91 Abstract The relationship between trade defi cit and fi scal defi cit has remained an important topic in the last decades. The literature reveals many attempts to investigate the relationship between the two defi cits, while terminologies like Keynes- ian Proposition, Ricardian Equivalence, and Targeting Current Account Defi cit are used for the relationship between the two defi cits. The literature also highlights the possibility of unidi- rectional or bidirectional causality between these defi cits for different countries and different time spans. These four possibilities have inspired us to investigate these possibilities in the light of many control variables like unemployment, urbaniza- tion, money supply, foreign direct investment, and economic development in the case of Paki- stan. The fi ndings of the study reveal that fi scal defi cit signifi cantly accelerates trade defi cit into both long term and short term; hence, it supports the Keynesian Proposition in Pakistan, while a bidirectional causality between fi scal defi cit and trade defi cit into both long term and short term was also observed. Moreover, the mean and variance of error term were also found to be structurally stable which confi rms the absence of structural break for the selected period in Pa- kistan. In the end, based on these fi ndings, this study has concluded that two possibilities prevail out of four on the relationship between fi scal and trade defi cits. The twin defi cit is relevant and it does prevail in a transition economy like Paki- stan. Keywords: Pakistan, fi scal and trade defi cit (twin defi cit), causality, cointegration. TESTING RELEVANCE OF TWIN DEFICIT FOR A TRANSI- TION ECONOMY LIKE PAKISTAN Muhammad Shahid HASSAN Ayesha WAJID Haider MAHMOOD Muhammad SHAHBAZ Muhammad Shahid HASSAN Assistant Professor, Department of Economics, School of Business and Economics, University of Management and Technology, Lahore, Pakistan Tel.: 0092-423-521.2801-10 E-mail: muhammadshahidhassan@yahoo.com Ayesha WAJID Department of Economics, School of Business and Economics, University of Management and Technology, Lahore, Pakistan Tel.: 0092-423-521.2801-10 E-mail: m.shassan@yahoo.com Haider MAHMOOD Assistant Professor, Department of Management Sciences, COMSATS Institute of Information Technology, Lahore, Pakistan Tel.: 0092-111-001.007 E-mail: haidermahmood@ciitlahore.edu.pk Muhammad SHAHBAZ Assistant Professor, Department of Management Sciences, COMSATS Institute of Information Technology, Lahore, Pakistan Tel.: 0092-111-001.007 E-mail: shahbazmohd@live.com Transylvanian Review of Administrative Sciences, No. 46 E/2015, pp. 91-106 92 1. Introduction The relationship between fi scal and trade defi cit can be summarized into the Keynesian Proposition and the Ricardian Equivalence Hypothesis. The Keynesian Proposition states that fi scal defi cit will have a signifi cant and positive impact on trade defi cit and argues that fi scal defi cit comes into being due to expansionary fi scal policy which enhances local expenditures or absorption for imports, therefore, the continuous increase in imports will start increasing the trade defi cit. It could be in- ferred that budget defi cit may positively create trade defi cit. There are studies which support the Keynesian Proposition, such as: Fleming (1962), Mundell (1963), Volcker (1987), Zaman and DaCosta (1990), Kearney and Monadjemi (1990), Bachman (1992), Smyth and Hsing (1995), Vamvoukas (1999), Aqeel and Nishat (2000), Lau and Haw (2003), Onafowora and Owoye (2006), Corsett i and Muller (2006), Mukhtar, Zakaria and Ahmed (2007), Kim and Roubini (2008), Muller (2008), Beetsma, Giuliodori and Klaassen (2008), Pantelidis et al. (2009), Bouhga-Hagbe et al. (2010), and Jawaid and Raza (2013). However, the fi ndings of Monacelli and Perott i (2007) revealed that fi s- cal defi cit will have a negative but signifi cant impact on trade defi cit. They further stated that as fi scal defi cit is becoming the reason of the current account or trade defi - cit, government regulations must be aimed at bringing the balance between volume of exports and volume of imports. Consequently, trade defi cit may decline and may achieve state of balance. The second view on the relationship between fi scal defi cit and trade defi cit is rec- ognized as the Ricardian Equivalence Hypothesis and it is proposed by Barro (1989). This view reveals that fi scal defi cit is not the cause of trade defi cit and simply both defi cits are neutral. The advocates stated the transmission mechanism for the neutral relationship between fi scal and trade defi cits that because of expansionary fi scal poli- cy government cuts taxes or may increase its expenditures. Consequently, as the dis- posable income of the masses increases, private savings will be enhanced and in turn this will encourage domestic investment; therefore, overall exports will increase in response to increase in domestic production and exhibiting no external defi cit in the country. This view is supported by the fi ndings of Miller and Russek (1989), Rahman and Mishra (1992), Evans and Hasan (1994), Wheeler (1999) and Kaufmann, Scharl- er and Winckler (2002). Moreover, there is another possibility of the relationship be- tween fi scal defi cit and trade defi cit: trade defi cit may have a signifi cant impact on fi scal defi cit or trade defi cit may cause fi scal defi cit. This relationship between the two defi cits is investigated by Summers (1988), Islam (1998), Khalid and Guan (1999), and Alkhatib-Alkswani (2000); they found that unidirectional causality running from current account defi cit to budget defi cit prevails. The prime reason behind investigat- ing the impact of trade defi cit on fi scal defi cit is to provide an answer for the question whether targeting current account defi cit aff ects fi scal/budget defi cit or not? Furthermore, the relationship between fi scal defi cit and trade defi cit could be bi- directional as well, meaning that both defi cits could cause each other and may con- tradict the Keynesian Proposition (Summers, 1988). The studies of Laney (1984), Dar- 93 rat (1990), Evans (1993), Ibrahim and Kumah (1996), Lau and Baharumshah (2004), Mukhtar, Zakaria and Ahmed (2007), Baharumshah (2007), Jayaraman and Choong (2007), Lau, Abu Mansor and Puah (2010), and Mehrara and Zamanzadeh (2011) have confi rmed bidirectional causal relationship between trade and fi scal defi cits. The co- existence of both defi cits is referred to as twin defi cit. The developing countries of the world have been experiencing twin defi cit in the past years and the co-movement of both defi cits is accelerating with time. Pakistan, as a transition economy, is also experiencing the simultaneous existence of both defi cits and these defi cits are also accelerating in Pakistan, further exhibiting many macroeconomic ills in the country. In the case of Pakistan, the causal relationship between trade and fi scal defi cits has been tested by Burney and Yasmeen (1989), Burney and Akhtar (1992), Kazmi (1992), Aqeel and Nishat (2000), Mukhtar, Zakaria and Ahmed (2007), and Hakro (2009). This study is an att empt to investigate whether the Keynesian Proposition or Ri- cardian Equivalence Hypothesis prevails in the case of Pakistan. This study will also test the causal relationship between fi scal defi cit and trade defi cit and will check which view point is more suitable or most relevant for Pakistan, so that it could help policy advisors in suggesting appropriate policy measures. In the present study we have considered unemployment, urbanization, money supply, foreign direct invest- ment and human development index as explanatory factors of trade defi cit along with fi scal defi cit. This study is diff erent from the other studies due to its control factors, the time span and the methodological framework used. This study applies Ng-Perron (2001) unit root test, ARDL Bounds Testing Approach, and VECM based Causality Test for investigating the relationship between trade defi cit and fi scal defi - cit in addition to various controlled factors for the dataset ranging from 1972 to 2012. Section 2 off ers a brief review of the previous researches, section 3 presents the methodological framework, section 4 reviews the empirical fi ndings of the study, and the fi nal section 5 reports the conclusions and policy implications. 2. Literature review Zaman and DaCosta (1990) investigated the causal relationship between budget defi cit and current account defi cit for the period ranging from 1971(Q1) to 1989 (Q4) and found unidirectional causality running from budget defi cit to current account defi cit. Bachman (1992), using VAR model on quarterly dataset for the period from 1974 to 1988, explored federal budget defi cit as a factor that explained variations into current account defi cit and confi rmed evidence of twin defi cit in the US. Beside Bach- man (1992), Vamvoukas (1999) also examined the causal relationship between budget defi cit and trade defi cit for the Greek economy for the period from 1948 to 1993 and found unidirectional causality running from budget defi cit to trade defi cit. This study concluded that the Keynesian Proposition prevailed on the long run and short run in the Greek economy. Aqeel and Nishat (2000) who investigated the twin defi cit hypothesis for Pakistan by considering a dataset from 1973 to 1998 found that fi scal defi cit positively and 94 signifi cantly caused a current account defi cit into long run, but it inversely caused a current account defi cit on the short run. The fi ndings further confi rmed the evidences of unidirectional causality running from economic growth to current account defi cit, running from money supply to current account defi cit, and running from exchange rate to current account defi cit in Pakistan. Lau and Haw (2003) also explored the twin defi cit for ASEAN economies like Malaysia and Thailand by applying Vector Au- toregressive model and Toda Yamamoto technique by covering a sample period for Thailand from 1976 (Q1) to 2000( Q4) and for Malaysia from 1976 (Q1) to 1998 (Q2) and found evidence of unidirectional causality running from budget defi cit to trade defi cit for the case of Thailand (validating the Keynesian Proposition) but evidence of bidirectional causality was found for the case of Malaysia. The study concluded that the budget defi cit consistently aff ects current account through exchange rate and interest rate channels. Lau and Baharumshah (2004) found evidence of twin defi cit in the case of Malaysia for the period from 1976 (Q1) to 2000 (Q4). In another study, Onafowora and Owoye (2006) found positive eff ects of budget defi cit on trade defi cit into both long run and short run, and also found an evidence of unidirectional cau- sality running from current account defi cit to budget defi cit in Nigeria for the period 1970-2001. The fi ndings further exposed that money supply, exchange rate, output growth and interest rate were negatively aff ecting trade defi cit in the long run. Be- sides this study, Mukhtar, Zakaria and Ahmed (2007) observed that budget defi cit increases trade defi cit in the long run but declines trade defi cit into short run and also found evidence of bidirectional causality between budget and trade defi cit for Paki- stan for a quarterly dataset from 1975 to 2005. Pantelidis et al. (2009) also investigated the twin defi cit hypothesis for the case of Greece for the 1960-2007 period and found an evidence of twin defi cit; however, this evidence was weak and they relate it with Quintos Terminology, therefore, they remained with Keynesian Proposition for their fi ndings regarding twin defi cit. Additionally, they came up with the fi ndings that public expenditures regarding aging will be a serious threat for the long run stability of social security fi nancing. Ozturkler and Colak (2010) explored the relationship between trade defi cit and unemployment and found that trade defi cit signifi cantly accelerates unemployment in Turkey for the period from 1960 to 2009. In another study, Waliullah et al. (2010) found income and money supply as important determinants of trade balance for both short and long term spans for Pakistan for the period from 1970 to 2005. The fi ndings further exposed that money supply signifi cantly decreases trade balance into both long run and short run. Mohammad (2010) found real eff ective exchange rate as more important determinants of trade defi cit than that of foreign income, do- mestic consumption and foreign direct investment for Pakistan for the period from 1975 to 2008. Syed, Hasnat and Li (2011) did not fi nd any signifi cant relationship between foreign direct investment and trade defi cit in Pakistan for the period from 1990 to 2010, whereas, the relationships between foreign direct investment and ex- ports and between foreign direct investment and imports were found to be signifi - 95 cant. Aurangzeb and Haq (2012) found the exchange rate, foreign direct investment, economic growth and remitt ances as signifi cant factors which aff ect trade defi cit in Pakistan in the long run for the period from 1981 to 2010. This study provides evi- dence of bidirectional causality between foreign direct investment but it also found unidirectional causality running from trade defi cit to economic growth in Pakistan. Hassan, Wajid and Ahmet (2012) found negative and signifi cant eff ect of urbaniza- tion on trade openness in the long run but positive and signifi cant eff ect of urbaniza- tion on trade openness in the short run in Pakistan for the period from 1975 to 2010. This study further found bidirectioanal causality between urbanization and trade openness in the short run but unidirectional causality running from urbanization to trade openness in the long run was evident in Pakistan; the authors concluded that urbanization could benefi t trade openness in the short run but it could harm trade openess in the long run. Anas (2013) found evidence of unidirectional causality running from current ac- count defi cit to fi scal defi cit for Morocco for the data set from 1980 to 2012, and con- cluded that current account defi cit left a negative impact on public defi cit in Morocco. Jawaid and Raza (2013) found signifi cant and positive eff ects of fi scal defi cit on trade defi cit for both long run and short run in Pakistan for the period from 1976 to 2010, and this study further found evidence of unidirectional causality running from fi scal defi cit to trade defi cit in the short run in Pakistan. Saysombath and Kyophilavong (2013) found evidence of bidirectional causal relationship between fi scal and trade defi cits for Lao PDR for the period from 1980 to 2010. Tufail et al. (2014) found a posi- tive and signfi cant eff ect of budget defi cit on trade defi cit and also confi rmed bidirec- tional causality between budget and trade defi cits for Pakistan for the sample period from 1972 to 2011. 3. Data source and methodology This part is designed to demonstrate the means through which the dataset is gath- ered up. Also, the aim of this section is to develop on the insights of the methodologi- cal framework that is going to be applied for the empirical investigation. 3.1. Methodological framework Many empirical studies were conducted to test Log Linear Models, like Ehrlich (1977), Layson (1983), Bowers and Pierce (1975), Cameron (1994) and Ehrlich (1996). All these studies revealed that the computed results from Log Linear models are more reliable and robust as compared with Linear Form models. Therefore, the objective of this study is to test the Log Linear Model which is given as below: lnHDI7lnFDI6 lnM5lnUB4lnUN3lnFD21lnTD ++ +++++= , whereas: 96 Table 1: Variables Variable Names Variable Representation Variable Transformation Data Source Sample Period Trade Defi cit lnTD ln [(Imports – Exports)/(Real GDP)] WDI, World Bank (2014) 1972 – 2012 Fiscal Defi cit lnFD Log [(Fiscal Expenditures – Fiscal Revenue)/(Real GDP)] WDI, World Bank (2014) 1972 – 2012 Unemployment lnUN ln [(Unemployment)/(Total Labour Force)] WDI, World Bank (2014) 1972 – 2012 Urbanization lnUB ln [(Urban Population)/(Total Population)] WDI, World Bank (2014) 1972 – 2012 Money Supply lnM ln [(M2/R.GDP = Monetary Assetas share of GDP)] WDI, World Bank (2014) 1972 – 2012 Foreign Direct Investment lnFDI ln [(Foreign Direct Investment Infl ows)/ (Real GDP)] WDI, World Bank (2014) 1972 – 2012 Economic Development lnHDI ln [Human Development Index] HDR (2009), UNDP 1972 – 2012 The relationship between trade defi cit and fi scal defi cit have been found in many studies such as Zaman and DaCosta (1990), Bachman (1992), Vamvoukas (1999), Aqeel and Nishat (2000), Lau and Haw (2003), Lau and Baharumshah (2004), On- afowora and Owoye (2006), Mukhtar, Zakaria and Ahmed (2007), Pantelidis et al (2009), and Jawaid and Raza (2013). Following the evidence of Ozturkler and Colak (2010), we would like to investigate the relationship between trade defi cit and un- employment for Pakistan. Researchers like Aqeel and Nishat (2000), Onafowora and Owoye (2006), and Waliullah et al. (2010) have explored the relationship between trade defi cit and money supply, and we intend to examine the relationship between trade defi cit and money supply in case of Pakistan. Studies conducted by Mohammad (2010), Syed, Hasnat and Li (2011), and Aurangzeb and Haq (2012) support us to test the relationship between trade defi cit and foreign direct investment for Pakistan. The relationship between economic growth and trade defi cit is also tested for Pakistan by Aurangzeb and Haq (2012). In the present study we will use human development index to represent economic development and we intend to test its impact on trade defi cit. We have not come across studies that explain the relationship between urban- ization and trade defi cit, but, Hassan, Wajid and Ahmet (2012) explored the impact of urbanization on trade openness for Pakistan, therefore, that study prompted us to consider urbanization as a factor which can determine trade defi cit. 3.2. Estimation methods The estimation procedure is divided into four parts: in the fi rst part the problem of Unit Root will be examined by applying Ng-Perron (2001). In the second part the long term cointegrating relation between outcome and predictors of this study will be examined by using ARDL Bounds Testing1 Approach. This same approach will facil- 1 The results for ARDL Bounds Testing Approach have been obtained using the Demo Version of Mircofi t 5.2 developed by Pesaran, Richard and Shin (2001). For more details on cointegration, please see Bannerjee, Dolado and Mestre (1998) or Engle and Granger (1987). 97 itate us to fi nd out long run and short run coeffi cients of the predictors for outcome variable and in the fi nal part the direction of causality for short run and long run will be scrutinized using VECM based causality test. The long run cointegration and long run coeffi cients will be estimated by using the following equation: 11 p 0i it lnHDI17 p 0i it lnFDI16 p 0i it lnM15 p 0i it lnUB14 p 0i it lnUN13 p 0i it lnFD12 p 1i it lnTD111tlnFDI171tlnFDI161tlnM15 1tlnUB141tlnUN131tlnFD121tlnTD1110C tlnTD + = − + = − + = − + = − + = − + = − + = − +−+−+−+ −+−+−+−+= The short run coeffi cients for the selected ARDL model will be estimated using the following equation: 111tECM11 p 0i it lnHDI17 p 0i it lnFDI16 p 0i it lnM15 p 0i it lnUB14 p 0i it lnUN13 p 0i it lnFD12 p 1i it lnTD11 10C tlnTD μ+−+= − + = − + = − + = − + = − + = − + = − += Moreover, Cumulative Sum of Recursive Residuals (CUSUM) as well as Cumu- lative Sum of the Squares of Recursive Residuals (CUSUMsq) will be used in order to explore stability of the mean and variance of the error term of the selected ARDL model. After discussing the section of methodology, now we would like to present the estimated results and their discussion in the following section: 4. Results and interpretation Table 2 reveals the fundamental information regarding the factors of the study. In the following table, the mean, median and standard deviation (among others) are re- ported. The J.B test was not signifi cant and showed that all the factors taken into this study are normally distributed. After the descriptive information, the problem of unit root has been addressed by applying the Ng-Perron (2001) unit root test. The results reported in Table 3 demon- strate that among all the variables only fi scal defi cit has found to be stationary at the same level, while other variables have found to be stationary at fi rst diff erence. Therefore, the dataset for this study has integrated order I(0) and I(1) or mixed order of integration. The literature on application of cointegration test has proposed that long term relationship between outcome and predictors can be examined by the Auto Regressive Distributed Lag (ARDL) model if data series are integrated at order I(0) and I(1). 98 Table 2: Descriptive statistics tlnTD tlnFD tlnUB tlnUN tlnFDI tlnHDI tlnM Mean -0.345107 -2.681915 344.6361 1.419921 0.000586 -9.739171 30.35993 Median -0.260032 -2.790124 344.4257 1.335001 0.000258 -9.838460 30.12166 Maximum 0.287562 -1.126162 364.6201 2.156854 0.003643 -6.805691 48.77433 Minimum -1.214745 -3.705248 324.6491 0.566228 2.44E-07 -13.09706 11.32993 Std. Dev. 0.430450 0.713870 11.22407 0.386482 0.000831 1.862924 11.11618 Skewness -0.500117 0.488674 0.022888 0.151717 2.222113 -0.015661 0.020760 Kurtosis 2.046594 2.008387 2.004647 2.390346 7.573392 1.912078 1.805771 Jarque-Bera 3.261979 3.311611 1.696074 0.792240 69.47288 2.023617 2.439339 Probability 0.195736 0.190938 0.428255 0.672926 0.000000 0.363561 0.295328 Sum -14.14940 -109.9585 14130.08 58.21678 0.024024 -399.3060 1244.757 Sum Sq. Dev. 7.411482 20.38441 5039.186 5.974730 2.76E-05 138.8194 4942.774 Observations 41 41 41 41 41 41 41 Std. = Standard; Sq. = Square, and Dev. = Deviation Table 3: Unit root test Ng-Perron Test Statistics Variable I (0) Variable I (1) MZa MZt MSB MPT MZa MZt MSB MPT tlnTD -1.68878 -0.72442 0.42896 11.6012 Δ tlnTD -31.5138*** -3.95587 0.12553 0.81888 tlnFD -18.6556*** -3.02172 0.16197 1.42978 Δ tlnFD -21.7022*** -3.28005 0.15114 1.17739 tlnUB 1.69168 1.51875 0.89777 64.9465 Δ tlnUB -7.31139* -1.89697 0.25945 3.40529 tlnUN -3.67021 -1.14572 0.31217 6.71918 Δ tlnUN -19.1433*** -3.05967 0.15983 1.40153 tlnFDI -0.39809 -0.35130 0.88245 40.7522 Δ tlnFDI -7.09732* -1.84147 0.25946 3.60186 tlnHDI 1.39698 1.40720 1.00731 75.9094 Δ tlnHDI -14.4573*** -2.66714 0.18448 1.77628 tlnM 0.41748 0.23132 0.55408 23.6233 Δ tlnM -9.64562** -2.17415 0.22540 2.62522 *; **, and *** reveals signifi cance level of test statistic at 10%, 5% and 1% respectively. To fi nd a long run relationship we have applied ARDL Bounds Testing Approach and the estimated results reported in Table 4 confi rm the existence of long run rela- tionships between trade defi cit and its factors as F – statistic has found to be greater than the upper critical bound at 5% level of signifi cance. The results of unit root test and cointegration are reported in Table 3 and Table 4. The estimated results from Table 4 show that the calculated value of F test is 6.0633 which is greater than its corresponding critical value 4.1892 at 5% level of sig- nifi cance, therefore, this confi rms the evidence of long run cointegration between trade defi cit and its factors like fi scal defi cit, unemployment, urbanization, foreign direct investment, economic development and money supply. The estimated proba- bility values of the chi-square tests for all the diagnostics tests were not found to be signifi cant which revealed that there are no serial correlation and heteroscedastici- ty problems in this study. Moreover, the error term of the selected ARDL model is 99 Table 4: ARDL bounds testing approach Estimated Models TDt = f(FDt, UBt, Ut, FDIt, HDIt, Mt) Optimal lags (1,0,0,0,0,1,1) F – statistics 6.0633** W – statistics 42.4432** Signifi cance Level Critical Bounds for F – Statistics Critical Bounds for W – Statistics Lower Critical Bound Upper Critical Bound Lower Critical Bound Upper Critical Bound 5 per cent 2.7985 4.1892 19.5892 29.3247 10 per cent 2.3499 3.5982 16.4495 25.1874 DIAGNOSTIC TESTS R2 0.9515 Serial Correlation 0.3506 [0.554] Adjusted - R2 0.9369 Functional Form 0.0024 [0.961] F – Statistics 65. 3742 Normality 0.9904 [0.637] P – Value [F – Statistics] [000] Heteroscedasticity 0.0593 [0.808] DW – Statistic 2.1353 Durbin H – Statistic -0.5753 [0.565] *;**, and *** demonstrates signifi cance level at 10%, 5% and 1% respectively. Also the values within [ ] represents Probability Values. normally distributed and the functional form of the selected ARDL model is also cor- rectly specifi ed. Afterwards, the long run and short run coeffi cients of fi scal defi cit along with other controls have also been estimated, and the results have been pre- sented in Table 5. Table 5: Long term and short term dynamics Estimated Long Term Coeffi cients using the ARDL Approach Error Correction Representation for the Selected ARDL Model Dependent Variable: tLTD Dependent Variable: tÄLTD Variables Coeffi cient[P - Value] Variables Coeffi cient [P - Value] tlnFD 0.1792 [0.078] Δ tlnFD 0.0896 [0.047] tlnUB 0.4250 [0.002] Δ tlnUB 0.2125 [0.000] tlnUN 0.5145 [0.008] Δ tlnUN 0.2573 [0.012] tlnFDI -0.11862 [0.999] Δ tlnFDI -0.0593 [0.999] tlnHDI -0.9016 [0.065] Δ tlnHDI -0.0778 [0.726] tlnM -0.2469 [0.001] Δ tlnM -0.0097 [0.802] C -148.6876 [0.003] 1-tECM -0.5001 [0.000] Diagnostics for ECM Model R-squared 0.6289 Mean Dependent Variable 0.0225 Adjusted R-squared 0.5175 S.D. Dependent Variable 0.1559 S.E. of Regression 0.1083 Akaike Information Criterion 27.9132 Sum Squared Residual 0.3518 Schwarz Bayesian Criterion 19.4688 Log Likelihood 37.9132 Durbin-Watson Stat 2.1353 F-statistic 7.2618 Prob. Value (F-statistic) [0.000] *; **, and *** reveals signifi cance level of test statistic at 10%, 5% and 1% respectively. 100 The estimates for long term and short term coeffi cients reported in Table 5 demon- strate that fi scal defi cit has a signifi cant and positive impact on trade defi cit in the long term and in the short term in Pakistan. This shows that fi scal defi cit begets trade defi - cit in Pakistan, and hence it also validates the existence of Keynesian Proposition in case of Pakistan. This fi nding is consistent with Fleming (1962), Mundell (1963), Vol- cker (1987), Zaman and DaCosta (1990), Kearney and Monadjemi (1990), Smyth and Hsing (1995), Vamvoukas (1999), Aqeel and Nishat (2000), Lau and Haw (2003), Ona- fowora and Owoye (2006), Corsett i and Muller (2006), Mukhtar, Zakaria and Ahmed (2007), Kim and Roubini (2008), Muller (2008), Beetsma, Giuliodori and Klaassen (2008), Pantelidis et al. (2009), Bouhga-Hagbe et al. (2010), and Jawaid and Raza (2013). Table 5 also shows that unemployment and urbanization are signifi cantly increas- ing the size of trade defi cit in both long term and short term. It has been generally ob- served that the purchasing power of the people declines as unemployment expands; consequently, aggregated demand declines, which reduces overall prices in the coun- try and hence the profi ts of entrepreneurs. Therefore, decline in the profi ts of entre- preneurs will make them reduce investments and manufacturing, which diminish- es exports and therefore increases the size of trade defi cit in the country. Moreover, the increase in the urban population puts pressure on the aggregated demand in the country and induces imports to increase, also stimulating trade defi cit in the country. It is also evident that money supply and economic development are signifi cantly curtailing trade defi cit into long term, but the coeffi cients of money supply and eco- nomic development have found to be negative and not signifi cant for short term. The transmission mechanism could be illustrated as that due to increase in money sup- ply in the hands of investors expand, which further induces investments to expand, production activities will fl ourish and these will further stimulate exports. Hence, it will reduce the trade defi cit in the country. The negative coeffi cient of economic de- velopment for trade defi cit demonstrates that as economic development takes place it improves infrastructure, the quality of human capital, and the effi ciency of the factors of production; as such, production expands due to a decline in the cost of production. The increase in production will ultimately improve the size of exports and hence it will shrink trade defi cit in the country. The impact of foreign direct investment on trade defi cit was also investigated and the coeffi cients were not found to be signifi - cant for both long term and short term. The coeffi cient for the fi rst-period lagged term of error was negative and signifi cant, which confi rms the evidence of convergence hypothesis in Pakistan, this means that long run and stable equilibrium will be re- stored by following a 50 percent speed of adjustment and this long run equilibrium will be restored in about 1.9996 (1/0.5001 = 1.9996) years. 4.1. VECM causality test After fi nding long term and short term dynamics, we are going to estimate VECM based causal relationship between trade defi cit and its factors for both long term and short term (the results are reported in Table 6). The estimates of the VECM based causality test reported in Table 6 have confi rmed the existence of bidirectional causal relationship between fi scal defi cit and trade 101 Table 6: Granger causality test Dependent Variable Short Run Causality Long Run Causality tlnTD tlnFD tlnUN tlnUB tlnM tlnFDI tlnHDI 1-tECM Δ tlnTD - 2.9328* 2.4733 1.1309 0.2960 0.5076 0.0766 -0.6585*** Δ tlnFD 3.8625** - 1.7913 0.8963 0.2087 1.9094 2.7839* -0.6014*** Δ tlnUN 3.2018* 1.1417 - 0.5291 1.5054 0.6165 6.9561*** -0.6845** Δ tlnUB 0.5962 1.3642 0.4273 - 2.2975 0.2287 0.6660 -0.085784 Δ tlnM 0.8793 0.4894 0.9885 2.2702 - 1.0157 0.9517 -0.6632*** Δ tlnFDI 2.0784 2.7897* 1.6307 0.1553 1.6341 - 1.5450 -0.2215** Δ tlnHDI 0.7574 1.2601 3.4756** 1.8570 1.2557 0.5445 - -0.7485** *; **, and *** reveals signifi cance level of test statistic at 10%, 5% and 1% respectively. defi cit in both short term and long term; thus, both defi cits are interdependent in Pakistan. This fi nding is supported by Laney (1984), Darrat (1990), Evans (1993), Ibra- him and Kumah (1996), Lau and Baharumshah (2004), Mukhtar, Zakaria and Ahmed (2007), Baharumshah (2007), Jayaraman and Choong (2007), Lau, Abu Mansor and Puah (2010), and Mehrara and Zamanzadeh (2011). Moreover, this study has also found bidirectional causal relationship between unemployment and trade defi cit in the long run, and unidirectional causality from trade defi cit to unemployment in the short run. The fi ndings further show that trade defi cit has a bidirectional causal rela- tionship with money supply, foreign direct investment and economic development in the long term. After discussing the estimates of the VECM based causality test, this study has also tested the structural stability of mean and variance of error term of the selected ARDL model and from the plots of CUSUM and CUSUM square we can conclude that both mean and variance of error term were found to be structural sta- ble, therefore, there are no problems of structural break in this study for the selected period (1972 -2012). CUSUM CUSUM SQUARE -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1973 1983 1993 2003 2012 The straight lines represent critical bounds at 5% significance level Plot of Cumulative Sum of Squares of Recursive Residuals -20 -10 0 10 20 1973 1983 1993 2003 2012 The straight lines represent critical bounds at 5% significance level Plot of Cumulative Sum of Recursive Residuals Figure 1: Stability test 102 5. Conclusion and policy implications 5.1. Conclusion This article aims at exploring the relationship between trade defi cit and fi scal defi - cit using unemployment, urbanization, money supply, foreign direct investment and economic development as control variables for Pakistan (from 1972 to 2012). The em- pirical results demonstrate that fi scal defi cit, unemployment and urbanization have a signifi cant and positive impact on trade defi cit into both long term and short term. However, money supply and economic development have a negative and signifi cant impact on trade defi cit only into the long term. Summarizing this, the signifi cant and positive impact of fi scal defi cit on trade defi cit confi rms the evidence of Keynesian Proposition for Pakistan. The empirical results regarding the direction of causality have confi rmed bidirectional causality between fi scal defi cit and trade defi cit in both short run and long run in Pakistan, meaning that both defi cits generate each other. We have also confi rmed that unemployment and economic development have a bidirectional relationship to each other into both short and long term in the case of Pakistan. The empirical fi ndings further reveal that in the short run there is unidi- rectional causality from trade defi cit to unemployment, while bidirectional causality prevails in the long run between these factors. Factors like trade defi cit, fi scal defi cit, unemployment, money supply, foreign direct investment and economic development have found to have a bidirectional causal relationship with each other in the long run, with the exception of urbanization. Unidirectional causality runs from urbanization to other factors like trade defi cit, fi scal defi cit, unemployment, money supply, foreign direct investment and economic development in the long run. The stability of both CUSUM and CUSUM square plots has confi rmed the absence of structural break. In the end, based on the fi ndings, this study confi rmed the Keynesian Proposition and twin defi cit in Pakistan for the selected sample (from 1972 to 2012). 5.2. Policy implications The fi ndings show that Pakistan is suff ering from twin defi cit, meaning that fi scal defi cit begets trade defi cit and trade defi cit begets fi scal defi cit. Att aining simultane- ous equilibrium into both public and trade fi nances is a highly diffi cult job; however, government could use a mix policy tool to deal with this issue. Firstly, government could curtail its non-development expenditures and should increase the volume of subsided inputs in the market. This att empt will curtail the cost of production but it will enhance domestic production, domestic employment and earnings of both the private sphere and government simultaneously. Moreover, increases in domestic production will also encourage the volume of exports and therefore, on the one side fi scal defi cit will come down because of an increase in gov- ernment income and on the other side trade defi cit would be reduced due to an in- crease in the volume of exports. Secondly, the government may also encourage an import substitution industry, so that domestic buyers will purchase import substitutes rather than imports. In this 103 way, the volume of imports will shrink and will decrease trade defi cit. The increase in import substitutes in the home country will increase home production, employment and earnings; therefore, it will also encourage home government earnings. A reduc- tion in the volume of imports will deteriorate trade defi cit and an increase in govern- ment revenues will squeeze fi scal defi cit in the country. References: 1. Alkhatib-Alkswani, M., ‘The Twin Defi cits Phenomenon in Petroleum Economy: Ev- idence from Saudi Arabia’, 2000, Economic Research Forum, Conference Paper No. 072000001, [Online] available at htt p://www.erf.org.eg/CMS/uploads/pdf/1185358196_ fi nance1.pdf, accessed on November 10, 2014. 2. Anas, M., ‘Twin Defi cits in Morocco: An Empirical Investigation’, 2013, International Journal of Business and Social Research, vol. 3, no. 7, pp. 160-172. 3. Aqeel, A. and Nishat, M., ‘The Twin Defi cit Phenomenon: Evidence from Pakistan’, 2000, The Pakistan Development Review, vol. 39, no. 4, pp. 535-550. 4. Aurangzeb and Haq, A.U., ‘Factors Aff ecting the Trade Balance in Pakistan’, 2012, Eco- nomics and Finance Review, vol. 1, no. 11, pp. 25-30. 5. Bachman, D.D., ‘Why is the US Current Account Defi cit So Large? Evidence from Vec- tor Autoregressions’, 1992, Southern Economic Journal, vol. 59, no. 2, pp. 232-240. 6. Baharumshah, A.Z. and Lau, E., ‘Dynamics of Fiscal and Current Account Defi cits in Thailand: An Empirical Investigation’, 2007, Journal of Economic Studies, vol. 34, no. 6, pp. 454-475. 7. Bannerjee, A., Dolado, J. and Mestre, R., ‘Error-Correction Mechanism Tests for Co-In- tegration in Single Equation Framework’, 1998, Time Series Analysis, vol. 19, no. 3, pp. 267-283. 8. Barro, R.J., ‘The Ricardian Approach to Budget Defi cits’, 1989, The Journal of Economic Perspectives, vol. 3, no. 2, pp. 37-54. 9. Beetsma, R., Giuliodori, M. and Klaassen, F., ‘The Eff ects of Public Spending Shocks on Trade Balances and Budget Defi cits in the European Union’, 2008, Journal of the Europe- an Economic Association, vol. 6, no. 2-3, pp. 414-423. 10. Bouhga-Hagbe, J., Ali Abbas, S.M., Fatas, A., Mauro, P. and Velloso, R.C., ‘Fiscal Poli- cy and the Current Account’, International Monetary Fund Working Papers no. 10/121, 2010, [Online] available at htt ps://www.imf.org/external/np/seminars/eng/2010/eui/ pdf/ABH.pdf, accessed on December10, 2014. 11. Bowers, W.J. and Pierce, J.L., ‘The Illusion of Deterrence in Isaac Ehrlich’s Work on the Deterrent Eff ect of Capital Punishment’, 1975, Yale Law Journal, vol. 85, no. 2, pp. 187-208. 12. Burney, N.A. and Yasmeen, A., ‘Government Budget Defi cits and Interest Rates: An Empirical Analysis for Pakistan’, 1989, The Pakistan Development Review, vol. 28, no. 4, pp. 971-980. 13. Burney, N.A. and Akhter, N., ‘Government Budget Defi cits and Exchange Rate Deter- mination: Evidence from Pakistan’, 1992, The Pakistan Development Review, vol. 31, no. 4, pp. 871-882. 14. Cameron, S., ‘A Review of the Econometric Evidence on the Eff ects of Capital Punish- ment’, 1994, The Journal of Socio-Economics, vol. 23, no. 1-2, pp. 197-214. 104 15. Corsett i, G. and Muller, G.J., ‘Twin Defi cits: Squaring Theory, Evidence and Common Sense’, 2006, Economic Policy, vol. 21, no. 48, pp. 597-638. 16. Darrat, A.F., ‘Structural Federal Defi cits and Interest Rates: Some Causality and Co-In- tegration Tests’, 1990, Southern Economic Journal, vol. 56, no. 3, pp. 752-759. 17. Ehrlich, I., ‘The Deterrent Eff ect of Capital Punishment: Reply’, 1977, The American Eco- nomic Review, vol. 67, no. 3, pp. 452-458. 18. Ehrlich, I., ‘Crime, Punishment, and the Market for Off ences’, 1996, The Journal of Eco- nomic Perspectives, vol. 10, no. 1, pp. 43-67. 19. Engle, R.F. and Granger, C.W.F., ‘Co-Integration and Error Correction: Representa- tion, Estimation, and Testing’, 1987, Econometrica, vol. 55, no. 2, pp. 251-276. 20. Evans, P., ‘Consumers are not Ricardian: Evidence from Nineteen Countries’, 1993, Economic Inquiry, vol. 31, no. 4, pp. 534-548. 21. Evans, P. and Hasan, I., ‘Are Consumers Ricardian? Evidence for Canada’, 1994, The Quarterly Review of Economics and Finance, vol. 34, no. 1, pp. 25-40. 22. Fleming, M.J., ‘Domestic Financial Policies Under Fixed and Under Floating Exchange Rates’, 1962, International Monetary Fund Staff Papers, vol. 9, no. 3, pp. 369-380. 23. Hakro, A.N., ‘Twin Defi cits Causality Link – Evidence from Pakistan’, 2009, Interna- tional Research Journal of Finance and Economics, no. 24, pp. 54-70. 24. Hassan, M.S., Wajid, A. and Ahmed, K., ‘Urbanization as a Way to Open Economy: Empirical Evidence from Pakistan’, 2012, World Applied Sciences Journal, vol. 20, no. 7, pp. 931-940. 25. Ibrahim, S.B. and Kumah, F.Y., ‘Comovements in Budget Defi cits, Money, Interest Rates, Exchange Rates and the Current Account Balance: Some Empirical Evidence’, 1996, Applied Economics, vol. 28, no. 1, pp. 117-130. 26. Islam, M.F., ‘Brazil’s Twin Defi cits: An Empirical Examination’, 1998, Atlantic Economic Journal, vol. 26, no. 2, pp. 121-128. 27. Jawaid, S.T. and Raza, A.S., ‘Dynamics of Current Account Defi cit: A Lesson from Pa- kistan’, 2013, Transition Studies Review, vol. 19, no. 3, pp. 357-366. 28. Jayaraman, T.K. and Choong, C.K., ‘Is the Twin Defi cits Hypothesis Relevant to Fiji?’, 2007, Fijian Studies: A Journal of Contemporary Fiji, vol. 5, no. 1, pp. 1-27. 29. Kaufmann, S., Scharler, J. and Winckler, G., ‘The Austrian Current Account Defi cit: Driven by Twin Defi cits or by Intertemporal Expenditure Allocation?’, 2002, Empirical Economics, vol. 27, no. 3, pp. 529-542. 30. Kazmi, A.A., ‘Ricardian Equivalence: Some Macro-econometric Tests for Pakistan’, 1992, The Pakistan Development Review, vol. 31, no. 4, pp. 743-758. 31. Kearney, C. and Monadjemi, M., ‘Fiscal Policy and Current Account Performance: In- ternational Evidence on the Twin Defi cits’, 1990, Journal of Macroeconomics, vol. 12, no. 2, pp. 197-219. 32. Khalid, M.A. and Guan, W.T., ‘Causality Tests of Budget and Current Account Defi - cits: Cross-Country Comparisons’, 1999, Empirical Economics, vol. 24, no. 3, pp. 389-402. 33. Kim, S. and Roubini, N., ‘Twin Defi cit or Twin Divergence? Fiscal Policy, Current Ac- count, and Real Exchange Rate in the U.S.’, 2008, Journal of International Economics, vol. 74, no. 2, pp. 362-383. 34. Laney, L., ‘The Strong Dollar, the Current Account, and Federal Defi cits: Cause and Eff ects’, 1984, Federal Reserve Bank of Dallas Economic Review, pp. 1-14. 105 35. Lau, E., Abu Mansor, S. and Puah, C.H., ‘Revival of the Twin Defi cits in Asian Crisis Aff ected Countries’, 2010, Economic Issues, vol. 15, no. 1, pp. 29-53. 36. Lau, E. and Haw, C.T., ‘Transmission Mechanism of Twin Defi cits Hypothesis: Evi- dence from Two Neighboring Countries’, 2003, INTI Journal, vol. 1, no. 3, pp. 159-166. 37. Lau, E. and Baharumshah, A.Z., ‘On the Twin Defi cits Hypothesis: Is Malaysia Diff er- ent?’, 2004, Pertanika Journal of Social Science and Humanities, vol. 12, no. 2, pp. 87-100. 38. Layson, S., ‘Homicide and Deterrence: Another View of the Canadian Time-Series Evi- dence’, 1983, The Canadian Journal of Economics, vol. 16, no. 1, pp. 52-73. 39. Mehrara, M. and Zamanzadeh, A., ‘Testing Twin Defi cits Hypothesis in Iran’, 2011, Interdisciplinary Journal of Research in Business, vol. 1, no. 9, pp. 7-11. 40. Miller, S.M. and Russek, F.S., ‘Are the Twin Defi cits Really Related’, 1989, Contempo- rary Policy Issues, vol. 7, no. 4, pp. 91-115. 41. Mohammad, S.D., ‘Determinant of Balance of Trade: Case Study of Pakistan’, 2010, European Journal of Scientifi c Research, vol. 41, no. 1, pp. 13-20. 42. Monacelli, T. and Perott i, R., ‘Fiscal Policy, the Trade Balance, and the Real Exchange Rate: Implications for International Risk Sharing’, 8th Jacques Polak Annual Research Conference, Washington, D.C., 2007, [Online] available at htt ps://www.imf.org/exter nal/np/res/seminars/2007/arc/pdf/tm.pdf, accessed on December 15, 2014. 43. Mukhtar, T., Zakaria, M. and Ahmed, M., ‘An Empirical Investigation for the Twin Defi cits Hypothesis in Pakistan’, 2007, Journal of Economic Cooperation, vol. 28, no. 4, pp. 63-80. 44. Muller, G.J., ‘Understanding the Dynamic Eff ects of Government Spending on Foreign Trade’, 2008, Journal of International Money and Finance, vol. 27, no. 3, pp. 345-371. 45. Mundell, R.A., ‘Capital Mobility and Stabilization Policy under Fixed and Flexible Ex- change Rates’, 1963, The Canadian Journal of Economics and Political Science, vol. 29, no. 4, pp. 475-485. 46. Ng, S. and Perron, P., ‘Lag Length Selection and the Construction of Unit Root Tests with Good Size and Power’, 2001, Econometrica, vol. 69, no. 6, pp. 1519-1554. 47. Onafowora, O.A. and Owoye, O., ‘An Empirical Investigation of Budget and Trade Defi cits: The Case of Nigeria’, 2006, The Journal of Developing Areas, vol. 39, no. 2, pp. 153-174. 48. Ozturkler, H. and Colak, O.F., ‘The Relationship between Current Account Defi cits and Unemployment in Turkey’, Turgut Ozal International Conference on Economics and Politics-1: Global Crises and Economic Governance, 2010, pp. 313-320, [Online] available at htt p://web.inonu.edu.tr/~ozal.congress/pdf/23.pdf, accessed on December 5, 2014. 49. Pantelidis, P., Trachanas, E., Athanasenas, A.L. and Katrakilidis, C., ‘On the Dynamics of the Greek Twin Defi cits: Empirical Evidence over the Period 1960-2007’, 2009, Inter- national Journal of Economic Sciences and Applied Research, vol. 2, no. 2, pp. 9-32. 50. Pesaran, M.H., Richard, J. and Shin, Y., ‘Bounds Testing Approaches to the Analysis of Level Relationships’, 2001, Journal of Applied Econometrics, vol. 16, no. 3, pp. 289-326. 51. Rahman, M. and Mishra, B., ‘Cointegration of US Budget and Current Account Defi cits: Twin or Strangers?’, 1992, Journal of Economics and Finance, vol. 16, no. 2, pp. 119-127. 52. Saysombath, P. and Kyophilavong. P., ‘Twin Defi cits in the Lao PDR: An Empirical Study’, 2013, International Business and Management, vol. 7, no. 1, pp. 62-68 106 53. Syed, S.H., Hasnat, H. and Li, J., ‘Foreign Direct Investment, Cost of War and Trade in Pakistan’, MPRA Paper No. 35598, 2011, [Online] available at htt ps://mpra.ub.uni- muenchen.de/35598/, accessed on December 5, 2014. 54. Smyth, D.J. and Hsing, Y., ‘In Search of an Optimal Debt Ratio for Economic Growth’, 1995, Contemporary Economic Policy, vol. 13, no. 4, pp. 51-59. 55. Tufail, M.K., Anwar, S., Raza, S.H. and Abbas, K., ‘Eff ect of Budget Defi cit on Trade Defi cit in Pakistan (A Time Series Analysis)’, 2014, Journal of Finance and Economics, vol. 2, no. 5, pp. 145-148. 56. Summers, L.H., ‘Tax Policy and International Competitiveness’, in Frenkel, J.A. (ed.), International Aspects of Fiscal Policies, Chicago: University of Chicago Press, 1988, pp. 349-386. 57. UNDP, Human Development Report, Overcoming Barriers: Human Mobility and Develop- ment, New-York: United Nation Development Program, 2009. 58. Vamvoukas, G.A., ‘The Twin Defi cits Phenomenon: Evidence From Greece’, 1999, Ap- plied Economics, vol. 31, no. 9, pp. 1093-1100. 59. Volcker, P.A., ‘Facing Up to the Twin Defi cits’, in Fink, R. and High, J. (eds.), A Nation in Debt: Economists Debate the Federal Budget Defi cit, Maryland: University Publications of America, 1987, pp. 154-161. 60. Waliullah, Kakar, M.K., Kakar, R. and Khan, W., ‘The Determinants of Pakistan’s Trade Balance: An ARDL Cointegration Approach’, 2010, The Lahore Journal of Econom- ics, vol. 15, no. 1, pp. 1-26. 61. Wheeler, M., ‘The Macroeconomic Impacts of Government Debt: An Empirical Anal- ysis of the 1980s and 1990s’, 1999, Atlantic Economic Journal, vol. 27, no. 3, pp. 273-284. 62. World Bank, World Development Indicators (CD Rom 2014), Washington, D.C.: World Bank, 2014. 63. Zaman, N.L. and DaCosta, M.N., ‘The Budget Defi cit and the Trade Defi cit: Insights into This Relationship’, 1990, Eastern Economic Journal, vol. 16, no. 4, pp. 349-354.