of expression in the field of digital governance, can significantly improve the working efficiency of the government and effectively improve the public image of the government [45,46]. It can be said that e-government is the key carrier of digital governance a significant part of digital governance, which can reflect the maturity of government digital services. Moreover, due to the differences in the degree, frequency, and way citizens use government websites, citizens’ perception of government transparency will vary, which will further affect their evaluation of government service capabilities and their political attitudes [47,48]. Thus, it can be seen that the Internet usage rate, as the fundamental support of digital governance, is an important condition for citizens to participate in government digital governance. Therefore, we combined the two indicators of e-government and Internet usage rate to construct a composite index to measure the level of digital governance.

The E-government indicator is derived from the Electronic Government Development Index (EDGI) constructed by the United Nations. EDGI is a comprehensive measurement index calculated by the weighted average of three standardized indicators: Online Service Index (OSI), Telecommunication Infrastructure Index (TII), and Human Capital Index (HCI). Three indicators are calculated as follows: OSI is an evaluation of government online services of countries, which divides online services into four development stages: information release stage, interaction stage, transaction processing stage, and seamless integration stage. The score of each country is standardized according to the stage reached by the government’s online services and the degree of completeness of the services, the value range is 0–1. TII collects the number of fixed telephone lines, the number of mobile phone subscriptions, the number of Internet users, broadband access speed, etc., normalizes the above data to eliminate the impact of dimensions, and assigns corresponding weights according to the importance of each indicator, then calculates the weighted average, and finally standardizes the score of each country, making its value range between 0 and 1. The HCI collects data on education indicators such as adult literacy rates, secondary mastery enrollment rates, and tertiary enrollment rates, as well as standardizing each country’s final score on a range of 0–1 [2].

Internet usage rate is derived from the “Facts and Figures” report constructed by the International Telecommunication Union (ITU). ITU obtains data from a variety of sources, including telecom operators, telecom/ICT regulators and national ministries of member States, demographic and health surveys, ICT Research Surveys in Africa, Asian LIRNE Survey, etc., and uses the population of each country as a weight, then calculate a weighted average of the proportion of individuals using the Internet in each country, estimates of regional and global internet use are produced [49].

The indicator of digital governance is a composite index calculated by the weighted average of standardized e-government and Internet usage rate, which is calculated as follows:

(4)

In model (4), DG represents the composite index – the digital governance level; weg and wiu are the weights of e-government and Internet usage rate, respectively; weg and wiu are standardized indicators of e-government and Internet usage.

3.2.3. Mediating variable.

This study used the anti-corruption efforts (ln AC) as the mediating variable, which is selected from the Corruption Perception Index (CPI) conducted by Transparency International (TI). Since 2012, the CPI has expanded access to data, including data from 13 different surveys and assessments from 12 agencies. Each country must be evaluated from at least three different sources before it can be included in the CPI. The data obtained will be standardized with a range between 0 and 100, with 0 being highly corrupt and 100 being very clean. On closer inspection, a score greater than 80 indicates relative integrity; A score between 50 and 80 indicates some level of corruption; A score between 25 and 50 indicates high levels of corruption; A score below 25 indicates high levels of corruption. Therefore, Consequently, an increase in the CPI score signifies improved anti-corruption performance.

3.2.4. Other variables.

Table 1 shows other variables, which include educational level (el), population size (ln ps), and the rule of law (ln rl), which are represented as control variables. To accurately disentangle the influence of digital governance level on the degree of political stability, this study refers to existing literature and includes the consciousness, social, and institutional factors that may have a potential impact on the degree of political stability in the control variables. Among them, EL represented the consciousness factor, and the data was selected from the World Bank; PS represented the social factor, again using data from the World Bank; RL is the judicial aspect of the system, and the data was selected from the WGI published by the World Bank.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Summary of measurements.

https://doi.org/10.1371/journal.pone.0338630.t001

4.2.1 The impact of digital governance on political stability.

This study used the individual fixed effect model to empirically test the relationship between the digital governance level and the degree of political stability. Specifically, this study used standardized weighted calculation of two indicators, which included e-government and Internet usage rate, to measure the digital governance level. Meanwhile, Political Stability and Absence of Violence/ Terrorism were to measure the degree of political stability. This study selected panel data from 112 countries during the period 2014–2023 to analyze the impact of digital governance level on the degree of political stability. Table 3 is the analysis result of benchmark regression. Column (1) lists the main explanatory variables used in this study to analyze how the digital governance level affects the degree of government stability, and other factors are not considered in this column. Columns (2) to (4) list the estimated results after each control variable is introduced based on the results in column (1). Column (5) lists the results of introducing a single fixed effect based on the above results.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Benchmark regression results.

https://doi.org/10.1371/journal.pone.0338630.t003

The estimated results in Table 3 showed that when the estimated coefficient of digital governance level is the only explanatory variable, the result is positively significant at the 10% level; When controlling variables are combined, the estimated coefficient of digital governance level was positively significant at 1% level; When combined with the individual fixed effect, the estimated coefficient of the digital governance level was positively significant at the 5% level, while the constant term is significant at the 1% level, indicating that the individual fixed effect captures part of the individual heterogeneity. Overall, the digital governance level was positively correlated with the degree of political stability, which verifies H1. In other words, as the digital governance level rises, the degree of political stability increases, which further supports the theoretical reasoning presented in this study. By the current international trend, strengthening the construction of digital infrastructure and improving the digital governance level will enhance the degree of political stability to a certain extent. At the same time, improving the digital governance level can improve public satisfaction by optimizing service supply, and it provides a foundation for countries to jointly address global challenges by deepening international cooperation. Therefore, accelerating the comprehensive and in-depth application of digital technologies, strengthening international digital cooperation, and narrowing the digital divide are international challenges that all countries should overcome in the future.

4.2.2 Robustness test.

This part tested the robustness of the influence of digital governance level on the degree of political stability. Specifically, this study tested the robustness of benchmark regression by adjusting the sample period, and the test results are shown in Table 4. The original sample time range of this study was 2014–2023. We shortened the original sample time range to 2015–2022, and the benchmark model was regressed again for the robustness test. Table 4 column (1) shows the regression result of the influence of digital governance level on the degree of political stability. The coefficient of the digital governance level is significantly positive at the 5% level, which is consistent with the benchmark regression result, and H1 has been verified.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Robustness test.

https://doi.org/10.1371/journal.pone.0338630.t004

Considering the changes in the global political stability situation in 2016, such as the US election, the turmoil in the Middle East, the political scandal in South Korea, the terrorist attacks in many places in Europe, and other events that have had an enormous impact on the degree of political stability, as well as the persistence of these events, we conducted a robustness test by eliminating samples. In detail, samples from 2016 to 2018 were removed, and the benchmark model was regressed again. The estimated results are shown in column (2) of Table 4. The influence of digital governance level on the degree of political stability is still positive and significant, which once again proves the robustness of the model.

4.2.3 Heterogeneity analysis.

There is a gap in the level of economic development of different countries. Therefore, this study divided the sample into four groups based on the World Bank’s income grouping criteria for economies in 2020 [50]. The four groups are: low-income economies ($1035), lower-middle-income economies ($1036 - $4045), upper-middle-income economies ($4046 - $12535), and high-income economies ($12536), and regression was performed for each group. The regression estimation results are listed in Table 5 columns (1) – (4). The results show that the effect of digital governance level on the degree of political stability is significant in lower-middle-income economies and high-income economies, but not in low-income economies and upper-middle-income economies. The reason for this outcome may be that lower-middle-income economies are in economic transition, and the usage of digital governance can bring about significant changes that affect political stability. At the same time, the digital governance level in high-income economies is more developed than in the other three economies. As a result, any adjustment could have a significant impact on the degree of political stability. In contrast, low-income economies have weak digital infrastructure and limited capacity for digital governance, so the effect on political stability is less. However, digital governance systems in upper-middle-income economies are relatively mature but have not reached a certain level, so the impact on the degree of political stability may level off.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Heterogeneity analysis of economic development levels.

https://doi.org/10.1371/journal.pone.0338630.t005

The impact of digital governance level on the degree of political stability shows regional heterogeneity. As the research sample consists of 112 countries, which vary greatly in culture, system and development stage, it is difficult to capture their complex and multi-dimensional relationships by directly conducting full sample regression or simple regional dummy variable interaction terms. Therefore, this study adopted a stratified systematic sampling method to strategically reduce the initial sample. First, we stratify the overall sample based on two dimensions highly relevant to theoretical research: 1. Geographical region. According to the United Nations geoscheme [51], countries are divided into five continents (Asia, Europe, Africa, America, and Oceania) to control the differences in macro historical and cultural backgrounds. 2. Digital governance level. Based on the existing data, the average value (0.0004932), minimum value (−2.023), and maximum value (1.382) of digital governance levels in various countries were obtained and segmented in an equal-width manner. Countries within each continent are classified into three levels: low digital governance level (within the range of [−2.023, −0.888]), medium digital governance level (within the range of [−0.888,0.247]), and high digital governance level (within the range of [0.247,1.382]). Secondly, within each “regionally Digital governance level” cell formed by the intersection of the above two dimensions (for example, “Asia - High Digital Governance Level”), we aim to extract as many equal numbers of countries as possible. This “equal division” principle ensures that our sub-samples are not dominated by a specific group (such as European countries with high digital governance), thereby guaranteeing that each level of digital governance is evenly represented across different regions. This is crucial for fairly testing regional differences. Thirdly, through the above procedures, we ultimately obtained a highly representative balanced panel sub-sample covering all regions and digital governance levels around the world, which is approximately one-third of the original sample. We admit that this sacrifices a portion of the sample size, but we believe that this trade-off brings a key advantage to the research, namely enhanced comparability. Because if only the full sample is used, the research finds that it might be dominated by specific regions with extremely high or low levels of digital governance and large sample sizes (such as Europe or Africa), while this method effectively ‘matches’ countries in different regions that are at similar stages of digital governance development. For instance, we can representatively compare “high digital governance countries in Africa” with “high digital governance countries in Asia”, thereby more purely identifying the “regional” effect while controlling for the variable of “digital governance level”.

The regression estimation results are listed in columns (1) to (3) of Table 6. The results show that the impact of digital governance level is “context-dependent” rather than universal. In column (1) of Table 6, the impact of digital governance level is negative but not significant. This indicates that in this group of countries, digital governance level has not effectively promoted stability and may even have a slightly negative impact by amplifying social conflicts or being abused. Another factor that also has a negative impact but is very significant is the education level, which conforms to the assumption of the “Malthusian trap” [52]: In the case of an improvement in the level of education but no improvement in the opportunity structure, civic awakening may give rise to dissatisfaction with the current situation. However, the influence of the rule of law in this context is positive but not significant, indicating that in these countries, the rule of law itself may not be sound enough to serve as a stable cornerstone. Therefore, this group of countries may be confronted with deep-seated problems such as weak national capacity, incomplete systems or social divisions. The fundamental social and political issues (such as fairness, opportunity, and security) could be the primary contradictions among these countries. The “superstructure” such as digital governance is difficult to play an active role and may even exacerbate the contradictions due to improper use. For example, in this group of samples, Haiti’s national conditions are consistent with this result. According to the EDGI report (2025), “As in 2022, Haiti remains at the lowest EGDI level in the region with ongoing political crises and conflicts severely undermining efforts to create a stable and effective digital infrastructure.” In column (2) of Table 6, the impact of digital governance level on political stability degree is positive and significant. However, unlike the previous group of countries, the influence of the rule of law level in this group of countries is no longer significant. This does not mean that the rule of law is unimportant; rather, it indicates that the rule of law has become a common “infrastructure” in this group of countries, with a relatively small degree of variation. Therefore, the role of digital governance as an “efficiency enhancer” has been highlighted. For instance, the situation of Bangladesh in this sample is rather typical. The country is orderly moving on the path of sustainable development, integrating digital technology with people’s livelihood and economic development, providing stable jobs, promoting industrial innovation and stabilizing infrastructure [53]. In column (3) of Table 6, the impact of digital governance level on national political stability is no longer significant. In other words, when all countries are mixed together to calculate the “average effect”, the net effect of digital governance level disappears. This precisely proves that its positive effect column (2) and negative effect column (1) cancel each other out in different types of countries. This is also the most powerful evidence to prove the existence of regional differences. This result also confirms the “threshold effect”. For digital governance to have a positive impact, it needs to cross a key “institutional threshold”. The countries in column (1) are below this threshold. The main contradiction lies in fundamental institutional issues (such as the lack of effective rule of law and social injustice), and the effect of digital governance is not significant. The countries in column (2) have crossed this threshold and thus can reap the “digital dividend”. Overall, whether the level of digital governance can enhance political stability depends on the institutional environment of the country.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Heterogeneity test of digital governance levels.

https://doi.org/10.1371/journal.pone.0338630.t006

4.3.1 The mediating effect of anti-corruption efforts.

Based on the theoretical analysis above, the rise in the digital governance level can not only promote the improvement of the degree of political stability but also may be conducive to anti-corruption, and thus affect the degree of political stability. In other words, not only does the digital governance level have a direct impact on the degree of political stability, but the digital governance level also has an indirect impact on the degree of political stability by influencing the anti-corruption efforts. Therefore, this research studied the mediating role of the anti-corruption efforts, and the test results of this analysis are shown in Table 7.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Mediating effect test.

https://doi.org/10.1371/journal.pone.0338630.t007

According to the test procedure of the stepwise regression method, it is shown in column (1) of Table 7 that the estimated coefficient of digital governance level is positive and significant, indicating that with the improvement of digital governance level, the degree of political stability increases accordingly. Therefore, the total impact of the digital governance level on the degree of political stability is significant, confirming H1. Digital governance is a means for governments to use technology to change governance methods and improve government efficiency. Consequently, through digital governance, governments can use digital tools such as artificial intelligence, big data, etc., to optimize decision-making and improve decision-making efficiency. At the same time, from the perspective of national development, digital governance can promote the participation of social organizations, citizens, and enterprises in political life and enhance social cohesion. In terms of a country’s security, digital governance can improve monitoring capabilities, strengthen early warning capabilities, and formulate effective measures in advance to avoid external attacks and internal disharmony. All of these effects of digital governance can enhance the degree of political stability.

Column (2) of Table 7 shows the impact of digital governance on the anti-corruption efforts. The result indicates that the higher the digital governance level in a country, the higher its integrity level. It can be said that digital governance can reduce corruption. The underlying logic is that digital governance through the use of information technology, enhances government transparency, optimizes government processes, breaks down information barriers among government departments, and promotes the development of the Internet and social media, thereby increasing information transparency, optimizing the operation of power, achieving data sharing, strengthening social supervision, and ultimately enhancing the intensity of anti-corruption efforts.

Column (3) of Table 7 shows the mediating effect of the anti-corruption efforts on the impact of the digital governance level on the degree of political stability. The anti-corruption efforts coefficient is positively significant, indicating that there is a mediating effect on the anti-corruption efforts, which verifies H2. To be precise, digital governance enhances the degree of political stability by raising the strength of anti-corruption. From an international perspective, corruption may undermine the international economic order [3]; From a national perspective, corruption can lead to economic instability in a country [54]; From a citizen’s point of view, corruption can undermine political trust [27], and all of these hazards listed above can contribute to political instability. However, digital governance can reduce the degree of political instability by mitigating these hazards.

4.3.2 Robustness test of mediating effect.

To test the robustness of the mediating effect of the anti-corruption efforts, this study conducted the Sobel test and Bootstrap test on the regression results respectively, and the test results are shown in Table 8. Column (1) in Table 8 is the Sobel test result, which indicates that the R-squared of the sober test is 0.631, and the imitative effect is relatively good. Moreover, the anti-corruption efforts coefficient is 0.379, and is significant at the 1% level, indicating that the anti-corruption efforts does have an intermediary effect and is robust.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Robustness test of mediating effect.

https://doi.org/10.1371/journal.pone.0338630.t008

The results of the Bootstrap test are listed in column (2) of Table 8. The results showed that Bootstrap 1 and Bootstrap 2 are significantly positive at the 1% level. Thus, digital governance can strengthen the degree of political stability by raising the strength of anti-corruption, once again verifying H2.

Although this study employed robust statistical tests such as Sobel and Bootstrap to identify mediating effects, a thorough discussion of the inherent limitations of the data itself is crucial for the correct interpretation of the research conclusions. First, the measurement of digital governance level faces inherent challenges. Many mainstream digital governance indices (such as the UN’s EGDI) rely on expert surveys and business questionnaires. These data often reflect the “perception” of business executives or specific expert groups, rather than the actual usage experience of ordinary citizens or the objective effectiveness of the government’s back-end operations. This “elite perspective” may lead to a systematic overestimation of the governance levels of countries that have invested more in digital propaganda but have insufficient actual inclusiveness. Meanwhile, existing indices usually focus on the breadth of online service supply and digital infrastructure, but they do not adequately measure the “quality” dimension of digital governance. As a result, our research may capture the “efficiency” aspect of digital governance, but fail to fully capture its “fairness” or “rights” aspect, the latter of which may have a more complex relationship with political stability. Second, the degree of political stability encompasses various aspects. For instance, the “political stability and no violence/terrorism” indicator in the WGI used in this study. However, such measurements may focus more on “the stability of order” or “the stability of regime”, and the capture of “the stability of the system” – that is, the long-term resilience and legitimacy of the political system itself – is relatively indirect. Therefore, our research finds that digital governance can promote stability, but the nature of this stability requires careful interpretation. The limitations of the above-mentioned data are not the fatal weakness of our research, but rather a common challenge faced by all cross-border quantitative political economy studies. They mainly affect the accuracy of our conclusion and the boundary conditions. Therefore, our conclusion should be understood as -- there is a connection between digital governance and political stability through a specific mechanism under the current mainstream measurement system.