Time-frequency connectedness between Thailand and international stock markets during crises

The Stock Exchange of Thailand (SET), which trades an average of USD 1.3 billion per day, has been the most liquid stock market in Southeast Asia since 2012. Between 2014 and 2019, it served as the region's preeminent fundraising forum. In the context of the rise in socially responsible investments, the number of listed firms on the SET recognized internationally for their outstanding sustainability performance has grown. Moreover, SET offers clients a comprehensive selection of attractive products that augment and diversify investment opportunities (SET, 2024). According to Bloomberg's survey, which used 10 distinct economic performance criteria to assess 17 emerging nations and their prospects, Thailand is the most developed. This is partly driven by Thailand's sizable foreign reserves and potential capital flows. However, few studies have evaluated how the pandemic has affected the Thai stock market. Therefore, this study explores the SET and its connectedness with major trading partners, including the USA, China and Japan. Thailand's largest export market is the USA, and China is considered Thailand's largest importer (World Bank, 2024). Moreover, Japan and the USA are the prominent foreign investors (OECD, 2024).

Building on prior research, this study analyzes the transmission of two major recent global financial crises to Thailand's stock markets: the COVID-19 epidemic and the Russia–Ukraine conflict. The pandemic disrupted global markets beginning in 2020, while the Russia–Ukraine conflict triggered renewed external shocks. In July 2022, a United Nations Development Program report indicated that the world was facing a precarious inflationary surge attributed to supply chain disruptions and price increases in essential commodities owing to the ripple effects of the conflict, particularly affecting the energy and food markets. At that time, the world was entrenched in a cost-of-living crisis characterized by unprecedented energy, fertilizer and food prices; inflationary pressures; trade restrictions due to conflict; and currency devaluations in emerging nations (Nolan, 2023). To assess this, we investigate the dynamic interconnectedness between the Thai stock market and major trading partners, including China, Japan and the USA, as well as the effects of economic policy ambiguity.

This study is significant for the following four reasons. First, it offers insight into how two different crises, with distinct genesis stories, led to varying degrees of risk transmission to markets beyond their nation of origin. Second, it assists regulators and policymakers in determining whether investor losses from COVID-19 pandemic shocks exceed those from a cost-of-living crisis. This comparative examination of shocks resulting from financial ties helps identify crises that may call for more insulation of the domestic economy. Third, international investors and portfolio managers can use the results to formulate strategies for diversifying their portfolios throughout Thailand's economy. In conclusion, incorporating short- and long-term connectedness dynamics into the analysis of a specific variable network enhances the accuracy of dynamic connectedness evaluation using the frequency-based framework for time-varying parameter vector autoregressive (TVP-VAR) models (Chatziantoniou et al., 2023).

Diebold and Yilmaz (2012) and Diebold and Yilmaz (2014) have been widely cited in the literature on volatility spillovers between financial markets. The majority of the literature observed spillover in the frequency connectedness framework (Ghazani et al., 2024; Nadeem et al., 2025; Sahoo, 2024) using Baruník and Křehlík (2018), an updated version of Diebold and Yilmaz (2012) and Diebold and Yilmaz (2014) technique. By removing the rolling window issue and using Bayesian techniques to estimate parameters at each time point to reflect dynamic connectivity among variables, the TVP-VAR model proposed by Antonakakis et al. (2020) overcame these problems. Because an arbitrary window size is not required, the entire sample data set is used instead. By combining the techniques of Baruník and Křehlík (2018) and Antonakakis et al. (2020) to address the rolling window problem, Chatziantoniou et al. (2023) presented a novel TVP-VAR-based framework for frequency connectedness to investigate static and dynamic interactions in both the time and frequency-domains. The current study used this novel approach to investigate volatility spillover patterns in the frequency and time domains.

The remainder of this paper is organized as follows. The subsequent section reviews relevant literature. Section 3 presents the results of unit root testing and the corresponding data sources. Section 4 explains the methodology of the TVP-VAR frequency connectedness model. Section 5 presents the empirical study's findings, and Section 6 outlines the conclusions and policy implications.

One financial market that significantly influences the economy is the stock market. Investors consider stock returns when making investment decisions. The financial systems of several nations may be connected to globalization. Financial integration is facilitated by the growth of shared markets, government liberalization initiatives, robust trade and economic relations and advancements in trade and telecommunications (Nguyen and Le, 2021). Globalization has increased the interdependence between capital markets. Numerous studies have demonstrated spillovers in volatility and stock market return transmission due to growing financial integration (Hung, 2021; Joo et al., 2023; Li and Chen, 2021; Mensi et al., 2017; Nammouri et al., 2022; Nguyen and Le, 2021; Vuong et al., 2022). Phiri et al. (2023) found that the co-movement between COVID-19 and G20 stock returns flipped between positive and negative correlations during the COVID-19 pandemic. Additional discoveries from the wavelet coherence analysis indicate that negative (positive) linkages are more mixed and predominantly manifest as lower (higher) frequencies in the context of cases and deaths. The results also indicate that the short-frequency components are linked to the period surrounding the initial declaration of the pandemic and subsequent variations in COVID-19. This study also covers the ramifications for markets and policies.

The conflict between Russia and Ukraine is another recent international catastrophe that has caused unimaginable suffering and presented serious hazards to the global economy. The Russian invasion of Ukraine on February 24, 2022, has hampered the slow economic recovery from the systemic shocks of the ongoing COVID-19 crisis. The result of this invasion was a fierce geopolitical battle that affected many markets and economies worldwide. The effects of the Russian–Ukrainian conflict have been compared to those of war due to their ferocity (Ari et al., 2025; Bossman and Gubareva, 2023; Xiao et al., 2023). Energy prices skyrocketed, and the world's financial markets collapsed after the invasion (Wang et al., 2022; Zhang and Sun, 2023). The conflict in Ukraine and the supply chain disruptions stemming from the COVID-19 outbreak imposed significant inflationary pressure on the global economy. In 2022, many countries experienced significant increases in inflation. For many individuals, salaries and welfare benefits failed to keep pace with inflation, resulting in a cost-of-living crisis (England et al., 2024; Ha, 2026; Morão, 2025; Nolan, 2023). The economy entered a recession in the latter part of 2022 and again in the latter half of 2023 (Fountas et al., 2025; Ihrig and Waller, 2024; Linnerud et al., 2025; Tsiaplias and Wang, 2023). The decline in the affordability of energy, food, fuel and housing resulted in detrimental short- and long-term health consequences. Historical evidence from economic crises indicates that economic shocks and rising living costs adversely impact mental health across the population, disproportionately affecting vulnerable groups (England et al., 2024).

Researchers use various methods to evaluate the interconnectedness of financial markets. Granger (1969) causality approach, the variance decomposition of the vector autoregression model, correlation analysis and the transfer entropy method are examples of methodologies that fall under this category. Both the pairwise correlation and Granger (1969) causality methods fail to evaluate systemic interconnections because they focus exclusively on bivariate interactions. The VAR technique allows a thorough investigation of the relationships present in a multivariate system. Diebold and Yilmaz (2009) and Diebold and Yilmaz (2012) authored the most prominent models of the VAR approach. A network architecture using the variance decomposition method was proposed by Diebold and Yilmaz (2014). They used variance decomposition to determine how changes in the stock returns of major US financial firms were linked over time, which allowed them to build stock return and volatility networks. Other methods have recently been developed to improve this strategy. To investigate how the eight stock indices were affected by the COVID-19 epidemic and how they are dynamically linked to each other, Youssef et al. (2021) utilized the TVP-VAR method developed by Diebold and Yilmaz. They investigated how the unpredictability of economic policies affected the interconnections among these various factors. Chatziantoniou et al. (2022) used a TVP-VAR connectivity methodology to explore the interconnectedness of sectoral stock markets in India. They discovered that interconnectedness peaked during the Global Financial Crisis. Zhou et al. (2023) built a network within the Chinese sectoral stock markets using a TVP-VAR model. Additionally, they utilized high-frequency data to investigate sectoral risk spillovers.

The TVP-VAR technique was developed by Baruník and Křehlík (2018) using the spectral representation of variance decompositions. This technique was used to evaluate the interconnectivity of financial variables arising from different reactions to disturbances. They indicated that financial markets typically respond to economic uncertainties at varying frequencies and that investors possess diverse expectations regarding investment returns based on their distinct investment horizons. Consequently, it is essential to investigate the frequency-domain dynamics of the spillover (connectedness) effects, specifically the short-, medium- and long-term proportions of future uncertainty induced by shocks. Consequently, they developed a spillover index using a frequency-domain framework, as predicted by the methodologies of Diebold and Yilmaz (2009) and Diebold and Yilmaz (2012). Based on the research of Antonakakis et al. (2020), it is deemed essential and logical to enhance the frequency connectedness measure of Baruník and Křehlík (2018) using the TVP-VAR approach, hereafter referred to as the TVP-VAR frequency connectedness model. This new model not only leverages the benefits of time-varying connectedness measurement established by Antonakakis et al. (2020) but also captures these connectedness effects across diverse temporal frequencies, namely short-, medium- and long-term horizons, thereby enhancing our understanding of the heterogeneous responses of various markets to economic shocks.

The daily data employed in this analysis for the period from January 1, 2020, to December 31, 2024, came from the Thai stock market, and the top three major traders included China, Japan, and the USA. The SET, Shanghai Composite (SSEC), Tokyo Price Index (TOPIX) and National Association of Securities Dealers Automatic Quotation System (NASDAQ) composite indices are proxies for the Thai, Chinese, Japanese and US stock indices, respectively. Moreover, we add the volatility index (VIX) and the price of West Texas Intermediate (WTI) crude oil to the model to control for market sentiment and oil price volatility, respectively. This finding is consistent with (Yahya et al., 2023). Chatziantoniou et al. (2023), Chang et al. (2023) and Yuan et al. (2021) also employed WTI in their analyses. Chang et al. (2023) specified that oil prices impact nations whose economies are largely dependent on oil imports and exports. Chatziantoniou et al. (2023) claimed that swings in WTI pricing could imply substantial changes in the markets that follow them. We obtained all the indexes from the London Stock Exchange Group (LSEG) Workspace and then computed the stock returns using r_t = Δln(p_t) × 100, where p_t is the closing price of a stock in country t. To investigate how these crises impact volatility spillover among these stock markets, we initially defined two separate sub-periods. The COVID-19 outbreak period begins on January 1, 2020, and continues until June 30, 2022, and the cost-of-living crisis period spans from July 1, 2022, to December 31, 2024. Table 1 presents the descriptive statistics and unit root test results of the returns, which show that the other variables were stationary at their levels in both sample periods.

Research by Baruník and Křehlík (2018) and Antonakakis et al. (2020) forms the basis of the methodology of TVP-VAR frequency connectedness, which is introduced by Chatziantoniou et al. (2023). Antonakakis et al. (2020) assert that they used a correlation technique grounded in generalized forecast error variance decomposition (GFEVD). Diebold and Yilmaz (2009, 2012, 2014) built a framework for connectivity, which they combined with the TVP-VAR model created by Koop and Korobilis (2014). Therefore, the TVP-VAR model is able to avoid the issue of arbitrarily selected rolling window sizes by effectively identifying temporal variations. The following is a summary of the suggestions made by the Bayesian information criterion (BIC) regarding TVP-VAR.

where r_t, r_t-1 and ε_t are k × 1 dimensional vectors, representing all index return series in t, t-1, and the corresponding error term, respectively, Φ_t and Q_t are k × k dimensional coefficient matrices and the time-varying variance-covariances. vec(Φ_t) and v_t are k ² × 1 dimensional vectors while R_t is k ² × k ² dimensional matrix.

The concept of GFEVD is based on an old representation theorem. The theorem was first presented by Koop et al. (1996) and Pesaran and Shin (1998). The following equivalence must be used to translate the projected TVP-VAR model into its TVP-vector moving average (VMA) method: $rt=∑i=1pΦi,trt−i+εt=∑j=0∞Ψj,tεt−j$⁠. Because the recovered findings are completely independent of variable ordering, we favor GFEVD over its orthogonal colleague. Furthermore, Wiesen et al. (2018) emphasized that the GFEVD should be applied when no theoretical framework is available to discover the error structure. The following formula represents the GFEVD, which may be understood as the impact of a shock in variable j on variable i in terms of its forecast error variance:

where $ϕ~ij,t(H)$ shows the forecast error variance share, or the pairwise directional connectivity between variable i and variable j, as the influence that variable j has on variable i at the horizon H. $∑i=1kϕ~ij,t(H)=1$⁠, $∑i,j=1kϕ~ij,t(H)=k$⁠, and e_i serves as a zero vector at location i that exhibits unity. Furthermore, the total directional connection TO others measures the extent to which a shock in variable i is transmitted to all other variables j.

The amount that variable i receives from shocks in all other variables j is measured by the total directional connectivity FROM others:

The influence variable i has on the network under study is represented by the NET total directional connectedness, which represents the disparity between the overall directional connectivity TO and FROM other:

Market interconnectivity is determined by the total connectedness index (TCI), which is established as

Previously, our assessment of connectedness was constrained to the identical domain. Nonetheless, the disintegration method allows us to ascertain the regularity reaction characteristic. The method is $Ψ(e−iω)=∑h=0∞e−iωhΨh$ , where $i=−1$ and ω represents the frequency. The spectral density of r_t at frequency ω, which can be seen as a Fourier transformation of the TVP-VMA, is the next subject of the investigation:

The frequency GFEVD must be normalized in the same manner as the temporal normalization of the GFEVD, and it can be expressed as

Here, $ϕ~ij,t(ω)$ is the fraction of the ith variable's spectrum at a specific frequency ω that induces an uncertainty in the jth variable. The following is an aggregation of all frequencies within a specified range, which is used to quantify connections across various frequency bands (both short- and long-term).

Within the context of this equation, d = (a, b), where a, b ∈ (−π, π) and a < b. It is possible to properly quantify the interconnectivity that exists between different industries, as Diebold and Yilmaz (2012, 2014) have proved, respectively. However, they looked at frequency connectivity measurements and analyzed the data loss that occurred within a specific frequency range denoted by d. We offer the following all-encompassing equations.

Ultimately, we demonstrate the correlation among the frequency-domain metrics established by Baruník and Křehlík (2018) and the time-domain metrics proposed by Diebold and Yilmaz (2012, 2014):

This study employs a TVP-VAR frequency connectedness model established by Chatziantoniou et al. (2023) to achieve its purpose. Furthermore, by separately examining interactions in the short- and long-term, we are able to discover events that have a sustained influence on different asset classes. Drawing from the work of Baruník and Křehlík (2018), the current study examined connections between high-frequency intervals ranging from 1 to 5 days and low-frequency zones that went all the way from 5 to infinite days.

This part presents the empirical findings and important points arising from our research.

Table 2 summarizes the whole sample period results. Panels A and B provide the mean measures of connectedness, and the high- and low-frequency values are presented in parentheses. Panel A covers the COVID-19 crisis period, and Panel B covers the cost-of-living crisis period. The shock own-variance shares are displayed on the main diagonal, while the off-diagonal sections represent the interaction among financial assets. This study investigated the effects of the pandemic on the average connections between financial markets. It is shown in Panel A that SET was responsible for 66.18% of the variance share spillovers throughout the time of the COVID-19 pandemic, according to the average dynamic connectivity report. Only 11.12% of them showed own-source variation spillovers over the long term, while 55.07% showed short-term effects. Consequently, all remaining factors accounted for 33.82% of the variance in the SET forecast error. A comprehensive analysis indicates that the SSEC, TOPIX, NASDAQ, VIX and WTI influence 5.94%, 9.24%, 10.80%, 5.45% and 2.39%, respectively. Short-term and long-term spillovers are the two categories into which these effects can be classified. SET is significantly influenced by NASDAQ, which is responsible for 7.81% of short-term spillovers, along with 2.99% of long-term spillovers. SET exerts an impact on other variables amounting to 30.29% and is affected by them to a degree of 33.82%. According to this, it is a shock contributor with a net transmission of −3.53%. It serves as a transmitter of shocks in both the short and long run, exhibiting a short-term net transmitter effect of 1.08% and a long-term net receiver effect of −4.61%. NASDAQ is the foremost contributor of shocks among all analyzed series at 19.42%, with a VIX of 4.69%. The emergence of NASDAQ as a significant conduit for shocks within the overall system is unsurprising, as fluctuations in the US stock market can generate ripple effects across the financial system, directly impacting market uncertainty and thereby influencing the VIX. This finding is comparable to that reported by Phiri et al. (2023) and Samitas et al. (2022). The VIX, with a robust shock transmission of 4.69%, serves as the primary short-run contributor at 3.98%. Conversely, WTI was the dominant shock absorber (−6.90%) and the main short-term net recipient (−6.97%) and long-term net transmitter (0.06%). The average TCI shows that short-term momentum is more than four times long-term spillover (23.86 vs. 4.76%). These numbers denote the average indicators of connections, perhaps concealing temporal variations and unique time impacts. Consequently, our analysis explored dynamic connectedness plots to enhance the comprehension of these interrelated interactions.

The median dynamic connectedness report for the cost-of-living throughout a crisis period is shown in Panel B of Table 2. This indicates that SET constitutes 83.75% of the variance share of spillovers. This allocation comprises 67.08% short-run own-source variance spillovers and 16.67% long-run own-source variance spillovers. Thus, the remaining factors jointly accounted for 16.25% of the variance in the SET forecast error. SET returns are influenced by several factors, with the SSEC, TOPIX, NASDAQ, VIX and WTI contributing 3.00%, 5.74%, 5.45%, 1.07% and 0.99%, respectively. These consequences are short- and long-term spillovers. TOPIX has the most substantial influence on SET, with 4.17% of short-term and 1.57% of long-term spillovers. Also, with 4.71% of the total, SET is the factor that influences TOPIX the most. SET's net absorption rate is −4.25%, implying it functions as a shock transmitter net. As a short- and long-term shock receiver, it contributes −1.76% to the former and −2.49% to the latter. NASDAQ is the principal conduit of shocks among all analytical series, significantly affecting TOPIX (16.78%) and VIX (7.52%). In contrast, the TOPIX is the principal shock absorber at −18.33%, serving as the principal net recipient of shocks for both the short-term and long-term, with values of −14.63% and −3.70%, respectively. It should be noted that while calculating the average TCI, the short-run momentum component (11.79%) is far higher than the long-run spillover component (2.93%).

In summary, the COVID-19 crisis era had a low level of connectedness at 28.62%, with 23.86% in the short-term and 4.76% in the long-term. Very low levels of total connectivity (14.73%), short-term connectedness (11.79%) and long-term connectedness (2.93%) were also observed during the cost-of-living crisis. This suggests that during the COVID-19 and cost-of-living crises, connectivity was driven mostly by short-term spillover effects. The median TCI of the COVID-19 crisis period somewhat exceeds that of the cost-of-living crisis period, suggesting that under typical market and economic conditions, the COVID-19 crisis period has greater risks and uncertainties than the cost-of-living crisis period. The shift toward greater market segmentation has fostered improved market resilience. As markets become less dependent on a single global supply chain, they develop localized strengths to withstand economic pressures. This improved resilience is particularly evident in how individual markets stabilized domestic supply and demand despite the cost-of-living crisis. In summary, the research highlights a paradoxical benefit of the current economic shift: the erosion of global interconnectedness has contributed to reduced systemic risk and improved market resilience, providing a much-needed buffer against the volatility of the cost-of-living crisis.

This section provides the total dynamic connectedness or the TCI's dynamic progression throughout the investigation. Figure 1 shows the results. To ensure comprehensiveness, Figure 1 examines the general development of the TCI (represented by the black-shaded region) and its short- and long-term phases (pink- and green-shaded areas). An extensive examination of the connectedness plot in Figure 1 uncovers several significant trends. The meaning of total, short-term and long-term dynamic connections was further investigated. Total spillovers began to increase significantly in early 2020, during the COVID-19 crisis period. Consequently, the evidence indicates that financial markets exhibited a significant degree of integration throughout the sample period, with the level of integration occasionally reaching notable milestones. Many studies reveal robust dynamic interconnectedness across global financial markets (Hung, 2021; Phiri et al., 2023; Yuan et al., 2021). Overall, short-term and long-term TCIs then dropped and stabilized at approximately 15% until the end of 2021. An increase was noted at the onset of 2022, coinciding with Russia's commencement of its invasion of Ukraine on February 24, 2022. Subsequently, overall TCI reverted to the prior level of approximately 25% during the Russia–Ukraine conflict. The short-term and overall TCIs both increased, but the long-term TCI remained relatively unchanged, suggesting that the present crisis had a primarily short-term impact on the total connectivity index.

Regarding the cost-of-living crisis period's short-run, long-run and overall TCI statistics, both the overall and short-run TCI exhibited a sustained increase akin to the COVID-19 crisis period but experienced numerous fluctuations in early 2023, coinciding with the energy crisis (Linnerud et al., 2025) and then decreased from 2023 to mid-2024. The long-term TCI has comparatively prolonged oscillation and stability, in contrast to the short-term TCI's considerable volatility. Therefore, it is essential to analyze the short- and long-term trends independently. Concentrating exclusively on the aggregate TCI may hide the origins of these variations. In late 2024, a significant increase was evident during the soft landing, suggesting that the global economy is projected to persist in its deceleration, with certain experts forecasting a soft landing in which inflation is managed without precipitating a significant recession. According to the frequency analysis, short-run dynamics, not long-run causes, were primarily responsible for the overall increase in TCI. Compared to long-term TCI dynamics, which are less volatile, short-term dynamics have a more significant impact on the overall TCI.

Moreover, the results are in line with Ari et al. (2025) and Xiao et al. (2023), who found notable TCI peaks, especially in early 2020 and early 2022. Long-term and short-term interconnectivity both increase, but their variations are not as noticeable as those of total TCI. TCI shows a steady decline starting in the second half of 2020 and stays at lower levels until 2023. Given that regional factors are increasingly influencing market dynamics rather than widespread global shocks, this drop points to a weakening of global financial interconnection. This illustrates a global contagious impact that transcends borders. These effects, however, were more noticeable in the short term than in the long term, indicating that markets typically responded to negative news with initial panic before decoupling and regaining equilibrium when asset-specific fundamentals took back control. The progressive fading of pandemic-related disruptions, a relative recovery in global economic conditions and a shift toward more localized financial dynamics are all reflected in the TCI stabilizing at lower levels by 2023. With volatility being driven more by local and regional reasons than by systemic global shocks, the drop in TCI suggests a weakening of global financial interdependence. The consistently low short-term TCI values in 2023 and later indicate that abrupt global shocks have less of an impact and that localized economic and geopolitical events mostly determine financial movements.

We then began by focusing on overall directional connectivity. Figures 2 and 3 display these results for the times of the COVID-19 pandemic and the cost-of-living periods, respectively. In addition to the overall directional connectedness (black shade), each panel in the figures shows the results for short-short (pink shade) and long-term (green shade) directional connectivity. Furthermore, for every other variable type in the network, the corresponding variable type serves as a net transmitter of price shocks. When the shaded area falls within the positive values, on the other hand, negative values represent net beneficiaries. It is also important to note that any variable can eventually adopt either state (i.e. net transmitter or net recipient).

As shown in Figure 2, net directional connectivity demonstrates clear temporal fluctuations. SET, SSEC, TOPIX and WTI are net recipients of short-term volatility spillovers during the COVID-19 crisis. The SET emerges as a net long-run shock recipient, particularly in 2020. Both the NASDAQ and the VIX exhibit contradictory tendencies over time; that is, they serve as shock transmitters in the short and long runs, respectively, except for 2022, when the VIX serves as a long-run shock transmitter. An increase in the natural gap in the US economy during that time was likely caused by this outcome, which was likely tied to the biggest increase in the US inflation rate in 40 years (Xiao et al., 2023). Moreover, in 2022, the stock market experienced noteworthy expansion, marked by substantial growth in major indices (Ihrig and Waller, 2024). This expansion was driven by multiple factors, the most important of which were the Federal Reserve's monetary policy, improvements in trade relations and strong performance in the technology sector.

Regarding Figure 3, considering the results for the cost-of-living crisis period, SET, SSEC and TOPIX are the net receivers of the connected network. At the same time, NASDAQ, VIX and WTI are the net transmitters. These results are in line with those of the COVID-19 crisis period for NASDAQ and VIX. Our study indicates that WTI is continuously classified as a net shock transmitter by long-run dynamics during the whole observation period. Nevertheless, it shifted to being a net recipient of short-run spillover shocks prior to 2024 and after 2024. Because of its relative stability, this lasting transmitter property gives investors and financial advisors crucial insights. The results highlight structural weaknesses in emerging markets by identifying Thailand as a net recipient of shocks. The main way the shock spread during the COVID-19 pandemic was through a decline in global travel demand. On the other hand, supply-side inflation and financial contagion from rising global interest rates were two ways the cost-of-living issue exerted pressure. These two examples highlight how Thailand continues to be a passive beneficiary of global volatility, despite de-globalization. This enduring situation highlights the need for greater market segmentation and greater market resilience to protect the domestic economy from foreign shocks it unintentionally absorbs.

In addition, we calculate the overall frequency connectedness of the returns of the Thai stock market and its main trading partners for two distinct frequency bands as a robustness test: short term (1–10 days) and long term (10 days to infinity). Appendix displays the findings.

To achieve our research goals, we consider a variety of asset classes, including the Thai stock market and major trading markets such as China, Japan and the USA, covering the periods before and after the coronavirus epidemic. Additionally, we included WTI crude oil prices and the VIX in the model. The main benefit of the TVP-VAR model is that it gives more accurate connectivity measures than the standard rolling-windows method, effectively making it an extension of Diebold and Yilmaz (2014), Koop and Korobilis (2014), Baruník and Křehlík (2018), Antonakakis et al. (2020) and Chatziantoniou et al. (2023) methodology. The results span January 1, 2020, to December 31, 2024, and depend on the daily price fluctuations of the variables listed above. For each time point, we first established four separate sub-periods. The timeframe for the COVID-19 epidemic began on January 1, 2020, and ended on June 30, 2022, and the cost-of-living crisis period was July 1, 2022, to December 31, 2024. This study's time-varying technique has several advantages over other methods (e.g. the standard rolling window method) because it does not include outlier data and does not arbitrarily limit the duration of the estimation windows.

The empirical results show a high level of connectedness between stock markets during the COVID-19 pandemic, with an overall dynamic connectivity of approximately 28.62%. The connectedness index declined during the cost-of-living crisis by approximately 14.73%. Considering the results for the COVID-19 crisis period, the US stock market sends signals to the connected network, which in turn receives signals from China, Japan and the Thai stock markets. Moreover, the study reveals that the co-movement between the Thai stock market and its three principal trading partners did not differ significantly across the two economic crises. Despite the expansion of China's influence in Thailand through trade and investment over many years, the role of the USA remains undiminished. Foreign investors must pay close attention to these findings. Global investors may consider these findings when making decisions regarding asset allocation or portfolio management.

These empirical findings indicate that the COVID-19 pandemic and the cost-of-living crisis altered the interrelationships among financial markets. The outcomes of this experiment shed light on the risk contagion between Thailand and global stock markets. In order to reduce portfolio risk, they are also crucial for investors in securities. Moreover, they indicate the assets that convey and absorb market shocks arising from economic uncertainty, and the impact of that uncertainty on their connections in terms of frequency. Consequently, our findings enhance our understanding of the influence of war on the dynamics of global financial markets. Investors, portfolio managers and governments can develop effective investment strategies and safeguards against uncertainty and engage in risk assessments. Our finding that the short-term influence on returns is more significant than the long-term effect highlights the necessity for prompt portfolio reallocation and the formulation of hedging strategies in response to economic uncertainty. Our discovery of a more pronounced long-term effect on volatility dynamics suggests that risk transmission from such uncertainty must be factored into long-term asset allocation strategies.

Additionally, this study proposes that the integration of Thailand's stock market with global equity markets has been driven by globalization. Spectral return spillover test results demonstrate that statistical test values vary with frequency. These findings show that there may be differences in the relationships between stock market performance across different frequency spectrum bands. Therefore, frequency-domain analysis should be performed to provide deeper insights into return spillovers among stock markets, as time-domain causality testing may not fully capture these relationships. Both short- and long-term investors will benefit from this approach, as it will provide them with more information to make needs-based investment decisions. Future studies should concentrate on quantifying myopic and intertemporal asset-allocation choices amid uncertainties.

The supplementary material for this article can be found online.