2.1 Uncertainty and GEPU

Economic uncertainty is quantified through the frequency of newspaper coverage, encompassing uncertainties surrounding policy decision-makers, anticipated economic policies, and their timing (^{Baker et al., 2016}). This captures the ambiguity surrounding government policies and regulatory frameworks in the short term (^{Al-Taqueb and Algharabali, 2019}). The GDP-weighted average index, incorporating the Economic Policy Uncertainty (EPU) of each country, is termed the Global Economic Policy Uncertainty (GEPU) index, developed by Davis (²⁰¹⁶). This index encompasses significant non-economic events impacting the economy, such as 9/11, the Iraq invasion, the European Immigration crisis, and Brexit (^{Dai et al., 2021}). Recent research has identified breakpoints in the relationship between the Geopolitical Risk Index (GPR) and GEPU due to events like the Russia-Ukraine conflict (^{Shen and Hong, 2023}). The EPU for each country reflects the relative frequencies of newspaper articles mentioning economic, policy, and uncertainty topics, while the GEPU is a normalized version of each country's EPU, weighted by GDP to compute the index, encompassing each country's GDP. This index covers 20 countries, identical to those considered in weighing GDP to construct the index.

Uncertainty exerts notable effects on various aspects including consumer spending, financial markets, corporate behavior, and risk management (^{Al-Taqueb and Algharabali, 2019}). In Latin American countries, uncertainty has been observed to impact stock markets and macroeconomic variables (^{Coronado et al., 2020}; ^{Coronado et al., 2022}). Additionally, in Latin American emerging markets, interest rates and inflation show positive responses to uncertainty in the long run (^{Aytaç and Saraç, 2022}). The effects of uncertainty extend to a wide array of variables including bitcoin, grain prices, agricultural commodities, crude oil, energy assets, as well as travel and leisure companies (^{Khan et al., 2021}; ^{Long et al., 2023}; ^{Hamidu et al., 2022}; ^{Lyu et al., 2021}; ^{Chen et al., 2021}; ^{Ersan et al., 2018}). Trade policies have been identified as a source of uncertainty affecting stock market stability (^{Davis, 2019}). The uncertainty of leading regional countries has been examined by Keddad (2024), while Caporale et al. (²⁰²²) demonstrate cointegrated relationships between ASEAN (Association of Southeast Asian Nations) countries and the U.S. and China. Causal effects stemming from GEPU have been investigated by Wu et al. (²⁰¹⁵) within the context of OECD countries. Pirgaip and Dinçergök (²⁰²⁰) explored policy uncertainty, energy consumption, and carbon emissions in G7 countries. Wu et al. (²⁰¹⁹) identified a causal relationship between uncertainty and International Tourism Receipts (ITR), while Olanipekun et al. (²⁰¹⁹) linked uncertainty to the exchange market in BRIC countries, considering Brazil, Russia, India, and China.

2.2. GEPU and economic growth

Damstra and Bourkes (²⁰²¹) studied the impact of economic news on public perceptions of the economy in the Dutch context. They distinguished between positive and negative changes in coverage as well as people's judgments. Hollanders and Vliegenthart (²⁰¹¹) also examined the economic coverage of Dutch media, highlighting the relationship between negative news and decreasing consumer confidence, particularly after the onset of the credit crisis. Danisman et al. (²⁰²⁰) conducted a panel data analysis considering Economic Policy Uncertainty (EPU), a variable similar to GEPU, and found a negative impact of EPU on credit growth for private and listed banks in the United Kingdom, Germany, Spain, Italy, and France. Chen et al. (²⁰¹⁹) used a VAR model and Granger causality test with GEPU and oil prices to analyze China’s economic growth, finding that GEPU harms it, while oil prices, combined with GEPU, Granger cause economic growth in the industrial sector. Lunde and Torkar (²⁰²⁰) performed an analysis for China, concluding that the predictive power of news sentiment improves GDP forecasts. Wu et al. (²⁰²¹) also identified a causal relationship between GEPU and tourism in Brazil, India, Indonesia, South Africa, and Turkey, noting that it changes over time. Hong et al. (²⁰²²) revealed a strong relationship between GDP and categorical economic policy uncertainty in the USA through a mixed frequency and Granger causality approach.

Ma et al. (²⁰²⁰) stated that GEPU contains predictive information about the gold futures markets, revealing that higher levels of GEPU lead to higher volatility. Adam et al. (²⁰²²) investigated the impact of uncertainty and volatility in ten Islamic countries and the predictive content of GEPU for forecasting and nowcasting GDP in emerging markets such as Brazil, Indonesia, Mexico, South Africa, and Turkey. They noted that including GEPU in benchmark and dynamic factor econometric models leads to superior GDP growth predictions and highlighted its capacity to reflect spillover effects in emerging markets. Luk et al. (²⁰¹⁸) established that economic policy uncertainty shocks affect real economic activities in small open economies such as Hong Kong, remarking on the deteriorating financial conditions due to international spillovers of uncertainty from major economies. Özyeşil (²⁰²²) explored a combination of the Baltic Dry Index (BDI), Volatility Index (VIX), and GEPU for the USA, UK, Germany, Italy, France, and Canada, stating that these indexes have a statistically significant effect on stock markets. He also remarked that these indexes are useful tools for efficient timing in stock markets by monitoring BDI, VIX, and GEPU. Asafo-Adjei et al. (²⁰²¹) examined the BRICS (Brazil, Russia, India, China, and South Africa) economies, noting a bidirectional and positive causality between the financial sector and economic growth in the presence of GEPU. They identified that South Africa’s financial markets and GDP are particularly vulnerable to GEPU. Xu (²⁰²⁰) also identified GEPU as a transmission channel that increases firms' cost of capital.

Nowzohour and Stracca (²⁰²⁰) highlighted the potential of sentiment as a driver for business cycles and concluded that there is a global factor for international spillovers of sentiment. This reasoning justifies using GEPU, the global index for EPU, as a common variable affecting economies. Effects of US economic policy uncertainty shocks on India's economy were assessed by Nyawo and van Wyk (²⁰¹⁸), who found that the contribution to Indian macroeconomic variables from US economic policy uncertainty is even larger than that from domestic Indian shocks. A similar conclusion was reached by Stockhammar and Österholm (²⁰¹⁶) in their study on the effects of US policy uncertainty on Swedish GDP growth, using a Bayesian VAR model, which concluded that there is a negative effect from US policy uncertainty on the Swedish economy. Yalçinkaya and Daştan (²⁰²⁰) went further by combining GEPU with geopolitical uncertainties for the Turkish economy in a VAR analysis, finding negative effects on macroeconomic indicators such as inflation, interest rates, unemployment, exchange rates, account balances, and economic growth in both the short and long run. Other studies involving GEPU as a determinant variable include Korus and Celebi (²⁰¹⁹), who found a relationship between Brexit and the depreciation of the British pound against the euro and the US dollar. The media's impact on the British hotel industry was assessed by Tajvidi and Karami (²⁰²¹), who found a positive relationship between marketing and firm performance.

The present research document contributes to the understanding of the role of global economic uncertainty and their impact on the economic growth of several countries by accounting for the effects of economic news (e.g. newspapers) as a transmitter of relevant information, which influences expectations about economic growth. Through rigorous analysis and comprehensive data evaluation, our research not only corroborates existing findings but also seeks to identify new relationships and effects, thereby contributing to a more nuanced and detailed comprehension of economic policy uncertainty's role in shaping global economic dynamics.

5.1. Results for America

The first VAR-MIDAS model did not show a high R-squared for any of the involved countries. The highest R-squared is present for Brazil with 0.39, and 0.38 for Mexico, but for the rest of the countries, the R-squared is not higher than 0.33. The model was estimated with 2 lags (Table 3). The second VAR-MIDAS was estimated with 2 lags and included China’s GDP as an extra endogenous variable. The R-squared improved significantly with the influence of China with Mexico reaching a 0.75 and with the lowest value being 0.49 (Table 4). Regarding the estimations for the American countries, the model that includes China replicates better the behavior of the series and we present Mexico as an example. It can be seen that the second model (Figure 2) follows better the ups and downs of economic downturns than the first model (Figure 1). The influence of China in the American region improved the estimations and its presence is indicative of an important relationship and economic integration with the region. Regarding the causality test in both models, GEPU has a causal influence only on Brazil, but jointly with the rest of the countries, influences the GDP of the region. Some causal relationships that appeared in the first model are those from the US to Brazil, from Brazil, Chile, and Mexico to Canada, with the latter being bidirectional. Chile also influences Mexico and the US, while Canada has a causal relationship with the US. The second model showed causal relationships from Mexico, the US, and China to Brazil, from China to Canada, and the US, while Mexico causes Chile’s GDP. China also has bidirectional relationships with Chile and Mexico.

The economics of causal GDP relationships between different countries involves understanding how economic activity in one country influences or is influenced by economic activity in another. This interconnectedness can be analyzed through various lenses, including trade, investment flows, financial linkages, policy decisions, and global economic conditions. Here are some key aspects:

Understanding these relationships could require sophisticated econometric tools and models, such as the ones applied in the present research document. Again, these are VAR (Vector Autoregression), VAR-MIDAS, and Granger causality tests, to identify and quantify the directions and magnitudes of these influences. By analyzing how GDP in one country responds to changes in another, it is recommended that economists do understand the global economic environment and to closely follow inform policy decisions. About these recommendations we make a suggestion to do it for future research.

Source: own elaboration using R computer language

Figure 1 Estimation for Mexico (fisrt model)  

Source: own elaboration using R computer language

Figure 2 Estimation for Mexico (second model)  

5.2. Results for Europe

For the model including only European countries, the appropriate number of lags was one (considering information criterion). The countries with the highest R-squared values are France (0.97), Italy (0.96), and Belgium (0.95). The lowest R-squared values are for Spain (0.24) and Ireland (0.31), while the rest of the countries had values of at least 0.72. (Tables 5a and 5b). Regarding the estimations produced by the first model, it is notably accurate for France (Figure 3) and Italy (Figure 4), while the model including China did not represent a difference (Figure 5 and Figure 6).

Source: own elaboration using EViews 12

Figure 3 Estimation for France (first model)  

Source: own elaboration using R computer language

Figure 4 Estimation for Italy (first model)  

Source: own elaboration using R computer language

Figure 5 Estimation for France (second model)  

Source: own elaboration using R computer language

Figure 6 Estimation for Italy (second model)  

The model including China did not show significant improvement but Spain and Ireland enhance their R-squared with 0.289 and 0.356 respectively (Tables 6a and 6b). The rest of the countries showed just marginal improvement. The third model for Europe (Tables 7a and 7b), which included European countries and the US showed significant improvement for Ireland but for the rest of the countries, there are similar results with the R-squared worsening notably for Italy with just 0.32.

The last model for Europe, which included both China and the US did not show better results in general (Tables 8a and 8b). The model including the US showed an acceptable fit to estimate Ireland (Figure 7) except for times with extreme perturbations, and reflecting the R-squared improvement described prior, but the same did not hold for the model including China and the US (Figure 8). Estimations for the rest of the countries look like the first model.

Source: own elaboration using R computer language.

Figure 7 Estimation for Ireland (third model)  

Source: own elaboration using R computer language.

Figure 8 Estimation for Ireland (fourth model)  

Regarding causal relationships in Europe, the first model showed a causal relationship from Spain to all countries in the model, but this relationship is bidirectional with Belgium. Spain is a special case in this model because it influences all other countries, but only Belgium influences it. All countries and GEPU jointly influence each country's GDP, except for Spain's. Bidirectional relationships are present between Germany and Belgium, France and Croatia, and Germany and the UK.

Results for the model that includes China showed a causal relationship from China to Croatia, France, Germany, Ireland, the Netherlands, Spain, and the UK. Again, Spain influences all countries' GDP, including China, in a bidirectional relationship. The difference with the first model is that in the second model, all countries and GEPU jointly influence Spain's GDP. Other bidirectional relationships are present between Spain and Belgium, Germany and Greece, the UK and France, the Netherlands and France, and Spain and the Netherlands.

The third model for Europe, which included the US, showed the same causal relationship from Spain to all countries, but Spain is no longer influenced by the rest of the countries and GEPU, except for the bidirectional relationship with Belgium, which remains. The US Granger-causes only the Netherlands and Sweden, while bidirectional relationships are present between Germany and France, and the Netherlands and Greece.

For the fourth model, which included both the US and China, findings indicate bidirectional relationships between Croatia and France, Germany and France, Germany and Greece, and China and Spain. The US and China influence only the Netherlands and the UK simultaneously, while Spain again influences all countries in the model but is influenced only by China. The joint influence of all countries and GEPU on a specific country's GDP is present for all economies except Spain.

Some insights about the casual and bidirectional relationships between uncertainty and economic growth can be, among others, exchange rate volatility. That is, uncertainty can lead to volatility in exchange rates, affecting competitiveness, trade balances, and ultimately GDP (^{Benavides, 2021}). Countries with volatile exchange rates may experience economic instability, impacting their growth. Also, stock market fluctuations. Considering that uncertainty can lead to increased volatility in stock markets, affecting wealth and consumption patterns. Declines in stock markets can reduce household wealth and consumer spending, thus impacting GDP in several economies (^{Benavides, 2023}).

5.3. Results for Asia

For the first VAR-MIDAS model only which only included Asian countries (Table 9), the country with the highest R-squared is India with 0.821 followed by Japan, with 0.68. The lowest R-squared value is China with 0.37. Regarding the forecast using this model, it was capable to capture the effects of the COVID-19 pandemic and generally for the rest of the sample. We present India (Figure 9) and Japan (Figure 10) as examples. This model was estimated with 5 lags. With the second model including influence from the US, the R-squared improved for all countries (Table 10) and also was estimated with 5 lags. The forecasts using the second model for Asia also showed better results, especially for China (Figure 11).

Source: own elaboration using R computer language.

Figure 9 Estimation for India (first model)  

Source: own elaboration using R computer language.

Figure 10 Estimation for Japan (first model)  

Source: Own elaboration using R computer language.

Figure 11 Estimation for China (second model)  

Regarding the causal effects appreciated in the first model, China showed influence over all countries in the model, while GEPU influences South Korea. As in the models for America, there is a joint causal effect from the set of variables to the GDP, but it does not hold for China, with the same results for the second model including the US. The second model showed a causal effect from China to all countries individually, and from Japan to South Korea. These results confirm China is a strong economic leader in its region and also confirm this influence does not change in the presence of external factors. Possible reasons for these causal relationships could be, for example, credit conditions. Uncertainty can tighten credit conditions, making it harder for businesses and consumers to borrow and spend, leading to slower economic growth. Also, Global Economic Integration i.e. in a highly globalized world, economies are interconnected through various channels. A shock in one part of the world can quickly transmit to other parts, affecting GDP through trade, investment, and financial linkages.