Volume 13 Issue 3 Article 2 12-30-2011 Are Pakistani consumer Ricardian? Muhammad Waqas Masood Sarwar Awan Follow this and additional works at: https://www.ebrjournal.net/home Recommended Citation Waqas, M., & Sarwar Awan, M. (2011). Are Pakistani consumer Ricardian?. Economic and Business Review, 13(3). https://doi.org/10.15458/2335-4216.1238 This Original Article is brought to you for free and open access by Economic and Business Review. It has been accepted for inclusion in Economic and Business Review by an authorized editor of Economic and Business Review. 161 ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011 | 161–177 ARE PAKISTANI CONSUMERS RICARDIAN? MuHaMMaD Waqas1 MasooD sarWar aWan1 aBstract: The purpose of this study is to check the Ricardian Equivalence Hypothesis in case of Pakistan by using annual data for the period of 1973-2009. Government expendi- ture, private consumption expenditure, tax revenue, government debt, disposable income, government budget deficit and wealth are the variables which are used for analysis. Cointe- gration results show a long run relationship among the variables. Results of structural form consumption function invalidate the Ricardian Equivalence Hypothesis in case of Pakistan. These results draw attention towards the significance of fiscal policies in boosting private consumption and controlling budget deficits, which are the prime goals of stabiliza- tion policies in Pakistan. keywords: fiscal policy, Ricardian equivalence, government debt. JeL classification: E62, E21. 1. introDuction In last decades most of the developing and developed economies are plagued by the budget deficits and government debt. These issues have fascinated the attention of public and politicians towards the minimization of government debt and reduction of budget deficit. In case of budget deficit government can finance its spending by three alternative ways; print new money, raising taxes and borrowing. Every option has its own conse- quences. Assume that government preferred borrowing to fulfill their needs instead of printing money and raising taxes. There are two schools of thought, regarding the rela- tionship between government debt and private consumption. Two centuries ago David Ricardo (1772-1823) introduced a theory regarding the rela- tionship between public deficit and private savings which has been invigorated by Robert Barro (1974) and hence called Ricardian Equivalence hypothesis (REh). The REh states that consumer deals government debt as future tax liabilities. Thus they are of view that reduction in taxes will not increase their consumption expenditure (aggregate demand will unaffected) but that will increase their savings because they believe that present bor- rowing will increase future tax on their generations. Consumers do this because after the maturity of borrowing government has to pay borrowing amount plus rate of inter- est so government imposes new taxes on their generation. Thus in order to protect new 1 University of Sargodha Pakistan, Department of Economics, E-mail: economist147@hotmail.com ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011162 generation from these taxes consumers buys bonds and does not consider them as a net wealth. hence private savings increase by same amount as budget deficit and national savings remain unaffected and there will be no crowding out of private investment2. Op- ponents of this theory, the Keynesians, are of view that consumers do not treat govern- ment bonds as a net wealth. On the response of tax cut consumers private consumption will increase (aggregate demand increases) and private saving will remain unaffected because consumers prefer present on past and does not consider the welfare of their generations in their mind. hence fiscal policy can affect the national output. These two approaches actually tell about the effectiveness of fiscal policy. If consumers are Ricard- ian fiscal policy is ineffective and if they behave like Keynesian fiscal policy is effective, but all this influence depends how consumer treat government debt in the context of net wealth. Therefore in order to design stabilization program a comprehensive research on the issue of REh is very essential. few studies highlight this issue in case of Pakistan and each of them has own limitations. This paper serves as an attempt to extend the exist- ing area of this research. Emphasizes is given to the use of less restrictive model for the investigation of REh. The rest balance of paper is designed as: part two explains the specification of the mod- el, part three explains the variables and data sources, part four discuss the empirical methodology, part five investigates and interprets the empirical results. finally, part six presents the conclusions of the study and also provides some policy implications. 2. sPecification of tHe MoDeL There are two types of consumption function, discussed in the literature, to check the validity of REh. After discussing those studies that extended the consumption function models, methodology for the present is discussed. REh can be checked by using two forms of consumption functions, Structural consumption function and Euler equation consumption function. Several studies validate REh3 and several invalidate4 it. for now structural consumption function is use to check the validity of REh and Euler equation consumption is on future agenda for researcher. 2 REh holds number of assumptions that must be satisfied for its validity (Giorgioni and holden, 2001). Like taxes and bonds must be perfect substitute, taxes must be used to pay interest on the debt, consumer invest same rate as government invest and consumer have perfect information about future and taxes are lump sum. Diamond (1965) said that this will be only possible if consumer lives forever, if consumer realizes that govern- ment will collect the tax after his death his consumption pattern definitely will changed. Bernheim (1987), King’s (1983) and Con and Jappeli (1990) results showed that consumer’s behavior is changed due to liquidity constraints. feldstein (1988) said that uncertainty in parent’s future income fails REh. 3 Khalid (1996), Rockerbie (1997), Cardia (1997), Lucke (1999), Drakos (2001, 2003), Sachsida & Carneiro (2001), Giorgioni & holden (2001), Walker (2002), Kaadu & Usukula (2004), Safa & Siddiq (2005), Cuaresma & Reitschuler (2007). 4 haq & Montiel (1987), Whelan (1991), Kazmi (1992, 1994), Ghatak & Ghatak (1996), Abimanyu (1998), Car- los (2001), Khan & Ashraf (2003), Ricciuti & Laurea (2003), Malengier & Pozzi (2004), Gray & Stone (2005), Gracia & Ramajo (2005), Nipple (2006), Apergis & Lyroudi (2006), Afonso (2008), Siddiki (2008), Waqas et al. (2011). M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 163 2.1 Structural Consumption Function Ricardian equivalence is rejected by feldstein (1982) by using following equation; Where C stands for total consumer expenditure, Y is current income, W indicates mar- ket value of privately owned wealth, SSW is value of future social security benefits, T symbolizes total tax revenue, TR shows government transfers to individuals,D is total government debt and et is error term. To check the validity of REh this function requires certain restrictions that must be fulfilled. Aschauer (1985) criticized feldstein model and argued that the use of current income as endogenous variable was the reason of endogenity in this model. No doubt, feldstein used one lagged values of income and taxes as instrumental variable to remove endog- enity but this problem may not be removed by using these instruments. Seater (1993) criticized the inconsistent criteria used by feldstein for inferring the results. Along with some weakness feldstein work provides sound simplification about REh. In 1983 Kormendi introduced “consolidated approach” which has a plus point that this model is based on permanent income hypothesis. Where PC is private consumption, Y stands for current total income, GS represents total government spending on goods and services, W symbolizes total wealth, TR is transfers, TX is tax revenue, REh is corporate retained earnings, GINT is government interest payment on outstanding debt, GB demonstrates market value of outstanding govern- ment debt and Ut is error term. following restrictions must be fulfilled for the validity of REh. After “Consolidated approach” Kormendi introduced a “Standard approach” which con- siders that consumption is determined by disposable income (Yd), total wealth plus gov- ernment debt (W+GB) and Ut is error term. The standard approach considers consump- tion as a function of disposable income via concept of private wealth. for REh subsequent conditions must be hold. own limitations. This paper serves as an attempt to extend the existing area of this research. Emphasizes is given to the use of less restrictive model for the investigation of REH. The rest balance of paper is designed as: part two explains the specification of the model, part three explains the variables and data sources, part four discuss the empirical methodology, part five investigates and interprets the empirical results. Finally, part six presents the conclusions of the study and also provides some policy implications. 2. Specification of the Model There are two types of consumption function, discussed in the literature, to check the validity of REH. After discussing those studies that extended the consumption function models, methodology for the present is discussed. REH can be checked by using two forms of consumption functions, Structural consumption function and Euler equation consumption function. Several studies validate REH 3 and several invalidate 4 it. For now structural consumption function is use to check the validity of REH and Euler equation consumption is on future agenda for researcher. 2.1 Structural Consumption Function Ricardian equivalence is rejected by Feldstein (1982) by using follo i g eq ation; ttttttttt eDaTRaTaGaSSWaWaYaaC ++++++++= 76543210 here C stands for total consumer expenditure, Y is current income, W indicates market value of privately owned wealth, SSW is value of future social security benefits, T symbolizes total tax revenue, TR shows government transfers to individuals,D is total government debt and t e is error term. 3 Khalid (1996), Rockerbie (1997), Cardia (1997), Lucke (1999), Drakos (2001, 2003), Sachsida & Carneiro (2001), Giorgioni & Holden (2001), Walker (2002), Kaadu & Usukula (2004), Safa & Siddiq (2005), Cuaresma & Reitschuler (2007). 4 Haq & Montiel (1987), Whelan (1991), Kazmi (1992, 1994), Ghatak & Ghatak (1996), Abimanyu (1998), Carlos (2001), Khan & Ashraf (2003), Ricciuti & Laurea (2003), Malengier & Pozzi (2004), Gray & Stone (2005), Gracia & Ramajo (2005), Nipple (2006), Apergis & Lyroudi (2006), Afonso (2008), Siddiki (2008), Waqas et al. (2011). To check the validity of REH this function requires certain restrictions that must be fulfilled. 4 5 6 3 2 7 0 , 0 , 0 , 0 ,a a a a a a< = = = = Aschauer (1985) criticized Feldstein model and argued that the use of current income as endogenous variable was the reason of endogenity in this model. No doubt, Feldstein used one lagged values of income and taxes as instrumental variable to remove endogenity but this problem may not be removed by using these instruments. Seater (1993) criticized the inconsistent criteria used by Feldstein for inferring the results. Along with some weakness Feldstein work pr vides sound simplification about REH. In 1983 Kormendi introduced “consolidated approach” which has a plus point that this model is based on permanent income hypothesis. tttttttttt UGBaGINTaREaTXaTRaWaGSaYaaPC +++++++++= 876543210 Where PC is private consumption, Y stands for current total income, GS represents total government spending on goods and services, W symbolizes total wealth, TR is transfers, TX is tax revenue, REH is corporate retained earnings, GINT is government interest payment on outstanding debt, GB demonstrates market value of outstanding government debt and t U is error term. Following restrictions must be fulfilled for the validity of REH. 2 4 5 6 7 8 0 , 0a a a a a a< = = = = = After “Consolidated approach” Kormendi introduced a “Standard approach” which considers that consumption is determined by disposable income (Yd), total wealth plus government debt (W+GB) and t U is error ter . The standard approach considers consumption as a function of disposable income via concept of private wealth. ttttttttttt UGBaGINTaREaTXaTRaWaGBWaYDaaPC ++++++++++= 876543210 )( For REH subsequent conditions must be hold. To check the validity of REH this function requires certain restrictions that must be fulfilled. 4 5 6 3 2 7 0 , 0 , 0 , 0 ,a a a a a a< = = = = Aschauer (1985) criticized Feldstein model and argued that the use of current income as endogenous variable was the reason of endogenity in this model. No doubt, Feldstein used one lagg d values of income and taxes as instrumental v riable t remove endogenity but this problem may not be removed by using these instruments. Seater (1993) criticized the inconsistent criteria used by Feldstein for inferring the results. Along with some weakness Feldstein work provides sound simplification about REH. In 1983 Kormendi introduced “consolidated approach” which has a plus point that this model is based on permanent income hypothesis. tttttttttt UGBaGINTaREaTXaTRaWaGSaYaaPC +++++++++= 876543210 Where PC is private consumption, Y stands for current total income, GS represents total government spending on goods and services, W sy bolizes total wealth, TR is transfers, TX is tax revenue, REH is corp ra retain d e rnings, GINT is government interest p yment on out t nding debt, GB demonstrates market value of outstanding government debt and t U is error term. Following restrictions must be fulfilled for the validity of REH. 2 4 5 6 7 8 0 , 0a a a a a a< = = = = = After “Consolidated approach” Kormendi introduced a “Standard approach” which considers that consumption is determined by disposable income (Yd), total wealth plus government debt (W+GB) and t U is error term. The standard approach considers consumption as a function of disposable income via concept of private wealth. ttttttttttt UGBaGINTaREaTXaTRaWaGBWaYDaaPC ++++++++++= 876543210 )( For REH subsequent conditions must be hold. To check the validity of REH this function requires certain restrictions that must be fulfilled. 4 5 6 3 2 7 0 , 0 , 0 , 0 ,a a a a a a< = = = = Aschauer (1985) criticized F ldstein model and arg ed that the use of current income as e d genous variable was the rea on of endogen ty in this model. No doubt, Feldstein u ed one lagged values of income and taxes as instr mental variable to remove endogenity but this problem may not be r moved by using these instrument . Seater (1993) criticized the inconsiste t c iteria used by Felds ein for inferring the results. Along with some weakness Feldstein work provides sound simplification about REH. In 1983 Kormendi introduced “consolidated approach” which has a plus point that this model is based on permanent income hypothesis. tttttttttt UGBaGINTaREaTXaTRaWaGSaYaaPC +++++++++= 876543210 Where PC is private consumption, Y stands for c rrent total income, GS represents total government spending on oods and services, W symbolizes total wealth, TR is transfers, TX is tax revenue, REH is corporate retained earnings, GINT is government interest payment on outstanding debt, GB demonstrates market value of outstanding government debt and t U is error term. Following restrictions must be fulfilled for the validity of REH. 2 4 5 6 7 8 0 , 0a a a a a a< = = = = = After “Consolidated approach” Kormendi introduced a “Standard approach” which considers that consumption is determined by disposable income (Yd), total wealth plus government debt (W+GB) and t U is error term. T e standard approach considers consumption as a function of disposable income via concept of private wealth. ttttttttttt UGBaGINTaREaTXaTRaWaGBWaYDaaPC ++++++++++= 876543210 )( For REH subsequent conditions must be hold. To check the validity of REH this function requires certain restrictions that must be fulfilled. 4 5 6 3 2 7 0 , 0 , 0 , 0 ,a a a a a a< = = = = Aschauer (1985) criticized Feldstein model and argued that the use of current income as endogenous variable was the reason of endogenity in this model. No doubt, Feldstein used one lagged values of income and taxes as instrumental variable to remove endogenity but this problem may not be removed by using these instruments. Seater (1993) criticized the inconsistent criteria used by Feldstein for inferring the results. Along with some weakness Feldstein work prov des sound simplification about REH. In 1983 Kormendi introduced “consolidated approach” which has a plus point that this model is based on permanent income hypothesis. tttttttttt UGBaGINTaREaTXaTRaWaGSaYaaPC +++++++++= 876543210 Where PC is private cons mption, Y stands for current total income, GS represents total government spendi g on goods a d services, W symbolizes tot l wealth, TR is transfers, TX is t x revenue, REH is corporate retained earnings, GINT is government interest payment on outstanding debt, GB demonstrates arket value of outstanding governme t debt a d t U is error term. Following restrictions must be fulfilled for the validity of REH. 2 4 5 6 7 8 0 , 0a a a a a a< = = = = = After “Consolidated ap roach” Kormendi introduced a “Standar approach” which considers that consumption is determined by disposable income (Yd), total wealth plus government debt (W+GB) and t U is error term. The st ndard approach considers consumption as a function of disposable income via concept of private wealth. ttttttttttt UGBaGINTaREaTXaTRaWaGBWaYDaaPC ++++++++++= 876543210 )( For REH sub equent conditions must be hold. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011164 Modigliani and Sterling (1986) criticized the low value of coefficient of income and high value of transfers variable in Kormendi’s approach. They claimed that a raise in transfers may be negative tax; therefore according to REh transfers should not have any effect on private consumption. Secondly, he used an unsuitable deflator (all variables were de- flated by implicit price deflator for Net National Product). Thirdly they claimed that that Second World War period must be debarred from the sample during the analysis done by Kormendi. feldstein and Elmendorf (1990) suggested that Kormendi must use of ratio specification to diminish co linearity among Net National Product (NNP) and fiscal variables. Sec- ondly, they suggested the use of instrumental variables in order to reduce the endogenity among NNP and fiscal variables. By using the model of Kormendi and past values of the endogenous variables lagged 2, 3 and 4 years, feldstein and Elmendorf results rejected REh. In 1986 Modigliani and Sterling introduced a consumption function by putting the ac- cent on life cycle theory and assumed the expectations as distributed lag of past vari- ables. Where L is equal to 5, TL indicates taxes net of transfers plus government net real ex-post domestic interest payments. DEf shows government budget deficits and for REh b1 = –b0 and ∑di = ∑ci must be hold. In 1987 Bernheim introduced two models to test REh, where C is real per capita con- sumption, X is vector of other exogenous variables, r is interest rate, Y-TX is disposable income, TX-G-rGB is government surplus and et is error term. In second equation he deals disposable income without subtracting for taxes, for REh a2 = a1 and (b2 = 0) and for Keynesian view a2 = 0 and b2 = –b1 must be hold. for inter- national comparison Bernheim introduced following equation, where Y is real gross do- mestic growth, Pop is population growth and GB is domestically held government debt. Pereleman and Pestieau (1993) used disposable income, government budget deficit, wealth and government debt in order to check the validity of REh. for REh, restric- 2 4 5 6 7 8 0 , 0a a a a a a= = − = = = > Modigliani and Sterling (1986) criticized the low value of coefficient of income and high value of transfers variable in Kormendi’s approach. They claimed that a raise in transfers may be negative tax; therefore according to REH transfers should not have any effect on private consumption. Secondly, he used an unsuitable deflator (all variables were deflated by implicit price deflator for Net National Product). Thirdly they claimed that that Second World War period must be debarred from the sample during the analysis done by Kormendi. Feldstein and Elmendorf (1990) suggested that Kormendi must use of ratio specification to diminish co linearity among Net National Product (NNP) and fiscal v riables. Secondly, th y suggested the use of instrumental variables in order to reduce the endogenity among NNP and fiscal variables. By using the model of Kormendi and past values of the endogenous variables l gged 2, 3 and 4 years, Feldstein and Elmendorf results rej cted REH. In 1986 Modigliani and Sterling introduced a consumption function by putting the accent on life cycle theory and assumed the expectations as distributed lag of past variables. ∑ ∑ = = −−− ++−+++= L i L i tittitittt UDEFdTLYCGBbWbaC 1 1 1110 )( Where L s equa to 5, TL i ic tes taxes n t of transfers plus government net real ex-post domestic interest payments. DEF shows government budget deficits and for REH 01 bb −= and ∑ ∑= ii cd must be hold. In 1987 Bernheim introduced two models to test REH, where C is real per capita consumption, X is vector of other exogenous variables, r is interest rate, Y-TX is disposable income, TX-G-rGB is government surplus and t e is error term. ttttttrttttt eXWGBGGBrGTXTXYC +++++−−+−+= _ 543210 )()( ααααααα _ 0 1 2 3 4 5 ( ) t t t t r t t t t t t C Y TX G r GB G GB W X eβ β β β β β β= + + − − + + + + + 2 4 5 6 7 8 0 , 0a a a a a= = − = = = > odigliani and Sterling (1986) criticized the low value of c efficient of income and high value of transfers variable in Kormendi’s approach. They claimed that a raise in transfers may be negative tax; therefore according to REH transfers should not have any effect on private consumption. Secondly, he used an unsuitable deflator (all variables w re deflated by implicit price deflator for Net National Product). Thirdly they claimed that that Second World War period must be debarred from the sample during the analysis done by Kormendi. Fel stein and Elmen orf (1990) suggested th t Kormend must use of ratio specificatio to diminish co linearity among Net National Product (NNP) and fiscal variables. Secondly, they suggested the use of instrumental variables in order to reduce the endogenity among NNP a d fiscal variables. By using the model of Kormendi and past values of the e dogenous variables lagged 2, 3 and 4 years, Feldstein and Elmendorf results rejected REH. In 1986 Modigliani and Sterling introduced a consumption function by putting the accent on life cycle theory and assumed the expectations as distributed lag of past variables. ∑ ∑ = = −−− ++−+++= L i L i tittitittt UDEFdTLYCGBbWbaC 1 1 1110 )( Where L is equal to 5, TL indicates taxes net of transfers plus government net real ex-post domestic interest payments. DEF shows government budget deficits and for REH 01 bb −= and ∑ ∑= ii cd must be hold. In 1987 Bernheim introduced two models to test REH, wher C is real per capita consumption, X is vector of other exogenous variables, r is interest rate, Y-TX is disposable i come, TX-G-rGB is gov ment surplus an t e is error term. ttttttrttttt eXWGBGGBrGTXTXYC +++++−−+−+= _ 543210 )()( ααααααα _ 0 1 2 3 4 5 ( ) t t t t r t t t t t t C Y TX G r GB G GB W X eβ β β β β β β= + + − − + + + + + 2 4 5 6 7 8 0 , 0a a a a a a= = − = = = > Modigliani and Sterling (1986) criticized the low value of coefficient of income and high value of transfers variable in Kormendi’s approach. They claimed that a raise in transfers may be negative tax; therefore according to REH transfers should not have any effect on private consumption. S condly, h used an u suitable deflator (all variables were deflated by implicit price deflator for Net National Product). Thirdly they claimed that that Second World War period must be debarred from the sample during the analysis done by Kormendi. Feldstein and Elmendorf (1990) suggested that Kormendi must use of ratio specification to diminish co linearity among Net National Product (NNP) and fiscal variables. Secondly, they suggested the use of instrumental variables in order to reduce the endogenity among NNP a fiscal vari bles. By using the model of Kormendi and past values of the endog nous variables lagged 2, 3 and 4 years, Feldstein and Elmendorf results rejected REH. In 1986 Modigliani and Sterling introduced a consumption function by putting the accent on life cycle theory and assumed the expectations as distributed lag of past variables. ∑ ∑ = = −−− ++−+++= L i L i tittitittt UDEFdTLYCGBbWbaC 1 1 1110 )( here L i l in ta net of transfers plus govern net real ex-post domestic interest pay ents. DEF shows government budget deficits and for REH 01 bb −= a d ∑ ∑= ii cd mus be hold. In 1987 Bernheim introduced two models to test REH, where C is real per capita consumption, X is vector of other exogenous variables, r is interest rate, Y-TX is disposable income, TX-G-rGB is government surplus and t e is error term. ttttttrttttt eXWGBGGBrGTXTXYC +++++−−+−+= _ 543210 )()( ααααααα _ 0 1 2 3 4 5 ( ) t t t t r t t t t t t C Y TX G r GB G GB W X eβ β β β β β β= + + − − + + + + + In second equation he deals disposable income without subtracting for taxes, for REH 2 1 2 ( 0)andα α β= = and for Keynesian view 2 0α = and 12 ββ −= must be hold. For international comparison Bernheim introduced following equation, where Y is real gross domestic growth, Pop is population growth and GB is domestically held government debt. 1 2 3 4 5 6 7 defC G GB W Y Pop e Y Y Y Y Y β β β β β β β= + + + + + + + P releman and Pestieau (1993) used disposable income, government budget deficit, wealth and government debt in order to check the validity of REH. For REH, restrictions 0 21 =+αα and 0 4 =α must be hold and for Keynesian view 0 2 =α must be fulfilled. eGBWDEFTXYC ++++−+= 43210 )( Study rejected both pure Ricardian and Keynesian view because coefficient of deficit is negative. After discussing the different structural consumption functions, their weaknesses their contributions in the literature, the present study estimates following structural consumption function. Dependant variable is private consumption (PC), while independent variables are disposable income (YD), government expenditure (GE), total wealth (W), tax revenue (TR), government debt (GD), government budget deficit (GBD) and t U is error term. This model is more familiar with Kormendi’s (1983) and Pereleman and Pestieau’s (1993) models. Keeping in the views of Modigliani and Sterling (1986) a transfer variable is not included in our model because they argued that transfers may be treated as negative tax. 0 1 2 3 4 5 6 t t t t t t t t C YD G E W T R G D G BD U α α α α α α α = + + + + + + + To hold REH following restrictions must be fulfilled. 2 4 5 1 6 3 5 0, 0, 0, 0,α α α α α α α< = = + = = 2 4 5 6 7 8 0 , 0a a a a a a= = − = = = > Modigliani and Sterling (1986) criticized the low value of coefficient of income and high value of transfers variable in Kormendi’s approach. They claimed that a raise in transfers may be negative tax; therefore according to REH transfers should not have any effect on private consumption. Secondly, he used an unsuitable deflator (all variables were deflated by implicit price deflator for Net National Product). Thirdly they claimed that that Second World War period must be debarred from the sample during the analysis done by Kormendi. Feldstein and Elmendorf (1990) suggested that Kormendi must use of ratio specification to diminish co linearity among Net National Product (NNP) and fiscal variables. Secondly, they suggested the use of instrumental variables i order to reduce the endogenity among NNP and fiscal variables. By using the model of Kormendi and past values of the endogenous variables lagged 2, 3 and 4 years, Feldstein and Elmendorf results rejected REH. In 1986 Modigliani and Sterling introduced a consumption function by putting the accent on life cycle theory and assumed the expectations as distributed lag of past variables. ∑ ∑ = = −−− ++−+++= L i L i tittitittt UDEFdTLYCGBbWbaC 1 1 1110 )( Where L is equal to 5, TL indicates taxes net of transfers plus government net real ex-post domestic interest payments. DEF shows government budget deficits and for REH 01 bb −= and ∑ ∑= ii cd must be hold. In 1987 Bernheim introduced two models to test REH, where C is real per capita consumption, X is vector of other exogenous variable , r is interest ate, Y-TX is disposable income, TX-G-rGB is government surplus and t e is error term. ttttttrttttt eXWGBGGBrGTXTXYC +++++−−+−+= _ 543210 )()( ααααααα _ 0 1 2 3 4 5 ( ) t t t t r t t t t t t C Y TX G r GB G GB W X eβ β β β β β β= + + − − + + + + + M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 165 tions a1 = a2 = 0 and a4 = 0 must be hold and for Keynesian view a2 = 0 must be ful- filled. Study rejected both pure Ricardian and Keynesian view because coefficient of deficit is negative. After discussing the different structural consumption functions, their weaknesses their contributions in the literature, the present study estimates following structural con- sumption function. Dependant variable is private consumption (PC), while independent variables are disposable income (YD), government expenditure (GE), total wealth (W), tax revenue (TR), government debt (GD), government budget deficit (GBD) and Ut is error term. This model is more familiar with Kormendi’s (1983) and Pereleman and Pes- tieau’s (1993) models. Keeping in the views of Modigliani and Sterling (1986) a transfer variable is not included in our model because they argued that transfers may be treated as negative tax. To hold REh following restrictions must be fulfilled. first restriction states that government expenditure must be less than zero which depicts that as government expenditure increases private consumption will decrease. Second restriction demonstrates that tax revenue must be equal to zero which means that deficit financing has no affect on private consumption. Third restriction shows that government debt must be equal to zero which affirms that government debt has no impact on private consumption. fourth restriction states that disposable income plus government budget deficit must be equal to zero. Moreover, wealth must be equal to government debt which describes that consumers purchase same amount of bonds as government do deficit fi- nancing. This restriction also depicts that in response to tax cut consumers not increase their consumption but increase their savings. 3. VariaBLes anD Data sources The study used time series data of Pakistan for the period of 1973-2009, collected from International financial Statistics (IfS) and different Economic Surveys of Pa- kistan. Government expenditure, private consumption expenditure, tax revenue, government debt, disposable income5, government budget deficit and wealth6 are 5 A proxy variable of Gross National Income. 6 By following Garcia and Ramajo (2003) this is a proxy variable computed by adding Government debt and M2. In second equation he deals disposable income without subtracting for taxes, for REH 2 1 2 ( 0)andα α β= = and for Keynesian view 2 0α = and 12 ββ −= must be hold. For international comparison Bernheim introduced following equation, where Y is real gross domestic growth, Pop is population growth and GB is domestically held government debt. 1 2 3 4 5 6 7 defC G GB Y Pop e Y Y Y Y Y β β β β β β β= + + + + + + + Pereleman and Pestieau (1993) used disposable income, government budget deficit, wealth and government debt in order to check the validity of REH. For REH, restrictions 0 21 =+αα and 0 4 =α must be hold and for Keynesian view 0 2 =α must be fulfilled. eGBWDEFTXYC ++++−+= 43210 )( Study rejected both pure Ricardian and Keynesian view because coefficient of deficit is negative. After discussing the different structural consumption functions, their weaknesses their contributions in the literature, the present study estimates following structural consumption function. Dependant variable is private consumption (PC), while independent variables are disposable income (YD), government expenditure (GE), total wealth (W), tax revenue (TR), government debt (GD), government budget deficit (GBD) and t U is error term. This odel is more familiar with Kormendi’s (1983) and Pereleman and Pestieau’s (1993) models. Keeping in the views of Modigliani and S erling (1986) a transfer variable is not included in o r model because they argued that transfers may be treated as negative tax. 0 1 2 3 4 5 6 t t t t t t t t C YD G E W T R G D G BD U α α α α α α α = + + + + + + + To hold REH following restrictions must be fulfilled. 2 4 5 1 6 3 5 0, 0, 0, 0,α α α α α α α< = = + = = In second equation he deals disposable income without subtracting for taxes, for REH 2 1 2 ( 0)andα α β= = and for Keynesian view 2 0α = and 12 ββ −= must be hold. For international comparison Bernheim introduced following equation, where Y is real gross domestic growth, Pop is population growth and GB is domestically held government debt. 1 2 3 4 5 6 7 defC G GB W Y Pop e Y Y Y Y Y β β β β β β β= + + + + + + + Pereleman and Pestieau (1993) used disposable income, government budget deficit, wealth and government debt in order to check the validity of REH. For REH, restrictions 0 21 =+αα and 0 4 =α must be hold and for Keynesian view 0 2 =α must be fulfilled. eGBWDEFTXYC ++++−+= 43210 )( ααααα y t t i ffi i t fi is negative. After discussing the different structural consumption functions, their weaknesses their contributions in the literature, the present tudy estimates following structural consumption function. Dependant variable is private consumption (PC), while independent variables are disposable income (YD), government expenditure (GE), total wealth (W), tax revenue (TR), government debt (GD), gove nm nt budg t defici (GBD) an t U is error term. This model is more familiar with Kormendi’s (1983) and Pereleman and Pestieau’s (1993) models. Keeping in the views of Modigliani and Sterling (1986) a transfer variable is not included in our model because they argued that transfers may be treated as negative tax. 0 1 2 3 4 5 6 t t t t t t t t C YD G E W T R G D G BD U α α α α α α α = + + + + + + + To hold REH following restrictions must be fulfilled. 2 4 5 1 6 3 5 0, 0, 0, 0,α α α α α α α< = = + = = In second equation he deals disposable income without subtracting for taxes, for REH 2 1 2 ( 0)andα α β= = and for Keynesian view 2 0α = and 12 ββ −= must be hold. For international comparison Bernheim introduced following equation, where Y is real gross domestic growth, Pop is population growth and GB is domestically held government debt. 1 2 3 4 5 6 7 defC G GB W Y Pop e Y Y Y Y Y β β β β β β β= + + + + + + + Pereleman and Pestieau (1993) used disposable income, government budget deficit, wealth and government debt in order to check the validity of REH. For REH, restrictions 0 21 =+αα and 0 4 =α must be hold and for Keynesian view 0 2 =α must be fulfilled. eGBWDEFTXYC ++++−+= 43210 )( Study rejected both pure Ricardian and Keynesian view because coefficient of deficit is negative. After discussing the different structural consumption functions, their weaknesses their co tributions in the literature, the presen study stim tes following structural consumption function. Dependant variable is p ivate consumption (PC), while independent variables are disposa le income (YD), government x enditure (GE), total wealth (W), tax revenue (TR), government debt (GD), government budget deficit (GBD) and t U is error term. This model is more familiar with Kormendi’s (1983) a d Pereleman and Pestieau’s (1993) models. Keeping in the views of Modigliani and Sterling (1986) a transfer variable is not included in our model b cause they a gued that transfers may be treated as negative tax. 0 1 2 3 4 5 6 t t t t t t t t P C YD G E W T R G D G BD U α α α α α α α = + + + + + + + To hold REH following restrictions must be fulfilled. 2 4 5 1 6 3 5 0, 0, 0, 0,α α α α α α α< = = + = = In second equation he deals disposable income without subtracting for taxes, for REH 2 1 2 ( 0)andα α β= = and for Keynesian view 2 0α = and 12 ββ −= must be hold. For international comparison Bernheim introduced following equation, where Y is real gross domestic growth, Pop is population growth and GB is domestically held government debt. 1 2 3 4 5 6 7 defC G GB W Y Pop e Y Y Y Y Y β β β β β β β= + + + + + + + Pereleman and Pestieau (1993) used disposable income, government budget deficit, wealth and government debt in order to check the validity of REH. For REH, restrictions 0 21 =+αα and 0 4 =α must be hold and for Keynesian view 0 2 =α must be fulfilled. eGBWDEFTXYC ++++−+= 43210 )( Study rejected both pure Ricardian and Keynesian view because coefficient of deficit is negative. After discussing the different structural consumption functions, their weaknesses their contributions in the literature, the present study estimates following structural consumption function. Dependant variable is private consumption (PC), while independent variables are disposable income (YD), government expenditure (GE), total wealth (W), tax revenue (TR), government debt (GD), government budget deficit (GBD) and t U is error term. This model is more familiar with Kormendi’s (1983) and Pereleman and Pestieau’s (1993) models. Keeping in the views of Modigliani and Sterling (1986) a transfer variable is not included in our model becau h y argued that transfers may be treated as negative tax. 0 1 2 3 4 5 6 t t t t t t t t C YD G E W T R G D G BD U α α α α α α α = + + + + + + + To hold REH following restrictions must be fulfilled. 2 4 5 1 6 3 5 0, 0, 0, 0,α α α α α α α< = = + = = ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011166 the variables used in this analysis. All the variables are transformed into real per capita. 4. eMPiricaL MetHoDoLoGy It is very important to check the long run and short run dynamics among the variables, before the estimation of any time series model. In econometric literature there are lots of uni-variate7 and multi-variate8 techniques to check the cointegration among the var- iables. Before applying any cointegration technique, firstly we have to detect order of integration. Mostly time series data is non-stationary and in order to beware spurious regression results researchers used different unit root test. 4.1 Unit Root Test 4.1.1 Augmented Dickey Fuller (ADF) unit root test Dickey and fuller, after Dicky fuller unit root test, suggested a new test to check unit root, ADf. In order to remove the autocorrelation this test includes additional lagged terms of the dependent variable as a one of the independent variable. Mostly the time series data have a trend, but ADf test give following three possibilities. (1) (2) (3) Equation 1 states the possibility when no trend and no intercept found in the data, equa- tions 2 states the possibility when data has intercept only 3 states the possibility when data has both intercept and trend. D eterministic elements a0 and a2t differentiate the above three equation form each other. While using ADf test there are two important things which a researcher has to keep in his mind. Specify the lagged first difference terms. If we select zero lagged difference this will be Df test. In ADf, in order to remove serial correlation among residuals, sufficient lags are included. Secondly, when we choose the different possibilities of ADf, discussed above, their critical values also changed. McKinnon (1991) table of critical values is used to check the acceptance or rejection of null hypothesis. 7 Engle&Granger, (1987) and Phillips& hansen’s fMOLS procedures (1990). 8 Johansen, (1988), Johansen & Juselius, (1990),and Johansen’s (1995) and Auto-Regressive Distributed Lag (ARDL), proposed by Pesaran& Shin, (1995, 1998), Pesaran et.al., (1996), and Pesaran et.al., (2001) t me eri s dat is non-stationary and i order to beware spurious regr ssion esults researchers used different unit root test. 4.1 Unit Root Test 4.1.1 Augmen ed Dickey Fuller (ADF) unit root test Dickey and Fuller, after Dicky Fuller unit root test, suggested a new test to check unit root, ADF. In order to remove the autocorrelation this test includes additional lagged terms of the dependent variable as a one of the indep ndent variable. Mostly the time series data have a trend, but ADF test give following three possibilities. 1 1 (1) t t i t t Z Z Z eφ γ − − ∆ = + ∆ +∑ KK 0 1 1 (2) t t i t t Z Z Z eα φ γ − − ∆ = + + ∆ +∑ KK 0 1 2 1 (3) t t i t t Z Z a t Z eα φ γ − − ∆ = + + + ∆ +∑ KK Equation 1 states the possibility when o trend and no interc pt found i the data, equations 2 states the possibility when data has intercept only 3 states the possibility when data has both intercept and trend. Deterministic elements 0 α and 2 a t differentiate the above three equation form each other. While using ADF test there are two important things which a res arch r has t keep n his mind. Specify th lagged first diffe en e terms. If we select z ro lagged difference this will be DF test. In ADF, in order to remove serial correlation among residuals, sufficient lags are included. Secondly, when we choose the different possibilities of ADF, discussed above, their critical values also changed. McKinnon (1991) table of critical values is used to check the acceptance or rejection of null hypothesis. 4.1.2 The Phillips-Perron Unit Root Test. The Dickey-Fuller test is based on the assumption that the error terms are statistically independent and have a constant variance. Phillips and Perron (1988) introduced a new test of unit root in which they used mild assumptions as compared to Dickey and Fuller. time series data is non-stationary and in order to beware spurious regression results researchers used different unit root test. 4.1 Unit Root Test 4.1.1 Augmented Dickey Fuller (ADF) unit root test Dickey and Fuller, after Dicky Fuller unit root test, suggested a new test to check unit root, ADF. In order to remove the autocorrelation this test includes additional lagged terms of the dependent variable as a one of the independent variable. Mostly the time series data have a trend, but ADF test give following three possibilities. 1 1 (1) t t i t t Z Z Z eφ γ − − ∆ = + ∆ +∑ KK 0 1 1 (2) t t i t t Z Z Z eα φ γ − − ∆ = + + ∆ +∑ KK 0 1 2 1 (3) t t i t t Z Z a t Z eα φ γ − − ∆ = + + + ∆ +∑ KK Equation 1 states the possibility when no trend and no intercept found in the data, equations 2 states the possibility whe data has intercept only 3 stat s the possibility when data has both intercept nd trend. Deterministic elements 0 α and 2 a t differentiat the above thre equation form each other. While using ADF test there are two important things which a researcher has to keep in his mind. Specify the lagged first difference terms. If we select zero lagged difference this will be DF test. In ADF, in order to remove serial correlation among residuals, sufficient lags are included. Secondly, when we choose the different possibilities of ADF, discussed above, their critical values also changed. McKinnon (1991) table of critical values is used to check the acceptance or rejection of null hypothesis. 4.1.2 The Phillips-Perron Unit Root Test. The Dickey-Fuller test is based on the assumption that the error terms are statistically independent and have a constant variance. Phillips and Perron (1988) introduced a new test of unit root in which they used mild assumptions as compared to Dickey and Fuller. ti e series data is non-stationary and in order to be are spurious regression results researchers used diff r nt unit root test. 4.1 nit oot Test 4.1.1 ug ented ickey Fuller ( F) unit root test ickey and Fuller, after icky Fuller u it root test, suggested a ne test to check unit root, A F. In order to re ove the autocorrelation this test includes additional lagged ter s of the dependent variable as a one of the independent variable. ostly the ti e series data have a trend, but A F test give follo ing three possibilities. 1 1 (1) t t i t t Z Z Z eφ γ − − ∆ = + ∆ + 0 1 1 (2) t t i t t Z Z Z eφ γ − − ∆ = + + ∆ + 0 1 2 1 (3) t t i t t Z Z a t Z eφ γ − − ∆ = + + + ∆ + Equation 1 states the possibility hen no trend and no intercept found in the data, equations 2 states the possibility hen data has intercept only 3 states the possibility hen data has both intercept and trend. eter inistic ele ents 0 and 2 a t differentiate the above three equation for each other. hile using A F test there are t o i portant things hich a researcher has to keep in his ind. Specify the lagged first difference ter s. If e select zero lagged difference this ill be F test. In A F, in order to re ove serial correlation a ong residuals, sufficient lags are included. Secondly, hen e choose the different possibilities of A F, discussed above, their critical values also changed. c innon (1991) table of critical values is used to check the acceptance or rejection of null hypothesis. 4.1.2 The hillips- erron nit oot Test. The ickey-Fuller test is based on the assu ption that the error ter s are statistically independent and have a constant variance. Phillips and Perron (1988) introduced a ne test of unit root in hich they used ild assu ptions as co pared to ickey and Fuller. M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 167 4.1.2 The Phillips-Perron Unit Root Test. The Dickey-fuller test is based on the assumption that the error terms are statistically in- dependent and have a constant variance. Phillips and Perron (1988) introduced a new test of unit root in which they used mild assumptions as compared to Dickey and fuller. Consider AR(1) process; (4) PP test is the modification of ADf test it just make a correction of the t-statistic of Z’s coefficient by using comparatively less restrictions than ADf, in order to remove serial correlation. McKinnon (1991) critical values are also used for this test. Moreover, this test also has the same three possibilities which ADf has; intercept, intercept and trend and no intercept and no trend. 4.1.3 The Kwiatkowski, Phillips, Schmidt, and Shin test (KPSS). This test is different from other unit root tests because it is based on the residuals obtain from ordinary least square method. Suppose we have endogenous variable zt and an ex- ogenous variable wt. (5) The LM statistic is; (6) Where at zero frequency f0 is an estimator of the residual spectrum and S(t) shows the cumulative residual function; , which is based on the residuals . The calculation of the estimator δ is based on the OLS method. 4.2 Johansen Co-Integration approach After the pioneer work of Granger (1981) about cointegration, many studies9 elaborated this concept. Johansen (1988) introduced a new approach of checking the cointegration between more than two series. It removes all the drawbacks, which Engle-Granger ap- 9 Engle and Granger (1987), Engle and Yoo (1987), Stock and Watson (1988), Phillips (1986& 1987), Phillips and Ouliaris (1990) and Johansen (1988, 1991, 1995) Consider AR(1) process; 1 0 1 (4) t t t Z Z eα γ − − ∆ = + + KK PP test is the modification of ADF test it just make a correction of the t-statistic of Z’s coefficient by using comparatively less restrictions than ADF, in order to remove serial correlati n. McKinnon (1991) critical val es are also used fo this test. Moreover, this test also has the same three possibilities which ADF has; intercept, intercept and trend and no intercept and no trend. 4.1.3 The Kwiatkowski, Phillips, Schmidt, and Shin test (KPSS). This test is different from other unit root tests because it is based on the residuals obtain from ordinary least square method. Suppose we have endogenous variable t z and an exogenous variable t w . (5) t t t z w uδ′= + KK The LM statistic is; ) 2 2 0 (6) t S t LM T f =∑ KK Where at zero frequency 0 is an estimator of the residual spectrum and S(t) shows the cumulative residual function; ( ) 1 ˆ t r r S t u = =∑ , which is based o ( ) ˆ ˆ 0 t t t u z wδ′= − The calculation of the estimator δ is based on the OLS method. onsider AR(1) process; 1 0 1 (4) t t t Z Z eα γ − − ∆ = + + KK test is the modific tio of F test it just make a correctio of t e t-statistic of Z’s coefficient by using comparatively less restrictions than ADF, in order to remove serial correlation. McKinnon (1991) critical values are also used for this test. Moreover, this test also has the same three possibilities which ADF has; intercept, intercept and trend and no intercept and no trend. 4.1.3 The Kwiatkowski, Phillips, Schmidt, and Shin test (KPSS). This test is different from other unit root tests because it is based on the residuals obtain from ordinary least square method. Suppose we have endogenous variable t z and an exogenous variable t w . (5) t t t z w uδ′= + KK The LM statistic is; ) 2 2 0 (6) t S t LM T f =∑ KK Where at zero frequency 0 is an estimator of the residual spectrum and S(t) shows the cumulative residual function; ( ) 1 ˆ t r r S t u = =∑ , which is based on the residuals ) ˆ ˆ 0 t t t u z wδ′= − . The calculation of the estimator δ is based on the OLS method. Consider AR(1) process; 1 0 1 (4) t t t Z Z eα γ − − ∆ = + + KK PP test is the modification of ADF test it just make a correction of the t-statistic of Z’s coefficient by using comparatively less restrictions than ADF, in order to remove serial correlation. McKinnon (1991) critical values are also used for this test. Moreover, this test also has the same three possibilities which ADF has; intercept, intercept and trend and no inte cept a d no tre d. 4.1.3 The Kwiatkowski, Phillips, Schmidt, and Shin test (KPSS). This test is different from other unit root tests because it is based on the residuals obtain from ordinary least square method. Suppose we have endogenous variable t z and an exogenous variable t w . (5) t t t z w uδ′= + KK The LM statistic is; ) 2 2 0 (6) t S t LM T f =∑ KK Where at zero frequency 0 is an estimator of the residual spectrum and S(t) shows the cumulative residual function; ( ) 1 ˆ t r r S t u = =∑ , which is based on the residuals ) ˆ ˆ 0 t t t u z wδ′= − . The calculation of the estimator δ is based on the OLS method. Consider AR(1) process; 1 0 1 (4) t t t Z Z eα γ − − ∆ = + + KK PP test is the modification of ADF test it just make a correction of the t-statistic of Z’s coefficient by using comparatively less restrictions than ADF, in order to remove serial correla ion. McKinnon (1991) critical values are also used fo this test. Moreover, this test also has the same three ossibilities which ADF has; intercept, intercept and trend and no intercept and no trend. 4.1.3 The Kwiatkowski, Phillips, S hmidt, and Shin test (KPSS). This test is different from other unit root tests because it is based on the residuals obtain from ordinary least square method. Suppose we have endogenous variable t z and an exogenous variable t w . (5) t t t z w uδ′= + KK The L statistic is; ( ) 2 2 0 (6) t S t LM T f =∑ KK here at zero fre uency 0 is an esti ator of t e resi l s ectr d (t) s o s t e cumulative residual function; ( ) 1 ˆ t r r S t u = =∑ , which is based on the residuals ) ˆ ˆ 0 t t t u z wδ′= − . The calculation of the estimator δ is based on the OLS method. Consider AR(1) process; 1 0 1 (4) t t t Z Z eα γ − − ∆ = + + PP test is the odification of ADF test it just ake a correction of the t-statistic of Z’s coef icient by using co paratively less restrictions than ADF, in order to re ove serial correlation. cKinnon (1991) critical values are also used for this test. oreover, this test also has the sa e three possibilities which ADF has; intercept, intercept and trend and no intercept and no trend. 4.1.3 The K iatko ski, Phil ips, Sch idt, and Shin test (KPSS). This test is dif erent fro other unit root tests because it is based on the residuals obtain fro ordinary least square ethod. Suppose we have endogenous variable t z and an exogenous variable t w . (5) t t t z w uδ′= + The L statistic is; ) 2 2 0 (6) t S t L T f = here at zero frequency 0 is an esti ator of the residual spectru and S(t) shows the cu ulative residual function; ( ) 1 ˆ t r r S t u = = , whic the residuals ) ˆ ˆ 0 t t t u z wδ′= − . The calculation of the esti ator δ is based on the OLS ethod. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011168 proach has. In case of Johansen approach the ECM also extended into Vector Error Cor- rection Model (VECM). Now suppose that we have three endogenous variables, L, M and N. In matrix form this can be written as; (7) (8) In the context of VECM we can written as (9) Whereas, (10) and (11) P shows the 3×3 matrix, which depicts the true long run relationship between Yt = [Lt,Mt,Nt]. The P = fχ′, in which f shows the speed of adjustment towards equilibrium and long run coefficients matrix is χ′. In single equation case χ′Yt–1 is error correction term. To find out for multivariate case now assumes k = 2. So the model is (12) or we can say that; (13) for simplicity just analyze the first equation’s error correction part. The first row of P matrix is; (14) This can also be written as; (15) 4.2 Johansen Co-Integration approach After the pioneer work of Granger (1981) about cointegration, many studies 9 elaborated this concept. Johansen (1988) introduced a new approach of checking the cointegration between more than two series. It removes all the drawbacks, which Engle-Granger approach has. In case of Johansen approach the ECM also extended into Vector Error Correction Model (VECM). Now suppose that we have three endogenous variables, L, M and N. In matrix form this can be written as; ], , (7) t t t t Y L M N= KK 1 1 2 2 ....... (8) t t t k t k t Y Y Y Yβ β β µ − − − = + + + + K In the context of VECM we can written as 1 1 2 2 1 1 1 ....... (9) t t t k t k t t Y Y Y Y Y µ − − − − − − ∆ = Γ ∆ +Γ ∆ + +Γ ∆ +Π + K Whereas, 1 2 (1 ......... )( 1, 2,....., 1) (10) i k i kβ β βΓ = − − − − = − K and 1 2 (1 ......... ) (11) k β β βΠ = − − − − − KK Π shows the 3 3× matrix, which depicts the true long run relationship between [ ], , t t t t Y L M N= . The φχ ′Π = , in which φ shows the speed of adjustment towards equilibrium and long run coefficients matrix is χ ′ . In single equation case 1t Y − ′ is error correction term. To find out for multivariate case now assumes 2k = . So the model is t t-1 t-1 t 1 t-1 t-1 t t-1 t-1 (12) t L L L M M M e N N N ∆ ∆ ∆            ∆ = Γ ∆ + Π ∆ +            ∆ ∆ ∆       KK or we can say that; 9 Engle and Granger (1987), Engle and Yoo (1987), Stock and Watson (1988), Phillips (1986& 1987), Phillips and Ouliaris (1990) and Johansen (1988, 1991, 1995) 4.2 Johansen Co-Integration approach After the pioneer work of Granger (1981) about cointegration, many studies 9 elaborated this concept. Johansen (1988) introduced a new approach of checking the cointegration between more than two series. It removes all the drawbacks, which Engle-Granger approach has. In case of Johansen approach the ECM also extended into Vector Error Correction Model (VECM). Now suppose that we have three endogenous variables, L, M and N. In matrix form this can be written as; [ ], , (7) t t t t Y L M N= KK 1 1 2 2 ....... (8) t t t k t k t Y Y Y Yβ β β µ − − − = + + + + K In the context of VECM we can written as 1 1 2 2 1 1 1 ....... (9) t t t k t k t t Y Y Y Y Y µ − − − − − − ∆ = Γ ∆ +Γ ∆ + +Γ ∆ +Π + K Whereas, 1 2 (1 ......... )( 1, 2,....., 1) (10) i k i kβ β βΓ = − − − − = − K and 1 2 (1 ......... ) (11) k β β βΠ = − − − − − KK Π shows the 3 3× matrix, which depicts the true long run relationship between [ ], , t t t t Y L M N= . The φχ ′Π = , in which φ shows the speed of adjustment towards equilibrium and long run coefficients matrix is χ ′ . In single equation case 1t Y − ′ is error correction term. To find out for multivariate case now assumes 2k = . So the model is t t-1 t-1 t 1 t-1 t-1 t t-1 t-1 (12) t L L L M M M e N N N             ∆ = Γ ∆ + Π ∆ +            ∆ ∆ ∆       KK or we can say that; 9 Engle and Granger (1987), Engle and Yoo (1987), Stock and Watson (1988), Phillips (1986& 1987), Phillips and Ouliaris (1990) and Johansen (1988, 1991, 1995) . -I t ti ft r t i r r f r r ( ) t c i t r ti , st i s 9 l r t t is c c t. s ( ) i tr c r c f c c i t c i t r ti t r t t s ri s. It r s ll t r c s, ic l - r r r c s. I c s f s r c t ls t i t ct r rr r rr cti l ( ). s s t t t r s ri l s, , . I tri f r t is c ritt s; ], , ( ) t t t t 1 1 2 2 ....... ( ) t t t k t k t− − − I t c t t f c ritt s 1 1 2 2 1 1 1 ....... ( ) t t t k t k t t− − − − − − r s, 1 2 ( ......... )( , ,....., ) ( ) i k i 1 2 ( ......... ) ( ) k s s t tri , ic icts t tr l r r l ti s i t [ ], , t t t t . ′ , i ic s s t s f j st t t r s ili ri l r c ffici ts tri is ′ . I si l ti c s 1t− ′ is rr r c rr cti t r . fi t f r lti ri t c s ss s . t l is t t-1 t-1 t 1 t-1 t-1 t t-1 t-1 ( ) t ∆ ∆ ∆                              r c s t t; le a ra er ( ), le a ( ), t c a ats ( ), illi s ( ), illi s a liaris ( ) a J a se ( , , ) 4.2 Johansen Co-Integration approach After the pioneer work of Granger (1981) about cointegration, any studies 9 elaborated this concept. Johansen (1988) introduced a new approach of checking the cointegration between ore than two series. It re oves all the drawbacks, which Engle-Granger approach has. In case of Johansen approach the EC also extended into Vector Error Correction odel (VEC ). ow suppose that we have three endogenous variables, L, and . In atrix for this can be written as; ], , (7) t t t t Y L= 1 1 2 2 ....... (8) t t t k t k t Y Y Y Yβ β β µ − − − = + + + + In the context of VEC we can written as 1 1 2 2 1 1 1 ....... (9) t t t k t k t t Y Y Y Y Y µ − − − − − − ∆ = Γ ∆ +Γ ∆ + +Γ ∆ + + ereas, 1 2 (1 ......... )( 1, 2,....., 1) (10) i k i kβ β βΓ = − − − − = − and 1 2 (1 ......... ) (11) k β β β= − − − − − shows the 3 3× atrix, which depicts the true long run relationship between [ ], , t t t t Y L= . The φχ ′= , in which φ shows the speed of adjust ent towards equilibriu and long run coefficients atrix is χ ′ . In single equation case 1t Y − ′ is error correction ter . To find out for ultivariate case now assu es 2k = . So the odel is t t-1 t-1 t 1 t-1 t-1 t t-1 t-1 (12) t L L L e             ∆ = Γ ∆ + ∆ +            ∆ ∆ ∆       or we can say that; 9 Engle and Granger (1987), Engle and Yoo (1987), Stock and atson (1988), Phillips (1986& 1987), Phillips and Ouliaris (1990) and Johansen (1988, 1991, 1995) 4.2 Johansen Co-Integration approach After the pioneer work of Granger (1981) about cointegration, many studies 9 elaborated this concept. Johansen (1988) introduced a new approach of checking the cointegration between more than two series. It removes all the drawbacks, which Engle-Granger approach has. In case of Johansen approach the ECM also extended into Vector Error Correction Model (VECM). Now suppose that we ave three endogenous variables, L, M and N. In matrix form this can be written as; ], , (7) t t t t Y L M N= KK 1 1 2 2 ....... (8) t t t k t k t Y Y Y Yβ β β µ − − − = + + + + K In the context of VECM we can written as 1 1 2 2 1 1 1 ....... (9) t t t k t k t t Y Y Y Y Y µ − − − − − − ∆ = Γ ∆ +Γ ∆ + +Γ ∆ +Π + K hereas, 1 2 (1 ......... )( 1, 2,....., 1) (10) i k i kβ β βΓ = − − − − = − K 1 2 (1 ......... ) (11) k β β βΠ = − − − − − KK Π shows the 3× matrix, w ich depic s the true long run relationship between [ ], , t t t t Y L M N= . The φχ ′Π = , in which φ shows he spe d of adjustment towards equilibrium and long run coefficients matrix is χ ′ . In single equation case 1t Y − ′ is error correction term. To find out for multivariate case now assumes 2k = . So the model is t t-1 t-1 t 1 t-1 t-1 t t-1 t-1 (12) t L L L M M M e N N N ∆ ∆ ∆            ∆ = Γ ∆ + Π ∆ +            ∆ ∆ ∆       KK or we can say that; 9 Engle and Granger (1987), Engle and Yoo (1987), Stock and Watson (1988), Phillips (1986& 1987), Phillips and Ouliaris (1990) and Johansen (1988, 1991, 1995) 4.2 Johansen Co-Integration approach After the pioneer work of Granger (1981) about cointegration, many studies 9 elaborated this concept. Johansen (1988) introduced a new approach of checking the cointegration between more than two series. It removes a l the drawbacks, which Engle-Granger approach has. In case of Johansen approach the ECM also extended into Vector E ror Co rection Model (VECM). Now suppose that we have three endogenous variables, L, M and N. In matrix form this can be wri ten as; ], , (7) t t t t Y L N= 1 1 2 2 . (8) t t t k t k t Y Y Y Yβ β β µ − − − = + + + + In the context of VECM we can wri ten as 1 1 2 2 1 1 1 . (9) t t t k t k t t Y Y Y Y Y µ − − − − − − ∆ = Γ ∆ +Γ ∆ + +Γ ∆ +Π + Whereas, 1 2 (1 . )( 1, 2, ., 1) (10) i k i kβ β βΓ = − − − − = − and 1 2 (1 . ) (11) k β β βΠ = − − − − − Π shows the 3 3× matrix, which depicts he true lo g run relationship between [ ], , t t t t Y L N= . Th φχ ′Π = , in which φ shows the speed of adjustment towards equilibrium and long run coe ficients matrix is χ ′ . In single equation case 1t Y − ′ is e ror co rection term. To find out for multivariate case now a sumes 2k = . So the model is t t-1 t-1 t 1 t-1 t-1 t t-1 t-1 (12) t L L L e N N N ∆ ∆ ∆            ∆ = Γ ∆ + Π ∆ +            ∆ ∆ ∆       or we can say that; 9 Engle and Granger (1987), Engle and Yoo (1987), Stock and Watson (1988), Phi lips (1986& 1987), Phi lips and Ouliaris (1990) and Johansen (1988, 1991, 1995) t t-1 t-111 12 11 21 31 t 1 t-1 21 22 t-1 12 22 32 31 32t t-1 t-1 (13) t L L L M M M e N N N φ φ χ χ χ φ φ χ χ χ φ φ                ∆ = Γ ∆ + +                 ∆ ∆        K For simplicity just analyze the first equation’s error correction part. The first row of Π matrix is; [ ][ ][ ] 1 1 1 11 11 12 12 11 21 12 22 11 31 12 32 1 1 ( ) (14) t t t t t L Y M e N φ χ φ χ φ χ φ χ φ χ φ χ − − − −     Π = + + + +       K This can also be written as; 1 1 11 11 1 21 1 31 1 12 12 1 22 1 32 1 ( ) ( ) (15) t t t t t t t Y L M N L M Nφ χ χ χ φ χ χ χ − − − − − − − Π = + + + + + K Equation clearly express the two cointegrating vectors and the terms of their speed of adjustment 11 φ and 12 φ . Regarding the rank of matrix, there are three cases which are as follow; i. The variables in t Y are I(0), if Π has a full rank. ii. There are no cointegrating relationships, when the Π is zero. iii. There are ( 1)r n≤ − cointegrating relationships, when Π has a reduced rank. To check the goodness of fit, diagnostic test like Serial correlation, functional form, normality and heteroskedasticity tests and stability test like Cumulative Sum of Recursive Residuals (CUSUM) and Cumulative Sum of Squares of Recursive Residuals (CUSUMsq.) are performed. t t-1 t-111 12 11 21 31 t 1 t-1 21 22 t-1 12 22 32 31 32t t-1 t-1 (13) t L L L M M M e N N N φ φ χ χ χ φ φ χ χ χ φ φ ∆ ∆               ∆ = Γ ∆ + +                 ∆ ∆        K For simplicity just analyze t e first equation’s error correction part. The first row of Π matrix is; [ ][ ][ ] 1 1 1 11 11 12 12 11 21 12 22 11 31 12 32 1 1 ( ) (14) t t t t t L Y M e N φ χ φ χ φ χ φ χ φ χ φ χ − − − −     Π = + + + +       K This can also be written as; 1 1 11 11 1 21 1 31 1 12 12 1 22 1 32 1 ( ) ( ) (15) t t t t t t t Y L M N L M Nφ χ χ χ φ χ χ χ − − − − − − − Π = + + + + + K Eq ation clearly express the two cointegrating vectors and the terms of their speed of adjustmen 11 φ d 12 φ . Regarding the r nk of matrix, there are three cases which are as follow; variables in t Y a e I(0), if Π ha a full rank. ii. There are no cointegrating relationships, when the Π is zero. iii. There are ( 1)r n≤ − cointegrating relationships, when Π has a reduced rank. To check the goodness of fit, diagnostic test like Serial correlation, functional form, normality and heteroskedasticity tests and stability test like Cumulative Sum of Recursive Residuals (CUSUM) and Cumulative Sum of Squares of Recursive Residuals (CUSUMsq.) are performed. t t-1 t-111 12 11 21 31 t 1 t-1 21 22 t-1 12 22 32 31 32t t-1 t-1 ( ) t ∆ ∆                                       r si licit j st l z t first ti ’s rr r c rr cti rt. first r f tri is; [ ][ ][ ] 1 1 1 11 11 12 12 11 21 12 22 11 31 12 32 1 1 ( ) ( ) t t t t t − − − −           is c ls ritt s; 1 1 11 11 1 21 1 31 1 12 12 1 22 1 32 1 ( ) ( ) ( ) t t t t t t t− − − − − − − ti cl rl r ss t t c i t r ti ct rs t t r s f t ir s f j st t 11 12 . r i t r f tri , t r r t r c s s ic r s f ll ; i. ri l s i t r I( ), if s f ll r . ii. r r c i t r ti r l ti s i s, t is z r . iii. r r ( )r c i t r ti r l ti s i s, s r c r . c c t ss f fit, i stic t st li ri l c rr l ti , f cti l f r , r lit t r s sticit t sts st ilit t st li l ti f c rsi si ls ( ) l ti f r s f c rsi si ls ( s .) r rf r . M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 169 Equation clearly express the two cointegrating vectors and the terms of their speed of adjustment f11 and f12. Regarding the rank of matrix, there are three cases which are as follow; i. The variables in Yt are I(0), if Π has a full rank. ii. There are no cointegrating relationships, when the Π is zero. iii. There are r ≤ (n – 1) cointegrating relationships, when Π has a reduced rank. To check the goodness of fit, diagnostic test like Serial correlation, functional form, nor- mality and heteroskedasticity tests and stability test like Cumulative Sum of Recursive Residuals (CUSUM) and Cumulative Sum of Squares of Recursive Residuals (CUSUM- sq.) are performed. 5. eMPiricaL finDinGs 5.1 Unit root results To ward off the spurious results the study tested the variables for unit root. Three meth- ods of unit root are adopted, ADf, PP, and KPSS. The study check the stationarity of the variables under two models, with intercept and trend and secondly with intercept and no trend. All the variables are I(1) under ADf test, except government expenditure. PP test result indicates that all the variables are I(1). This time government expenditure is stationary at first difference. In the next model, which considers no trend in data, all the variables are I(1) under ADf and PP tests. Under KPSS in the first model, with intercept and trend, all the variables are stationary I(1). In the second model, with intercept but no trend, government expenditures, debt, budget deficit and wealth are stationary at I(1). Keeping in view the results of three unit roots tests the study deals the variable at I(1). (See table 5.1) Prior to the estimation of the main model it is necessary to check that whether the said variables have long run or short relationship or not? for this purpose different cointegra- tion techniques are used in literature10. After checking the stationarity of data we come to know that all the variables are I(1), so Johansen and Juselius (1990) cointegration tech- nique is applied. In JJ approach the first step is to identify the order of VAR. On the basis of AIC and SBC lag length of VAR is selected. Both criterions selected three lag length of VAR (See table 5.2) 10 however, not in case of Pakistan. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011170 Ta bl e 5. 1: U ni t r oo t r es ul ts Va ri ab le s A D F PP KP SS W it h tr en d Le ve l P* D iff er en ce P* Le ve l Q * D iff er en ce Q * Le ve l K* D iff er en ce K* PC -0 .8 58 1 -4 .5 15 ** * 1 -0 .5 84 4 -5 .1 47 ** * 3 0. 40 6 4 0. 12 8* ** 0 G E -1 .3 42 2 -2 .7 84 3 -2 .3 04 3 -8 .8 70 ** * 2 0. 25 0 4 0. 07 8* ** 6 YD -2 .7 47 2 -4 .5 22 ** * 3 -3 .2 18 * 4 -7 .2 97 ** * 3 0. 21 6 3 0. 04 0* ** 3 TR -1 .2 71 1 -3 .6 59 ** 1 -1 .5 61 1 -6 .4 20 ** * 1 0. 27 1 5 0. 09 6* ** 2 D EF -2 .6 83 2 -4 .2 30 ** * 3 -2 .9 83 2 -7 .11 0* ** 3 0. 23 4 3 0. 13 0* ** 9 D EB T -1 .6 13 1 -4 .5 18 ** * 2 -1 .5 88 4 -4 .7 95 ** * 2 0. 21 7 4 0. 12 9* ** 5 W EA LT H -1 .6 50 2 -4 .7 27 ** * 3 -1 .6 54 3 -4 .8 13 ** * 3 0. 22 9 4 0. 14 3* ** 5 W it ho ut tr en d PC 1 .1 84 2 -4 .0 54 ** * 1 1 .2 84 5 -4 .8 20 ** * 3 0. 70 5 5 0. 28 2 2 G E -1 .6 32 1 -2 .7 44 ** 3 -2 .2 40 3 -8 .8 75 ** * 2 0. 34 3 4 0. 13 6* ** 5 YD -1 .9 58 3 -4 .5 83 ** * 2 -2 .4 65 4 -7 .3 96 ** * 3 0. 50 1 4 0. 52 1 3 TR -1 .8 99 2 -3 .3 80 ** * 3 -2 .0 53 1 -6 .2 35 ** * 1 0. 38 2 5 0. 27 3 3 D EF -2 .7 27 3 -4 .2 91 ** * 1 -3 .0 33 3 -7 .2 26 ** * 4 0. 16 7* ** 3 0. 13 9* ** 9 D EB T -1 .2 23 1 -4 .4 14 ** * 3 -0 .8 91 1 -4 .8 57 ** * 3 0. 56 4 5 0. 13 3* ** 5 W EA LT H -1 .1 80 2 -4 .5 45 ** * 4 -0 .7 66 4 -4 .8 37 ** * 4 0. 63 1 5 0. 14 6* ** 4 N ot es : P C is re al p er c ap ita p ri va te c on su m pt io n; G E is re al p er c ap ita G ov er nm en t e xp en di tu re ; Y D is re al p er c ap ita d is po sa bl e in co m e; T R is re al pe r c ap ita ta x re ve nu e; D Ef is re al p er ca pi ta b ud ge t d efi ci t; D EB T is re al p er ca pi ta d eb t; W EA LT h is re al p er c ap ita w ea lth . P * s ho w s t he m ax im um la g le ng th , a s d et er m in ed b y us in g A IC . U nd er P P te st Q * a nd K * i n K PS S te st sh ow s N ew ey -W es t B an dw ith , a s d et er m in ed b y Ba rt le tt- K er ne l. ** * s ho w s 1 % si gn ifi ca nc e le ve l; ** sh ow s 5 % si gn ifi ca nc e le ve l a nd * re pr es en ts 1 0% si gn ifi ca nc e le ve l. M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 171 Table 5.2: Lag length selection criterion Order LL AIC SBC LR test Adjusted LR 0 -928.22 -935.22 -940.57 757.90[0.00] 267.49[0.00] 1 -746.28 -802.28 -845.02 394.01[0.00] 139.06[0.00] 2 -673.59 -778.59 -858.02 248.64[0.00] 87.75[0.00] 3 -549.27 -703.27 -802.80 ------ ------ By using Pantula Principal the model with unrestricted intercept and no trend is select- ed, among the five cointegration models. Both Eigen value and Trace statistic reject the null hypothesis of no cointegration because the value of trace test (207.10) is grater then 5% and 1% critical values. Result reveals that there is one cointegrating vector, based on the Eigen values and Trace statistics. Table 5.3: Johansen Maximum Likelihood Test for cointegration Hypotheses Trace test 5% critical values 10% critical values Hypotheses Max- Eigen Statistic 5% critical value 10% critical values R = 0 207.10 124.62 119.68 R = 0 92.76 45.63 42.700 R ≤ 1 114.34 95.87 91.40 R = 1 48.33 39.83 36.84 R ≤ 2 66.00 70.49 66.23 R = 2 28.53 33.64 31.02 R ≤ 3 37.47 48.88 45.70 R = 3 20.14 27.42 24.99 R ≤ 4 17.32 31.54 28.78 R = 4 10.71 21.12 19.02 R ≤ 5 6.61 17.86 15.75 R = 5 5.52 14.88 12.98 R ≤ 6 1.08 8.07 6.50 R = 6 1.08 8.07 6.50 After investigating the long run relationship among variables, it is important to investi- gate the short run dynamics. Error correction term shows the speed of convergence to- wards equilibrium. It is significant and negative in sign. The speed of correction towards equilibrium depends upon the value of error correction term. Table 5.4: ECM regression results Variables Coefficients t-values Prob-value Constant 28.82 5.045 0.000 DYD -0.0157 -1.983 0.001 DGE 0.0291 0.092 0.366 DDEF 0.112 0.383 0.704 DWEALTH -0.032 -1.095 0.283 DTR -0.033 -0.605 0.550 DDEBT 0.044 2.268 0.000 DECM(-1) -0.812 -2.583 0.000 R-Squared 0.681 Adjusted R-Squared 0.583 S.E. of Regression 5.040 DW-statistic 2.15 Log-likelihood 731.8864 F-stat 6.948 [0.000] Notes: DPC is dependant variable. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011172 Brown et al. (1975) proposed two tests Cumulative Sum and Cumulative Sum of Square, to check the structural stability. CUSUM test captured the systematic changes in regres- sion coefficients, while CUSUMSQ detain the departure of parameters from constancy. hence, parameter consistency is checked by using these two tests. following graphs shows the stability of model for whole sample because the residuals are within 5% criti- cal bonds. fig 5.1: Cumulative Sum of Recursive Residual Plot of Cumulative Sum of Recursive Residual The straight line represent critical bonds at 5% significance level fig 5.2: Cumulative Sum of Square Recursive Residual Plot of Cumulative Sum of Square Recursive Residual The straight line represent critical bonds at 5% significance level Under structural consumption function we want to test that government expenditures are negatively effect private consumption; taxes, deficit financing, and debt has no im- pact on private consumption; budget deficit and disposable are equal; and wealth is equal to government debt. These restriction are reject by the data so, there is no evidence in favor of REh in case of Pakistan. Restrictions are rejected by the Wald test. According to REh government expenditures and private consumption must inversely related to each other but in results government expenditure is positively related with 17 Fig 5.1: Cumulative Sum of Recursive Residual Plot of Cumulative Su of Recursive Residual The straight line represent critical bonds at 5% significance level Fig 5.2: Cumulative Sum of Square Recursive Residual Plot of tive Sum of Squar Recursive Residual The straight line represent critical bonds at 5% significance level Under structural onsumption function we want to test t at government expenditures are negatively effect private consumptio ; axes, deficit fina cing, and debt has no impact on private consumption; budget deficit and disposable are equal; and wealth is equal to government debt. These restriction are reject by the data so, there is no evidence in favor of REH in case of Pakistan. Restrictions are rejected by the Wald test. 17 Fig 5.1: Cumulative Sum of Recursive Residual Plot of Cumulative Sum of Recursive Residual The straight line represent critical bonds at 5% significance level Fig 5.2: Cumulative Sum of Square Recursive Residual Plot of Cumulative Sum of Square Recursive Residual The straight line represent critical bonds at 5% significance level Under structural consumption function we want to test that government expenditures are negatively effect private consumption; taxes, deficit financing, and debt has no impact on private consumption; budget deficit and disposable are equal; and wealth is equal to government debt. These restriction are reject by the data so, there is no evidence in favor of REH in case of Pakistan. Restrictions are rejected by the Wald test. M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 173 private consumption, hence we reject REh. Moreover, results depict that taxes and debt is negatively related with private consumption. Disposable income is positively effect private consumption, which means that when person’s disposable income increases he increases his consumption expenditures. These results are contradictory with the theory of REh. The theory states that when disposable income increases a person will decrease its consumption expenditures and save more in order to protect his children. The results are in line with the existing literature of REh in case of developing countries. In case of Pakistan Kazmi (1992, 1994) rejected the REh and concluded that REh is a rough and oversimplified approximation of consumer behavior. Table 5.5: Results of REH Variables Coefficients t-value Prob-value Constant 3.574 2.836 0.000 DYD 0.047 3.916 0.000 DGE 0.105 2.100 0.000 DWEALTH 0.882 3.785 0.000 DTR -1.190 1.931 0.021 DDEBT -1.000 3.597 0.000 DDEF 0.355 1.082 0.285 a2 < 0, a4 = 0, a5 = 0, a1 = 0, a6 = 0, a3 = a5 l2(5)= 16.36 [0.005] R-square 0.520 D.W 2.046 Adjusted R-square 0.495 F-statistic 2.98 [0.018] SER 5.838 The correlation matrix in table 5.6 describes the degree of association between the variables. It is assumed that two variables will be highly correlated if the correlation coefficient is greater than 0.5, or it lies between 0.3 and 0.49. Moreover, if this value lies 0.2 to 0.29 than it is moderate correlation and if it lies 0.1 to 0.10 it is weak cor- relation. Table 5.6: Results of Correlation Matrix Variables DEBT DEF GE PC TR WEALTH YD DEBT 1.0000 DEF 0.3789 1.0000 GE 0.6582*** 0.4782** 1.0000 PC 0.6660*** 0.3792** 0.4431** 1.0000 TR 0.5893*** 0.4572** 0.8606*** 0.4739** 1.0000 WEALTH 0.3429** 0.0450* 0.0389* 0.6057*** 0.0702* 1.0000 YD 0.0975* 0.3683** 0.5616*** 0.5726*** 0.6868*** 0.4361* 1.0000 Note: *** Strong Correlation ** Moderate Correlation *Weak Correlation ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011174 6. suMMary anD concLusion The aim of this study is to examine the REh by using the annual data of Pakistan from 1973-2009. The study used variables, government expenditure, private consumption ex- penditure, tax revenue, government debt, disposable income, government budget deficit and wealth to meet the objectives of the study. Results of ADf, PP and KPSS unit root tests show that all the variables are I(1). JJ approach of cointegration shows a long run re- lation among the variables. Under the results of Structural consumption function there is no evidence in favor of REh in case of Pakistan. Restrictions are significantly rejected by the Wald test. The findings of the study validate the effectiveness of fiscal policy because consumers treat government debt as a net wealth. Thus fiscal policies should be used as major policy instruments in order to boost private consumption and control trade deficits, which are prime goal of stabilization policies in Pakistan. references Abimanyu, Y. (1998). Using Indonesia’s Real Exchange Rate to Test Ricardian Equivalence. International Economic Journal,12, 17-29. Apergis, N., & Lyroudi, K., (2008). Private and Government Consumption in Transitional Economies: A Panel Data Analysis. Journal of Economic Development, 33, 17-27. Aschauer, D. (1985). fiscal Policy and Aggregate Demand. The American Economic Review, 75, 117-127. Aschauer, D. (1988). The Equilibrium Approach to fiscal Policy. Journal of Money, Credit and Banking, 20, 41-62. Aschauer, D. (1989). Is Public Expenditure Productive?. Journal of Monetary Economics, 23, 177-200. Aschauer, D. (1993). fiscal Policy and Aggregate Demand: Reply. The American Economic Review, 83, 667- 669. Barro, R. (1974). Are Government Bonds Net Wealth?. The Journal of Political Economy, 82, 1095-1117. Barro, R. (1979). On the Determination of the Public Debt. The Journal of Political Economy, 87, 940-971. Barro, R. (1980). federal Deficit Policy and the Effects of Public Debt Shocks. Journal of Money, Credit, and Banking, 12, 747–761. Barro, R. (1989), The Ricardian Approach to Budget Deficits. The Journal of Economic Perspectives, 3, 37-54. Bernheim, D. (1987). Ricardian Equivalence: An Evaluation of Theory and Evidence. NBER, Macroeconomics Annual, 263-315. Cardia, E. (1997). Replicating Ricardian equivalence Tests with Simulated Series. The American Economic Review, 87, 65-79. Cuaresma, J. C. & Reitschuler, G. (2007). Is Ricardian Equivalence Ricardian Proposition an ‘Aerie fairy Theory for Europe?. Economica, 74, 682-694. M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 175 Diamond, P. (1965). National Debt in a Neoclassical Growth Model. The American Economic Review, 55, 1126-1150. Drakos, K. (2001). Testing Ricardian Equivalence Theorem: Time Series Evidence from Greece. Journal of Economic Development, 26, 1-12. Drakos, K. (2003). Testing the Ricardian equivalence proposition: Panel data Evidence From the EU. Working Paper, Department of Economics. University of Rome. Eisner, E. (1989). Budget Deficits: Rhetoric and Realities. The Journal of Economic Prespectives, 3, 73-93. Evans, P. (1985). Do Large Deficits Produce high Interest Rates?. The American Economic Review, 75, 66- 87. Evans, P. (1988). Are Consumers Ricardian? Evidence for the United States. The Journal of Political Economy, 96, 583-1004. Evans, P. (1993). Consumers are not Ricardian: Evidence from Nineteen Countries. Economic Inquiry, 31, 534–548. Evans, P. (2001). Consumer Behaviour in the United States: Implications for Social Security Reform. Eco- nomic Inquiry, 39, 568–582. feldstein, M. (1982). Government Deficits and Aggregate Demand. Journal of Monetary Economics, 9, 1-20. feldstein, M. (1988). The Effects of fiscal Policies When Incomes Are Uncertain : A Contradiction to Ricard- ian Equivalence. The American Economic Review, 78, 14-23. feldstein, M. & Douglas, E. (1990). Government Debt, Government Spending, and Private-Sector Behavior Revisited: Comment. The American Economic Review, 80, 589-599. Garcia, A. & Ramajo, J. (2003). fiscal Policy and Private Consumption Behaviour: The Spanish Case. Empiri- cal Economics, 30, 115–135. Ghatak, A., & Ghatak, S. (1996). Budgetary Deficits and Ricardian Equivalence: The Case of India. Journal of Public Economics, 60, 267-282. Giorioni, G. & holden, K. (2001). Does the Ricardian Equivalence Proposition hold in Less Developed Coun- tries. International Review of Applied Economics, 17, 209-221. Giorioni, G. & holden, K. (2001). Some Further International Evidence on Ricardian equivalence: A VECM Approach. Working Paper, Center for International Banking, Economics and finance. Graham, f. & himarios, D. (1993). Consumption, Wealth, and finite horizons: Tests of Ricardian Equiva- lence. Economic Inquiry, 34, 527-544. Graham, f. (1993). fiscal Policy and Aggregate Demand: Comment. The American Economic Review, 83, 659-666. Graham, f. (1995). Government Debt, Government Spending, and Private-Sector Behavior: Comment. The American Economic Review, 85, 1348-1356. Gray, J. & Stone, J. A. (2005). Ricardian equivalence For Sub National States. University of Oregon Eugene. hall, R. (1978). Stochastic Implications of the Life Cycle/Permanent Income hypothesis: Theory and Evi- dence. Journal of Political Economy, 86, 971 987. ECONOMIC AND BUSINESS REVIEW | VOL. 13 | No. 3 | 2011176 haug, A. (1996) Blanchard’s Model of Consumption: An Empirical Study. Journal of Business and Economic Statistics,14, 169–177. haque, N. U. & Montiel, P. (1987). Ricardian Equivalence, Liquidity Constraints, and the Yari-Blanchard Ef- fect: Tests for Developing Countries. IMf Working Paper. haque, N. U. & Montiel, P. (1989). Consumption in Developing Countries: Tests for Liquidity Constraints and finite horizons. The Review of Economics and Statistics, 71, 408-415. hayashi, f. (1985). The Effect of Liquidity Constraints on Consumption: A Croos-Sectional Analysis. Quar- terly Journal of Economics, 100, 183-206. hayashi, f. (1985). Tests for Liquidity Constraints: A Critical Survey. NBER, Working Paper, (1720) 1-46. himarios, D. (1995). Euler Equations Test of Ricardian Equivalence. Economic Letters, 48, 165–171. Kaadu, h. & Uuskula, L. (2004). Liquidity Constraints and Ricardian Equivalence in Estonia. Working Papers of Eesti Pank, No.7. Kazmi,A. A. (1991). Saving, Consumption and Ricardian equivalence: A Macro Econometric Analysis of Paki- stan. Ph.D. Dissertation, Boston University, Boston, U.S.A. Kazmi, A. A. (1992). Ricardian equivalence: Some Macro Econometric Tests for Pakistan”. The Pakistan Development Review, 31, 743-758. Kazmi, A. A. (2001). Ricardian Equivalence hypothesis: Some Empirical Tests for Pakistan Based on Blan- chard-Evans Models. The Lahore Journal of Economics, 6, 75-92. Khalid, A. M. (1996). Ricardian Equivalence: Empirical Evidence from Developing Economies. Journal of Development Economics, 51, 413-432. Kormendi, R. (1983). fiscal Policy and Aggregate Demand. The American Economic Review, 73, 994-1010. Kormendi, R. & Meguire, P. (1986). Government Debt, Government Spending, and Private Sector Behaviour: Reply. American Economic Review, 76, 1180–1187. Kormendi, R. & Meguire, P. (1990). Government Debt, Government Spending, and Private Sector Behaviour: Reply and Update. American Economic Review, 80, 604–617. Kormendi, R. & Meguire, P. (1986). Government Debt, Government Spending, and Private Sector Behaviour: Reply. American Economic Review, 85, 1357–1361. Lucke, B. (1999). Econometric Tests of Ricardian Equivalence: Results for Germany. Universität hamburg. fachbereich Wirtschaftswissenschaft. Institut für Wachstum und Konjunktur. Leiderman, L. (1987). Testing Ricardian Neutrality with an Intertemporal Stochastic Model. NBER, Working Paper No. 2258. Marinheiro, C. f. (2001). Ricardian equivalence:An Empirical Application to Portuguese Economy. Katholieke University Leuven Workshop in International Economics. Modigliani, f. & Sterling, A. (1986). Government Debt, Government Spending, and Private Sector Behaviour: Comment. American Economic Review, 76, 1168–1179. Modigliani, f. & Arlie, S. (1990). Government Debt, Government Spending, and Private Sector Behavior: A further Comment. The American Economic Review, 80, 600-603. M. WAQAS, M. SARWAR AWAN | ARE PAKISTANI CONCUMERS RICARDIAN? 177 Malengier, G. & Poozi, L. (2004). Examining Ricardian Equivalence by Estimating and Bootstrapping a Non- linear Dynamic Panel Model. Money Macro and finance Research Group Confrence. Niple, K. (2006). Are We Ricardian? Evidence from U.S Countries. Issues in Political Economy, 15, 1-10. Pemberton, J. (2004). Ricardian Consumers with Non-Keynesian (and possible Ricardian) Propensities. Scot- tish Journal of Political Economy, 51, 95-104. Reitschuler, G. & Jesús, C. C. (2004). Ricardian Equivalence Revisited: Evidence from OECD countries. Eco- nomics Bulletin, 5, 1−10. Ricciuti, R. & Laurea, D. D. (2003). An Experimental Analysis of Two Departures form Ricardian equiva- lence. Eonomic Bulletin, 8, 1-11. Robert, R. B., Mankiw, N. G. & Zeldes, S. P. (1987). Ricardian Consumer with Keynesian Propensities. Ameri- can Economic Review, 76, 676-691. Rockerbie, D. W. (1997). Are Consumers Ricardian When Some Are Liquidity Constrianed?. Applied Eco- nomics, 29, 821-827. Sachsida, A. & Carneiro, P. E. (2001). Capital Stock, Saving and Public Debt: Are Brazilians Ricardian? Catho- lic University of Brasilia; Institute for Applied Economic Research. Safa, M. f. & Siddiq, f. K. (2005). Assessment of the Canadian Experience. Third Annual Meeting of The At- lantic Canada Economic Association. Seater, J. (1993). Ricardian Equivalence. Journal of Economic Literature, 31, 142-190. Seater, J. & Roberto, M. (1985). New Tests of the Life Cycle and Tax Discounting hypotheses. Journal of Mon- etary Economics, 15, 198-215. Siddiki, J. U. (2008). The Ricardian Equivalence hypothesis: Evidence from Bangaldesh. Applied Economics, 42, 1419–1435. Vamvoukas, A. G. & Gargalas, N. V. (2008). Testing Keynesian Proposition and Ricardian Equivalence: More Evidence on the Debate. Journal of Business and Economics Research, 6, 67-76. Walker, W. C. (2002). Ricardian Equivalence and fiscal Policy Effectiveness in Japan. Asian Economic Jour- nal, 16, 285-302. Whelan, K. (1991). Ricardian Equivalence and the Irish Consumption function: The Evidence Re-examined. The Economic and Social Review, 22, 229-238. Waqas, M., Awan, M. S. & Aslam, M. A. (2011). We Are Living on the Cost of our Children. Interdisciplinary Journal of Contemporary Research in Business, 2, 607-623.