Economic Journal of Emerging Markets The impact of real effective exchange rate on revealed comparative advantage and trade balance of Pakistan

This study estimates the effects of devaluation and appreciation of real effective exchange rate ( REER ) on revealed comparative advantage ( RCA ) at Harmonized System 2-digit level of exports in Pakistan. A nonlinear Autoregressive Distribute Lag ( ARDL ) technique is applied to test the asymmetric evidence. This study employs two models to explore the export performance. Findings/Originality: The results of model 1 estimation confirm the proof of asymmetric ARDL and concludes that devaluation has a positive effect on selected RCA ’s index value and helps enhance exports of Pakistan. Meanwhile, the appreciation of REER is having an adverse impact. Model-2 estimates the effect of these selected RCA ’s , REER , and world aggregated income ( Yw ) on the trade balance ( TB ) of Pakistan. The results estimate that an increase in selected RCA ’s index values, world aggregated income, and REER depreciation is useful to decrease in deficit TB of Pakistan.


Introduction
In the current era of globalization, a decrease in trade barriers, technological improvement, and exports determine the revealed comparative advantage. Most of the developing countries, including Pakistan, intend to expand their markets by reducing their production costs and achieving economies of scale to compete with their giant competitors in the region (Saeid Mahdavi & Sohrabian, 1993). Moreover, exports are also one of the key channels to transfer technology and skills to the developing economies (Kale, 2001). The exports led-based policies should focus on boosting exports to generate foreign reserves and balance of payment surpluses (Holmes & Panagiotidis, 2009;Onafowora, 2003). There may be numerous factors that can determine the trade performance of any country. As an example, absolute advantage, relative advantage, and differences in product cost, terms of trade (TOT), the concentration of advanced-technology, revealed comparative advantage (RCA), real effective exchange rate (REER), world-aggregated income, etc. (Bahmani-Oskooee, 2001;Bahmani-Oskooee & Fariditavana, 2016;Smith, 1976).
In the presence of giant-exports competitors, developing countries like Pakistan are conscious of boosting their export share in the world. The way to survive for any developing country is to capture more product share in the world markets if she has a comparative advantage in certain products. One of the critical factors like devaluation of the real effective exchange rate may help assess the world markets by reducing export prices. The REER has a significant impact on exports and imports of any economy (Kemal & Qadir, 2005). An appreciation or depreciation in the domestic currency may result in a change in the balance of trade of any county, e.g., RCA1 hides & skins. Figure-5 elucidates that RCA2 textiles and clothes contain the highest portion of export commodities at the HS-2-digit level of products of Pakistan (see Annexure-I).
One of the critical purposes of this research study is to explore Pakistan's export competitiveness by using Balassa (1965) revealed comparative advantage index (RCA) for selected sectors at HS 2digit level of products. Primarily, to check whether the impact of fluctuations in REER (i.e., positive shock & negative shock) have significant effects on our selected RCA's index values or not. Secondly, to check the impact of REER, world aggregated income, and our selected RCA's indexes on Pakistan's trade balance.
This research is novel in some ways. Waliullah, Kakar, Kakar, and Khan (2010), Shahid and Afzal (2013), and Hussain and Bashir (2013) examined the impact of the real effective exchange rate on the trade balance of Pakistan. However, there is no prior research studies have highlighted the importance of selected RCA's in the external sector of Pakistan by taking into account the positive and negative shocks of REER by applying a new technique (asymmetric autoregressive distributed lag). Secondly, this research study considers the Balassa (1965) RCA index, which helps explain Pakistan's specialization in a particular sector or industry in the world. Thirdly, this research will try to improve Pakistan's trade balance by boosting its untapped export demand in the world through a favorable exchange rate regime.

Real Effective Exchange Rate: Theoretical Framework
The theoretical background of the real effective exchange rate (REER) is a proportion of prices of non-tradable goods (Pnt) to tradable goods (Pn). Due to this logic, policymakers and economists often depend upon the REER that is occupying by the index ratio of local and foreign consumer rates, to proximate the REER (Henry & Longmore, 2003). The REER can be expressed as e = (r/r * ) (p/p * ), where r is the local exchange rate depicted in terms of international currency, r * is a compound of trading partners exchange rate, P is a local rate index, and P * is an external rate index.

Various Measurement of Revealed Comparative Advantages (RCA's) Indexes
Many studies define the measurement of comparative advantage. For example, RCA was empirically studied the first time by Liesner (1958) by the following method: where Xpj represents the exports of country p for commodity j and Xnj; n is the set of countries for j commodities or sectors. Balassa (1965) introduced the following measure. It has two dimensions, i.e., greater than one and less than one. If the RCA is greater than 1, the country has revealed comparative advantage in that commodity or industry. When the value of RCA is less than one, the result is reversed, i.e., the comparative disadvantage in that industry or sector. Balassa (1965) RCA index is calculated as follow: XPL: exports of product L by country P, XP: total exports from country P. XWL: total exports of product L by the rest of the world, XW: total exports from the world. Greenaway and Milner (1993) introduced to measure RCA is − = the difference in exports & imports of the country p for j sector. The RCA3 ranges as -1 ≤ RCA3 ≤ +1. However, the closer the value to +1, the greater the revealed comparative advantage will be according to that RCA index. Vollrath's (1991) RCA in logarithmic form can be written as This index represented the difference between the logarithmic revealed export advantage and revealed import advantage.
In short, some of the researchers, including Hoen and Oosterhaven 2006), Laursen (1998), and Proudman and Redding (2000) tried their best to overcome the shortcomings of the Balassa RCA index and introduced many alternative RCA indices. However, no one could succeed in rule out all these deficiencies. Therefore, Balassa (1965) RCA index is still widely recognized as one of the standard measures to estimate the revealed comparative advantage as a standard measure.

Model Specification
To examine the degree of trade-specialization of a specific country, Balassa (1965) Model 1 is constructed to check the impact of REER movement on RCA of these commodities, as suggested by Shin et al. (2011), by using the bivariate model: where i; 1, 2, 3 and t; 1980 to 2018, RCA1, RCA2, and RCA2 are Balassa (1965) index value of RCA's, hides & skins, textiles & clothes, and vegetables at HS-2-digit level of products respectively ln REER +,-; the natural log of the positive and negative shock of the real effective exchange rate of Pakistan. Model 2 is constructed to evaluate the impact of the real effective exchange rate, selected revealed comparative advantage indexes and world's aggregated income on the trade balance of Pakistan, use the following, adopted by Rose and Yellen (1989): where lnYw is world aggregated income, and other variables have been explained in model M-1. Annual data for Pakistan's export commodities are in the U.S. dollar and used to examine the non-linear relationship between each selected RCA and REER at HS 2-digit level products from 1980 to 2018. Data of RCA Balassa index is obtained from United Nations Commodity Trade Statistics Database (UN-COMTRADE), and World Integrated Trade Solution (WITS), data for REER and world aggregated income is taken from World Development Indicator (WDI), and State Bank of Pakistan (SBP). Table 1 presents a brief description of the variables. The nature of this research is two-folded. So, we assume the following symmetrical model by adding a REER variable as a determinant of revealed comparative advantage (RCA). The proposed model is: A Non-linear Autoregressive Distributed Lag (NARDL) technique developed by Shin et al. (2011) will be followed to observe the impact of positive shock (depreciation) and negative shock (appreciation) of the REER on revealed comparative advantage (RCA) index value. NARDL technique also has the advantage of solving the multi-collinearity problem by taking the appropriate lag order for the variables. Furthermore, Shin, Yu, and Greenwood-Nimmo (2014) suggested a methodology to differentiate one variable into two variables in the following manners: Equations (9) and (10) represent the partial sum of positive and negative movements in the real effective exchange rate (REER), respectively. If we put the variables calculated from equations (2) and (10) into equation (1), we can get the non-linear autoregressive distributed lag (NARDL), as suggested by Shin et al. (2014). This NARDL or asymmetric ARDL is the addition of the linear or symmetrical ARDL model of Pesaran, Shin, and Smith (2001) and expressed in the following way: First model-1, we have to find that there is a long-run relationship. For that purpose, the null hypothesis (H0) for no cointegration can be written as H0: (φ= β = γ + = γ -= 0). If estimated Fvalues from the Wald test are greater than the upper critical values calculated by Pesaran et al. (2001), we may reject the null hypothesis (H0) and conclude the existence of cointegration in equation (10). After confirming the cointegration, we can check the long-run effects of our explanatory variables, i.e., REERt and NEERt on RCAit by generating normalized coefficients of REERt-1 and NEERt-1, standardized by RCAt-1.
ECM model can be expressed by adding the lag of error term generated from the equation (4). This may have the following expression: From equation (11), a negative sign of the coefficient of Error Correction Model (ECM) is the proof for the short-run association in the model. It can also be an alternative confirmation for long-run relationships or cointegration in the model, as Pesaran et al. (2001) suggested. The maximum lag lengths are measured by the Schwarz Information Criterion (SIC).

Results and Discussions
As autoregressive distributed lag (ARDL) technique is useful in the presence of mixed order of integration, i.e., integration of order zero I (0) or one I (1). Our series is stationary at level zero or first difference, and found none of our series is required to check at their second difference. Table 2 shows the results obtained from Augmented Dickey-Fuller (ADF-test) for unit root (at 5% and 10% level of significance) at the level and first difference (Dickey & Fuller, 1979). It is clear that the RCA1 index value, real effective exchange rate (REER), and world aggregated income are stationary at level, and RCA2 and RCA3 are stationary at first difference. -----6.538*** (0.000) -5.831*** (0.010) -----3.013*** (0.004) ***, **, * indicate rejection of the null hypothesis at the 1%, 5% and 10%, respectively From Table 2, we can see that RCA1, RCA2, and RCA3 are stationary at the first difference. Non-linear Auto Regressive Distributed Lag (NARDL) Wald-test is used to check the existence of long-run relationship among the endogenous and exogenous variables in Model 1. The projected models of revealed comparative advantage (RCA's) are in Table 3.  Table 3, it is clear that long-run relationship exists in our examined model as calculated F-values are larger than the critical F-values, i.e., (upper bound critical values 3.2, 4.08, and 4.66 at 10%, 5%, and 1% respectively). Our models are free from autocorrelation, nonnormality, and heteroskedasticity problems. The Ramsey RESET test is also predicting the normality of the export models. Most significantly, the estimated NARDL parameters are stable in estimating CUSUM and CUSUM sq. Therefore, we can calculate the short-run and long-run effects of the projected models.   Table 4, we can see that the coefficients of positive real significant exchange rate shock (PREER) for RCA1 Hides & skins, RCA2 Textiles and Clothes, and RCA3 Vegetables. These coefficients represent the positive movement of REER, i.e., depreciation of the domestic currency. The PREER coefficient for RCA1 Hides & skins, 0.099, means that a 1% increase in REER brings a 0.0987% increase in revealed comparative index value (RCA1) Hides & skins. Similarly, the coefficient of negative real effective exchange rate movement (NREER) depicts the appreciation of PKR. The coefficient for RCA1 Hides & skins is -0.0433, which elaborates that a 1% decrease in REER will decrease the 0.043% index value of revealed comparative advantage for RCA1 Hides & skins. Furthermore, the coefficients of PREER for RCA2 Textiles and Clothes and RCA3Vegetables are 0.1233 and 0.025, respectively. 1% increase in real effective exchange rate brings 0.1233% and 0.025% increase in RCA index value for Textiles & Clothes, and Vegetables. However, coefficients of NREER -0.0951 and -0.0122 depict appreciation of PKR that elucidates a 1% decrease in REER may decrease 0.09% and 0.012% index value of revealed comparative advantage for RCA2 Textiles & Clothes, and RCA3 Vegetables respectively. Table 5 elaborates on the results of the short-run and Error Correction Model (ECM). The parameters of ECMt-1 are negative and significant in the short run. This is evidence that all the models have a short-run association. It may be the alternative verification of cointegration. Furthermore, due to depreciation, export performance improves in all the categories of export commodities, and appreciation has an adverse effect on all selected RCA's at the HS-2-digit level of products. The speed of convergence from the short run to the long run for selected RCA1, RCA2, and RCA3 are 1/0.2676*12=44 months, 23 months, and 29 months, respectively.  In overall estimation, we examined that positive and negative shocks of REER have expected impacts on selected RCA at HS-2-digit level of exports of Pakistan. Furthermore, all the models are predicting the non-linear effects in the long and short-run analysis as PREER and NREER are showing different signs of their parameters. Our research findings contrast the following papers, such as Ahmad and Kalim (2014), Ahmad and Kalim (2013), and Karamelikli (2016).
Model 2 applies a bound test to check the presence of a long-run relationship between the dependent and independent variables, and the outcomes are as follows. Table 7 indicates a longrun relationship between the variables as Ftestis greater than the upper bound value both at 5% and 10% significance levels.
After estimating the bound test ARDL technique, to calculate the cointegration form, an appropriate lag length is selected based on the Akaike Information Criterion (AIC); the results are shown in the following table. Moreover, the increase in ln (X/M) represents an improvement in the trade balance, while the decline shows an increase in the deficit trade balance of Pakistan. -0.020* 0.087 C 0.407 0.798 ***, **, * indicate rejection of the null hypothesis at the 1%, 5% and 10%, respectively Table 7 shows that coefficients of RCA1 Hides & skins, RCA2 Textiles and Clothes, and RCA3 Vegetables are negative and significant. The coefficient of RCA1 Hides & skins is -0.06544, which indicates that 1% increase in revealed comparative advantage index value will reduce the deficit in the balance of trade of Pakistan by 0.06%. Similarly, coefficients of RCA2 Textiles and Clothes and RCA3 Vegetables are -1.0077 and -0.04587 that represent 1% increase in revealed comparative advantage index value will lead to a decline by 1% and 0.045% in the deficit trade balance of Pakistan at HS 2-digit level products. Further, the real effective exchange rate (REER) has a significant impact on the balance of trade of Pakistan. Ln (REER) is -0.00332, which indicates that a 1% increase in real effective exchange rate brings a 0.0033% decrease in the deficit balance of trade of Pakistan. Moreover, the coefficient of world aggregated income (Yw) is 0.020, which indicates that a 1% increase in world aggregated income may decrease in deficit BOT of Pakistan by 0.020%. Hence, it can be concluded that RCA indexes, Yw, and REER are helpful in reducing the deficit trade balance of Pakistan. A similar relationship has been observed by many researchers such as Chinn (1991Chinn ( , 2006, Duasa (2007), and Moffett (1989). Table 8, ECM for the balance of trade equation represents the speed of adjustment in the short run to the long run (ECM speed of convergence is 1/0.3890*12 months=31 months).  (-1) -0.389*** 0.004 ***, **, * indicate rejection of the null hypothesis at the 1%, 5% and 10%, respectively Table 9 shows that the value of R-squared (0.8870) indicates that 88.70% variation in the TB equation is due to independent variables REER, selected RCA index values, and world aggregated income. The lower part of table 9 represents the different diagnostic-tests to examine the rationality of the ARDL model of the TB equation. The results confirm the absence of heteroscedasticity, serial correlation, and conclude that the error term is not generally distributed in the model. The value of the LM version is > 0.05, which indicates rejecting the null hypothesis and accept the alternative hypothesis.
The CUSUM (cumulative summations) and CUSUMsq (cumulative sum of the square of residuals across time trend) are used to validate the stability of the ARDL model. These tests were developed by Brown, Durbin, and Evans (1975) and also suggested by Pesaran and Shin (1998) in the previous literature.

Conclusions
This study estimates the effects of devaluation and appreciation of the REER on RCA at the HS-2-digit level of exports of Pakistan. Using annual data from 1980 to 2018, this study found the long run and short run cointegration for the products in which Pakistan has revealed comparative advantage. From the results of model-1, appreciation of REER has adverse effects on all selected RCA models, i.e., RCA1 hides & skins, RCA2 textiles & clothes, and RCA3 vegetables; likewise, devaluation has positive effects on these selected RCA models. Therefore, depreciation helps boost exports of commodities at the HS-2-digit level of the industry. From model-2, the outcomes reveal that the real effective exchange rate (REER) depreciation of domestic currency (PKR) has a significant impact on the trade balance (TB) of Pakistan. The world aggregated income (Yw; is assumed to assess the export demand of Pakistan) selected RCA's indexes at HS 2-digit level of products can help reduce deficit balance of trade by seeking potential market so that the demand for Pakistani exports may enhance in the world market. Moreover, this paper suggests that the government should adopt such an exchange rate policy that leads to increase exports and may cause a reduction in the deficit trade balance of Pakistan. Furthermore, this paper also suggests that policymakers should make effective monetary and fiscal policies to enhance the trade potential of Pakistan to capture the export share in world markets.
A clear limitation of this study is that it takes only three index values of selected RCA's in which Pakistan has a comparative advantage in the world. Other RCA's may be included to predict better impacts on the trade of potential of Pakistan, but unluckily, Pakistan has no revealed comparative advantage in those commodities at HS 2-digit level of products.