Do the changes in macroeconomic variables have a symmetric or asymmetric effect on stock prices? Evidence from Pakistan
Abstract
Purpose
The purpose of this paper is to examine whether macroeconomic variables have a symmetric or asymmetric effect on stock prices (SP) of Karachi Stock Exchange 100 index in the context of Pakistan. It also examines whether the asymmetric impact of macroeconomic variables on SP has been affected by tail events such as the global financial crisis.
Design/methodology/approach
This study uses linear and nonlinear autoregressive distributed lag models for the full sample period as well as in pre- and post-crisis periods. The whole sample period covers the data from June 2004 to June 2016 which include 145 observations in total. The pre-crisis period covers data from June 2004 to December 2007 and the post-crisis period covers the data from January 2009 to June 2016 where these periods include 43 and 90 observations, respectively.
Findings
The findings suggest that the relationship between macroeconomic variables and SP is asymmetric in the short run whereas this effect is symmetric in the long run when the whole sample period is selected. However, when pre- and post-crisis periods are selected this effect becomes asymmetric in the long run as well; that is, positive and negative shocks in macroeconomic variables do not affect the SP in the same way.
Practical implications
Investors, governments and other stakeholders are advised to consider the asymmetric behavior of macroeconomic variables and SP while making an investment or other decisions. They may consider the financial crisis as well since the asymmetric behavior of the underlying variables change as a result of the financial crisis.
Originality/value
This study extends previous studies by examining the asymmetric effect of macroeconomic variables and also contributes to the existing literature by discussing how this relationship changes as a result of the financial crisis.
Keywords
Citation
Chang, B. and Rajput, S. (2018), "Do the changes in macroeconomic variables have a symmetric or asymmetric effect on stock prices? Evidence from Pakistan", South Asian Journal of Business Studies, Vol. 7 No. 3, pp. 312-331. https://doi.org/10.1108/SAJBS-07-2018-0077
Download as .RISPublisher
:Emerald Publishing Limited
Copyright © 2018, Emerald Publishing Limited
1. Introduction
Stock markets are usually considered as performance indicators for any economy since they react quickly to policy changes and shocks (Shiskin and Moore, 1968). Financial economists and macroeconomists have, therefore, been taking a keen interest in examining the effect of changes in macroeconomic variables on stock prices (SP) (Ajaz et al., 2017). The global financial crisis has further motivated the researchers to examine the link between macroeconomic variables and SP (Ajaz et al., 2017).
Different theoretical frameworks such as efficient market hypothesis proposed by Fama (1970) and arbitrage pricing theory proposed by Ross (1976) have been used to examine the effect of changes in macroeconomic variables on SP. Some of the recent studies (see, e.g., Rahman and Uddin, 2009; Yang et al., 2014; Kuosmanen et al., 2015; Peiró, 2016) also study the impact of macroeconomic variables such as exchange rate, interest rate, consumer price index (CPI) and industrial production index (IPI) on SP. However, these studies do not provide the consistent results.
The major limitation of these studies is that they take into account the symmetric effect only by using the standard co-integration techniques such as linear autoregressive distributed lag (ARDL) model. Anoruo (2011) argues that one limitation of most of the studies is that they assume time series to be linear in nature but in reality they are nonlinear. Granger and Yoon (2002) propose a term “hidden co-integration” to describe the equilibrium relationship between negative and positive changes among the underlying variables in the long run. Thus, we use both linear and nonlinear (asymmetric) ARDL models to explore the hidden co-integration since the use of these both models can help to examine whether the results change when the nonlinear ARDL (NARDL) model has been taken into consideration. Using the asymmetric ARDL model is also justified since SP react more quickly to bad news as compared to good news (Koutmos, 1998, 1999). Another limitation of the existing literature is that it has not considered how the global financial crisis changes the asymmetric behavior of the underlying variables. Global financial crisis severely affects all asset classes (Fratzscher, 2009; Melvin and Taylor, 2009). Moreover, the financial crisis affects the asymmetric behavior of the underlying variables as well (Fosten, 2012). Time series plots depicted in Figure 1 also indicate larger variations in macroeconomic variables and SP during the crisis period. The shaded areas in Figure 1 indicate that, during the crisis period, there is a significant decrease in the stock price index. There are also larger variations in other macroeconomic variables during the crisis period, which further justifies investigating the impact of the financial crisis on the underlying variables.
Bahmani-Oskooee and Saha (2015) extend previous studies by examining the asymmetric effect of exchange rate changes on SP in the context of the USA by using monthly data. Bahmani-Oskooee and Saha (2016) replicate the same study in other countries such as the UK, Mexico, Malaysia, Korea, Japan, Indonesia, Chile, Canada and Brazil. Findings of these studies suggest that exchange changes have an asymmetric effect on SP. This study extends the works of Bahmani-Oskooee and Saha (2015, 2016) in three ways. First, it replicates those studies in the context of emerging economy like Pakistan since, as suggested by Yusof and Majid (2007), risk and returns characteristics of developing countries are different from those of developed countries. Second, this study investigates CPI, IPI and interest rate in order to investigate whether the effect of these variables on SP is symmetric or asymmetric. Third, it examines whether the behavior of the underlying variables changes as a result of the tail events such as the global financial crisis by dividing the data into pre- and post-crisis periods.
In particular, we focus on Pakistani stock market since, as indicated by Sohail and Hussain (2009), it is a highly volatile market. The major cause of volatility is instability and political uncertainty such as terrorist attacks, and judiciary crisis. Moreover, the global financial crisis also severely affected the Karachi Stock Exchange (KSE) 100 index as shown in Figure 1. During 2008 and 2009, KSE 100 index point dropped by more than 10,000 points. Due to this severe decrease in prices, the Board of Directors of Pakistan Stock Exchange placed a floor in August 2008 and later removed in December 2008. It, therefore, further motivates the authors to examine whether the association between macroeconomic variables and Pakistani stock market changes as a result of tail events such as financial crisis.
To achieve this purpose, this study uses both linear and NARDL models. These models have various advantages as compared to other standard models such as Engle–Granger and Johansen’s co-integration techniques. They provide robust results regardless of whether the variables are integrated of order 0 or 1 or combination of both and when there is a small sample size. Moreover, the NARDL model can test the hidden co-integration which other models fail to test (Lahiani et al., 2016).
For the full sample period, overall results indicate that in the short-run macroeconomic variables asymmetrically affect the SP whereas this effect is symmetric in the long run. However, the results change when the data are divided into the pre- and post-crisis periods. The NARDL findings in the pre-crisis period suggest that all macro variables have an asymmetric effect on SP both in short run and long run with the exception of interest rate in the long run. The NARDL findings for post-crisis period suggest that, in the short run, the interest rate and exchange rate have asymmetric effect whereas in the long run all macro variables have an asymmetric effect on SP except CPI.
Overall, the findings suggest the asymmetric behavior of the underlying variables in the short run. Investors and other stakeholders may need to consider this asymmetric behavior while making the short-term investment or other decisions. In addition, macroeconomic variables asymmetrically affect both in the long run and short run, while the tail events such as financial crisis have been taken into consideration; therefore, making an investment or other decisions without considering the financial crisis may lead to unfavorable consequences. Governments and other stakeholders may, therefore, need to consider the financial crisis as well while formulating the policies at the aggregate level. Moreover, researchers also need to take into account the financial crisis while investigating the asymmetric behavior of the underlying variables since ignoring the financial crisis may lead to spurious results.
The rest of the paper is organized as follows: Section 2 discusses the literature review, Section 3 discusses the data and methodology, Section 4 discusses the results related to the ARDL and nonlinear ARDL models, and Section 5 provides the concluding remarks of this paper.
2. Literature review
There is a vast growing literature focusing on the relationship between macroeconomic variables and SP. Below sections present the literature related to the main variables. Hypotheses have then been developed based on the literature review.
2.1 Exchange rate and stock prices
By using daily data from East-Asian countries, Granger et al. (2000) examine the relationship between exchange rate and SP and find that exchange rate significantly affects SP in eight out of nine countries. Nieh and Lee (2002) study the relationship between exchange rate and SP in G7 countries by using daily data from 1993 to 1996. Results of these studies conclude that there is no long-run relationship between exchange rate and SP in all G7 countries; however, the significant relationship is found in the short run. Rahman and Uddin (2009) apply the method of Granger causality test and Johansen’s (1991) co-integration test by using monthly data from 2003 to 2008 in the contexts of Pakistan, India and Bangladesh. Results neither support long-run relationship nor any causal association of exchange rate with SP. Yang et al. (2014) also study the association between these two variables for Thailand, Taiwan, Singapore, Philippines, Malaysia, Korea, Japan, Indonesia and India by using daily data from 1997 to 2010. They use Granger causality test by dividing the data into different quintiles and find the feedback relationship between exchange rate and SP during Asian financial crises for all countries except Thailand.
Some of the recent studies that investigate the relationship between exchange rate and SP include Ajaz et al. (2017), Delgado et al. (2018) and Roubaud and Arouri (2018). Ajaz et al. (2017) discuss that changes in the exchange rate can affect the SP in either way. When the local currency depreciates, the domestic products become cheaper for foreigners and, hence, increase the demand for local products. This increase in demand results in an increase in profits for export-oriented firms and hence increases the SP. On the other hand, depreciation of local currency can increase the cost of imported inputs and it ultimately increases the cost of production for the firms which are not export-oriented. This increase in cost decreases the profits for those firms and hence decreases SP. Due to this fact, change in currency level can affect the SP in either direction. Delgado et al. (2018) conduct a study in the context of Mexico by using the VAR model and find that exchange rate negatively and significantly affects the stock market index which concludes that appreciation of currency increases the SP. Roubaud and Arouri (2018) conduct a study to investigate the interaction between exchange rate and SP by using the VAR and multivariate Markov switching vector autoregressive models. Findings of their study support the nonlinear relationship between exchange rate and SP and suggest that the interaction between these variables changes from one regime to another regime.
2.2 Inflation and stock prices
There are a number of studies that have studied the relationship between inflation and SP as well. The studies such as Miller et al. (1976), Geske and Roll (1983), Chen et al. (1986) and Marshall (1992), and conclude that there is a negative correlation between inflation and SP. By using Johansen’s (1991) co-integration techniques, Mukherjee and Naka (1995) examine the negative co-integration between changes in Japanese inflation and Tokyo Stock Exchange index movement. Maysami and Koh (2000) also find evidence that these two variables are negatively co-integrated. However; there are some other studies such as Nasseh and Strauss (2000), Ibrahim (2003), Ibrahim and Aziz (2003) that suggest that SP are positively co-integrated with inflation. Dritsaki (2005) study the causal relationship between SP and inflation and find evidence in favor of unidirectional causal relationship running from inflation to SP. Delgado et al. (2018) conduct study in the context of Mexico by using the VAR model and find that inflation negatively and significantly affects the stock market index. In addition, there are some other studies that indicate that inflation affects the SP (see, e.g., Rapach, 2002; Bjørnland and Leitemo, 2009; Valcarcel, 2012; Bjørnland and Jacobsen, 2013).
2.3 Interest rate and stock prices
The interest rate has also been considered in the literature as one of the key determinants of SP. Most of the studies such as Bulmash and Trivoli (1991), Abdullah and Hayworth (1993) and Maysami and Koh (2000) investigate that there is a negative relationship between interest rate and SP. Mukherjee and Naka (1995) and Nasseh and Strauss (2000) find mixed evidence between these two variables of interest. Nasseh and Strauss (2000) indicate the SP are positively co-integrated with short-term interest rate but negatively co-integrated with the long-term interest rate. Mukherjee and Naka (1995) conclude that the SP are positively co-integrated with short-term call money rate but negatively co-integrated with long-term government bond rate. Humpe and Macmillan (2009) conduct a study in the context of USA and examine that there is a negative effect of long-term interest rate on US SP. Andrieș et al. (2014) conduct study in the context of India and find that interest rate and SP are linked with each other. Huang et al. (2016) investigate the relationship between these two variables in the context of the USA and find the negative effect of interest rate on SP. Finally, Peiró (2016) conduct study in the context of the UK, Germany and France and conclude that interest rate examines stock returns.
2.4 Industrial production and stock prices
There are also several studies including Canova and De Nicolo (1995) and Nasseh and Strauss (2000) which have examined the association between industrial production and SP. The most of the studies have found the positive relationship between these two variables. Humpe and Macmillan (2009) find that industrial production positively affects the US SP. Singh (2010) examines the causal relationship between macroeconomic variables and SP in the context of India and finds the bidirectional causal relationship between industrial production and SP. Singh (2014) also conducted a study in the context of India and found a significant effect of industrial production on the stock market index. Kuosmanen et al. (2015) also addressed the relationship between these two variables in four Nordic countries and concluded that this relationship is stronger in Sweden and Finland than in Norway and Denmark. Peiró (2016) conducted a study in the context of the UK, Germany and France and conclude that industrial production examines stock returns.
However, the main limitation of above studies is that they have examined the linear relationship between these macroeconomic variables such as interest rate, real effective exchange rate (REER), IPI and CPI and the SP by employing econometric techniques such as Engle–Granger and Johansen’s (1991) test. Anoruo (2011) argue that one limitation of these models is that they assume time series to be linear in nature but in reality they are nonlinear. As discussed earlier, Bahmani-Oskooee and Saha (2015) and Bahmani-Oskooee and Saha (2016) extend previous studies by examining the asymmetric effect of exchange rate changes on SP in the USA and the other developed countries, respectively. This study further extends these two studies in three ways: first, it focuses on the asymmetric effect of exchange rate changes on the SP in the context of developing country such as Pakistan since the risk and returns profiles of developing markets are different from those of developed markets (Yusof and Majid, 2007). Second, it investigates the asymmetric effect of other determinants of SP such as CPI, interest rate and IPI. Finally, it examines whether the asymmetric behavior of the underlying variables changes as a result of tail events such as global financial crisis since the financial crisis severely affects all asset classes (Fratzscher, 2009; Melvin and Taylor, 2009).
We mainly focus on Pakistani stock market since it is a highly volatile market (Sohail and Hussain, 2009). The major cause of this volatility is political uncertainty and instability such as terrorist attacks and judiciary crisis. Figure 1 also presents large variations in Karachi stock index and other macro variables especially during financial crisis period. In addition, Figure 1 further indicates that interest rate and SP move upward before the crisis period; however, after crisis period stock price index is moving upward whereas interest rate is moving downward. In the same way, SP and real exchange rate show the similar trend; however, during 2013, real exchange rate is showing downward trend whereas SP show upward trend. It overall indicates that Pakistani stock market does not move in the same way as the macroeconomic variables which further motivates the authors to investigate the asymmetric relationship between these variables in the context of Pakistan.
In order to fulfill the objectives of this study, the following hypotheses have been proposed which are tested during the full sample, as well as in the pre- and the post-crisis periods:
Exchange rate changes have an asymmetric impact on SP.
Interest rate changes have an asymmetric impact on SP.
CPI changes have an asymmetric impact on SP.
IPI changes have an asymmetric impact on SP.
3. Data and methodology
3.1 Data sources and measurement
This study focuses on the data with a monthly frequency which covers the period from June 2004 to June 2016 which includes 145 observations in total. Taking the advantage of NARDL model, which is robust to a small sample size, this study further divides the data into pre- and the post-crisis periods for checking whether the global financial crisis of 2008 has brought significant changes in the asymmetries of the data. The pre-crisis period covers the data from June 2004 to December 2007, whereas the post-crisis period covers data from January 2009 to June 2016[1]. The pre-crisis period consists of 43 observations whereas the post-crisis period consists of 90 observations in total. Data for all independent variables such as CPI, IPI, interest rate and REER are taken from International Financial Statistics, a database of International Monetary Fund. CPI indicates consumer price index whereas IPI indicates industrial production index which is used as a proxy for GDP since the data for GDP are not available in the monthly frequency. Irate and REER indicate interest rate and the REER, respectively. T-bill rates of six months are used as a proxy for the interest rate. All the independent variables, except interest, have been represented in indexes, where the base year is 2010. Finally, data for a dependent variable, that is, SP of KSE 100 index are taken from Bloomberg. Following the works of Bahmani-Oskooee and Saha (2015) and Bahmani-Oskooee and Saha (2016), all variables are represented in a natural logarithm.
3.2 Methodology
This study focuses on linear ARDL model proposed by Pesaran et al. (2001) and NARDL model proposed by Shin et al. (2014) to examine whether changes in macroeconomic variables have a symmetric or asymmetric effect on SP both in the long and short run. Both models have been used to compare whether results change when we take into consideration the NARDL model. Further, we divide the data into pre- and the post-crisis periods for examining whether the behavior of the underlying variables changes as a result of the tail events such as the global financial crisis.
NARDL model is an extension of linear ARDL model which takes into account the asymmetries present in the data. This framework jointly tests the co-integration as well as nonlinear asymmetries and gives better results as compared to other co-integration techniques when the sample size is small (Romilly et al., 2001). Another advantage of this model is that it gives better results regardless of whether the variables are integrated of order zero I(0) or one I(1) (Nusair, 2016). Moreover, it helps in examining the hidden co-integration (Shahzad et al., 2017). Hence, we can say that the NARDL model can help in assessing whether the co-integration is linear, nonlinear (asymmetric) or there is no co-integration.
As discussed above, NARDL can be applied when the variables are integrated of order zero I(0), one I(1) or a combination of both. However, NARDL model cannot be applied when the variables are integrated of order two I(2). Therefore, in order to confirm the order of integration of the variables, the ADF test has been used in this study.
Since NARDL model is basically an extension of linear ARDL model. Therefore, we first present the general form of unrestricted error correction model for linear ARDL which is as given below:
Since in our case, few variables are integrated of order zero whereas other variables are integrated of order one, therefore, we can consider both upper bounds and lower bounds critical values. In this case, if test statistics falls above the upper bounds critical values the null hypothesis is rejected, and if test statistics falls below the lower bounds then the null hypothesis is not rejected whereas the results become inconclusive if the obtained test statistic is in between the upper and lower bounds critical values. Linear ARDL model assumes that all exogenous variables have a symmetric effect on the dependent variable. Linear ARDL model used in our study is as follows:
For taking into account the asymmetric relationship, Shin et al. (2014) build NARDL model based on following asymmetric long-run equilibrium relationship:
When we combine Equation (3) with the linear ARDL model given in Equation (1), asymmetric error correction model, as mentioned below, can be obtained:
The NARDL model is implemented in the same way as the linear ARDL model. The first step is to estimate the error correction model given in Equation (7) through OLS. The second step is to estimate the asymmetric long-run relationship with the help of bounds test where the variables are used at level. F-statistics proposed by Pesaran et al. (2001) which is denoted by F_{PSS} is used to test the null hypothesis (ρ=θ^{+}=θ^{−}= 0) of no co-integration. Decision for rejecting or not rejecting the null hypothesis of no co-integration is made in the same way by considering upper and lower bounds as discussed for the linear ARDL model. The third step is to consider whether the effect of exogenous variables is symmetric or asymmetric on the dependent variable both in the long run (θ^{+}=θ^{−}) and short run
The NARDL model used in this study is given below:
4. Results analysis and discussion
Keeping in view the limitations of previous studies, we focus on both linear and NARDL models to examine whether the changes in macroeconomic variables have a symmetric or asymmetric effect on SP in the context of Pakistan. Moreover, we focus whether this asymmetric behavior changes as a result of the financial crisis. Since both of these models require that none of the variables should be integrated of order two I(2). So we first examine the order of integration of each variable by using the data for the whole sample period, pre-crisis period and post-period. Results of the ADF test for checking the order of integration of variables are given in Table I. Results suggest that SP, CPI, IPI and REER are integrated of order one, for all three periods of the study. Interest rate (Irate) is integrated of order zero for the whole sample and the pre-crisis period whereas it is integrated of order one when the post-crisis period is used. Since none of our variables is integrated of order two I(2) so we proceed further with our analysis.
Both linear and nonlinear models given in Equations (2) and (8) are first estimated by using maximum four lags in differenced variables, then general to specific approach is used for selecting the best model by using Akaike’ information criterion (AIC) criteria. Different models are estimated by changing the lags of each differenced variables and the models with lowest AIC values are finally retained which are given in Tables II–VII. This study provides a detailed discussion of linear and NARDL model results for the whole sample period in Tables II and III. The findings for pre-crisis and post-crisis period analysis are presented in Tables IV–VII.
Table II provides the results of linear ARDL model estimated for the whole sample period. Results are provided in three panels. Panel A provides the results for short-run estimation. Panel B provides results for long-run coefficients and finally Panel C provides results for various diagnostic tests used. Results indicate that in the short run, REER, interest rate and CPI are significant determinants of SP; whereas in the long run, only interest rate and CPI are significant determinants of SP. However, a precondition for long-run coefficients to be valid is that there must exist a co-integration. Wald test statistics are used to test the joint significance of lagged level variables. The obtained value of F-statistics is 5.36 which is above the upper bound critical values suggested by Pesaran et al. (2001); hence we can conclude that there exists long-run co-integration between macroeconomic variables and SP. Moreover, there is an alternative test for co-integration based on error correction mechanism (ECM). The basic requirement is that the ECM coefficient should be negative and significant. In our case, ECM is significant which further substantiates the previous findings that there exists a long-run equilibrium relationship. The size of the ECM coefficient (−0.07) indicates that the system gets back to long-run equilibrium at a speed of 7 percent per month. In addition, some other diagnostics tests have been applied as well. Lagrange multiplier (LM) test has been used to test whether there is no serial correlation present in the data. Obtained LM statistic (3.4) is insignificant in our case since the critical value at 5 percent significance level is 21.03 so we can say that there is no issue of serial correlation in our data. Moreover, Ramsey’s RESET test has been used in order to analyze whether the model is correctly specified. The test statistic for the RESET suggests that the model is not correctly specified since the obtained value is much higher than the critical value of 3.84. Moreover, adjusted R^{2} have also been reported in order to analyze whether the model enjoys a good fit. Adjusted R^{2} of 0.52 indicates that model enjoys a good fit.
Since the above results indicate that in the short-run REER, interest rate and CPI are significant determinants of SP whereas in the long run, only interest rate and CPI are significant determinants of SP. For example, the coefficient of interest rate in the long-run is −1.29 which indicates that a 1 percent increase in interest rate will decrease SP by 1.29 percent and vice versa. But the question is whether, in reality, an increase in interest rate affects the SP in the same way as for decrease in interest rate. The same question remains unexplored for other variables such as CPI and IPI.
Table III provides results for each exogenous variable by dividing it into positive and negative shocks in order to explore whether the increase in each exogenous variable affects the SP in the same way as for decrease in that variable. In the short run, results indicate that an increase in interest rate and CPI significantly affects SP whereas a decrease in these variables does not significantly affect SP. Moreover, both decrease and increase in IPI significantly affect SP but in the opposite direction. Hence, it can be concluded that there is an asymmetric effect of these variables on SP. However, in the long run, positive and negative changes in each of the exogenous variables provide insignificant results suggesting that in the long run neither positive changes nor negative changes in exogenous variables explain variation in SP. The overall findings suggest that, when the whole sample period is selected, macroeconomic variables asymmetrically affect the SP in the short run whereas this effect becomes symmetric in the long run. A possible reason is that in the short run, the investors and other stakeholders differently react to positive and negative shocks because of asymmetric information available to them. However, in the long run, they are absorbed by the full information available so they equally react to the positive and negative news.
Like in linear ARDL model, the precondition in NARDL model is that for long-run coefficients to be valid, the co-integration must exist. The value for obtained F-statistics for joint significance is 10.87 which is above the upper bound critical values suggested by Pesaran et al. (2001). Hence, we can say that there exists co-integration in the long run. In our case, ECM is also significant which further substantiates the previous findings that there exists a long-run equilibrium relationship. The size of the ECM coefficient (−0.13) indicates that the system gets back to long-run equilibrium at a speed of 13 percent per month. In addition, some other diagnostics tests have been applied as well. Obtained LM statistic suggests that there is no problem of serial correlation. Ramsey’s RESET test suggests the model is correctly specified since the obtained value is lower than the critical value of 3.84. Adjusted R^{2} for NARDL model is 0.69 as compared to 0.52 in the linear ARDL model, indicating that the NARDL model enjoys a better fit than the linear ARDL model. Formal test statistics for asymmetry suggest that only IPI has an asymmetric effect on SP in the short run.
Tables IV and V provide the results for the pre-crisis period. Table IV provides the linear relation between macroeconomic variables and SP where results indicate that in the long run, IPI and CPI significantly and positively affect SP, whereas interest rate negatively affects SP. On the other hand, the results given in Table V suggest that all macro variables have an asymmetric effect on SP both in the short run and long run except interest rate in the long run.
Finally, Tables VI and VII provide the results for a post-crisis period. Table VI indicates the linear relation between macroeconomic variables and SP where results suggest that, in the long run, CPI significantly and positively affect SP whereas interest rate negatively affects SP. The results in Table VII indicate that, in short run, interest rate and exchange rate have an asymmetric effect on SP whereas IPI and CPI have a symmetric effect on SP. In the long run, interest rate, IPI and exchange rate have an asymmetric effect on SP whereas CPI has a symmetric effect on SP.
Bahmani-Oskooee and Saha (2016) conducted a study on countries like Brazil, Canada, Chile, Indonesia, Japan, Korea, Malaysia, Mexico and the UK by using monthly data. Their study only examined the asymmetric impact of exchange rate changes on SP and find that it asymmetrically affects the SP; however, for majority of the countries this asymmetric effect is in the short run only. The findings of our study are consistent with Bahmani-Oskooee and Saha (2016) when the whole sample period is used whereas those findings change when the data are divided into the pre- and the post-crisis periods. The changes in results are due to the fact that the tail events such global financial crisis severely affect all asset classes (Fratzscher, 2009; Melvin and Taylor, 2009). Moreover, as suggested by Fosten (2012), the financial crises affect the asymmetric behavior of the underlying variables as well. These findings also indicate that asymmetric behavior of the underlying variables depends upon the sample period selected. Without considering the sample, the selected period may lead to spurious results. Investors and other stakeholders may, therefore, consider the financial crisis and sample of data selected before taking any policy or investment decisions.
Finally, Figures 2 and 3 present the graphs of stability tests. Figure 2 presents the CUSUM and CUSUMQ results for the linear ARDL model, whereas, Figure 3 presents the results for the NARDL model. Graphs indicate that both linear and NARDL models are stable for all sample periods except CUSUMQ test for linear ARDL model in the whole sample period.
5. Conclusion
There are various studies that investigate the relationship between macroeconomic variables and SP. Some of the macroeconomic variables included in these studies are inflation, exchange rate, interest rate and industrial production. Bahmani-Oskooee and Saha (2015) provide a review article which indicates that macroeconomic variables mainly affect the SP in the short run and not in the long run. Moreover, majority of the studies reviewed assume that the relationship between these variables is symmetric. However, some of the recent studies indicate that exchange rate changes have an asymmetric effect on SP. Bahmani-Oskooee and Saha (2015) use the data in the context of the USA to investigate that exchange rate changes asymmetrically affect the S&P 500 index in the USA and recommend extending this study in context of other countries.
Following the recommendation of Bahmani-Oskooee and Saha (2015), we conduct this study in context of Pakistan and further extend it by examining the asymmetric effect of other macroeconomic variables such as interest rate, CPI and IPI on SP. We also investigate whether this relationship changes as a result of tail events such as global financial crisis. We use both linear and NARDL models of this analysis purpose.
Findings from linear ARDL model indicate that the interest rate has a negative and significant effect on SP in the long run given all sample periods. CPI has also significant, but positive effect on SP for all sample periods. Whereas IPI has significant and positive effect during pre-crisis period only and finally exchange rate has insignificant effect for all sample periods. The NARDL model is used to check the asymmetry both in the long run and short run and for all sample periods. In the short run, results indicate that interest rate has an asymmetric effect on SP during pre- and post-crisis period only, IPI for the whole and pre-crisis period only, CPI for the pre-crisis period only and exchange rate in pre- and post-crisis periods only. In the long run, results indicate that interest rate and IPI have an asymmetric effect on SP during the post-crisis period, CPI for the pre-crisis period and exchange rate during the post-crisis period only. These results, therefore, conclude that overall findings are dependent upon the selected sample period. Any policy or investment decisions without considering the financial crisis may lead to unfavorable consequences. While devising policies based on the asymmetric behavior of the underlying variables sample period selected may be taken into consideration as well.
Finally, all diagnostics test statistics, used in this study, suggest that NARDL models are better specified, enjoy a better fit and are more stable than linear ARDL models.
Figures
ADF test results at level and first difference
Whole sample | Pre-crisis period | Post-crisis period | ||||
---|---|---|---|---|---|---|
Variable | p-value at level | p-value at first difference | p-value at level | p-value at first difference | p-value at level | p-value at first difference |
Panel A: Pakistan | ||||||
Ln SP | 0.82 | 0.00* | 0.29 | 0.00* | 0.92 | 0.00* |
Ln CPI | 0.33 | 0.00* | 1.00 | 0.00* | 0.05 | 0.00* |
Ln IPI | 0.43 | 0.00* | 0.13 | 0.00* | 0.93 | 0.00* |
Ln Irate | 0.00* | 0.00* | 0.04* | 0.00* | 0.96 | 0.00* |
Ln REER | 0.83 | 0.00* | 0.82 | 0.00* | 0.67 | 0.00* |
Notes: Whole sample period covers data from June 2004 to June 2016, the pre-crisis period from June 2004 to December 2007 and the post-crisis period is from January 2009 to June 2016. *Significant at 5 percent critical level
ARDL model results: whole sample period
Panel A: Short run | |||||
Variables | Lags | ||||
0 | 1 | 2 | 3 | 4 | |
Δ ln SP | |||||
Δ ln REER | 0.96 (2.16**) | ||||
Δ ln Irate | −0.27 (−2.00**) | ||||
Δ ln IPI | |||||
Δ ln CPI | −0.07 (−2.73***) | ||||
Panel B: Long run | |||||
Ln REER | Ln Irate | Ln IPI | Ln CPI | C | |
−0.52 (−0.21) | −1.29 (−2.38***) | 0.72 (0.76) | 0.81 (1.82*) | 8.05 (0.43) | |
Panel C: Diagnostics | |||||
ECMt-1 | Wald (joint significance) | Adj. R^{2} | RESET | LM | |
−0.07 [0.00***] | 5.36*** | 0.52 | 15.21 | 3.4 |
Notes: This table reports the results of linear autoregressive distributed lag (ARDL) model regressions of stock prices for the whole sample period – June 2004 to June 2016. Panel A represents the results in short run, and Panel B in the long run, where SP, REER Irate, CPI, and IPI indicate stock prices, real effective exchange rate, interest rate, consumer price index, and industrial production index, respectively; whereas values in the brackets represent t-values. In the short run, only significant coefficients have been presented. Finally, Panel C represents the results of diagnostic tests. ECM and Wald (joint significance) tests are used to indicate whether there is co-integration in the long run. Adj. R^{2}, RESET and LM tests indicate whether the model enjoys a good, model is correctly specified and whether there is no problem of serial correlation, respectively. *,**,***Indicate the rejection of the null hypothesis at 10, 5, and 1 percent significant levels, respectively
Nonlinear ARDL model results: whole sample period
Panel A: Short run | ||||
Variables | Lags | |||
0 | 1 | 2 | ||
Δ ln SP | 0.26 (2.36**) | 0.01 (0.05) | ||
Δ ln Irate^{−} | −0.13 (−0.42) | 0.07 (0.22) | −0.12 (−0.39) | |
Δ ln Irate^{+} | −0.13 (−0.49) | −0.36 (−1.68*) | −0.22 (−1.00) | |
Δ ln IPI^{−} | −0.51 (−1.76*) | 0.42 (1.69*) | −0.15 (−0.55) | |
Δ ln IPI^{+} | 0.30 (1.79*) | 0.05 (0.25) | 0.35 (1.46) | |
Δ ln CPI^{−} | 1.51 (0.87) | −1.88 (−0.45) | −1.04 (−0.98) | |
Δ ln CPI^{+} | −0.44 (−0.34) | 2.36 (1.74*) | 0.16 (0.11) | |
Δ ln REER^{−} | −0.33 (−0.27) | 0.07 (0.06) | 0.07 (0.06) | |
Δ ln REER^{+} | −1.13 (−1.27) | 0.46 (0.52) | 0.46 (0.53) | |
Panel B: Long run | ||||
Ln SP | Ln Irate^{−} | Ln Irate^{+} | Ln IPI^{−} | Ln IPI^{+} |
−0.34 (−3.19***) | −0.02 (−0.25) | −0.10 (−1.13) | −0.28 (−1.26) | −0.25 (−1.42) |
Ln CPI^{−} | Ln CPI^{+} | Ln REER^{−} | Ln REER^{+} | |
1.41 (0.49) | 0.02 (0.03) | 0.35 (0.51) | 0.45 (0.76) | |
Panel C: Diagnostics | ||||
F | ECMt-1 | LM | Adj. R^{2} | RESET |
1.59* | −0.13*** | 2.5 | 0.69 | 1.66 |
Joint significance | Ln Irate_{SR} | Ln IPI_{SR} | Ln CPI_{SR} | Ln REER_{SR} |
10.87*** | −0.01 [0.99] | −2.11 [0.03**] | 0.85 [0.39] | 0.38 [0.70] |
Ln Irate_{LR} | Ln IPI_{LR} | Ln CPI_{LR} | Ln REER_{LR} | |
0.47 [0.63] | −0.12 [0.90] | 0.41 [0.68] | −0.08 [0.93] |
Notes: This table reports the results of nonlinear autoregressive distributed lag (NARDL) model regressions of stock prices for the whole sample period – June 2004 to June 2016. Panel A represents the results in the short run, and Panel B in the long run, where SP, Irate, CPI, REER, and IPI indicate stock prices, interest rate, consumer price index, real effective exchange rate and industrial production index, respectively; whereas values in the brackets represent t-values. Moreover, a positive and negative sign with each variable indicates that each variable has been decomposed into positive and negative changes. These positive and negative change variables are used to indicate whether there is the asymmetric effect of these variables on stock prices. Finally, Panel C represents the results of diagnostic tests. ECM and Wald (joint significance) tests are used to indicate whether there is co-integration in the long run. Wald Adj. R^{2}, RESET and LM tests indicate that whether the model enjoys a good, model is correctly specified and whether there is no problem of serial correlation, respectively. Moreover, Panel C provides results for short-run and long-run asymmetries where subscript “SR” with each variable indicates Wald test for short-run asymmetry and subscript “LR” with each variable indicates Wald test for long-run asymmetry. Values in large brackets in Panel C are the p-values. Significant p-values suggest that there is an asymmetric relationship. *,**,***Indicate the rejection of the null hypothesis at 10, 5, and 1 percent significant levels, respectively
ARDL model results: pre-crisis period
Panel A: Short run | |||||
Variables | Lags | ||||
0 | 1 | 2 | 3 | 4 | |
Δ ln SP | |||||
Δ ln REER | 2.05 (1.90*) | 1.94 (1.78*) | |||
Δ ln Irate | −0.95 (−2.93***) | −0.40 (−1.68*) | |||
Δ ln IPI | |||||
Δ ln CPI | −4.80 (−3.63***) | ||||
Panel B: Long run | |||||
Ln REER | Ln Irate | Ln IPI | Ln CPI | C | |
−0.14 (−0.05) | −0.95 (−2.99***) | 1.45 (4.23***) | 2.83 (3.76***) | −6.27 (−0.51) | |
Panel C: Diagnostics | |||||
ECMt-1 | Wald (joint significance) | Adj. R^{2} | RESET | LM | |
−0.39 [0.00***] | 5.12*** | 0.61 | 1.03 | 1.11 |
Notes: This table reports results of linear autoregressive distributed lag (ARDL) model regressions of stock prices for the pre-crisis period – June 2004 to December 2007. Panel A represents the results in the short run, and Panel B in the long run, where SP, REER Irate, CPI, and IPI indicate stock prices, real effective exchange rate, interest rate, consumer price index, and industrial production index, respectively; whereas values in the brackets represent t-values. In short run, only significant coefficients have been presented. Finally, Panel C represents the results of diagnostic tests. ECM and Wald (joint significance) tests are used to indicate whether there is co-integration in the long run. Adj. R^{2}, RESET and LM tests indicate that whether the model enjoys a good, model is correctly specified and whether there is no problem of serial correlation, respectively. *,**,***Indicate the rejection of the null hypothesis at 10, 5, and 1 percent significant levels, respectively
Nonlinear ARDL model results: pre-crisis period
Panel A: Short run | ||||
Variables | Lags | |||
0 | 1 | 2 | ||
Δ ln SP | 1.11 (6.45***) | 0.41 (3.00***) | ||
Δ ln Irate^{−} | 137.24 (2.33**) | −7.13 (−1.74**) | ||
Δ ln Irate^{+} | 0.05 (0.12) | −0.46 (−1.21) | ||
Δ ln IPI^{−} | −1.22 (−2.84**) | −1.62 (−3.13***) | −1.60 (−3.99***) | |
Δ ln IPI^{+} | 0.51 (1.93*) | 0.03 (0.15) | 0.51 (2.62**) | |
Δ ln CPI^{−} | −3.17 (−2.86**) | −29.9 (−4.45***) | ||
Δ ln CPI^{+} | −3.79 (−3.61***) | 3.25 (1.91*) | ||
Δ ln REER^{−} | 0.95 (0.91) | −2.38 (−1.26) | ||
Δ ln REER^{+} | −1.78 (−1.25) | 4.12 (1.13) | ||
Panel B: Long run | ||||
Ln SP | Ln Irate^{−} | Ln Irate^{+} | Ln IPI^{−} | Ln IPI^{+} |
−0.74 (−2.48**) | 133 (1.79*) | −0.69 (−3.00***) | −0.001 (−0.01) | 0.27 (0.62) |
Ln CPI^{−} | Ln CPI^{+} | Ln REER^{−} | Ln REER^{+} | |
−11.52 (−1.91*) | −0.41 (−0.23) | 4.77 (1.62) | 9.62 (1.87*) | |
Panel C: Diagnostics | ||||
F | ECMt-1 | LM | Adj. R^{2} | RESET |
2.33* | −0.86 [0.09*] | 0.95 | 0.69 | 3.40 |
Joint significance | Ln Irate_{SR} | Ln IPI_{SR} | Ln CPI_{SR} | Ln REER_{SR} |
18.92*** | 2.49 [0.03**] | −2.53 [0.03**] | −4.32 [0.02**] | −2.52 [0.08*] |
Ln Irate_{LR} | Ln IPI_{LR} | Ln CPI_{LR} | Ln REER_{LR} | |
1.69 [0.13] | −0.42 [0.03**] | −1.69 [0.09*] | −0.90 [0.001***] |
Notes: This table reports the results of nonlinear autoregressive distributed lag (NARDL) model regressions of stock prices for the pre-crisis period – June 2004 to December 2007. Panel A represents the results in the short run, and Panel B in the long run, where SP, Irate, CPI, REER, and IPI indicate stock prices, interest rate, consumer price index, real effective exchange rate and industrial production index, respectively; whereas values in the brackets represent t-values. Moreover, a positive and negative sign with each variable indicates that each variable has been decomposed into positive and negative changes. These positive and negative change variables are used to indicate whether there is the asymmetric effect of these variables on stock prices. Finally, Panel C represents the results of diagnostic tests. ECM and Wald (joint significance) tests are used to indicate whether there is co-integration in the long run. Wald Adj. R^{2}, RESET and LM tests indicate that whether the model enjoys a good, model is correctly specified and whether there is no problem of serial correlation, respectively. Moreover, Panel C provides results for short-run and long-run asymmetries where subscript “SR” with each variable indicates Wald test for short-run asymmetry and subscript “LR” with each variable indicates Wald test for long-run asymmetry. Values in large brackets in Panel C are the p-values. Significant p-values suggest that there is an asymmetric relationship. *,**,***Indicate the rejection of the null hypothesis at 10, 5, and 1 percent significant levels, respectively
ARDL model results: post-crisis
Panel A: Short run | |||||
Variables | Lags | ||||
0 | 1 | 2 | 3 | 4 | |
Δ ln SP | |||||
Δ ln REER | |||||
Δ ln Irate | −0.31 (−1.70*) | ||||
Δ ln IPI | |||||
Δ ln CPI | 2.83 (3.86***) | ||||
Panel B: Long run | |||||
Ln REER | Ln Irate | Ln IPI | Ln CPI | C | |
−1.15 (−0.92) | −0.55 (−1.72*) | 0.81 (1.40) | 2.64 (5.31***) | −0.12 (−0.02) | |
Panel C: Diagnostics | |||||
ECMt-1 | Wald (joint significance) | Adj. R^{2} | RESET | LM | |
−0.16*** | 12.16*** | 0.68 | 2.91 | 0.43 |
Notes: This table reports results of linear autoregressive distributed lag (ARDL) model regressions of stock prices for the post-crisis period – January 2009 to June 2016. Panel A represents the results in the short run, and Panel B in the long run, where SP, REER Irate, CPI, and IPI indicate stock prices, real effective exchange rate, interest rate, consumer price index, and industrial production index, respectively; whereas values in the brackets represent t-values. In the short run, only significant coefficients have been presented. Finally, Panel C represents the results of diagnostic tests. ECM and Wald (joint significance) tests are used to indicate whether there is co-integration in the long run. Adj. R^{2}, RESET and LM tests indicate that whether the model enjoys a good, model is correctly specified and whether there is no problem of serial correlation, respectively. *,**,***Indicate the rejection of the null hypothesis at 10, 5, and 1 percent significant levels, respectively
Nonlinear ARDL model results: post-crisis
Panel A: Short run | ||||
Variables | Lags | |||
0 | 1 | 2 | ||
Δ ln SP | 0.14 (1.38) | |||
Δ ln Irate^{−} | −0.21 (−0.71) | −0.05 (−0.14) | ||
Δ ln Irate^{+} | 0.87 (1.74*) | 0.10 (0.25) | ||
Δ ln IPI^{−} | −0.13 (−0.61) | −0.01 (−0.02) | ||
Δ ln IPI^{+} | 0.12 (0.80) | −0.01 (−0.16) | ||
Δ ln CPI^{−} | 3.95 (1.21) | 5.41 (1.33) | ||
Δ ln CPI^{+} | 1.79 (1.31) | 0.10 (0.06) | ||
Δ ln REER^{−} | −1.53 (−1.13) | 1.10 (0.76) | ||
Δ ln REER^{+} | 1.86 (2.10**) | −0.06 (−0.08) | ||
Panel B: Long run | ||||
Ln SP | Ln Irate^{−} | Ln Irate^{+} | Ln IPI^{−} | Ln IPI^{+} |
−0.58 (−5.50***) | 0.01 (0.16) | −0.30 (−1.99**) | 0.01 (0.12) | 0.14 (0.60) |
Ln CPI^{−} | Ln CPI^{+} | Ln REER^{−} | Ln REER^{+} | |
−0.85 (−0.30) | 1.22 (1.21) | −3.55 (−3.45***) | 0.51 (1.12) | |
Panel C: Diagnostics | ||||
F | ECMt-1 | LM | Adj. R^{2} | RESET |
1.85** | −0.16*** | 9.25*** | 0.73 | 3.60 |
Joint significance | Ln Irate_{SR} | Ln IPI_{SR} | Ln CPI_{SR} | Ln REER_{SR} |
4.78*** | −1.7 [0.09*] | −0.95 [0.24] | 0.64 [0.52] | −1.85 [0.05**] |
Ln Irate_{LR} | Ln IPI_{LR} | Ln CPI_{LR} | Ln REER_{LR} | |
1.95 [0.05**] | 1.8 [0.07*] | −0.59 [0.55] | −2.8 [0.00***] |
Notes: This table reports the results of nonlinear autoregressive distributed lag (NARDL) model regressions of stock prices for the post-crisis period – January 2009 to June 2016. Panel A represents the results in the short run, and Panel B in the long run, where SP, Irate, CPI, REER, and IPI indicate stock prices, interest rate, consumer price index, real effective exchange rate and industrial production index, respectively; whereas values in the brackets represent t-values. Moreover, a positive and negative sign with each variable indicates that each variable has been decomposed into positive and negative changes. These positive and negative change variables are used to indicate whether there is the asymmetric effect of these variables on stock prices. Finally, Panel C represents the results of diagnostic tests. ECM and Wald (joint significance) tests are used to indicate whether there is co-integration in the long run. Wald Adj. R^{2}, RESET and LM tests indicate that whether the model enjoys a good, model is correctly specified and whether there is no problem of serial correlation, respectively. Moreover, Panel C provides results for short-run and long-run asymmetries where subscript “SR” with each variable indicates Wald test for short-run asymmetry and subscript “LR” with each variable indicates Wald test for long-run asymmetry. Values in large brackets in Panel C are the p-values. Significant p-values suggest that there is an asymmetric relationship. *,**,***Indicate the rejection of the null hypothesis at 10, 5, and 1 percent significant levels, respectively
Note
Following the reviewer’s suggestion, year 2008 has been excluded from the pre-crisis and the post-crisis period samples, since the pre-crisis period is prior to 2008 and the post-crisis period is after 2008.
References
Abdullah, D.A. and Hayworth, S.C. (1993), “Macroeconometrics of stock price fluctuations”, Quarterly Journal of Business and Economics, Vol. 32 No. 1, pp. 50-67.
Ajaz, T., Nain, M.Z., Kamaiah, B. and Sharma, N.K. (2017), “Stock prices, exchange rate and interest rate: evidence beyond symmetry”, Journal of Financial Economic Policy, Vol. 9 No. 1, pp. 2-19.
Andrieș, A.M., Ihnatov, I. and Tiwari, A.K. (2014), “Analyzing time–frequency relationship between interest rate, stock price and exchange rate through continuous wavelet”, Economic Modelling, Vol. 41, pp. 227-238.
Anoruo, E. (2011), “Testing for linear and nonlinear causality between crude oil price changes and stock market returns”, International Journal of Economic Sciences and Applied Research, Vol. 4 No. 3, pp. 75-92.
Bahmani-Oskooee, M. and Saha, S. (2015), “On the relation between stock prices and exchange rates: a review article”, Journal of Economic Studies, Vol. 42 No. 4, pp. 707-732.
Bahmani-Oskooee, M. and Saha, S. (2016), “Do exchange rate changes have symmetric or asymmetric effects on stock prices?”, Global Finance Journal, Vol. 31, pp. 57-72.
Bjørnland, H.C. and Jacobsen, D.H. (2013), “House prices and stock prices: different roles in the US monetary transmission mechanism”, The Scandinavian Journal of Economics, Vol. 115 No. 4, pp. 1084-1106.
Bjørnland, H.C. and Leitemo, K. (2009), “Identifying the interdependence between US monetary policy and the stock market”, Journal of Monetary Economics, Vol. 56 No. 2, pp. 275-282.
Bulmash, S.B. and Trivoli, G.W. (1991), “Time–lagged interactions between stocks prices and selected economic variables”, The Journal of Portfolio Management, Vol. 17 No. 4, pp. 61-67.
Canova, F. and De Nicolo, G. (1995), “Stock returns and real activity: a structural approach”, European Economic Review, Vol. 39 No. 5, pp. 981-1015.
Chen, N.-F., Roll, R. and Ross, S.A. (1986), “Economic forces and the stock market”, Journal of Business, Vol. 59 No. 3, pp. 383-403.
Delgado, N.A.B., Delgado, E.B. and Saucedo, E. (2018), “The relationship between oil prices, the stock market and the exchange rate: evidence from Mexico”, The North American Journal of Economics and Finance, Vol. 45, pp. 266-275.
Dritsaki, M. (2005), “Linkage between stock market and macroeconomic fundamentals: case study of Athens Stock Exchange”, Journal of Financial Management & Analysis, Vol. 18 No. 1, pp. 38-47.
Fama, E.F. (1970), “Efficient capital markets: a review of theory and empirical work”, The Journal of Finance, Vol. 25 No. 2, pp. 383-417.
Fosten, J. (2012), “Rising household diesel consumption in the United States: a cause for concern? Evidence on asymmetric pricing”, Energy Economics, Vol. 34 No. 5, pp. 1514-1522.
Fratzscher, M. (2009), “What explains global exchange rate movements during the financial crisis?”, Journal of International Money and Finance, Vol. 28 No. 8, pp. 1390-1407.
Geske, R. and Roll, R. (1983), “The fiscal and monetary linkage between stock returns and inflation”, The Journal of Finance, Vol. 38 No. 1, pp. 1-33.
Granger, C. and Yoon, G. (2002), “Hidden cointegration”, economics working paper, University of California, San Diego, CA.
Granger, C.W., Huangb, B.-N. and Yang, C.-W. (2000), “A bivariate causality between stock prices and exchange rates: evidence from recent Asian flu”, The Quarterly Review of Economics and Finance, Vol. 40 No. 3, pp. 337-354.
Huang, W., Mollick, A.V. and Nguyen, K.H. (2016), “US stock markets and the role of real interest rates”, The Quarterly Review of Economics and Finance, Vol. 59, pp. 231-242.
Humpe, A. and Macmillan, P. (2009), “Can macroeconomic variables explain long-term stock market movements? A comparison of the US and Japan”, Applied Financial Economics, Vol. 19 No. 2, pp. 111-119.
Ibrahim, M. (2003), “Macroeconomic forces and capital market integration: a VAR analysis for Malaysia”, Journal of the Asia Pacific Economy, Vol. 8 No. 1, pp. 19-40.
Ibrahim, M.H. and Aziz, H. (2003), “Macroeconomic variables and the Malaysian equity market: a view through rolling subsamples”, Journal of Economic Studies, Vol. 30 No. 1, pp. 6-27.
Johansen, S. (1991), “Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models”, Econometrica: Journal of the Econometric Society, Vol. 59 No. 6, pp. 1551-1580.
Koutmos, G. (1998), “Asymmetries in the conditional mean and the conditional variance: evidence from nine stock markets”, Journal of Economics and Business, Vol. 50 No. 3, pp. 277-290.
Koutmos, G. (1999), “Asymmetric price and volatility adjustments in emerging Asian stock markets”, Journal of Business Finance & Accounting, Vol. 26 Nos 1/2, pp. 83-101.
Kuosmanen, P., Nabulsi, N. and Vataja, J. (2015), “Financial variables and economic activity in the Nordic countries”, International Review of Economics & Finance, Vol. 37, pp. 368-379.
Lahiani, A., Hammoudeh, S. and Gupta, R. (2016), “Linkages between financial sector CDS spreads and macroeconomic influence in a nonlinear setting”, International Review of Economics & Finance, Vol. 43, pp. 443-456.
Marshall, D.A. (1992), “Inflation and asset returns in a monetary economy”, The Journal of Finance, Vol. 47 No. 4, pp. 1315-1342.
Maysami, R.C. and Koh, T.S. (2000), “A vector error correction model of the Singapore stock market”, International Review of Economics & Finance, Vol. 9 No. 1, pp. 79-96.
Melvin, M. and Taylor, M.P. (2009), “The crisis in the foreign exchange market”, Journal of International Money and Finance, Vol. 28 No. 8, pp. 1317-1330.
Miller, K.D., Jeffrey, F.J. and Mandelker, G. (1976), “The ‘Fisher effect’ for risky assets: an empirical investigation”, The Journal of Finance, Vol. 31 No. 2, pp. 447-458.
Mukherjee, T.K. and Naka, A. (1995), “Dynamic relations between macroeconomic variables and the Japanese stock market: an application of a vector error correction model”, Journal of Financial Research, Vol. 18 No. 2, pp. 223-237.
Nasseh, A. and Strauss, J. (2000), “Stock prices and domestic and international macroeconomic activity: a cointegration approach”, The Quarterly Review of Economics and Finance, Vol. 40 No. 2, pp. 229-245.
Nieh, C.-C. and Lee, C.-F. (2002), “Dynamic relationship between stock prices and exchange rates for G-7 countries”, The Quarterly Review of Economics and Finance, Vol. 41 No. 4, pp. 477-490.
Nusair, S.A. (2016), “The effects of oil price shocks on the economies of the Gulf Co-Operation Council countries: nonlinear analysis”, Energy Policy, Vol. 91, pp. 256-267.
Peiró, A. (2016), “Stock prices and macroeconomic factors: some European evidence”, International Review of Economics & Finance, Vol. 41, pp. 287-294.
Pesaran, M.H., Shin, Y. and Smith, R.J. (2001), “Bounds testing approaches to the analysis of level relationships”, Journal of Applied Econometrics, Vol. 16 No. 3, pp. 289-326.
Rahman, M.L. and Uddin, J. (2009), “Dynamic relationship between stock prices and exchange rates: evidence from three South Asian countries”, International Business Research, Vol. 2 No. 2, pp. 167-174.
Rapach, D.E. (2002), “The long-run relationship between inflation and real stock prices”, Journal of Macroeconomics, Vol. 24 No. 3, pp. 331-351.
Romilly, P., Song, H. and Liu, X. (2001), “Car ownership and use in Britain: a comparison of the empirical results of alternative cointegration estimation methods and forecasts”, Applied Economics, Vol. 33 No. 14, pp. 1803-1818.
Ross, S.A. (1976), “The arbitrage theory of capital asset pricing”, Journal of Economic Theory, Vol. 13 No. 3, pp. 341-360.
Roubaud, D. and Arouri, M. (2018), “Oil prices, exchange rates and stock markets under uncertainty and regime-switching”, Finance Research Letters.
Shahzad, S.J.H., Nor, S.M., Ferrer, R. and Hammoudeh, S. (2017), “Asymmetric determinants of CDS spreads: US industry-level evidence through the NARDL approach”, Economic Modelling, Vol. 60, pp. 211-230.
Shin, Y., Yu, B. and Greenwood-Nimmo, M. (2014), “Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework”, Festschrift in Honor of Peter Schmidt, Springer, New York, NY, pp. 281-314.
Shiskin, J. and Moore, G.H. (1968), “Composite indexes of leading, coinciding, and lagging indicators”, National Bureau of Economic Research, Inc., New York, NY, pp. 1-8.
Singh, D. (2010), “Causal relationship between macro-economic variables and stock market: a case study for India”, Pakistan Journal of Social Sciences, Vol. 30 No. 2, pp. 263-274.
Singh, P. (2014), “An empirical relationship between selected Indian stock market indices and macroeconomic indicators”, International Journal of Research in Business Management, Vol. 2 No. 9, pp. 81-92.
Sohail, N. and Hussain, Z. (2009), “Long-run and short-run relationship between macroeconomic variables and stock prices in Pakistan: the case of Lahore Stock Exchange”, Pakistan Economic and Social Review, Vol. 47 No. 2, pp. 183-198.
Valcarcel, V.J. (2012), “The dynamic adjustments of stock prices to inflation disturbances”, Journal of Economics and Business, Vol. 64 No. 2, pp. 117-144.
Yang, Z., Tu, A.H. and Zeng, Y. (2014), “Dynamic linkages between Asian stock prices and exchange rates: new evidence from causality in quantiles”, Applied Economics, Vol. 46 No. 11, pp. 1184-1201.
Yusof, R.M. and Majid, M.S.A. (2007), “Macroeconomic variables and stock returns in Malaysia: an application of the ARDL bound testing approach”, Savings and Development, Vol. 31 No. 4, pp. 449-469.