Document Type : Original Article
Authors
^{1} PhD Candidate in Islamic Economy, Department of Economics, Firoozkooh Branch, Islamic Azad University, Firoozkooh, Iran.
^{2} Department of Economics, Firoozkooh Branch, Islamic Azad University, Firoozkooh, Iran.
^{3} Assistant Prof., Department of Economics, Central Tehran Branch, Islamic Azad University, Tehran, Iran.
Abstract
Keywords
Introduction
The complexity of financial markets and their interconnections make a loss in a sector or a country spread rapidly to other sectors of the economy of other countries. The empirical evidence has shown that markets are not separated from each other and their movements will not take place in an isolated space. Then, fluctuations in different asset markets are strongly related to each other and understanding the relationship between financial assets is necessary for investors in order to make better decisions.
Securities and exchange markets have ever been the sensitive parts of financial markets. These two markets rapidly absorb fluctuations in commercial cycles of the economy and reflect the economic changes. At the same time, distress in one or two markets creates concerns among policymakers and investors. The dynamic interaction between these two markets has encouraged researchers, policymakers and analysts to conduct accurate detailed analyses. Despite the lack of consensus in theory about the interaction between stock prices and exchange rate, it should be noted that the mediating variables such as wealth, demand for money and interest rates play significant roles in the establishment of such linkages (Saleh, 2008).
Financial markets suddenly change their behavior and the resulting behavior for most of the asset prices remains the same for a long period of time. During certain periods like financial crisis, the average, instability and correlation patterns in asset returns reveal dramatic changes. Recessions and booms in financial markets will lead to an increased likelihood of abnormal movements in the market which couldn't be ignored by labeling them as outliers. Since the summer of 2011 and due to imposed economic sanctions, unilaterally withdrawal of the United States from the joint comprehensive plan of action, liquidity growth during many years and other factors, the Iranian economy has witnessed an unprecedented situation. After a long period of controlling the exchange rate, the market experienced instability, leading to considerable fluctuations in other financial markets. The Recordbreaking Tehran Stock Index began in 2012 and reached its peak in July 2020. As a result of the recessions in economic activities and high inflation, stock market or foreign exchange market as alternatives for investments were ahead of investors who had hot money[1] in their hands.
In such turmoil in the mentioned asset markets, the questions are: had the exchange rate fluctuations strong asymmetric correlations with the stock price of export and importoriented industries[2] over the time? Which of the listed industries will have the highest weight in the optimal portfolio?
Diversification, hedging and risk management are not possible without a good deal of knowledge on behavior and relationship between targeted assets, and finding answers for aforementioned questions can provide a clear insight about analysis and prediction of return assets developments and using them to make appropriate investment policies.
To achieve our goals, this paper is organized into five sections: introduction, literature review, methodology, findings, and finally, conclusion and remarks.
Literature Review
So far, no consensus is reached on the dynamic relationship between exchange rate and stock price, so that two attitudes could be distinguished in this regard: Dornbusch and Fisher (1980) by developing floworiented models assume that a country's current account and current balance are two important determinants of the exchange rate. Accordingly, changes in the exchange rate affect international competitiveness and balance of trade and then it may affect real variables of the economy such as real production and income, the firms' future and current liquidity flow and thus the stock price. This model suggests that depreciation of local currency (increased exchange rate) makes the local firms more competitive and their exporting goods become cheaper at the international level. The increased advantage of domestic products and consequently, higher exports lead to higher income which will increase the firms' stock price on its own. Then in these models, the exchange rate affects stock price in a positive direction.
Another attitude is stockoriented models. These models assume that capital account mainly determines the exchange rate. They are composed of a portfolio balanced model and a monetary model. In the portfolio balanced model, Branson (1983) posits that there is a negative relationship between exchange rate and stock price. Accordingly, decreases in a stock price reduce the wealth of local investors, leading to lower demand for money with ensuing lower interest rates. Lowering the interest rate encourages the capital outflow, ceteris paribus, and causes depreciation of local currency and appreciation of foreign exchange rate. In the monetary model of Gavin (1989), as opposed to the aforementioned models, there is no relationship between exchange rate and stock price. Therefore, with regard to these three models, it could be concluded that theoretical studies have not come to a clear outcome on the relationship between the exchange market and stock price. On the other hand, Bhattacharya and Mukherjee (2002) theory suggests that until the 1950s risk was considered a qualitative factor and then it was Markowitz efforts that made the risk quantifiable and standard deviation of cash flows in different conditions was introduced as risk measurement quantity. He postulated that investors do not necessarily look for maximization of expected return, because if they do, they will only choose the asset with the maximum expected return; but in practice, investors own a portfolio containing different securities and thus, they compare different investments with respect to expected return (average return) and standard deviation (squared variance), picking out the best investment. As the exchange rate is an integral part of a portfolio, Markowitz theory implies that the rate of this asset may affect its demand and subsequently, causes changing the stock price. For example, in case of depreciation of exchange rate, the product costs will be reduced and as result, the profit margin will change in a positive direction (Heidari&Bashiri, 2012). Therefore, investors get to change their attitude on risk management. Many hedging strategies are emerging in an attempt to respond to this question that how other markets may be used for purpose of hedging. Given that the ways of finding optimal portfolio weights have become a critical problem for investors, making a portfolio to hedge with other assets like foreign exchange, gold, oil and so on has found its way as a risk management strategy. In addition, volatility spillover between two markets is essential to build hedge ratios and optimal portfolios (Chang et al, 2011). According to previous studies, it is important to estimate optimal hedge ratio and portfolio weights, such that investors have more information while adopting hedge strategy (Huisman et al, 2009; Yao & Wu, 2012).
Researchers have studied the relationship between exchange rate and the stock market, using numerous methods, to find an appropriate approach for the implementation of hedging strategies in these markets. For example, Yousaf et al (2021) examine the hedging role of gold against Asian stock markets during the Covid19 outbreak. Their results suggest that during the Covid19 outbreak, gold is an appropriate asset to hedge stock prices in most Asian stock markets[3]. The optimal weights of goldstock portfolios are lower during the Covid19 subperiods than the preCovid19 period. As result, investors should increase their investment in gold during the Covid19 subperiod. Hedging effectiveness for most Asian stock markets is higher during the Covid19 subperiod.
Akhtaruzzaman et al (2021) investigate financial contagion through financial and nonfinancial firms between China and G7 countries during the Covid19 pandemic. Their results suggest that financial and nonfinancial firms have experienced a significant increase in the conditional relationship between their stock returns. Moreover, the magnitude of increase in correlations is significantly higher for financial firms during the Covid19 outbreak, highlighting the importance of their role in financial contagion. They show that optimal weights ratios increase considerably in most of the cases, implying higher costs of hedging during the Covid19 outbreak.
Arash Habibi & Chin Lee (2019) investigate asymmetric effects of exchange rate on stock prices in G7 countries (Canada, France, Germany, Italy, Japan, United Kingdom and the United States) during 1997206; and for this purpose, they use NARDL model. They show that in all G7 countries, change of exchange rate has an asymmetric shortterm effect on stock prices; while in the longterm, this result is not true, except for Germany.
Simona MoagarPoladian et al (2019) examine the linkage between exchange rates and stock markets in four Central and Eastern European (CEE) countries using the DCCMIDAS model. The estimates suggest significant differences between correlation patterns in four sample countries. Moreover, they find a relatively low degree of convergence among exchange rates markets in CEE.
Sosa et al (2018) analyze the dynamic links between the stock market and exchange rate in MILA countries (Colombia, Chile, Mexico, and Peru) using Markov SwitchingAutoregressive and Markov SwitchingVector Autoregressive models. The results reveal that stock returns in these countries evolve according to low volatility and high volatility regimes. Moreover, the effect of stock markets on exchange rates is more than the effect of the exchange rate on stock markets. These results confirm that contagion exists between stock markets and exchange rates in Peru and Chile.
Wang et al (2016) examined whether gold was used to hedge against the global exchange rate. Their results revealed that exchange rate fluctuations result in threshold effects and influence the gold hedging effectiveness.
Arouri et al (2015) studied the relationship between world gold prices and stock return in China to develop hedging and diversification strategies using multivariate GARCH. They analyzed optimal weights and hedge ratios for stock and gold and indicated that how empirical results could be used to adopt a hedging strategy.
Basher et al (2015) investigate the conditional correlation between emerging markets of stock, oil price, VIX, gold price and bonds using DCC, ADCC and GOGARCH. The results suggest that in most situations, oil is the best hedge for stock prices in emerging markets. The hedge ratios resulted from ADCC is preferred over other methods to hedge emerging markets stock prices with oil, VIX or bonds. In some cases, hedge ratios resulted from GOGARCH have the highest hedging effectiveness to hedge emerging markets stock prices with gold.
YenHesienLee et al (2014) examine portfolio construction between the stock price of G7 countries and West Texas Intermediate crude oil. In addition, to study volatility spillovers between stock price and oil price with DCC, CCC and BEKK models, they obtain optimal portfolio weights. The results suggest that the hedge effectiveness of the DCC model is better than the CCC model and BEKK model, and the hedging effect in Canada is the highest and it is lowest in Japan. Moreover, the results show that Japan has the biggest optimal portfolio weight and the lowest hedge ratio.
Elmi et al (2014) investigate the effect of structural breaks in volatilities on momentum transfer and volatility spillover between gold and stock markets in Iran over a period of 20072013, using the conventional iterated cumulative sum of squares (ICSS) algorithm, as well as modified iterated cumulative sum of squares algorithm. The results suggest that momentum and volatility spillovers between studied markets are bidirectional and misspecification of structural breaks made the evaluation of momentum and spillover misleading.
Jahangiri& Hekmati (2015) studied the relationships between Tehran Stock Exchange, foreign exchange and gold coin markets (as domestic markets) on the one hand, and world oil and gold markets, US stock market and European Stock Market Index (as international markets) on the other hand, and calculated volatility spillover effect during 20012013, using Markov Switching Regime. The results reveal that in a low return regime, there are no significant spillover effects between markets. Moreover, when stock markets are in regime zero (low return regime), the gold market acts as an intermediary for the transmission of shocks from global markets to asset markets in Iran. In contrast, when stock markets are in regime one (high return regime), the oil market acts as a transmission channel of shocks to domestic asset markets.
Botshekan et al (2017) address spillovers effects of foreign exchange (dollar and euro) on banks and financial institutes' indexes to explain systemic elements of volatility spillover from exchange market to monetary and financial sectors. To do so, they considered log of return, dynamic conditional correlation and volatility spillover by using four commonly used multivariate GARCH models during 12 years (from the beginning of 2005 until the end of 2016). The results show positive conditional correlations of the dollar in the shortterm, volatilities of the euro in the longterm, the existence of spillover effects of exchange rate on banks and financial institutes' index.
Sarrafi et al (2018) examine the symmetric or asymmetric effect of positive and negative shocks of the dollar in the freemarket on chemicals and basic metals indices, as two listed industries in the stock exchange with the highest share of nonoil exports in the Iranian economy. They apply weekly data from 2007 to 2016. The results obtained from the primary model, i.e. The NonLinear Autoregressive Distributive Lag (NARDL) indicate that the effect of increases in dollar rate on both indices in the short and long term is positive and significant, and the effect of its decreases is nonsignificant. Moreover, the derived coefficients imply a higher effect of positive dollar shock in the free market on the chemicals index over the basic metals index. The control variable, i.e. OPEC crude oil price has a direct and significant effect on both indices in the short and long term.
Hatami et al (2018) investigate dynamics of optimal hedge ratio in stock and gold markets. Their results suggest that dynamics of optimal hedge rate have increased during the 20092013 period and regimeswitching has occurred from 2013 to 2016. Moreover, investors should use gold to hedge their investment in the stock market and consider gold as an item with the stock asset in their portfolio.
Elham Farzanegan (2018) estimates and addresses the performance of hedging performance of cash and future markets of gold coins from 20101027 to 2016721, using the new generation of multivariate volatility models. The results reveal the superiority of hedging ratios derived from the GOGARCH model over competitive models, to hedge against volatilities of cash prices with the future price of gold coin. Moreover, cash and future prices tend to comovement during tensions in the gold coin market.
Nikoumaram et al (2014) investigate exchange and gold markets volatility spillovers on capital market in both EOIs and IOIs during 2007 2014, using multivariate GARCH model. The results confirm the spillover effect of EOIs from the foreign exchange parallel market; however, the results do not show spillovers from another parallel market, i.e. gold. In this regard, the spillover effect of IOIs from foreign exchange and gold has not been established.
Firouz Fallahi et al (2013) analyze the correlation between volatilities of stock, exchange and gold coin markets in Iran during 20112013, using dynamic conditional correlation (DCCGARCH). The results imply a high conditional correlation between exchange rate return and gold coin, as well as a low conditional correlation between stock index return with exchange and gold coin. Moreover, the equity market is the best place for investment compared to other asset markets.
And finally, Hassan Heidari et al (2012) investigate the relationship between volatilities of real exchange rate and stock price index in Tehran Stock Exchange over the period of 1999 to 2011, using Bivariate Generalized Autoregressive Conditional Heteroskedasticity model (VARGARCH). The results suggest that there is a negatively significant relationship between the uncertainty of real exchange rate and stock price index, and no significant relationship is seen between the uncertainty of stock price and exchange rate.
Research Methodology
Symmetric dynamic conditional correlation (DCC) developed by Engle (2002) and asymmetric dynamic conditional correlation (ADCC) developed by Cappiello et al (2006) to investigate asymmetry of dynamic correlations between financial markets. DCC models involving constant conditional correlation (CCC) developed by Bollerslev (1990) assume that conditional correlations are timedependent. This model could be estimated even for multidimensional datasets using a twostep procedure. At the first step, conditional variances are derived by estimating a series of univariate GARCH models. In the second step, intercepts of conditional correlations are estimated.
Let be a n*1 vector of a multivariate stochastic process and be the log of stock return and log of exchange rate return.
The innovative process of conditional correlation , has n*n conditional covariance matrix, H_{t}.
Where,
α is the vector of constants. 𝜑 and 𝜈 refer to an n×1 vector of coefficient for AR (1) and MA (1), respectively. is the vector of residuals. is a process with n*1 vector so that gives multivariate studentdensity function:
Where,
is the gamma function and v is the degree of freedom for v>2.
The reason for using tdistribution is that it makes modeling with a fat tail possible. Proposed DDCGARCH by Engle (2002) may successfully be estimated for timedependent covariance matrices. This covariance matrix of DCCGARCH could be decomposed as:
Where,
is a diagonal matrix with conditional standard deviations along the diagonals. It means:
And is conditional correlations matrix. Estimation method is composed of two steps. First, conditional variances, , for i=1,…,n assets are estimated using univariate GARCH(1,1) proposed by Bollerslev (1986):
In which, parameters should be estimated.
Second, conditional correlations are estimated using standardized residuals derived in the first step. Specifically, timevarying correlation matrix which is:
And correlation matrix, , is calculated as:
In above equation, is standardized residuals by conditional standard deviation, i.e. ; is standardized residuals of unconditional covariance, and is diagonal matrix with reverse toots of arrays in diagonal ; i.e. . Then, correlation coefficients, are given by:
As asymmetry is not included in Equations 6 and 7, Cappiello et al (2006) extended DCC to build leverage effect on conditional correlation of asset returns and news impact curve. Asymmetric generalized DCC (AGDCC) is expressed as:
Where,
The indicator function n*1, is equal to 1 when this argument is valid, and 0 otherwise. ‘°' indicates Hadamard Product and and are unconditional matrices of and . For , when becomes positive definite with probability 1 that is positive definite. If matrices A, B, G are replaced with scalarsθ_{1},θ_{2} and θ_{3}, ADCC (1,1) are distinguished from AGDCC (1,1). This study focuses only on the asymmetric effect, ignoring the effect of assetspecific news. Therefore, the developed models are used to answer this question that whether conditional correlations between exchange rate return and stock return of EOIs and IOIs tend to change over time, and if is it possible these markets rise during the periods with high volatility due to exchange shocks.
Structural breaks models have been well documented in macroeconomy and finance. Due to the importance of detecting accurate number and time of structural changes in the variance of financial timeseries, numerous methods have been proposed for this purpose. The commonly used method to detect indigenously breakpoints in the variance is Iterated Cumulative Sum of Square (ICSS) algorithm, designed by Inclan and Tiao (1994). ICSS searches for significant changes in variance, created by a structural break in process of generating volatility in time series. This algorithm assumes that the studied time series contain T observation with normal, uniform and independent distribution (Arago, V, & Fernandez, M.A, 2007).
It is assumed that the studied time series has constant unconditional variance in the initial period, and suddenly due to an unexpected financial, economic or political event a shock is introduced to the system so that a structural change would occur in the variance of time series. In other words, the occurrence of this shock increases the deviance of current variance from the previous one to the extent that it implicates structural change of market volatility. Then, unconditional variance again turns back to the constant state in a new surface, until in the next shock it would experience another structural change. This process is repeated over time and a time series with breakpoints in unconditional variance is generated (Kang, S.H & Cheong C. & Yoon, S.M, 2011).
ICSS takes the variance between each pair of breakpoints as fixed and the kurtosis as normal. First, this algorithm is defined for the state that conditional heteroskedasticity exist, while, the empirical evidence shows that most financial and economic time series have often variable variance (or heteroskedasticity). Therefore, conventional ICSS is not well enough in case that a dependent process like GARCH exists (Malik, F. & Ewing, B.T & Payne, J.E, 2005; Ewing et al, 2013).
Second, ICSS assumes that time series has a normal distribution. However, financial time series often have fat tail distribution with additional kurtosis (>3) (Arago& Fernandez, 2007). Sanso et al (2003) applied some additional assumptions on and indicate that for financial data which are not normally distributed and have conditional heteroskedasticity, the validity of IT test results would come into question and if critical values had not been adjusted correctly, erroneous rejection of null hypothesis should be expected. As such, some identified structural breaks in the ICSS algorithm may likely be spurious and the number of real structural breaks invariance would be less than what was detected in conventional ICSS. Thus, they modified Inclan and Tiao (IT) test in order to be applicable when error terms are followed by a dependent process like GARCH (Malik et al, 2005). The number of structural changes resulted from the implementation of Sanso and colleagues' method is remarkably lower than Inclan and Tiao approach (Arago& Fernandez, 2007).
In this section, we examine how weights of the optimal portfolio (WSEs) are extracted between the return of exchange rate and index of EOIs and IOIs during the currency crisis period. The optimal weights of exchange – stock in a oneRial portfolio for export and importoriented industries are given as follows:
Where,
is the weight of optimal holding of currencies in a 1Rls portfolio in time t. is conditional covariance between the stock price index and exchange price, is the conditional variance of stock price index, and is the conditional variance of exchange price in time t.
Research Findings
In this study, the weekly data of the unofficial exchange rate and the daily price index of EOIs and IOIs from December 2008 to September 2020 were used. The logarithmic difference of consecutive prices of stock index and the exchange rate is used to calculate price return. The natural log of net price return is defined as follow:
Tehran Stock Exchange data were extracted from tse.ir. Daily dollar price in the free market is not announced by any official reference and the author gathered these data by using the published data on the website of reliable Tehran currency exchanges. In the process of research, Augmented DickeyFuller (ADF) and Breakpoint ADF tests were used to evaluate the stationarity of variables. The results of these tests show that all variables are stationary at the level of 99%.
Table 1. Results of Unit Root test
ADF 
Breakpoint ADF 

Exchange rate 
52.80^{***} 
54.37^{***} 
Car 
23.61^{***} 
38.38^{***} 
Computer 
37.31^{***} 
38.72^{***} 
Cement 
14.93^{***} 
32.15^{***} 
Chemicals 
23.56^{***} 
36.41^{***} 
Petroleum 
24.40^{***} 
40.51^{***} 
Basic metals 
23.28^{***} 
38.07^{***} 
Tile and ceramics 
22.94^{***} 
36.32^{***} 
Metal ores 
24.22^{***} 
38.75^{***} 
Pharmaceutical 
21.76^{***} 
30.12^{***} 
Source: research findings
As two currency crises occurred in Iran during the studied period, then in each crisis, hedging models were estimated independently. Several econometrics methods can be used to detect structural breakpoints, and subsequently the critical points. In this study, as the exchange time series has a fat tail and conditional heteroskedasticity, the modified iterated cumulative sum of squares algorithm, presented in Section 3.2, has been used and the results are reported in Table 2.
Table 2. Results of structural break test for the currency
ICSS (IT) 
ICSS(κ_{1}) 
ICSS(κ_{2})[4] 

1388/04/03 
)130( 
1390/09/05 
)720( 
1390/09/05 
)720( 
1389/07/17 
)444( 
1392/04/04 
)1096 
1397/01/15 
)2252( 
1392/07/08 
)1162( 
1397/10/09 
)2432( 

1392/12/07 
)1263( 



1393/02/28 
)1314( 

1393/09/04 
)1442( 

1393/09/25 
)1456( 

1396/10/06 
)2190( 

1397/01/18 
)2253( 

1397/10/09 
)2432( 
Source: research findings
Structural break tests estimate the peak of crisis; in fact, they detect the point in which the highest changes occurred in the exchange rate. Results of the ICSS test detected two breakpoints in the exchange rate, contemporaneous to two foreign exchange shocks that occurred during the 2010s in Iran. According to the results of the modified ICSS, the starting point of the first and second exchange crises was assumed November 26, 2011, and April 4, 2017, respectively.
Table 3. Summary of descriptive statistics
Currency crisis 
Exchange rate 
Cement 
Chemicals 
Petroleum 
Basic metals 
Tile and ceramics 
Metal ores 
Car 
Computer 
Pharmaceuticals 

Mean 
1 
0.09 
0.12 
0.22 
0.24 
0.19 
0.17 
0.19 
0.11 
0.22 
0.16 
2 
0.12 
0.17 
0.18 
0.19 
0.20 
0.19 
0.18 
0.22 
0.19 
0.22 

Maximum 
1 
22.88 
10.79 
8.98 
25.66 
8.54 
9.75 
9.69 
8.02 
18.98 
6.27 
2 
18.38 
5.35 
4.94 
6.59 
6.48 
13.31 
9.13 
13.51 
9.78 
5.76 

Minimum 
1 
11.52 
2.99 
6.76 
10.19 
4.76 
15.50 
17.01 
7.55 
3.98 
1.87 
2 
20.98 
4.00 
5.84 
8.56 
8.14 
6.64 
4.90 
6.13 
4.93 
4.35 

S. E 
1 
1.82 
0.81 
1.07 
1.77 
1.22 
1.19 
1.54 
1.45 
1.52 
0.59 
2 
1.81 
1.29 
1.30 
1.81 
1.49 
1.52 
1.65 
2.17 
1.29 
1.22 

Skewness 
1 
2.11 
3.35 
0.64 
4.67 
0.56 
0.70 
0.39 
0.53 
3.41 
3.25 
2 
0.34 
0.61 
0.40 
0.10 
0.43 
1.06 
0.80 
0.38 
0.60 
0.65 

Kurtosis 
1 
37.58 
34.35 
12.12 
65.19 
6.45 
35.15 
19.10 
5.25 
38.81 
23.59 
2 
34.74 
4.17 
6.29 
4.18 
5.55 
9.52 
5.57 
4.29 
8.16 
6.02 

JarqueBera test statistic 
1 
58640 (0.00) 
49667 (0.00) 
4101 (0.00) 
191170 (0.00) 
636 (0.00) 
50069 (0.00) 
12559 (0.00) 
298 (0.00) 
64240 (0.00) 
22542 (0.00) 
2 
70091 (0.00) 
198 (0.00) 
796 (0.00) 
100 (0.00) 
506 (0.00) 
3272 (0.00) 
640 (0.00) 
156 (0.00) 
1947 (0.00) 
755 (0.00) 

LMARCH test statistics 
1 
14.8 (0.00) 
11.8 (0.00) 
8.3 (0.00) 
0.02 (0.98) 
5.2 (0.00) 
0.001 (0.99) 
4.6 (0.00) 
34.0 (0.00) 
11.9 (0.00) 
84.7 (0.00) 
2 
50.3 (0.00) 
289.6 (0.00) 
304.0 (0.00) 
161.8 (0.00) 
132.3 (0.00) 
38.6 (0.00) 
144.3 (0.00) 
17.3 (0.00) 
111.5 (0.00) 
387.5 (0.00) 

Spearman Correlation 
1 
100% 
0/012 
0/059 
0/025 
0/04 
0/012 
0/035 
0/009 
0/007 
0/026 
2 
100% 
0/000 
0/005 
0/04 
0/004 
0/016 
0/015 
0/019 
0/04 
0/026 

Pearson Correlation 
1 
100% 
0.000 
0.046 
0.005 
0.031 
0.005 
0.026 
0.044 
0.032 
0.022 
2 
100% 
0.035 
0.017 
0.042 
0.033 
0.015 
0.020 
0.059 
0.047 
0.028 

Kendall's Correlation 
1 
100% 
0.008 
0.041 
0.017 
0.028 
0.009 
0.024 
0.005 
0.005 
0.017 
2 
100% 
0.003 
0.005 
0.027 
0.002 
0.013 
0.011 
0.013 
0.028 
0.017 
Source: research findings
Table 3 summarizes descriptive statistics of return series of EOIs and IOIs with exchange rates during the first and second exchange crises. For all indices, the standard deviation is bigger than the mean. The means are positive for all assets; such that in the first exchange crisis, the minimum and maximum value of mean belongs to automobile and oil products industries. Moreover, the highest standard deviation is seen in oil products and the lowest one in the pharmaceutical industry. In the second exchange crisis, the lowest mean belongs to cement and the highest value to automobile, pharmaceuticals and multidisciplinary industries. Considering the LMARCH test, in the first crisis, the existence of heteroskedasticity is not rejected for all industries except oil products and tile and ceramics. However, in the second crisis, this possibility is not rejected for all industries. As the kurtosis ratio of returns is higher than the kurtosis of the normal density function, then, the density function of these assets in both crises are fattailed with a high peak. JarquBera test statistics suggest that distribution functions of the studied returns in both crises would not be normal.
In this study, the timevarying asymmetric dynamic correlation pattern was estimated for two groups of industries, i.e. exportoriented and import oriented, and if variables didn't follow the pattern, the symmetric dynamic correlation model would be estimated. Given the nonnormality of returns, DCC and ADCC[5] models are estimated with multivariate tdistribution. In tables 4 and 5, the results of symmetric DCC and asymmetric ADCC between stock returns of EOIs and IOIs with exchange rates are reported. The reason for this categorization goes back to the fact that whether different categories respond differently to the exchange shocks.
Table 4. Results of DCC and ADCC parameters in IOIs
ADCC _ First crisis 
DCC _ First crisis 
ADCC _ Second crisis 
DCC _ Second crisis 



Coef. 
t 
prob 
Coef. 
t 
prob 
Coef. 
t 
prob 
Coef. 
t 
prob 
Exchange rate 
µ 
0.012 
1.900 
0.056 
0.011 
1.640 
0.100 
0.004 
0.556 
0.577 
0.003 
0.430 
0.660 
𝜑 
0.891 
14.230 
0.000 
0.893 
11.970 
0.000 
0.600 
10.800 
0.000 
0.602 
11.540 
0.000 

𝜈 
0.876 
13.670 
0.000 
0.877 
11.330 
0.000 
0.596 
10.600 
0.000 
0.599 
10.730 
0.000 

𝜔 
0.005 
1.750 
0.081 
0.004 
1.820 
0.069 
0.003 
1.180 
0.239 
0.003 
1.200 
0.231 

α 
0.157 
2.190 
0.029 
0.135 
2.770 
0.006 
0.103 
4.070 
0.000 
0.199 
4.290 
0.000 

β 
0.805 
21.740 
0.000 
0.809 
23.320 
0.000 
0.739 
14.950 
0.000 
0.731 
15.700 
0.000 

γ 
0.021 
0.130 
0.897 
0.068 
1.030 
0.302 

Car 
µ 
0.020 
1.590 
0.110 
0.015 
0.921 
0.357 
0.043 
0.833 
0.400 
0.024 
0.486 
0.626 
𝜑 
0.180 
2.180 
0.029 
0.180 
2.348 
0.019 
0.313 
2.220 
0.026 
0.312 
2.278 
0.023 

𝜈 
0.200 
2.680 
0.007 
0.211 
3.084 
0.002 
0.558 
4.330 
0.000 
0.556 
4.450 
0.000 

𝜔 
0.001 
0.620 
0.534 
0.001 
0.670 
0.505 
0.061 
1.180 
0.240 
0.052 
1.080 
0.281 

α 
0.108 
3.060 
0.002 
0.116 
3.090 
0.002 
0.114 
4.490 
0.000 
0.105 
4.820 
0.000 

β 
0.874 
25.120 
0.000 
0.867 
25.050 
0.000 
0.712 
14.530 
0.000 
0.816 
48.700 
0.000 

γ 
0.115 
0.970 
0.332 
0.029 
1.220 
0.221 

Computer 
µ 
0.013 
1.320 
0.185 
0.013 
1.290 
0.196 
0.036 
2.067 
0.039 
0.017 
1.055 
0.290 
𝜑 
0.623 
6.850 
0.000 
0.595 
6.430 
0.000 
0.330 
4.140 
0.000 
0.338 
4.311 
0.000 

𝜈 
0.316 
2.765 
0.005 
0.283 
2.520 
0.012 
0.037 
0.418 
0.675 
0.025 
0.309 
0.757 

𝜔 
0.001 
0.700 
0.485 
0.001 
0.930 
0.354 
0.026 
3.700 
0.000 
0.026 
3.240 
0.001 

α 
0.166 
2.100 
0.036 
0.112 
2.280 
0.023 
0.179 
5.460 
0.000 
0.284 
4.840 
0.000 

β 
0.825 
19.100 
0.000 
0.830 
19.080 
0.000 
0.705 
16.410 
0.000 
0.687 
13.580 
0.000 

γ 
0.307 
1.510 
0.130 
0.294 
4.160 
0.000 

Pharmaceuticals 
µ 
0.038 
2.190 
0.028 
0.022 
1.300 
0.194 
0.008 
0.504 
0.610 
0.001 
0.039 
0.970 
𝜑 
0.750 
14.410 
0.000 
0.760 
11.300 
0.000 
0.499 
8.280 
0.000 
0.486 
8.238 
0.000 

𝜈 
0.416 
4.964 
0.000 
0.425 
3.950 
0.000 
0.058 
0.786 
0.432 
0.046 
0.649 
0.516 

𝜔 
0.020 
1.410 
0.158 
0.027 
1.650 
0.099 
0.004 
2.220 
0.027 
0.003 
1.990 
0.047 

α 
0.196 
1.510 
0.133 
0.306 
2.040 
0.042 
0.124 
3.600 
0.000 
0.130 
3.790 
0.000 

β 
0.756 
5.110 
0.000 
0.648 
5.160 
0.000 
0.860 
25.570 
0.000 
0.861 
25.420 
0.000 

γ 
0.608 
1.840 
0.067 
0.114 
2.260 
0.024 

Correlation 
θ1 
0.006 
0.300 
0.767 
0.017 
2.120 
0.034 
0.007 
3.720 
0.000 
0.007 
3.610 
0.000 
θ2 
0.880 
3.250 
0.001 
0.890 
13.960 
0.000 
0.991 
266.500 
0.000 
0.990 
397.300 
0.000 

θ2 
0.120 
0.750 
0.456 
0.006 
1.190 
0.234 

λ 
2.542 
11.380 
0.000 
2.551 
11.660 
0.000 
4.229 
15.540 
0.000 
4.170 
15.900 
0.000 

Akaike 
10.072 
10.064 
10.0519 
10.045 

Schwarz 
10.238 
10.207 
10.1758 
10.153 

Shibata 
10.070 
10.062 
10.0508 
10.0451 

HannanQuinn 
10.135 
10.118 
10.0978 
10.0858 

LL 
4455.3 
4474.26 
8435.24 
8456.02 

LiMcLeod (5) 
84.4546 (0.289) 
79.8551 (0.420) 
39.968 (0.99) 
37.5044 (0.99) 
Source: research findings
Table 4 indicates that in both tables is insignificant. It implies a symmetric correlation over the time between exchange rate and index of industries in this category. DCC results suggest that α, β coefficients are insignificant in all variables and for both crises. As result, shortterm persistence (α) and longterm persistence (β) is confirmed. As it is seen, α is less than β for all variables. The estimated and are positive for both shocks, and significant at a 1% level. The sum of these coefficients is less than 1, indicating symmetric dynamic conditional correlations are meanreverting[6].
Table 5. Results of DCC and ADCC parameters in EOIs
ADCC _ First crisis 
ADCC _ Second crisis 

Coef. 
t 
prob 
Coef. 
t 
prob 

Exchange rate 
µ 
0.006 
0.904 
0.366 
0.010 
1.219 
0.223 
𝜑 
0.782 
1.963 
0.054 
0.559 
8.278 
0.000 

𝜈 
0.767 
0.954 
0.340 
0.549 
7.520 
0.000 

𝜔 
0.007 
1.861 
0.063 
0.002 
1.053 
0.292 

α 
0.223 
2.028 
0.043 
0.192 
3.254 
0.001 

β 
0.744 
14.680 
0.000 
0.776 
14.070 
0.000 

γ 
0.118 
0.581 
0.561 
0.112 
1.825 
0.068 

Cement 
µ 
0.025 
1.044 
0.297 
0.072 
3.369 
0.001 
𝜑 
0.624 
9.563 
0.000 
0.202 
3.068 
0.001 

𝜈 
0.199 
2.330 
0.020 
0.192 
3.012 
0.003 

𝜔 
0.041 
2.329 
0.020 
0.006 
1.462 
0.144 

α 
0.444 
1.770 
0.077 
0.136 
3.412 
0.001 

β 
0.713 
12.880 
0.000 
0.917 
27.040 
0.000 

γ 
0.769 
1.669 
0.095 
0.108 
4.198 
0.000 

Chemicals 
µ 
0.117 
3.493 
0.001 
0.027 
1.687 
0.092 
𝜑 
0.118 
1.783 
0.072 
0.249 
4.598 
0.000 

𝜈 
0.221 
1.748 
0.081 
0.089 
1.653 
0.099 

𝜔 
0.359 
0.724 
0.469 
0.013 
2.877 
0.004 

α 
0.282 
3.114 
0.001 
0.162 
5.124 
0.000 

β 
0.723 
5.637 
0.000 
0.864 
29.780 
0.000 

γ 
0.261 
1.708 
0.076 
0.068 
2.698 
0.007 

Basic metals 
µ 
0.058 
0.910 
0.363 
0.040 
1.749 
0.081 
𝜑 
0.124 
8.483 
0.000 
0.229 
2.693 
0.007 

𝜈 
0.132 
0.747 
0.455 
0.027 
0.310 
0.756 

𝜔 
0.084 
1.140 
0.255 
0.025 
2.312 
0.021 

α 
0.185 
2.335 
0.020 
0.116 
3.668 
0.000 

β 
0.894 
13.230 
0.000 
0.897 
30.710 
0.000 

γ 
0.185 
2.890 
0.003 
0.061 
2.769 
0.006 

Metal ores

µ 
0.016 
1.118 
0.264 
0.114 
5.212 
0.000 
𝜑 
0.295 
4.421 
0.000 
0.361 
5.477 
0.000 

𝜈 
0.007 
0.777 
0.438 
0.066 
0.951 
0.342 

𝜔 
0.002 
0.777 
0.438 
0.011 
1.913 
0.056 

α 
0.314 
3.973 
0.000 
0.113 
3.950 
0.000 

β 
0.682 
8.236 
0.000 
0.912 
33.510 
0.000 

γ 
0.102 
0.530 
0.597 
0.072 
3.303 
0.001 

Petroleum 
µ 
 
 
 
0.004 
0.387 
0.699 
𝜑 
 
 
 
0.054 
0.477 
0.633 

𝜈 
 
 
 
0.222 
2.004 
0.045 

𝜔 
 
 
 
0.000 
0.895 
0.371 

α 
 
 
 
0.074 
6.641 
0.000 

β 
 
 
 
0.886 
59.750 
0.000 

γ 
 
 
 
0.005 
0.232 
0.817 

Tile and ceramics 
µ 
 
 
 
0.069 
3.312 
0.001 
𝜑 
 
 
 
0.274 
2.929 
0.003 

𝜈 
 
 
 
0.042 
0.416 
0.677 

𝜔 
 
 
 
0.006 
1.709 
0.088 

α 
 
 
 
0.185 
3.975 
0.000 

β 
 
 
 
0.854 
32.930 
0.000 

γ 
 
 
 
0.085 
2.032 
0.042 


θ1 
0.037 
1.760 
0.079 
0.007 
5.891 
0.000 
θ2 
0.827 
18.750 
0.000 
0.980 
511.700 
0.000 

θ2 
0.113 
2.715 
0.007 
0.007 
2.212 
0.027 

λ 
2.622 
11.100 
0.000 
4.800 
15.960 
0.000 

Akaike 
11.094 
11.100 

Schwarz 
11.308 
11.341 

Shibata 
11.091 
11.096 

HannanQuinn 
11.175 
11.189 

LL 
6437 
14744 

LiMcLeod (5) 
181.201 (0.99) 
168.159 (0.99) 
Source: research findings
Table 5 indicates that in both tables is significant. It implies an asymmetric correlation over the time between exchange rate and index of industries in this category (i.e. EOIs). ADCC results suggest that α, β coefficients are significant in all variables and for both crises. As result, shortterm persistence (α) and longterm persistence (β) is confirmed. As it is seen, α is less than β for all variables. In the first crisis, estimated γ[7] is negative and statistically insignificant for all variables except the exchange rate, implying positive residuals often increase the variance (conditional volatility) more than the negative shocks with the same magnitude. The estimated and are positive for both shocks, and statistically significant at 1% level. The sum of these coefficients is less than 1, indicating asymmetric dynamic conditional correlations are meanreverting.
In this section, the WSEs between the return of exchange rate and index of EOIs and IOIs are calculated separately for both crisis and noncrisis periods according to Equation 17, and the results are reported in Table 8.
Table 8. Summary of statistics for stockexchange optimal weights in exportoriented industries
Cement 
Chemicals 
Petroleum 
Basic metals 
Tile and ceramics 
Metal ores 

Panel A: The first crisis 







Pre Mean Optimal Weights 
0.644 
0.749 
 
0.801 
 
0.857 

Period Mean Optimal Weights 
0.082 
0.162 
 
0.182 
 
0.251 

The difference in Optimal weights 
0.562 
0.587 
 
0.619 
 
0.606 

t–stat difference 
193.84^{***} 
145.38^{***} 
 
156.9^{***} 
 
138.76^{***} 

Panel B: The second crisis 







Pre Mean Optimal Weights 
0.648 
0.599 
0.786 
0.725 
0.694 
0.710 

Period Mean Optimal Weights 
0.364 
0.345 
0.452 
0.393 
0.455 
0.446 

The difference in Optimal weights 
0.284 
0.254 
0.334 
0.333 
0.239 
0.264 

t–stat difference 
20.8^{***} 
19.94^{***} 
25.89^{***} 
25.97^{***} 
16.69^{***} 
19.33^{***} 


*, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively.
Source: research findings
The results show that in both currency crises, the optimal weights of different industries during the noncrisis period is higher than 50%, while in the crisis period, they are less than 50%. Then, investors in the noncrisis period should invest more than half of a onerial portfolio to the dollar exchange rate. Moreover, in the crisis period, they should invest less than half of a onerial portfolio to a dollar rate, reducing the risk without any reduction in the return of the portfolio.
The optimal weights vary over time and experience a prominent change during currency crisis periods. For example, the highest weighted mean of the optimal portfolio during the first noncrisis period, i.e. 85%, belongs to metallic minerals. Thus, for a 1000 rials portfolio, 150 rials should be invested in a stock of metallic minerals, and 850 rials in the exchange market. In the first currency crisis period, the weighted mean of the optimal portfolio in this industry was reduced to 25%. In the second crisis, the highest weighted mean of the optimal portfolio is 78% and belongs to the oil production industry. And thus, for a 1000 rials portfolio, an amount of 320 rials should be invested in stock of oil products, and 780 rials in the exchange market. In the second crisis period, the weighted mean of the optimal portfolio in this industry was reduced to 45%.
Table 9. Summary of statistics for stockexchange optimal weights in importoriented industries
Car 
Computer 
Pharmaceutical 

Panel A: The first crisis 



PreMean Optimal Weights 
0.727 
0.870 
0.347 
Period Mean Optimal Weights 
0.241 
0.230 
0.255 
The difference in Optimal weights 
0.486 
0.641 
0.092 
t–stat difference 
44.32^{***} 
179.87^{***} 
8.23^{***} 
Panel B: The first crisis 



PreMean Optimal Weights 
0.890 
0.633 
0.357 
Period Mean Optimal Weights 
0.494 
0.350 
0.493 
The difference in Optimal weights 
0.396 
0.284 
0.136 
t–stat difference 
36.42^{***} 
18.96^{***} 
9.13^{***} 
Source: research findings
The results indicate that in both currency crises, the optimal weights of different industries except pharmaceuticals is higher than 50% during the noncrisis period, while they are less than 50% during the crisis period. Then, investors in the noncrisis period should invest more than half of a onerial portfolio to the dollar exchange rate. Moreover, during the crisis period, they should invest less than half of a onerial portfolio to dollar exchange to reduce the risk of the portfolio, without any reduction in its return.
The optimal weights of these industries also vary over time and experienced a prominent change during the currency crises periods. For example, the highest weighted mean of the optimal portfolio during the first noncrisis period, i.e. 87%, belongs to the computer industry. Then, for a 1000 rials portfolio, 150 rials should be invested in stock of computer, and 870 rials in the exchange market. In the first currency crisis period, the weighted mean of the optimal portfolio in this industry was reduced to 23%. In the second crisis, the highest weighted mean of the optimal portfolio is 89% and belongs to the automobile industry. And thus, for the 1000 rials portfolio, the amount of 110 rials should be invested in the stock of automobiles and 890 rials in the exchange market. In the second crisis period, the weighted mean of the optimal portfolio in this industry was reduced to 49%. The notable in this industry is the optimal weight of pharmaceuticals. As such, during the second crisis period, the weighted mean of the optimal portfolio increased from 35% in the noncrisis period to 49% in the crisis period. Since this industry is perceived as an import oriented industry, the increase in exchange rate leads to an increase in its prime cost and consequently, affects the profitmaking in this group
Conclusion
The exchange rate can affect the capital market through different channels. As the scope of this study covers the correlational structure of exchange rate and the different listed industries to provide a stockexchange portfolio with regard to the currency crisis, it is critical to pay attention to the simultaneous fluctuations between both markets. The listed industries react differently to the currency crises and it depends on their activities. In this way, exportoriented and importoriented industries are affected directly by the currency crisis.
The results of this study confirm the existence of a symmetric correlation between importoriented industries and exchange rates in both currency crises. As such, among these industries, pharmaceutical has the lowest value of weighted mean in the optimal portfolio during a currency crisis. Moreover, the results suggest an asymmetric correlation of exportoriented industries with exchange rates during both currency crises. As such, of these industries, the highest change of weight in the optimal portfolio is related to basic metals, implying that this industry is in good condition to reduce risk during a currency crisis.
Although there are differences in dynamics and type of the used GARCH model, the results of this study show inconsistency with the results of Nikoomaram et al (2014) who showed a lack of correlation between importoriented industries and exchange rate; however, our results are consistent with the results of Basher et al (2015). Regarding the decrease in optimal portfolio weight during the crisis, the results of this study fall in line with the results obtained from Yousaf et al (2021), Yen Hsin Lee et al (2014) and Hatami et al (2016).
Our results will be applicable in three major fields: first, as stock of different industries is listed in the capital market, identifying the correlational structure and extracting optimal weights in the listed export and importoriented and other industries is vital for stakeholders, including governance, regulatory agencies, managers and shareholders, and it can play an influential role in strategic decision making.
Second, a deep understanding of the correlation between different industries and exchange markets during currency crisis periods reveals information about aspects of fundamental financial analysis, focusing on exchange rate shocks, the missing loop of financial analyses. Finally, identification of correlation in a currency crisis is important for firms in managing stock prices, because having information on how to exchange crisis is transmitted, can be useful in choosing strategy, diversification and scenario analysis to reduce investment risk in different projects or to get results of financial performance and exchange rate in the form of portfolio adjustments.
In case of the currency crisis, it is suggested that investors invest in the stock of basic metals, because this industry is a pioneer in attracting currency crisis and increasing stock value of the industry through future cash flow and replacement value, and reduce the stock of pharmaceuticals and computers in their portfolio, due to attracting negative effects of the exchange market.
Regarding the applicability of the present study and its wide research scope, it is recommended to the future researchers:
As the source of the studied currency crises in this paper was the exchange market, the contagion of the Covid19 crisis to the different listed industries is suggested for future studies.
[1] Hot money refers to the flow of funds from one market to another in order to earn high shortterm profits.
[2] Exportoriented industries included in this study are basic metals, metallic minerals, chemicals, oil products and cement. Daily numerical values of their indices were used as data in the analysis. As these industries export a portion of their products and receive the common currency of importing countries (dollar or euro), they are significantly affected by exchange rate fluctuations. Importoriented industries included in this study are pharmaceuticals, computers and car industries. Daily numerical values of their indices were used as data in the analysis. As these industries import their raw materials or premanufactured parts and products during the production process, they need outlay of foreign exchange and thus are significantly affected by fluctuations of foreign exchange. Among these industries, pharmaceuticals have the highest volume of raw materials import.
[3] China, Indonesia, Singapore, Vietnam, Pakistan and Thailand stock markets
[4]The reader is referred to Sanso et al (2003) to get more information.
[5] .To estimate these models, the G@GARCH Package of OxMetric8.02 software is used.
[6] is satisfied for each GARCH model
[7].If γ is significant and negative, the effects of positive shocks are higher than negative shocks. If γ is significant and positive, then the effects of negative shocks are higher than positive ones (Soori, 2018)