Reexamining the economic globalization-welfare state nexus: a Bayesian mixed approach to linear and non-linear dynamics

Nga, Nguyen Thi Ngoc; Thach, Nguyen Ngoc Article Reexamining the economic globalization-welfare state nexus: a Bayesian mixed approach to linear and non-linear dynamics Cogent Economics & Finance Provided in Cooperation with: Taylor & Francis Group Suggested Citation: Nga, Nguyen Thi Ngoc; Thach, Nguyen Ngoc (2024) : Reexamining the economic globalization-welfare state nexus: a Bayesian mixed approach to linear and non-linear dynamics, Cogent Economics & Finance, ISSN 2332-2039, Taylor & Francis, Abingdon, Vol. 12, Iss. 1, pp. 1-14, https://doi.org/10.1080/23322039.2024.2416987 This Version is available at: https://hdl.handle.net/10419/321632 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Cogent Economics & Finance ISSN: 2332-2039 (Online) Journal homepage: www.tandfonline.com/journals/oaef20 Reexamining the economic globalization-welfare state nexus: a Bayesian mixed approach to linear and non-linear dynamics Nguyen Thi Ngoc Nga & Nguyen Ngoc Thach To cite this article: Nguyen Thi Ngoc Nga & Nguyen Ngoc Thach (2024) Reexamining the economic globalization-welfare state nexus: a Bayesian mixed approach to linear and non-linear dynamics, Cogent Economics & Finance, 12:1, 2416987, DOI: 10.1080/23322039.2024.2416987 To link to this article: https://doi.org/10.1080/23322039.2024.2416987 © 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Published online: 17 Oct 2024. Submit your article to this journal Article views: 485 View related articles View Crossmark data Citing articles: 1 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=oaef20 GENERAL & APPLIED ECONOMICS | RESEARCH ARTICLE Reexamining the economic globalization-welfare state nexus: a Bayesian mixed approach to linear and non-linear dynamics Nguyen Thi Ngoc Nga and Nguyen Ngoc Thach Department of Business Administration, Ho Chi Minh University of Banking, Ho Chi Minh City, Vietnam ABSTRACT Previous research on the globalization-welfare state nexus has typically been conducted within a linear monotonic framework, focusing on two primary hypotheses. The efficiency hypothesis predicts that globalization shrinks the welfare state’s size. In contrast, the compensation hypothesis argues that globalization increases demand for social security, leading to an expanded welfare state. Empirical evidence on this relationship is mixed, with multicollinearity suggested as one possible explanation. This study explores the non-linear, non-monotonic aspects of this nexus using a yearly panel of 10 ASEAN countries from 1970–2019, analyzed through a Bayesian mixed regression. The study focuses specifically on the effects of economic globalization. The results demonstrate a non-linear, non-monotonic relationship characterized by an inverted U-shape, where economic globalization initially expands the welfare state but, after reaching a certain point, leads to its retrenchment. This outcome supports a combination of both the compensation and efficiency hypotheses, aligning with the principles of non-linear and complex system sciences. The study offers a solid and reliable foundation for predicting long-term globalization processes and formulating comprehensive integration policies in ASEAN in the context of declining public expenditures. IMPACT STATEMENT The study aims to explore the monotonic and non-monotonic relationships between economic globalization and the welfare state in ASEAN. Unlike similar research, this study employs a Bayesian hierarchical approach to handle multicollinearity, revealing a non-linear and non-monotonic effect of economic globalization on the welfare state, consistent with the Armey curve. The findings provide a robust foundation for formulating long-term globalization policies for ASEAN countries. ARTICLE HISTORY Received 11 July 2024 Revised 9 September 2024 Accepted 10 October 2024 KEYWORDS Linear monotonic; nonlinear non-monotonic; economic globalization; welfare state; Bayesian mixed regression; ASEAN SUBJECTS Economics; Finance; International Relations; International Political Economy JEL CODES F62 1. Introduction There are two prevailing perspectives on how economic globalization, or globalization more broadly, impacts fiscal autonomy, often representative of the welfare state in theoretical and empirical analyses. The first perspective, known as the efficiency hypothesis, documents that free trade and the mobility of production factors increase competition among national governments, pressuring them to reduce tax rates and fiscal power (Ursprung, 2008). On the contrary, the compensation hypothesis posits a positive relationship between economic liberalization and government expenditure (Rodrik, 1998), arguing that globalization heightens external risks, necessitating higher public spending as a form of social insurance. Empirical studies investigating the link between economic globalization and the welfare state have typically relied on a linear, monotonic approach, resulting in findings consistent with one of the aforementioned hypotheses. However, these studies often produce mixed outcomes due to limitations such as small sample sizes, a narrow focus on specific dimensions of economic globalization, or, more importantly, the biases inherent in Newtonian classical sciences. First, prima facie evidence suggesting a CONTACT Nguyen Ngoc Thach [email protected] Ho Chi Minh University of Banking, Ho Chi Minh City, Vietnam This article has been corrected with minor changes. These changes do not impact the academic content of the article. ß2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent. COGENT ECONOMICS & FINANCE 2024, VOL. 12, NO. 1, 2416987 https://doi.org/10.1080/23322039.2024.2416987 negative relationship between globalization and government expenditure is evident from the rapid increase in international trade and capital mobility during the period of ‘High globalization’(Milanovic, 2016)or‘New globalization’(Baldwin, 2016) starting from 1990. This period coincided with an aggregate decline in government consumption as a share of GDP. These observed trends indicate a possible reversal in the relationship between globalization and the welfare state. Thus, by utilizing a larger data sample and embracing the philosophy of nonlinear science and complex systems, we anticipate a nonlinear, non-monotonic relationship between globalization and the welfare state. It is noteworthy that the linear, monotonic assumptions of classical science have been fundamental to the social sciences since Newton’s publication of Principia. Due to flaws in models for forecasting, the uncertainties of linearity prompted the development of the first mathematical model of nonlinear phenomena in the 1960s (Burke, 2009). To our knowledge, aside from the recent assessment by Yong & Chi (2022), no studies have explored the non-linear, non-monotonic relationship between economic globalization and the welfare state. Additionally, no research has specifically focused on ASEAN countries (Anderson & Obeng, 2021; Busemeyer, 2009; Cameron, 1978; Dreher et al., 2008; Garrett & Mitchell, 2001; Rodrik, 1997; Yong & Chi, 2022, and many others). The ASEAN community is attracting growing interest for several reasons, including the involvement of many countries in the region in global trade agreements, the implementation of the ASEAN Economic Community (AEC), high economic growth rates, and the region’s proximity to two major economic centers, China and India (Jamhari et al., 2021). Second, economic indicators of globalization are popularly utilized due to their perceived strong impact on government size and the welfare state. It is worth mentioning that economic globalization is a complex phenomenon that cannot be sufficiently represented by a single indicator like financial or trade liberalization. This complexity prompted the creation of the KOF globalization indices by the Swiss Institute of Economics (KOF stands for ‘Konjunkturforschungsstelle’). Finally, most importantly, earlier frequentist research encounters issues with multicollinearity, where some covariates are strongly correlated, leading to bias. To address multicollinearity, this research employs a Bayesian mixed (or hierarchical) regression. The superiority of the Bayesian framework over frequentist analysis in solving multicollinearity is highlighted by Block et al. (2011), Winship & Western (2016), and recently, Dan & Thach (2024). Furthermore, the mixed-effects approach allows for capturing variations across units (Nezlek, 2008). From the above discussion, we pose a substantial question: Does a non-linear, non-monotonic or a linear, monotonic relationship exist between economic globalization and the welfare state in an ASEAN sample? The study aims to provide insights into both the linear, monotonic and non-linear, non-monotonic effects of economic globalization on the welfare state using a panel of 10 ASEAN countries from 1970 to 2019. Compared to previous research, we utilize a longer time frame to examine the relationship beyond the traditional monotonic framework. The findings reveal a non-linear, non-monotonic linkage between economic globalization and the welfare state, supporting both the compensation and efficiency hypotheses at different stages of globalization. This aligns with modern philosophical perspectives that emphasize the non-linearity and non-monotonicity in relations among phenomena and processes in an increasingly globalized world (Burke, 2009). The results obtained in this study on non-linear, nonmonotonic patterns contribute significantly to the in-depth understanding of globalization processes’ impacts on national economies. The rest of the study is structured as follows: Section I discusses the motivation, aim, and contribution of the study. Section II reviews the analytic framework and empirical literature. Section III introduces the Bayesian mixed technique, describes the data, and specifies the baseline model. The main results are presented and discussed in Section IV, followed by the conclusion in Section V. 2. Literature review Theoretical and empirical literature on the globalization-welfare state relations is characterized by conflicting, inconsistent and inconclusive arguments and findings. Conceptually, there are two contradictory hypotheses: compensation and efficiency. Empirical analyses of the association between globalization and the welfare state (using government size as a proxy for the welfare state) provide mixed evidence. The first generation of studies, emerging in the 1990s, focused primarily on the net effect of globalization on the welfare state employing aggregate data, yielding ambiguous outcomes. In the meantime, 2 N. NGUYEN THI NGOC AND T. NGUYEN NGOC the second generation of studies applied more advanced quantitative techniques and achieved more convincing results, utilizing disaggregated data and new variables representing the scale of the welfare state and globalization. However, no final consensus has been reached regarding the relationship between globalization and the welfare state. Reviewing 19 research papers published between 1995 and 2006, Gemmell et al. (2008) found an equal number of studies reporting positive and negative correlations between trade or capital flow openness and government expenditure. More recently, Anderson & Obeng (2021), reviewing 13 empirical works published since 2006, revealed similarly mixed results. Several factors contribute to these conflicting results, such as small sample sizes, various model specifications (Anderson & Obeng, 2021), and different measures of globalization (Ashraf et al., 2017; Kimakova, 2009; Meinhard & Potrafke, 2011), and diverse samples and observations (Jetter & Parmeter, 2015). Most crucially, conventional frequentist techniques implemented in previous studies suffer from multicollinearity, where the key independent variable, globalization, and primary covariates are strongly correlated. Multicollinearity can lead to biased estimation (Dan & Thach, 2024). Additionally, previous analyses focused on linear, monotonic connections, ignoring the possibility of non-linear, non-monotonic ones. A monotonic relationship is one where variables move in a consistent direction relative to each other: as one variable increases or decreases, the other variable also increases or decreases, but not necessarily at a constant rate. In contrast, a non-monotonic relationship allows the direction of the relationship between the variables to change. This means that as one variable increases or decreases, the other variable does not consistently increase or decrease. A relationship, whether linear or non-linear, can be either monotonic or non-monotonic. Negative linear, monotonic relations align with the efficiency hypothesis. As emphasized by the classical approach of political economy, exemplified by Adam Smith and David Ricardo, increasing trade and financial liberalization enhances competition among national governments. The efficiency hypothesis posits that trade openness and factor mobility lead to a reduction in government size, lower tax revenue, and decreased government expenditure, ultimately eroding the welfare state (Schulze & Ursprung, 1999). During the globalization process, governments strategically compete in international markets to attract foreign direct investment (FDI), skilled labor, and high-tech advancements. Consequently, globalization imposes constraints on national fiscal autonomy, narrows taxation bases, reduces government expenditure, and alters budget composition. By its nature, globalization renders all nations ‘smaller.’ Small, open economies experience more intense international competition compared to larger countries, leading them to depend less on capital gains and corporate taxes. Consequently, as predicted by tax competition literature, a global race to the bottom in tax rates ensues (Bretschger & Hettich, 2002; Devereux & Griffith 1998). From this perspective, a negative linear, monotonic connection exists between economic globalization and the welfare state. Empirical research supporting the efficiency point of view reveals negative relationships between globalization and the welfare state. For example, Cusack & Garrett (1992) identified a negative relationship between international financial integration and the growth of government spending (excluding defense spending) in a panel sample of 16 OECD countries from 1955 to 1989. Rodrik (1997), utilizing a yearly panel dataset from OECD countries spanning 1966– 1991, explored a negative correlation between lagged trade liberalization and both social expenditure and total government expenditure. His findings suggested that in countries and periods with fully open capital accounts, trade openness has a more significantly negative correlation with social spending and total government expenditure compared to countries with capital restrictions. Similarly, Garrett & Mitchell (2001), analyzing a sample of 18 OECD countries during 1961–1993, concluded that trade liberalization negatively affects government expenditure, though not government consumption and social security transfers. Using a sample of 17 OECD countries from 1961 to 1993, Kittel & Winner (2005) found that FDI is negatively related to the welfare state, as measured by total government expenditure as a share of GDP. However, the effects of other variables, such as trade openness and imports from lowwage countries, are non-significant. More recently, Busemeyer (2009) also discovered a negative relationship between trade liberalization and government expenditure in an OECD sample. Conversely, the compensation perspective suggests a positive linear, monotonic relationship, indicating that government scope is broader in more open economies. This perspective posits that increased international integration heightens demand for social security, prompting governments to expand social welfare expenditures (Rodrik, 1998). Greater international integration increases external risks, leading to COGENT ECONOMICS & FINANCE 3 instability and job fluctuations, thereby raising the demand for social insurance and compensation for those adversely affected by globalization (Kim, 2009; Rodrik, 1998). Politicians respond to these demands by increasing social welfare spending to secure re-election (Dreher et al., 2008; Ursprung, 2008). Providing strong support for this hypothesis, Cameron (1978) highlighted that increased trade dependence significantly contributed to the growth of the public sector in 18 industrial countries between 1960 and 1975. Similarly, Rodrik (1998) identified a strong positive influence of trade liberalization on government consumption in his analysis of 125 countries from 1985–1989 and 103 countries from 1990–1992. Vaubel (2000,2005), Meinhart & Potrafke (2011), and Ferreira & Oliveira (2019) also found positive correlations between globalization and social spending. Specifically, Meinhart & Potrafke (2011) revisited the evidence by analyzing an annual panel dataset of 186 countries from 1970 to 2004, utilizing data from the Penn World Tables on government sector size and the KOF index of globalization. Their results indicate that globalization led to an increase in government sectors worldwide. In addition to the two aforementioned directions in the literature, some studies report mixed or nonsignificant findings. Kaufman & Segura-Ubiergo, utilizing data from 14 Latin American countries between 1973 and 1997, demonstrated that trade openness and capital liberalization negatively impacted social security spending. However, they found that capital liberalization positively influenced education and health spending. Bretschger & Hettich (2002) concluded that while the efficiency hypothesis explains tax patterns, the compensation hypothesis explains spending patterns in 14 OECD countries from 1967– 1996. In a reevaluation of Garrett & Mitchell (2001) using advanced methodologies, Kittel & Winner (2005) found no support for either the efficiency or compensation hypothesis. Instead, their findings indicate that domestic economic and demographic factors, such as unemployment rates and dependency ratios, have a more significant influence on government expenditures. Dreher et al. (2008) found that globalization does not impact budget composition, using data from 60 countries (1971–2001) and 10 OECD countries (1991–2000) with four different proxies for globalization. Recent findings showed that there is no substantial evidence for a global race to the bottom in tax rates (Pl€ umper et al., 2009). Ashraf et al. (2017) found that greenfield FDI had a positive effect on government consumption, while M&A FDI had no effect, based on a sample of 130 developed and developing countries from 2003 to 2011. In a study by Kim et al., trade globalization positively influenced total government spending, whereas financial globalization had a negative impact, in an analysis of 53 OECD and non-OECD countries from 1980 to 2011. Anderson & Obeng (2021) revealed that de jure trade globalization generally increased consumption spending, aligning with the compensation hypothesis. In contrast, de jure financial globalization tended to decrease consumption spending, supporting the efficiency hypothesis. Most recently, Yong & Chi (2022) reassessed the impact of trade openness and financial liberalization on the welfare state, finding an inverted U-shaped effect of trade liberalization on the welfare state but no large influence from financial liberalization. Tekin et al. (2023) found support for both the compensation and efficiency hypotheses in 11 transition economies in the Central and Eastern European (CEE) region from 1996 to 2021. From the above analyses, three gaps are identified: i. Earlier studies have primarily focused on linear, monotonic relationships between overall globalization or its sub-indices and the welfare state, overlooking potential non-linear, non-monotonic interactions. The divergent and even contradictory trends of economic globalization and government spending—with increasing economic openness and a shrinking share of government consumption in GDP over recent decades—suggest a non-linear, non-monotonic relationship between these processes. Thus, we hypothesize a non-linear relationship following an inverted U-shaped pattern, as modern complex sciences suggest. ii. The conventional frequentist techniques employed in a series of prior globalization-spending studies cannot effectively handle multicollinearity issues, leading to biased and non-robust outcomes. ‘Just as exact collinearity leads to non-identification of parameters, high collinearity can be considered as weak identification of the parameters’(Pesaran & Smith, 2019). iii. There is a lack of empirical analysis on the link between economic globalization and the welfare state in the ASEAN context, making it challenging to design a comprehensive globalization strategy for these nations. 4 N. NGUYEN THI NGOC AND T. NGUYEN NGOC 3. Bayesian mixed regression, model and data Although the implementation of Bayesian techniques for data analysis has revolutionized many fields, from genetics to macroeconomics since the 1990s, the Bayesian framework has been notably absent in empirical research on the nexus between globalization and the welfare state. In general, our work introduces researchers in this area to Bayesian estimation to address model uncertainty arising from statistical complexities. Bayesian analysis offers a more flexible and intuitive framework for dealing with uncertainty and overcomes several drawbacks of the frequentist approach (Thach et al., 2019; Briggs, 2023). Most notably, Bayesian inferences provide more robust estimates by combining prior knowledge with available data and offering an entire posterior distribution of the parameters. In Bayesian inferences, the data x, which a realization of XfðxjHÞ, is integrated with prior information encapsulated in a prior distribution with density pH ðÞ :This integration results in the posterior distribution pHjx ðÞ , summarizing the updated beliefs. The posterior distribution is derived from the joint distribution fðxjHÞpH ðÞ using Bayes’ formula: pHjx ðÞ ¼fxjH ðÞ pH ðÞ ÐfðxjHÞpH ðÞ dH(1) Where mx ðÞ ¼ÐfðxjHÞpH ðÞ dH is the marginal density of X(see Berger, 1985; Bernardo & Smith, 1994; Robert & Casella, 2004). Specifically, Bayesian methods handle multicollinearity more effectively than frequentist methods (Block et al., 2011; Dan & Thach, 2024; Pesaran & Smith, 2019; Winship & Western, 2016). In frequentist regression, exact collinearity between regressors makes their individual coefficients unidentified (Pesaran & Smith, 2019), leading to unstable coefficient estimates and inflated standard errors. With collinearity, small deviations from this assumption can cause significant changes in estimates (Winship & Western, 2016). As emphasized by Pesaran & Smith, (2019), given an informative prior, Bayesian posterior means are well defined. Furthermore, a Bayesian mixed (hierarchical) approach with specific priors effectively addresses non-stationarity in data. Non-stationarity, which can result from trends, seasonal effects, or structural breaks, may lead to spurious regression, producing misleading results with artificially high Rsquared values and significant t-statistics that do not reflect true relationships. Bayesian mixed models can manage non-stationarity by allowing parameters to vary over time or across different regimes. Additionally, incorporating informative priors in Bayesian inference helps regularize estimates and integrate prior knowledge, making it a valuable tool for handling non-stationarity (Spall, 1988). The study adopts a Bayesian mixed regression to investigate the impact of economic globalization on the welfare state. The baseline model is specified as follows: Expenditureit ¼a0þa1GlobalizationEcit þa2Xit þReffects þeit:(2) To capture the non-linear and non-monotonic nature in the globalization-spending relationship, we incorporate the quadratic term of GlobalizationEc into the model (2). Expenditureit ¼b0þb1GlobalizationEcit þb2GlobalizationEcit2þb3Xit þReffects þeit:(3) where Expenditure is the welfare state, GlobalizationEc is the economic measure of the KOF globalization, Xis a vector of controls, eis the random error, Reffects is random effects, and t,iare year and country, respectively. Our analysis utilizes the 2023 KOF globalization dataset, covering the period from 1970 to 2021, along with the 2023 World Development Indicators (World Bank) and annual panel data from the most recent Penn World Tables, version 10.01, updated only to 2019. These datasets include the dependent variable, welfare state (Expenditure), the key predictor, economic globalization (GlobalizationEc), and various baseline control variables available from 1950 to 2019 (data on the KOF globalization indices has been available since 1970). Hence, our panel dataset encompasses 10 ASEAN countries (Brunei, Indonesia, Cambodia, Laos, Myanmar, Malaysia, the Philippines, Singapore, Thailand, and Vietnam) from 1970 to 2019. In Table 1, the measure of the welfare state is defined as the government’s share of real GDP per capita, as compiled by the Penn World Tables, excluding transfer payments. The key predictor, COGENT ECONOMICS & FINANCE 5 GlobalizationEc, is the economic dimension of the KOF globalization index. This measure is employed to concentrate on the economic impacts of globalization. Note that the 2023 KOF index synthesizes 24 variables into a comprehensive index and three sub-indices, each representing the economic, social, and political dimensions of globalization. These variables are organized into six groups: (1) actual flows of trade, investment, and income payments to foreign nationals, (2) restrictions on international trade and capital accounts, (3) data on personal contacts with individuals in foreign countries, (4) data on information flows, (5) data on international cultural integration, and (6) data on international political integration. The economic globalization sub-index comprises the first two groups, the social globalization sub-index includes the next three groups, and the political globalization sub-index consists of the last group. Collectively, these components form the overall index (Gygli et al., 2019). Drawing on Alesina & Wacziarg (1998), Dreher et al. (2008), and Potrafke (2009), the study incorporates real per capita GDP (IncomePc), age dependency ratio (Dependence), total population (Population), and urbanization rate (Urbanization). Real per capita GDP provides insight into the overall socio-economic condition. According to Wagner’s law, increases in the levels of GDP per capita lead to increases in government expenditure as a share of GDP. As a country becomes wealthier, the demand for public services such as education, healthcare, infrastructure, and social welfare tends to increase. Higher income levels enable citizens to expect and demand better and more comprehensive public services. Besides, economic growth often brings about greater complexity in economic activities and social structures, necessitating more government intervention in terms of regulation, redistribution, and provision of public goods. The age dependency ratio adjusts for demographic structure and development. As the age dependency ratio increases, meaning there are more dependents (young and elderly) relative to the working-age population, it leads to increased government spending for several reasons. Older populations typically require more healthcare and social services, leading to higher public expenditure on these services. Furthermore, a higher proportion of young dependents necessitates more investment in education and related services. Lastly, increased numbers of elderly dependents require more spending on pensions and social security programs. Overall, more dependents generally increase the need for various welfare programs to support non-working populations. We measure urbanization by the proportion of the population living in urban areas. As expected, government expenditure will increase with rising urbanization. The rationale is straightforward: higher urbanization can lead to congestion, potentially decreasing citizen welfare, which may necessitate increased government expenditure to offset the welfare loss. However, spending on non-rival public goods like roads and street lighting may decrease with growing urbanization due to economies of scale (Anderson & Obeng, 2021). Total population serves as a proxy for country size, reflecting that larger populations tend to correlate with a smaller government share of GDP (Alesina & Wacziarg, 1998). Regarding total population, Alesina & Wacziarg (1998) suggest that countries with larger populations tend to have lower government consumption spending as a share of GDP for several reasons. Firstly, the per capita cost of providing non-rival goods is reduced in larger populations due to economies of scale (see Rodrik, 1998; Jetter & Parmeter, 2015). Additionally, larger Table 1. Interpreting variables in the Bayesian mixed model. Variables Proxies Period Notation Data source Dependent Welfare state Natural logarithm of share of government consumption in real GDP 1970–2019 Expenditure PWT 10.01 dataset Independent Economic globalization Natural logarithm of the economic sub-index of the KOF globalization index 1970–2019 GlobalizationEc Swiss Institute of Economics Real per capita income Natural logarithm of real per capita GDP 1970–2019 IncomePc PWT 10.01 dataset Age dependence rate Natural logarithm of the sum of the population in the ages 0–14 years and the population in the ages 65þas a percentage of the total population 1970–2019 Dependence 2023WDI dataset Country size Natural logarithm of total population 1970–2019 Population PWT 10.01 dataset Urban population Natural logarithm of urbanization rate 1970–2019 Urbanization 2023 WDI dataset Source: Authors. 6 N. NGUYEN THI NGOC AND T. NGUYEN NGOC populations exhibit more heterogeneous preferences for public goods provision. The overall effect depends on balancing the costs of increased heterogeneity in preferences against the benefits of lower per capita costs in public goods provision, with the latter potentially outweighing the former. Consequently, the impact of total population on government spending can be either positive or negative (Alesina & Wacziarg, 1998; Rodrik, 1998). Log transformations are applied to all variables to normalize skewed data, reduce the impact of outliers, and stabilize variance (Table 1). Moreover, taking the logarithm can stabilize the variance. For setting priors in the model, standard normal priors are used. Given that all parameters can take either positive or negative values, we assume a normal distribution with a mean of 0 and a mildly informative variance of 1 for the GlobalizationEc regressor, similar to the remaining covariates (Block et al., 2011). A non-informative inverse gamma (0.01, 0.01) is assigned to the variance components, including the variance of random effects and the overall variance, as suggested by Van de Schoot et al. (2015). Prior distributions for the model parameters are as follows: aNormal 0, 1 ðÞ , r02Igammað0:01, 0:01Þ, r12Igamma 0:01, 0:01 ðÞ , where ais a vector of structural parameters and r02,r12are the variance of random effects and the overall variance, respectively. 4. Simulation outcomes 4.1. Descriptive statistics and pairwise correlations The descriptive statistics table (Table 2) provides a comprehensive overview of the key variables, highlighting their average values, variability, and range, which is crucial for understanding the distribution and characteristics of the data, as well as for interpreting subsequent analytical results. Expenditure, representing government consumption spending as a share of GDP, has an average value of approximately 15.77%. The standard deviation of 5.63% indicates moderate variability around the mean. The minimum value is 4.82%, and the maximum is 31.88%, showing a broad range of government expenditure shares across different countries. The Economic Globalization Index (GlobalizationEc) has a mean score of 50.93, suggesting a moderate level of economic globalization on average among the sampled countries. The standard deviation of 19.70 shows substantial variation. The index ranges from 17.68 to 94.58, indicating that some countries are highly globalized while others are much less so. Per capita income (IncomePc) has a mean of $14,058.82, but the high standard deviation of $22,210.67 indicates a significant disparity in income levels across countries. The minimum value is $533.03, and the maximum is $99,005.20, highlighting the wide income range from very low-income to very high-income countries. The dependence ratio (Dependence) has a mean of 64.23, suggesting that, on average, there are 64 dependents for every 100 working-age individuals. The standard deviation of 17.77 indicates moderate variability. The dependence ratio ranges from 27.31 to 98.21, reflecting a substantial range of dependency burdens across countries. Population size (in millions) has an average of approximately 48.94 million people, with a high standard deviation of 58.46 million, indicating considerable variation in country sizes. The smallest population is around 0.22 million, and the largest is about 270.63 million, demonstrating a vast range of population sizes among the countries studied. Urbanization, expressed as the percentage of the Table 2. Descriptive statistics. Variable Obs Mean Std. dev Min Max Expenditure 486 0.1576643 0.0563471 0.048244 0.318799 GlobalizationEcI 486 50.92878 19.69739 17.6842 94.58176 IncomePc 486 14058.82 22210.67 533.028 99005.2 Dependence 486 64.2271 17.76848 27.31119 98.2091 Population 486 48.94011 58.46113 0.218175 270.6256 Urbanization 486 41.869 25.94483 4.477 100 Source: Authors. COGENT ECONOMICS & FINANCE 7 Nezlek, J. B. (2008). An introduction to multilevel modeling for social and personality psychology. Social and Personality Psychology Compass,2(2), 842–860. https://doi.org/10.1111/j.1751-9004.2007.00059.x Pesaran, M. H., & Smith, R. P. (2019). A Bayesian analysis of linear regression models with highly collinear regressors. Econometrics and Statistics,11,1–21. https://doi.org/10.1016/j.ecosta.2018.10.001 Pl€ umper, T., Troeger, V. E., & Winner, H. (2009). Why is there no race to the bottom in capital taxation? Tax competition among countries of unequal size, different levels of budget rigidities and heterogeneous fairness norms. International Studies Quarterly,53(3), 761–786. https://doi.org/10.1111/j.1468-2478.2009.00555.x Potrafke, N. (2009). Did globalization restrict partisan politics? An empirical evaluation of social expenditures in a panel of OECD countries. Public Choice,140(1–2), 105–124. https://doi.org/10.1007/s11127-009-9414-2 Robert, C. P., & Casella, G. (2004). Monte Carlo statistical methods. Springer Science þBusiness Media, LLC. second edition. Rodrik, D. (1997). Has globalisation gone too far?. Institute for International Economics. Rodrik, D. (1998). Why do more open Economies have bigger governments? Journal of Political Economy,106(5), 997–1032. https://doi.org/10.1086/250038 Rudra, N. (2002). Globalization and the decline of the welfare state in less-developed Countries. International Organization,56(2), 411–445. https://doi.org/10.1162/002081802320005522 Shalizi, C. R. (2019). The Truth about Linear Regression. Retrieved from the internet: https://www.stat.cmu.edu/ cshalizi/TALR/ Schulze, G. G., & Ursprung, H. W. (1999). Globalization of the economy and the nation-state. The World Economy, 22(3), 295–352. https://doi.org/10.1111/1467-9701.00205 Shalizi C.R. (2019). The Truth about Linear Regression. Accessed 18 April, 2014. https://www.stat.cmu.edu/cshalizi/ TALR/TALR.pdf Spall, J. (1988). Bayesian analysis of time series and dynamic models. 1st edition. CRC Press. Tekin, A., C¸ınar, _ I. T., Sa gdıc¸, E. N., & Yıldız, F. (2023). Trade openness and sustainable government size: Evidence from Central and Eastern European Countries. Sustainability,15(15), 11836. https://doi.org/10.3390/su151511836 Thach, N. N., Anh, L. H., & An, P. T. H. (2019). The effects of public expenditure on economic growth in Asia Countries: A Bayesian model averaging approach. Asian Journal of Economics and Banking,3(1), 126–149. Ursprung, H. W. (2008). Globalization and the welfare state. In Durlauf, S. N., and Blume, L.E. (Eds.) The New Palgrave Dictionary of Economics. Second Edition. Palgrave Macmillan (K€ oln). Van de Schoot, R., Broere, J. J., Perryck, K. H., Zondervan-Zwijnenburg, M., & Van Loey, N. E. (2015). Analyzing small data sets using Bayesian estimation: The case of posttraumatic stress symptoms following mechanical ventilation in burn survivors. European Journal of Psychotraumatology,6.https://doi.org/10.3402/ejpt.v6.25216 Vaubel, R. (2000). Internationaler politischer Wettbewerb: Eine europ€ aische Wettbewerbsaufsicht f€ ur Regierungen und die empirische Evidenz. In Schenk, K.-E., Schmidtchen, D., Streit, M. E. and Vanberg, V. (Eds.) Jahrbuch f€ ur Neue Politische € Okonomie. (pp.280–309). Mohr-Siebeck. Vaubel, R. (2005). Sozialpolitische Konsequenzen der Globalisierung-Theorie und Empirie. In Freytag, A. (Ed.) Wirtschaftlicher Strukturwandel, nationale Wirtschaftspolitik und politische Rationalit€ at. (pp. 143–158). K€ olner Universit€ atsverlag GmbH. Verbeke, G., & Molenberghs, G. (2000). Linear Mixed Models for Longitudinal Data. Springer. VNTR. (2022). ASEAN - A successful economic region model in globalization. Retrieved from https://vntr.moit.gov.vn/ news/asean-a-successful-economic-region-model-in-globalization Wagner, A. (1893). Grundlegungderpolitschen okonomie. C. F. Winter. Winship, C., & Western, B. (2016). Multicollinearity and model misspecification. Sociological Science,3, 627–649. https://doi.org/10.15195/v3.a27 Yong, M., & Chi, Y. (2022). Openness and government size: Revisiting the relationship using a large cross-country panel. International Review of Economics & Finance,79, 448–465. https://doi.org/10.1016/j.iref.2022.02.056 14 N. NGUYEN THI NGOC AND T. NGUYEN NGOC