Stock prices, uncertainty and risks: Evidence from developing and advanced economies

Sánchez-Gabarre, Mary Elena Article Stock prices, uncertainty and risks: Evidence from developing and advanced economies European Journal of Government and Economics (EJGE) Provided in Cooperation with: Universidade da Coruna Suggested Citation: Sánchez-Gabarre, Mary Elena (2020) : Stock prices, uncertainty and risks: Evidence from developing and advanced economies, European Journal of Government and Economics (EJGE), ISSN 2254-7088, Universidade da Coruna, A Coruña, Vol. 9, Iss. 3, pp. 265-279, https://doi.org/10.17979/ejge.2020.9.3.6999 This Version is available at: https://hdl.handle.net/10419/298628 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-nc/4.0/ European Journal of Government and Economics 9(3), December 2020, 265-279 European Journal of Government and Economics ISSN: 2254-7088 Stock prices, uncertainty and risks: Evidence from developing and advanced economies Mary Elena Sánchez-Gabarre a* a University of A Coruña, Spain * Corresponding author at: [email protected] Abstract. This paper studies the relationship between stock prices and three types of uncertainty: economic policy uncertainty, stock market volatility, and geopolitical risks. In particular, our aim is to determine whether these forms of uncertainty play the same role in developed and developing countries. With this purpose, we take Spain and Brazil as representative cases. In order to provide new insights into the abovementioned relationship, a cointegration approach is applied, specifically an ARDL model, using monthly data from the period January 2006-December 2019 for a series of financial and macroeconomic variables. The results obtained reveal that there is no uniform effect of uncertainty in stock markets of developing and developed countries. First, in Spain, there is a high perception of uncertainty in economic policy and stock market volatility, which impact negatively in share prices, both in the short and long term. Regarding Brazil, the global uncertainty in the stock markets has effects on share prices, in both time horizons. By contrast, geopolitical risks do not show any significant impact on Brazilian and Spanish share returns. Keywords. economic policy uncertainty; geopolitics; stock markets; uncertainty; volatility JEL Codes. C22; G12; G18 DOI. https://doi.org/10.17979/ejge.2020.9.3.6999 1. Introduction The main objective of this paper is to determine the relationship between uncertainty and the performance of stock market indices in the long term, to measure its effects and evaluate the possible differential impacts depending on the type of economy. According to literature, one would expect an increase in uncertainty to be associated in the long term with a decrease in stock market indices (Malkiel & Xu, 2006; Durnev, 2010; Caldara & Iacoviello, 2019), in other words, the uncertainty-price ratio would be negative. Therefore, the initial hypothesis we analyse is to what extent uncertainty (considered according to three perspectives that are explained afterwards) could affect stock prices. In addition, we study whether those different perspectives of uncertainty have different effects depending on the type of economy. Likewise, the level of the short-term impact of an uncertainty shock on the stock market is analysed, as well as the adjustment time required to return to the previous situation in the stock markets. To study this relationship, as a novel aspect in the analysis, as far as uncertainty is concerned, we use three representative indicators at a world level, rarely taken into account together: economic policy uncertainty, stock markets volatility and geopolitical risks. Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 266 This way, we can study the impact of this uncertainty at a global level on two economies with different features (a developed country and a developing country) and evaluate the possible differential impacts depending on the type of economy. Uncertainty in the markets has often been seen as a factor affecting the functioning of the economy as a whole. Among the main drivers of this uncertainty both market volatility and policymakers' decisions are usually considered (Carney, 2016). There is a strong tendency towards financial instability, which has made globalisation a phenomenon highly determined by uncertainty. In order to fill the gap, this paper analyses the importance of uncertainty in the evolution of stock market indices, more specifically the possible long-term relationship between uncertainty and two relevant indices, corresponding to a representative developing country and a representative developed country, respectively: BOVESPA [from Brazil, which has the largest capitalization of Latin American stock markets (Coleman, Leone & Medeiro, 2018; OECD, 2019)] and IBEX 35, which is the main stock index of Spain. Thus, we can compare the results obtained for these two cases and determine whether uncertainty has different effects in both stock markets, corresponding to two different economies in terms of degree of development. This paper is organized as follows. In section 2, to contextualise the object of study, we begin by considering the relationship between uncertainty and stock markets. In section 3, we present and justify the data, variables and econometric methods used, and we also show and discuss the results of our analysis. Finally, the main conclusions of this paper are summarised. 2. Literature review The relationship between uncertainty and the economy is not a new issue to be studied. Uncertainty has intensified in the wake of the global financial crisis, crises in the euro area and partisan political disputes. Evidence of this is provided by the research of the International Monetary Fund (IMF, 2012 and 2013) and the Federal Open Market Committee (2009), which highlight that uncertainty about fiscal, regulatory and monetary policies in the United States and Europe contributed to a sharp economic decline in the years of the financial crisis in 2008 and a slowdown in the subsequent economic recovery. Due to the negative impact of economic policy uncertainty on the stock markets, it is important to include it in this section. Both in the US and in Europe it has been found that it leads to a worsening of macroeconomic performance, in addition to effects on the volatility of share prices (Baker, Bloom & David, 2016). Furthermore, Carney (2016) includes geopolitical risk, together with the economic and political uncertainty, in an "uncertainty trinity", as these three factors could have significant adverse economic effects. In recent years, the European Central Bank - in the Economic Bulletin -, the International Monetary Fund and the World Bank - in the World Economic Outlook - have systematically highlighted and monitored these risks to the outlook posed by geopolitical uncertainties. Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 267 In this respect, authors such as Erb, Harvey and Viskanta (1996) find a relationship between political risk, as measured by the International Country Risk Guide, and future stock returns. It is also important to note the abnormally bullish stock market in the weeks leading up to major elections, especially those characterised by high levels of uncertainty (Pantzalis, Stangeland & Turtle, 2000; Li & Born, 2006). This evidence is consistent with a positive relationship between risk premium and political uncertainty. In this sense, we must expect a positive relationship between the risk premium and their measure of economic policy based on uncertainty in an international scenario (Brogaard & Detzel, 2015) and relate the risk premium to political cycles (Santa-Clara & Valkanov, 2003; Belo, Gala & Li, 2013). In addition, some authors (Bittlingmayer, 1998; Voth, 2003; Boutchkova, Doshi, Durnev & Molchanov, 2012) have studied the relationship between political uncertainty and volatility in stock markets. Therefore, three perspectives are taken into account together throughout this paper: economic policy uncertainty, geopolitical risk and stock market volatility. Firstly, economic policy uncertainty provide additional information on the stock market effects. This economic policy uncertainty comprises the actions and decisions that the authorities of each country take within the scope of the economy. Focusing explicitly on the existing literature on policy uncertainty, we find, for example, Friedman (1968), Rodrik (1991), Higgs (1997) and Hassett and Metcalf (1999), who consider the detrimental economic effects of uncertainty on monetary, fiscal and regulatory policies. More recently, Born and Pfeifer (2014) and Fernández-Villaverde, Guerrón-Quintana, Kuester and Rubio-Ramírez (2015) study policy uncertainty by means of DSGE (Dynamic Stochastic General Equilibrium) models, finding moderately negative effects, whereas Pastor and Veronesi (2012, 2013) model the theoretical links between their fluctuations, policy uncertainty and stock market volatility. Accordingly, some recent papers studied the negative relationship between economic policy uncertainty and economic activity (Bloom, 2009; Giavazzi & McMahon, 2012; Julio & Yook, 2012; Bloom, 2014; Leduc & Liu, 2015; Koijen, Philipson & Uhlig, 2016; Scotti, 2016; Basu & Bundick, 2017). Secondly, geopolitical risks owe its relevance to the fact that entrepreneurs, market participants and central banks have in recent years considered geopolitical risks to be key determinants of investment decisions and stock market dynamics. Investors express concern about the economic impact of the various military and diplomatic conflicts around the world, even indicating increased concern in this area with regard to political and economic uncertainty (Carney, 2016). The study of the influence of the geopolitical environment on the economy has not been widely analised because of the limitation resulting from the absence of an indicator of geopolitical risk that is consistent with the perception of press, public, investors and policy makers. These issues were taken into account by Caldara and Iacoviello (2019), who constructed a geopolitical risk index from 1985 and show that a negative geopolitical shock induces persistent declines in Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 268 investment, employment, consumer confidence and stock market price returns. The decline in equity prices is slightly larger than that in investment, suggesting that risk premia are increasing, although only to a limited extent. These results are supported by theoretical models such as those of Ilut and Schneider (2014) and others in which high levels of uncertainty lead to declines in employment and investment, e.g. Dixit and Pindyck (1994) and Bloom, Bond and Van Reenen (2007). Thirdly, stock market volatility is a measure of uncertainty in the equity markets. Volatility is one way of measuring this price risk, probably the most widely used. In fact, this dispersion of the prices on an asset has long been considered an indicator of risk since Markowitz (1959). In this line and taking the CAPM model as a reference, some authors (Merton, 1973; Campbell, 1993, 1996; Chen, 2003; Ang, Hodrick, Xing & Zhang, 2006 and 2009; Malkiel & Xu, 2006; Chen, Ghysels & Wang, 2015; Farago & Tédongap, 2018; Hollstein & Prokopczuk, 2018; Kaeck, 2018) argue that an increase in volatility can be interpreted as a worsening of overall investment opportunities and this predicts a negative relationship. Although there is a growing body of literature regarding the adverse influence of uncertainty on stock markets, to the best of our knowledge, there is not a joint analysis of the three most studied uncertainty forms. Furthermore, it is interesting to compare whether the influence of uncertainty behaves in the same way in developed and developing countries. 3. Empirical analysis 3.1. Data and variables In the study of the literature, it was possible to clearly see the relationship of the stock markets with macroeconomic variables, and several necessary control variables were taken into account. In this research, such variables are represented by the Gross Domestic Product at Market Prices (GDP), the Consumer Price Index (CPI), the Interest Rate of the 12-month National Bond (IR) and the Real Effective Exchange Rate index (ER) of Brazil and Spain respectively. In addition, uncertainty is represented by the following variables, respectively: the EPU or Economic Policy Uncertainty (Baker et al., 2016), the GPR or Geopolitical Risk Index (Caldara & Iacovello, 2019) and the VIX or Chicago Board Options Exchange Market Volatility Index (CBOE, 2019). The choice of the EPU for our study is due to the fact that, on the one hand, the macroeconomic factor increasingly seeks to anticipate investors' expectations regarding the evolution of fundamental economic variables. It is therefore not surprising that this variable associated with macroeconomic fundamentals has been the main determinant of European stock market dynamics over the last two decades (CaixaBank Research, 2016). Furthermore, although the GPR is correlated with the EPU, it shows a remarkable amount of additional and independent variation that predicts a lower percentage of economic activity. In relation to existing proxy indicators of uncertainty that tend to increase during recessions, this index points to episodes that are independent of the economic cycle. And we have to take into Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 269 account the benchmark indicator in the major stock markets around the world, the VIX. Although it is linked to the S&P 500 index, its use is increasing due to the importance of the US index in the rest of the world's stock markets. Therefore, it should be pointed out that in both global situations, it seems plausible to argue that the direction of causality ranges from geopolitical events to stock market volatility and political uncertainty. They do not all have to behave in the same way in the face of geopolitical risks. For example, the invasion of Iraq by the United States in 2003 appears to cause increased economic policy uncertainty, although it does not induce financial volatility. Furthermore, the three indices also show a large number of independent variations, as the GPR index does not vary much during periods of economic and financial difficulties, as can be seen in the period of the dot-com bubble (1997-2001) and during the world financial crisis in 2007, when the VIX and the EPU showed increases. The GPR index also does not move around the presidential elections, periods characterised by high political uncertainty. Therefore, compared to the VIX and the EPU, the GPR index captures those events that are most likely to be exogenous to business and financial cycles, and could lead to greater financial volatility and policy uncertainty. The monthly time series used in our empirical analysis covers the period from January 2006 to December 2019 and thus consists of 156 data for each variable. This is due to the availability of consistent data for all variables. A descriptive statistic of the data, as well as their sources, is presented in Table 1. It gives an initial idea of the data to be analysed. First and secondly, appear the minimum and maximum value that the variables have had throughout the period of study. In third and fourth place, respectively, are the mean and standard deviation of the variables over the 14 years observed. As we can see, the variables with the greatest standard deviations and extreme values are those of uncertainty. It is also worth mentioning the values of the interest rate for the case of Spain, since in certain periods it has a very marked variation. Table 1. Descriptive statistics and data sources. Variables Max. Min. Mean SD Source EPU 57.099 -39.112 1.5366 17.192 https://www.policyuncertainty.com/ VIX 90.751 -38.490 1.8067 21.725 EIKON GPR 120.71 -58.516 3.4295 30.581 https://www.matteoiacoviello.com/gpr.htm BRAZIL BOVESPA 24.918 -26.636 0.7316 8.7593 EIKON GDP 2.6903 -2.5557 0.7255 1.2277 https://www.ibge.gov.br/ IR 12.337 -15.391 -0.6538 4.5457 EIKON ER 10.167 -13.901 0.0038 3.1427 https://ec.europa.eu/eurostat/data/database CPI 1.3472 -0.3339 0.4412 0.2947 https://ec.europa.eu/eurostat/data/database SPAIN IBEX 16.625 -17.033 0.0612 5.4932 EIKON GDP 3.1493 -3.5299 0.1912 1.8530 https://ec.europa.eu/eurostat/data/database IR 1347.4 -1340.0 -0.6800 154.27 EIKON ER 2.6182 -3.1091 -0.0012 0.8454 https://ec.europa.eu/eurostat/data/database CPI 1.4106 -1.9070 0.1298 0.6060 https://ine.es/ Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 270 3.2. Methodology As we have been pointing out, in this study, an approach based on econometric techniques of cointegration was adopted. In particular, we used the ARDL (or Pesaran-Shin-Smith) method, since as we will see in our case, it is preferable to other methodologies, such as that of Engle and Granger (1987) or that of Gregory and Hansen (1996). In this method, a bound test is first carried out to determine the direction of co-integration [which variable(s) could act as a dependent], and, once the optimum equation has been selected, the estimation of the equilibrium model is carried out in both the short and long term. Since the possibilities of using the different methods of co-integration depend on the characteristics of the variables in relation to their stationarity or non-stationarity, before proceeding with the application of such methods, we must determine these characteristics, a task that we will address in the following section. 3.2.1. Unit root test To try to make our analysis more robust, we opted to apply different methods, such as the DickeyFuller test (ADF) and Kwiatkowski-Phillips-Schmidt-Shin test (KPSS), to level values. In this case, an analysis with constant has been used and 12 lags have been selected due to the monthly periodicity of the data; it should also be noted that these tests have been carried out using Akaike information criterion (AIC). Table 2. Results of unit root tests. Variables ADF levels KPSS levels EPU I(0) I(0) VIX I(0) I(0) GPR I(0) I(0) BRAZIL BOVESPA I(0) I(0) GDP I(1) I(1) IR I(0) I(0) ER I(0) I(0) CPI I(1) I(0) SPAIN IBEX I(0) I(0) GDP I(1) I(0) IR I(0) I(0) ER I(0) I(0) CPI I(1) I(0) Considering all the information provided by the unit root tests (Table 2), it can be seen that the results do not coincide optimally between both tests. The only non-stationary variable in all the tests would be safely the GDP in the Brazil case. Therefore, we see contradictory outcomes with respect to CPI (Brazil and Spain) and GDP (Spain). When carrying out the co-integration analysis, these results should be taken into account, as different methods could be applied depending on the stationary properties of our variables. The ARDL (Autoregressive Distributed Delays) method, also called PSS (Pesaran-ShinSmith) test (Pesaran & Shin, 1999; Pesaran, Shin & Smith, 2001), differs from other methods in Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 271 that it allows a more in-depth study, in a less restrictive way. For this reason, in the present study, an empirical analysis will be carried out applying the ARDL method, among other reasons because the Engle-Granger and Gregory-Hansen methods, require that to carry out the co-integration analysis absolutely all the variables considered (dependent and regressors) are integrated of order 1, I(1), that is, they have a unitary root. 3.2.2. Optimal number of lags Before applying the ARDL method, the number of lags with which to work must be established. A priori, according to Pesaran and Pesaran (2009), as these are monthly data, the recommended number of lags would be 12, which for our model would imply that the number of regressions required would be no less than 815.73 million by each country, which would be unmanageable in practice. Therefore, we decided to use the procedure followed by Khan and Khan (2018), in order to know the number of optimum lags. According to this procedure, for this case the appropriate number of lags would be 12 according to AIC and R2 criteria, whereas according to FPE (Final Prediction Error) and HQ (Hannan-Quinn Information Criterion) criteria this number would be 1. Given that, as we have seen above, in our model it is very complicated in practice to work with 12 lags, according to these results we opted to establish a maximum of 1 lag. 3.2.3. Bound test As the unit-root analysis was already carried out previously, we can now continue with the next phase of the estimation of the ARDL models, which is the bound test by means of an F-statistic. This test analyses the causality of the variables, that is, whether and how (dependent/explanatory variable) each of the variables considered should be included in the model. Its null hypothesis is the non-existence of cointegration. We must point out that when analysing the results of the corresponding statistics, there are problems due to their non-standard distribution, a situation that Toda and Yamamoto (1995) have tried to justify and for which Pesaran et al. (2001) indicate a special method to carry out the test with two theoretical values, a lower limit (Li) and an upper limit (Ls). Both limits must be sought in the tables provided by Pesaran and Pesaran (2009, p. 300), the search criteria being the number of variables of the model to be studied (in our case, k=8: BOVESPA or IBEX, EPU, VIX, GPR, GDP, IR, ER and CPI), the characteristics of the model (intercept, trend) and the significance level we want to apply. In addition to contrasting the F-statistic, whose null hypothesis would indicate the nonexistence of joint significance of the first lag of the variables in levels used in the analysis, a complementary test could be made in the event that the Fstatistic does not give conclusive results: a t-test, which would contrast the individual significance of the first lag of the variables. Since we are contrasting whether or not there is a long-term relationship between the variables, it must be taken into account that if we reject the null hypothesis, it would imply the existence of co-integration. Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 272 Table 3. PSS test. Dependent variable F-test t-test Regressors BOVESPA 17.550 140.403 EPU, VIX, GPR, GDP, IR, ER, CPI Significance level I(0) I(1) I(0) I(1) 5% 2.3907 3.6049 19.1256 28.8395 10% 2.0812 3.2004 16.6500 25.6036 Dependent variable F-test t-test Regressors IBEX 18.732 149.855 EPU, VIX, GPR, GDP, IR, ER, CPI Significance level I(0) I(1) I(0) I(1) 5% 2.3907 3.6049 19.1256 28.8395 10% 2.0812 3.2004 16.6500 25.6036 The result of the bound test for our models can be seen in Table 3, where the BOVESPA and IBEX indexes are the dependent variable (the regressors of the model being the variables EPU, VIX, GPR, GDP, IR, ER and CPI). The critical values of the intervals in the contrast of F and t come from the tables of Pesaran et al. (2001) for unrestricted constant and without trend (k = 8). In addition, it should be noted that the results reflected in Table 3 point out that there is no evidence against co-integration with a significance level of 5%, so we can proceed to the analysis of the ARDL models (Castellanos-García, Pérez-Díaz-del-Río & Sánchez-Santos, 2014). 3.2.4. Model estimation Once having applied the bound test, in which we have contrasted the relationship of our variables over time, the estimation of the unrestricted error correction model will be carried out. In this way, the ARDL model for the case of Brazil, could be expressed as follows: 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝑡𝑡=𝛼𝛼0 +�𝜔𝜔𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝑡𝑡−𝑖𝑖 +�𝛿𝛿𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝐵𝐵𝐸𝐸𝑡𝑡−𝑖𝑖 + �𝛽𝛽𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝑉𝑉𝑉𝑉𝑡𝑡−𝑖𝑖 + �𝜂𝜂𝑖𝑖 1 𝑖𝑖=0 𝐺𝐺𝐵𝐵𝐺𝐺𝑡𝑡−𝑖𝑖 +�𝛾𝛾𝑖𝑖 1 𝑖𝑖=0 𝐺𝐺𝐺𝐺𝐵𝐵𝑡𝑡−𝑖𝑖 + �𝜓𝜓𝑖𝑖 1 𝑖𝑖=0 𝑉𝑉𝐺𝐺𝑡𝑡−𝑖𝑖 + �𝜑𝜑𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝐺𝐺𝑡𝑡−𝑖𝑖 +�𝜛𝜛𝑖𝑖 1 𝑖𝑖=0 𝐶𝐶𝐵𝐵𝑉𝑉𝑡𝑡−𝑖𝑖 +𝜀𝜀𝑡𝑡 Where the variables have already been defined, εt would correspond to the random disturbance, α0 to the independent term of the equation and ωi, δi, βi, ηi, γi, ψi, φi and ϖi to the regressor coefficients. Once the estimation of the band test is done, we obtain the estimation of the ARDL method, being the optimal model for the case of Brazil [ARDL(1, 0, 0, 1, 1, 0, 0)]. We will replicate the model for the case of Spain, which could be expressed as follows: 𝑉𝑉𝐵𝐵𝐵𝐵𝑉𝑉𝑡𝑡=𝛼𝛼0 +�𝜔𝜔𝑖𝑖 1 𝑖𝑖=0 𝑉𝑉𝐵𝐵𝐵𝐵𝑉𝑉𝑡𝑡−𝑖𝑖 +�𝛿𝛿𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝐵𝐵𝐸𝐸𝑡𝑡−𝑖𝑖 + �𝛽𝛽𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝑉𝑉𝑉𝑉𝑡𝑡−𝑖𝑖 + �η𝑖𝑖 1 𝑖𝑖=0 𝐺𝐺𝐵𝐵𝐺𝐺𝑡𝑡−𝑖𝑖 +�γ𝑖𝑖 1 𝑖𝑖=0 𝐺𝐺𝐺𝐺𝐵𝐵𝑡𝑡−𝑖𝑖 + �ψ𝑖𝑖 1 𝑖𝑖=0 𝑉𝑉𝐺𝐺𝑡𝑡−𝑖𝑖 + �φ𝑖𝑖 1 𝑖𝑖=0 𝐵𝐵𝐺𝐺𝑡𝑡−𝑖𝑖 +�ϖ𝑖𝑖 1 𝑖𝑖=0 𝐶𝐶𝐵𝐵𝑉𝑉𝑡𝑡−𝑖𝑖 +𝜀𝜀𝑡𝑡 Mary Elena Sánchez-Gabarre / European Journal of Government and Economics 9(3), December 2020, 265-279 279 Pesaran, M.H., Shin, Y. and Smith, R.J. (2001). Bounds Testing Approaches to the Analysis of Level Relationships. Journal of Applied Economics 16(3), 289-326. https://doi.org/10.1002/ jae.616 Rodrik, D. (1991). Policy Uncertainty and Private Investment, Journal of Development Economics. 36 (2), 229-242. https://doi.org/10.1016/0304-3878(91)90034-S Santa-Clara, P. and Valkanov, R. (2003). The Presidential Puzzle: Political Cycles and the Stock Market. The Journal of Finance 58(5), 1841-1872. https://doi.org/10.1111/1540-6261.00590 Scotti, C. (2016). Surprise and Uncertainty Indexes: Real-time Aggregation of Real-Activity Macro-Surprises. Journal of Monetary Economics 82, 1-19. https://doi.org/10.1016/ j.jmoneco.2016.06.002 Toda, H. and Yamamoto, T. (1995). Statistical Inference in Vector Autoregressions with Possibly Integrated Processes. Journal of Econometrics 66(1-2), 225-250. https://doi.org/10.1016/ 0304-4076(94)01616-8 Voth, H.J. (2003). Stock Price Volatility and Political Uncertainty: Evidence from the Interwar Period. MIT Department of Economics. Working Paper No. 02-09. https://doi.org/10.2139/ ssrn.342641