The Long Run Determinants of the U.S. Trade Balance: A Reexamination Using Bi & Multivariate Cointegration Approach

Transcription

1 The Long Run Deerminans of he U.S. Trade Balance: A Reexaminaion Using Bi & Mulivariae Coinegraion Approach Amarendra Sharma Dep. of Economics, Elmira College One Park Place, Elmira, NY 14901, USA Tel: asharma@elmira.edu Received: December 20, 2011 Acceped: February 29, 2012 Published: March 17, 2012 doi: /rae.v4i URL: hp://dx.doi.org/ /rae.v4i Absrac This paper exends our undersanding of he long-run deerminans of he U.S. curren accoun and rade balances. Earlier sudies have recognized he long-run effecs of fiscal, moneary, and commercial policy variables on he exernal accoun balances. Bu none of hem repored he ineres rae measures o be coinegraed wih he exernal accoun balances. This sudy provides srong evidence o sugges ha he various ineres rae measures are also he long-run deerminans of he U.S. curren accoun and rade balances. Keywords: Trade Defici, Curren Accoun, Fiscal Policy, Moneary Policy, Commercial Policy, Ineres Rae 16

2 1. Inroducion The primary objecive of his paper is o reexamine he long-run deerminans of he U.S. rade balance. This paricular issue assumes added significance due he fac ha he U.S. has been running rade defici for he mos par of he las four decades in spie of changes in he fiscal, moneary, and commercial policies over he years o deal wih i, direcly or indirecly. I is no unusual for he poliicians o pu he blame of he economic problems facing he counry squarely on he win deficis of inernal and exernal accouns. I is hoped ha a beer undersanding of he long-run deerminans will allow he policy makers o use he policy ools a heir disposal in a more argeed and susained fashion o deal wih he issue of curren accoun deficis. The sudies ha have been carried ou so far o undersand he long-run relaionships among he policy ools measures and he exernal accoun balances are inadequae in he sense ha hey eiher do no apply he recen advances in ime series economerics on coinegraion and uni roo analysis (earlier sudies rely on he srucural models or reduced form models ha do no ake saionariy of he ime series involved ino accoun (see for example, Magee, 1973; Krugman & Baldwin, 1987; Rosensweig & Koch, 1988; Darrra, 1988, Miller & Russek, 1989; Bahmani-Oskooee, 1989), or are marred by he relaively small ime horizons (see for example Alse & Bahmani-Oskooee, 1992; Bahmani-Oskooee, 1992, 1995; Kim, 1995). One sudy ha sands ou in erms of is modern approach o his opic is by Bahmani-Oskooee (92), who invesigaes his issue by looking a he quarerly daa colleced from various sources for he ime period 1971:1-1988:4, a oal of eigheen years. Based on his sample he concludes ha he full employmen budge defici and o some exen he money supply as measured by M2 are he only wo possible long-run deerminans of he U.S. rade balance. His analysis relies on examining he saionariy of he individual ime series and bilaeral coinegraion a la Engle & Granger (1987). Kim (1995) in a follow-up o his paper, which essenially draws aenion o he inadequacies of he echniques used, comes up wih a conclusion ha he money supply as measured by M2, he erms of rade, and he nominal effecive exchange rae, are all coinegraed wih he rade and curren accoun balances, and herefore, have he equilibrium (long-run) relaionships. His sudy does no suppor he full employmen budge defici as one of he long run deerminans of he U.S. rade balance, as he does no find he rade balance o be coinegraed wih he full employmen budge. In his paper, I revisi he issue and look a he quarerly daa from 1974:1-2007:4. My sample includes curren governmen expendiure and receips as he measures of fiscal policy sance, and no he full employmen budge defici used by Bahmani-Oskooee (1992, 1995) and Kim (1995). Based on my analysis, I confirm ha on he fiscal policy side only he governmen s curren expendiure and on he moneary policy side only he M2 measure of money are coinegraed wih he balance of rade and he curren accoun balance in a bivariae seing. Bu when I consider he mulivariae coinegraion seings hen he governmen s curren expendiure, receips, and ne savings, M1 & M2 measures of money sock, he reasury bill rae, he ineres raes on he medium and long erm governmen bonds, he erms of rade, and he nominal effecive exchange rae are all found o have a leas one coinegaring vecor wih he exernal accoun balances, hereby confirming he exisence of he long-run 17

3 relaionships among hem. Earlier sudies, i appears deliberaely excluded ineres rae measures from he analysis on he grounds ha he federal reserve bank s moneary policy can be represened by eiher he ineres rae or he measures of money supply. Bu as is well known ha a change in money supply does no auomaically resul in a change in ineres rae especially if he financial insiuions do no have enough confidence in he borrower s abiliy o repay he loan and hey jus si on he excess reserves as was he case in he recen financial crisis. Clearly, we canno simply assume ha moneary aggregaes alone are sufficien o capure he long run relaionship beween he curren accoun balance and moneary policy. The direcion of flow is like his: An increase in M1 or M2 will decrease he ineres rae (which may or may no happen), which in urn will simulae domesic spending and invesmen causing GDP o rise and ne expors o decrease. On he oher hand, a fall in he ineres rae will resul in a fligh of funds in he foreign exchange marke causing he dollar o depreciae and he ne expors o improve in urn. This sudy lends credence o he earlier findings of Bahmani-Oskooee (1992, 1995) and Kim (1995), and exends he long run deerminans o include he ineres rae variables, which was no menioned as saisically significan by he earlier sudies. 2. Daa This sudy uses quarerly daa from 1974:1-2007:4. The daa is colleced from various sources including he publicaions from he Bureau of Economic Analysis (BEA), he Inernaional Financial Saisics (IFS), and he Bank for Inernaional Selemens (BIS). The variables BOT (real balance of rade on goods and services) and BCA (real balance on he curren accoun) represen he wo measures of he U.S. exernal accoun balance. The fiscal policy measures are represened by GEXP (real seasonally adjused governmen curren expendiure), GRECPT (real seasonally adjused governmen curren receips), and GNS (real seasonally adjused ne governmen savings), respecively. The variables M1 & M2 (real seasonally adjused money sock as measured by M1 & M2) represen he moneary policy measures. The commercial policy measures are represened by he variables TOT (real erms of rade) and NEX (nominal effecive exchange rae). I also have hree measures of ineres rae, namely, TB (real Treasury bill rae), GBMT (real ineres rae on he medium-erm governmen bond), and GBLT (real ineres rae on he long-erm governmen bond). The variable GNP (real seasonally adjused gross naional produc) measures he overall economic aciviy. This sudy uses he GNP deflaor wih 2005 as he base year. 3. Mehodology Following he prescripion of Engle & Granger (1987) I esed all hireen ime series for uni roo. This issue is considered imporan because he sandard economeric mehodologies assume saionariy of he underlying ime series. If he ime series is no saionary, which is he case for many of hem; he inferences based on he usual saisical ess become meaningless. For example, he ordinary leas squares (OLS) esimaion ha uses he non-saionary ime series, resuls in a spurious regression, unless hese ime series are coinegraed (Granger & Newbold, 1974). Due o he generally low power of uni roo ess, I decided o perform hree differen uni roo ess: he Augmened Dickey-Fuller es (ADF), 18

4 The Dickey-Fuller Generalized Leas Squares es (DFGLS), and he Phillips-Perron es (PP). I do no perform he simple Dickey-Fuller es because i is valid only when he series is generaed by a firs-order auoregressive process. The ADF es, on he oher hand, makes a parameric correcion for higher-order auocorrelaion by including he lagged differences of he regressand as he regressors. The ADF es is based on esimaing he following regression: y 1 p ' M y y (1) j 1 j j where M is a vecor of deerminisic erms (consan, rend, ec.). The p lagged difference erms, y j, are used o approximae he ARMA srucure of he errors, and he value of p is se such ha he error homoskedasic. is serially uncorrelaed. The error erm is also assumed o be The DFGLS es is a simple modificaion of he ADF es in which he daa is derended. The es regression for he DFGLS es is: y d y d 1 p (2) j 1 j y d j where y d is he derended daa. I has been shown ha he DFGLS es has a higher power han he DF ess. The Phillips-Perron (PP) uni roo es differs from he ADF es primarily in erms of is reamen of error specific serial correlaion and heeroskedasiciy. In paricular, he PP es ignores any serial correlaion in he es regression, whereas he ADF es ries o mimic he ARMA srucure of he error erms hrough parameric auoregression, The es regression for he PP es is: y ' M y u (3) 1 where u is I(0) and may be heeroskedasic. The PP es correcs for any serial correlaion and heeroskedasiciy in he error u by direcly modifying he es saisics. One advanage of he PP es over he ADF es is ha i is robus o general forms of heeroskedasiciy in he error erm regression. u. Addiionally, he user does no have o specify he lag lengh for he es 19

5 I performed hese hree ess on all hireen variables in levels and heir firs differences. These ess raise wo pracical problems. The firs involves he choice of exogenous variable in he regression (a consan and a linear ime rend, drif, and none). For he ADF es I repor he regression resuls for he cases wih a linear ime rend and he ones wih a drif. For he DFGLS es I repor he resuls for wo cases: (a) he series is saionary around a mean, and (b) he series is saionary around a linear ime rend. For he PP es I repor he resuls from cases where a rend is included and he ones wihou i. The second pracical problem arises due o he choice of he lag lengh in he regression. Tha is he number of he lagged differences o be added as regressors. Following Kim (95) I use four lags of he differences in he regressions for he ADF es as he daa is quarerly. For he DFGLS ess I use he opimal lag generaed by he Ng-Perron sequenial-t mehod. I is well documened in he lieraure ha mos macro variables are non-saionary in heir levels bu saionary in heir firs-differences. If individual ime series urn ou o be non-saionary in heir levels, ha is, conain sochasic rend, i is sill possible ha hese sochasic rends are common across he series, leading o one or more saionary combinaions of he level series. The exisence of a long-run equilibrium (saionary) relaionship among economic variables is referred o as coinegraion. To find ou if he variables are coinegraed, I perform wo maximum likelihood ess, he race and maximum eigenvalue ess of Johansen (1988) and Johansen and Juselius (1990). The firs es ess he null hypohesis of a mos r coinegraing vecors agains he alernaive hypohesis of more han r coinegraing vecors. The second es ess he null hypohesis of exacly r coinegraing vecors agains he alernaive hypohesis of r 1coinegraing vecors. I firs repor he esimaes from he bivariae cases which are followed by he mulivariae cases. The maximum lag lengh for hese wo ess was se a eigh lags. For breviy, I repor only he resuls from he models conaining linear rends in he coinegraing regressions for 2, 4, 6, and 8 lags, respecively. 4. Resuls A preliminary look a he ime series plos (see figure 1) suggess ha TOT, M1, and GNS are poenially saionary in he levels. However, he hree uni roo ess do no provide irrefuable evidence of saionariy in he levels for any series. There is some evidence ha GNS, GBMT, TOT, and M1 are I (0). Since here is no clinching evidence in favor of hese variables being I (0), I err on he side of cauion and conclude ha all he variables used in his sudy are I (1). The hree ess srongly sugges ha all variables become saionary afer firs differencing (See ables 1, 2, and 3). 20

6 0 TB GBMTGBLT Research in Applied Economics TIME realb realgbl realgbm NEX TOT TIME realermsofrade exchbis GEXP GRECPT GNP GNS 0 M1 M BOT & BCA TIME real bo real balance on curren accoun TIME realm1sa realm2sa TIME oalgovreceips oalgovexpendiure TIME real gnp sa Negovsaving Figure 1: Time series plos 21

7 Table 1: Augmened Dickey Fuller Uni Roo Tes Resuls Variable (1) (2) (3) (4) BOT *** *** BCA *** *** TB *** *** GBMT *** *** GBLT *** *** NEX *** *** TOT ** *** *** M * *** M *** GNP *** *** GEXP ** GRECPT ** * GNS ** ** * N Column1 presens he ADF wih drif z() es saisics for he variables in levels. Column2 presens he ADF wih rend z() es saisics for he variables in levels. Column 3 presens he ADF wih drif z() es saisics for he variables in firs differences. Column 4 presens he ADF wih rend z() es saisics for he variables in firs differences. *** denoes significance a he 1% level according o MacKinnon s approximae p-value for z() saisics. ** denoes significance a he 5% level according o MacKinnon s approximae p-value for z() saisics. * denoes significance a he 10% level according o MacKinnon s approximae p-value for z() saisics. 22

8 Table 2: Dickey Fuller Generalized Leas Squares Uni Roo Tes Resuls Research in Applied Economics Variable (1) (2) (3) (4) BOT (3) 2.374(1) (3)*** (2)*** BCA (3) (3) (2)*** (2)*** TB (3) (3) (4)*** (4)** GBMT (1)** (2) (4)** (4)* GBLT (3) (1) (2)*** (2)*** NEX (3) 0.796(3) (3) (3) TOT (1) (1) (2) (2) M (3) (2) (1)** (2)** M (3) 1.549(3) (4) (4) GNP (2) 2.800(2) (1)*** (1)*** GEXP 0.145(4) 2.468(4) (3)** (3) GRECPT (3) 1.854(3) (2)*** (2)*** GNS (3)** (3)** (2)*** (2)*** N Column1 repors he mu saisics for he DFGLS es for he series being saionary around a mean. The variables are in he levels. Column2 repors he mu saisics for he DFGLS es for he series being saionary around a linear ime rend. The variables are in he levels. Column3 repors he mu saisics for he DFGLS es for he series being saionary around a mean. The variables are in he firs differences. Column4 repors he mu saisics for he DFGLS es for he series being saionary around a linear ime rend. The variables are in he firs differences. The figures in he parenheses are he opimum lag (Ng-Perron seq ). *** denoes he significance a he 1% level according o Ellio, Rohenberg, and Sock (1996) inerpolaed criical values. ** denoes he significance a he 5% level according o Ellio, Rohenberg, and Sock (1996) inerpolaed criical values. * denoes he significance a he 10% level according o Ellio, Rohenberg, and Sock (1996) inerpolaed criical values. 23

9 Table 3: Phillips Perron Uni Roo Tes Resuls Variable (1) (2) (3) (4) BOT *** *** BCA *** *** TB *** *** GBMT *** *** *** GBLT *** *** NEX *** *** TOT ** *** *** M *** *** M *** *** GNP *** *** GEXP *** *** GRECPT *** *** GNS *** *** Column1 repors he z() saisics from he PP es in levels wihou including he rend. Column2 repors he z() saisics from he PP es in levels wih rend. Column3 repors he z() saisics from he PP es in he firs differences wihou including he rend. Column4 repors he z() saisics from he PP es in he firs differences wih rend. *** denoes significance a he 1% level according o MacKinnon s approximae p-value for z() saisics. ** denoes significance a he 5% level according o MacKinnon s approximae p-value for z() saisics. * denoes significance a he 10% level according o MacKinnon s approximae p-value for z() saisics. I repor he esimaes from he bivariae coinegraion ess in ables 4 and 5. According o he bivariae coinegraion ess based on he race and maximum eigenvalue saisics, he U.S. balance of rade is coinegraed wih he reasury bill rae (6 lags), M2 measure of money supply (4 & 8 lags), governmen expendiure (2, 4, 6, & 8 lags), and ne governmen savings (6 & 8 lags). Whereas, he real balance on curren accoun is coinegraed wih he reasury bill rae (6 lags), nominal effecive exchange rae (6 & 8 lags), he real erms of rade (6 & 8 lags), M1 measure of money supply (6 & 8 lags), M2 measure of money supply (4, 6, & 8 lags), governmen expendiure (2 & 8 lags), and ne governmen savings (6 lags). Following 24

10 Kim (1995), when I consider a rivariae coinegraion seing, he resuls of which are presened in ables 6 and 7, he balance of rade and real GNP are found o be coinegraed wih he erms of rade (6 lags), M2 measure of money supply (8 lags), governmen expendiure (8 lags), and governmen receips (8 lags). Whereas, he balance on curren accoun and real GNP are coinegraed only wih he M2 measure of money supply (8 lags) and governmen expendiure (2, 8 lags). In an exended version of he coinegraion model a la Bahmani-Oskooee (1995), he resuls of which are presened in ables 8 and 9, where I add he M2 measure of money supply along wih he real GNP, I find ha he real erms of rade (6, 8 lags), he nominal effecive exchange rae (2, 8 lags), governmen expendiure (2, 4, 6, 8 lags), governmen receips (2, 6, 8 lags), reasury bill rae (6, 8 lags), GBMT (6, 8 lags), and GBLT (4, 6, 8 lags) are all coinegraed. Whereas, in a similar seing wih BOT replaced wih he balance on curren accoun, I find evidence o suppor he coinegraion hypohesis for he real erms of rade (6, 8 lags), he nominal effecive exchange rae (2, 4, 6, 8 lags), governmen expendiure (8 lags), governmen receips (2, 6, 8 lags), reasury bill rae (6, 8 lags), GBMT (6, 8 lags), and GBLT (4, 8 lags). Table 4: Johansen-Juselius Bivariae Coinegraion Resuls (Balance of Trade) Variable Rank Trace Max Trace Max Trace Max Trace Max Lags TB * GBMT GBLT NEX TOT M M * 18.23* ** 29.50** GNP GEXP ** 25.90** 22.65* 20.56* 20.22* 17.98* 19.35* GRECP T GNS * 22.88** 19.35* 18.77* ** denoes significance a he 1% level, * denoes significance a he 5% level 25

11 Table 5: Johansen-Juselius Bivariae Coinegraion Resuls (Balance on Curren Accoun) Variable Rank Trace Max Trace Max Trace Max Trace Max Lags TB * 17.48* GBMT GBLT NEX * * ** 8.28** TOT * * ** 6.67** M * 18.78* 21.68* 18.16* M * 18.73* 21.12* ** 26.22** * 5.22* 6.38* 6.38 GNP GEXP ** 23.53** * GRECP T GNS * 18.95* ** denoes significance a he 1% level, * denoes significance a he 5% level 26

12 Table 6: Johansen-Juselius Trivariae Coinegraion Resuls Variables in he coinegraing vecor TOT NEX M1 M2 GEXP GRECPT GNS TB GBMT GBLT R Trace Max Trace Max Trace Max Trace Max a n k Lag 2 Lag 2 Lag 4 Lag 4 Lag 6 Lag 6 Lag 8 Lag * ** * * ** denoes significance a he 1% level, * denoes significance a he 5% level 27

13 Table 7: Johansen-JuseliusTrivariae Coinegraion Resuls Variables in he R Trace Max Trace Max Trace Max Trace Max coinegraing vecor a n Lag 2 Lag 2 Lag 4 Lag 4 Lag 6 Lag 6 Lag 8 Lag 8 k BCA, GNP, TOT * 4.90* 5.90* 5.90* BCA, GNP, NEX BCA, GNP, M BCA, GNP, * 28.72* M BCA, GNP, GEXP * * 27.25* BCA, GNP, GRECPT BCA, GNP, GNS BCA, GNP, TB BCA, GNP, GBMT BCA, GNP, GBLT ** denoes significance a he 1% level, * denoes significance a he 5% level 28

14 Table 8: Johansen-Juselius Mulivariae Coinegraion Resuls Variables in coinegraing vecor M2, TOT M2, NEX M2, GEXP M2, GRECPT M2, TB M2, GBMT M2, GBLT R Trace Max Trace Max Trace Max Trace Max a n k Lag 2 Lag 2 Lag 4 Lag 4 Lag 6 Lag 6 Lag 8 Lag ** ** 43.03** ** * * 4.98* 6.67** 6.67** ** 38.31** ** 35.78** ** 31.71* 55.87* * ** 32.60* * 26.04* * 4.4* 4.51* 4.51* ** * 36.85** 55.04* * ** 54.85** * 33.76* 72.86** 43.97** * * ** 41.15** ** 24.64* ** denoes significance a he 1% level, * denoes significance a he 5% level 29

15 Table 9: Johansen-Juselius Mulivariae Coinegraion Resuls Variables in coinegrai ng vecor BCA GNP M2 TOT BCA GNP M2 NEX BCA GNP M2 GEXP BCA GNP M2 GRECPT BCA GNP M2 TB BCA GNP M2 GBMT BCA GNP M2 GBLT R Trace Max Trace Max Trace Max Trace Max a n k Lag 2 Lag 2 Lag 4 Lag 4 Lag 6 Lag 6 Lag 8 Lag * ** 42.00** * ** 6.73* ** 43.05** 60.29* 32.20* 55.36* ** 35.22* ** 32.16* ** 26.91* * * 37.91** 56.43* * ** 58.05** * ** 45.87** * ** 41.17** * ** denoes significance a he 1% level, * denoes significance a he 5% level 5. Conclusion In his paper I revisi he long-run deerminans of he U.S. exernal accoun balances. Previous sudies have esablished ha he fiscal, moneary, and commercial policy measures are coinegraed wih he U.S. exernal accoun balances. To he bes of my knowledge, none of hese sudies have esablished he ineres rae measures o be coinegraed wih he exernal accoun balances. The primary reason behind leaving he ineres rae measures migh have been he premise ha changes in money supply will ake care of he ineres rae 30

16 in he desired fashion. Based on an expanded sample, no only I confirm he findings of he previous sudies, bu also provide srong evidence o show ha he various ineres rae measures are also coinegraed wih he U.S. exernal accoun balances. This finding has imporan policy implicaions, because in order o ackle he problem of rade defici, on he moneary policy side he use of moneary aggregaes alone is no sufficien, as i migh no influence he ineres rae as desired by he Fed due o he money muliplier no realizing is full poenial, especially in a recessionary environmen where he commercial banks are unwilling o lend funds o he public. This paper relies on he coinegraion ess o deermine he long-run relaionships among several variables. Sudies ha use coinegraion ess do suffer from some weaknesses. The primary weakness is ha he coinegraing vecor may no remain consan over he ime span used in his sudy, which violaes he assumpion of consan coinegraing vecors. This violaion may be due o such facors as echnological change, naural calamiies, changing poliical environmen, ec. This shif in he coinegraing vecor will make he long-run relaionship uncerain. References Alse, J., & Bahmani-Oskooee, M. (1992). Are he Twin Deficis Really Relaed?: A Commen. Conemporary Economic Policy, hp://dx.doi.org/ /j b00218.x Bahmani-Oskooee, M. (1989). Effecs of he U.S. Governmen Budge on is Curren Accoun: An Empirical Enquiry. Quarerly Review of Economics and Business, Bahmani-Oskooee, M. (1992). Wha are he Long Run Deerminans of he U.S. Trade Balance? Journal of Pos Keynesian Economics, hp:// Bahmani-Oskooee, M. (1995). The Long Run Deerminans of he U.S. Trade Balance Revisied. Journal of Pos Keynesian Economics, hp:// Dara, A.F. (1988). Have Large Budge Deficis Caused Rising Trade Deficis? Souhern Economic Journal, hp://dx.doi.org/ / Engle, R.F., & Granger, C.W.J. (1987). Co-inegraion and Error Correcion: Represenaion, Esimaion, and Tesing. Economerica hp://dx.doi.org/ / Granger, C.W.J., & Newbold, P. (1974). Spurious Regressions in Economerics. Economerics, hp://dx.doi.org/ / (74) Journal of Johansen, S. (1988). Saisical Analysis of Coinegraion Vecors. Journal of Economic Dynamics and Conrol, hp://dx.doi.org/ /j mp x Johansen, S., & Juselius, K. (1990). Maximum Likelihood Esimaion and Inference on 31

17 Coinegraion Wih Applicaions o he Demand for Money. Oxford Bullein of Economics and Saisics, hp://dx.doi.org/ /j mp x Kim, K.H. (1995). On he Long-Run Deerminans of he U.S. Trade Balance: A Commen. Journal of Pos Keynesian Economics, 17(3) hp:// Krugman, P.R., & Baldwin, R.E. (1987). The Persisence of he U.S. Trade Defici. Brookings Papers on Economic Aciviy, hp://dx.doi.org/ / Magee, S.P. (1973). Currency Conracs, Pass Through and Devaluaion. Brookings Papers on Economic Aciviy, hp://dx.doi.org/ / Miller, S.M., & Russek, F.S. (1989). Are he Twin Deficis Really Relaed? Conemporary Economic Policy, hp://dx.doi.org/ /j b00577.x Perron, P. (1989). The Grea Crash, The Oil Price Shock, and he Uni Roo Hypohesis. Economerica, hp://dx.doi.org/ / Perron, P. (1990). Tesing for a Uni Roo in a Time Series wih a Changing Mean. Journal of Business and Economic Saisics, hp://dx.doi.org/ / Rosensweig, J.A., & Koch, P.D. (1988). The U.S. Dollar and he Delayed J-Curve. Economic Review. Federal Reserve Bank of Alana Copyrigh Disclaimer Copyrigh reserved by he auhor(s). This aricle is an open-access aricle disribued under he erms and condiions of he Creaive Commons Aribuion license (hp://creaivecommons.org/licenses/by/3.0/). 32