Exchange volatility and trade performance in Morocco and Tunisia: what have we learned so far?

Transcription

1 MPRA Munch Personal RePEc Archve Echange volaly and rade performance n Morocco and Tunsa: wha have we learned so far? amal Bouoyour and Refk Selm CATT, Unversy of Pau, France., ESC, Tuns Busness School, Tunsa. 04 Onlne a hps://mpra.ub.un-muenchen.de/660/ MPRA Paper No. 660, posed 7 January 05 08:33 UTC

2 Echange volaly and rade performance n Morocco and Tunsa: wha have we learned so far? Forhcomng n: Macroeconomcs and Fnance n Emergng Marke Economes hp:// Jamal BOUOIYOUR CATT, Unversy of Pau, France E-mal : amal.bouoyour@yahoo.fr Refk SELMI b ESC, Tuns Busness School, Tunsa. E-mal : s.refk@yahoo.fr Absrac: Ths paper aemps o assess wo neresng ssues for wo small open economes (Morocco and Tunsa). Frs, analyses he hsorcal behavour of nomnal echange rae, dfferenal prce and real echange rae unceranes. Second, nvesgaes he sably of he neracon beween echange volaly and epors n nomnal and real erms. Our man resuls reveal ha he effec of dfferenal prce volaly on epors eceeds ha of nomnal echange rae by a large margn n erms of duraon of perssence, ARCH and GARCH effecs and nensy of shock. The relaonshp appears comple. In Morocco, s negave and sgnfcan n 75.8% (as average) of cases n nomnal erms and n 77.% n real erms. Ths lnk s sronger n Tunsa wh averages, respecvely, equal o 85.88% and 89.99%. We assocae he apparenly med resuls o he dfferenal prce uncerany self sensve o ups and down ol prce movemens, swchng regme and leverage effecs. Keywords: echange volaly; oal epors; secoral epors; GARCH. JEL classfcaon: F3; F; C3.

3 . Inroducon The effec of echange rae uncerany on rade performance has been and connues o be a keen neres o economss, especally, because n he shor run, s sll dffcul o denfy he reasons behnd ecessve echange volaly. Throughou he hsorcal leraure, ecessve ups and down echange rae movemens have played a crcal role on epors (e.g. Baley e al. (986) and Koray and Lasrapes (989)). Gven he aenon o hs lnkage, a consderable leraure has been devoed o sudy (e.g. Brooks and McKenze (997), Dell Arcca (999) and McKenze (999), ec ). Some sudes have found a negave neracon beween currency rsk and epors (e.g. Baum e al (00), Vergl (00), ec ). Ohers have found posve effecs (e.g. De Grauwe (99) and Achy and Sekka (003), among ohers). Despe hs large srand of leraure on he consdered ssue, he heory upon hs relaonshp vares and here s no clear-cu lnkage o be found. Up o now, very few sudes advance convncng argumens on he conroversal effec of echange rae uncerany on rade. Earler, o reconcle he med resuls of pror researches, usng mea-regresson analyss, Corc and Pugh (008) provde evdence ha he effec of echange volaly on rade s lkely o be adverse when measured n real raher han nomnal erm, when low frequency raher han hgh frequency varaons are consdered and when less developed raher han developng counres are consdered. Bouoyour and Selm (03) add usng he same echnque (.e. mea-analyss) ha he effec of echange volaly on epors performance dffers dependng o modelng sraeges. Followng hs vewpon, we hough o revs he echange rae volaly effecs on epors performance by carryng ou dfferen GARCH specfcaons o deermne he volaly (.e. symmercal versus asymmercal, lnear versus nonlnear, wh hreshold order versus wh level shf and wh power effec versus componen effec, ec ). By dong so, we ry o hghlgh addonal eplanaons of conflcng resuls wdely epeced eher heorecally or emprcally.

4 Imporanly, hs paper provdes suppor for he vew ha n counres whch adop managed floa regme, real echange rae s more affeced by he volale behavor of dfferenal commody prce raher han ha of nomnal echange rae. Hence, he remander of he paper s srucured as follows: Secon presens an overvew of echange and rade polces n Morocco and Tunsa. Secon 3 presens a recall of he noon of volaly followed by he dfferen measures used n hs work. Secon 4 derves and develops our sraegy o evaluae he lnk beween echange rae uncerany and oal and secoral epors n boh nomnal and real erms.secon 5 s devoed o robusness check k. Secon 6 concludes our paper.. Echange and rade polces n Morocco and Tunsa Followng he demse of he Breon Woods sysem n 973, Morocco and Tunsa have chosen very smlar developmen pahs. Indeed, hey oped for lberalsm snce her ndependence. They also mplemened adusmen programs almos a he same me (.e. md- 80s). Then, hey sgned a he same me a sgnfcan assocaon agreemen wh he European Unon n order o creae a free rade zone. Regardng our subec, we also noe ha he wo counres have a close echange polces. Unl he early 980s, hese wo small open economes mananed he nomnal effecve echange rae whn a sable band. In 996, hese counres chose o adop managed floa regme o preserve eernal compeveness. Hence, he nomnal echange rae movemens were lmed n hs perod (see Fgure ). However, we depc n Fgure, a grea varably of dfferenal prce. Ths can be eplaned by he concenraon of parners desnaon (see Fgure 3) and he specalzaon n more volale producs (see Fgure 4), whch can ulmaely affec he prce of commodes a maor source of real echange volaly. Thereafer, parcularly n he perod from 999 o 003, he epors n percenage of GDP of boh consdered counres have no flucuaed largely. However, from 004 o 009, hese slgh movemens have evolved o connuous ncrease. I appears clearly from Fgure 5 The real echange rae s defned as he dfferenal prce of a baske of raded and non-raded goods beween he domesc and he foregn economy leadng o a grea vulnerably o he volaly of commody prces. In The concenraon of foregn rade and he dsrbuon of Moroccan and Tunsan epors are almos smlar (Baccouche e al. 008). 3

5 ha he oal epors conrbued o he producon of GDP wh a marked superory of Tunsa. In ha cone, should be added ha manufacured producs accouned he larges share represenng almos 50 percen of overall epors (see Fgure 6). Dependens srongly o manufacurng and mnng epors, Morocco and Tunsa may be poenally nfluenced by he commody prces uncerany. Despe hs smlary worhy observable n erms of echange and rade polces 3, he resuls assocaed o prce compeveness and economc growh can dffer. Ths creaes a need o assess wheher echange polcy plays crucal role n provdng ncrease n eporng ncenves n Morocco and Tunsa. 3. The noon of echange volaly 3.. Defnon A large srand of leraure defne echange volaly as he absolue percenage of changes n echange raes or he movng average sandard devaon of he echange growh rae; For eample, Baley e al. (986), Chowdhury (993) and Dell Arcca (999), among ohers. Theorecally, hese models may gnore he nformaon on sochasc processes hrough whch echange raes are generaed. Sascally, fnancal markes daa ofen ehb volaly cluserng, where me seres show perods of hgh volaly and perods of low volaly. In fac, me-varyng volaly s more common han consan volaly. In ha cone, GARCH eensons are consdered more approprae o defne volaly (e.g. Bollerslev e al. 993). These models are very useful for descrbng he volaly of he condonal varance by akng no accoun he characerscs of seres usng he pas errors n esmaes. Whle a varey of echange rae volaly measures have been used n he emprcal leraure (e.g. McKenze (998), Nabl and Varoudaks (00), Sandkov e al. (004), ec ), here s sll no consensus on whch measure s he mos approprae. In hs sudy, we perform robusness check usng 880 GARCH eensons 4 n order o choose he bes model able o deermne he volaly of nomnal echange rae, ha of 3 For deals abou echange rae regmes and rade reforms, see Append A. 4 The measures of volaly are repored n Append B. We used 7 eensons GARCH. For each specfcaon, we apply 04 combnaons ((,) ; (,) ; (,) ; (,)), followng 03 dsrbuons (Gaussan, Suden and GED). For each eenson, we apply a few opons. Thus, each specfcaon can be zero Mean (u =0), consan mean (u = u 0 ) ou n mean (u = u 0 + u σ - ) 4

6 dfferenal prce and ha real effecve echange rae. Hence, s useful o recall ha GARCH modelng naed by Engle (98) and generalzed by Bollerslev (986) assume ha he condonal varance follows an ARMA process. Subsequenly, has undergone a remarkable evoluon of GARCH eensons (e.g. Tong (990), Nelson (990), Dng e al. (993) and Zakon (994), ec ). The sandard GARCH can be wren as follows: q p () Where,,, e are parameers o be esmaed and are successvely ARCH effec, GARCH effec, errors esmaes, condonal varance and reacon o he shock. GARCH models may be lnear or nonlnear 5, symmercal or asymmercal 6. These specfcaons are summarzed n append B 7. Snce he qualy of he resuls depends no only on he relevance of he model, bu also he effcency of he algorhm, we consder dfferen opons (.e. models wh zero mean, consan mean and n mean, wh movng average process and whou movng average process and followng Maquard and Bernd- Hall-Hall-Hausman algorhmc opmzaons) based on he sudy of Koksal (009) and Bouoyour e al. (0). 3.. Esmaes 3... Descrpve sascs and perhaps also wh MA erm (wh movng average process) or whou MA erm (whou movng average process). These models are appled o algorhmc opmzaons Maquard and Bernd-Hall-Hall-Hausman (For more deals, see appendces C and D). 5 Nonlnear models are hose wh funcon ndcaor ha ake he value f he resdue of he prevous perod s negave and 0 oherwse. The condonal varance follows wo dfferen processes dependng on he sgn of he error erms or accordng o he dynamcs of he condonal sandard devaon of reurns (Threshold). I s pecewse lnear funcons dependng on he sgn of he shock (Zakon, 994). 6 Symmerc models were nroduced by Engle (98) and Bollerslev (986). The formulaon of hese eensons GARCH mposes a sensvy of he rsk premum volaly. These models do no ake no accoun cyclcal behavor or sudden shocks seres ha s why hey are raher resrcve. Insead, asymmerc models descrbe he behavor of he condonal varance usng good or bad news. The asymmery of he volaly can be eplaned, for eample, by he nervenon of he moneary auhores (Engle, 990). 7 For dealed analyss of GARCH specfcaons, see Bouoyour e al. (0). 5

7 To measure he volaly of nomnal echange rae, dfferenal prce and real echange rae, seems mporan o choose he opmal models able o deermne he hsorcal behavors of her unceranes. To sar, he REER s consruced by dvdng he rade-weghed foregn prce level nde (P *) by he correspondng domesc prce level nde (P ), afer pror converson o a common numerare usng nomnal effecve echange rae (NEER ). REER =NEER (P */P ) () Fgure 7 allows us o hghlgh changes n echange rae n boh nomnal and real erms ha wll neres us hroughou hs sudy. We clearly observe a large and permanen varably for hese varables. In addon, he resuls repored n Table reveal ha he coeffcen of kuross s greaer han 3 for he echange rae and he dfferenal prce of consdered counres, ecep for he Moroccan REER. Ths ndcaes ha he dsrbuon s less flaened han he normal n he frs cases and nversely n he las one. The skewenss s posve n he case of Morocco across all me seres. Ths shows ha he asymmercal dsrbuon s more plausble n Moroccan case, unlke Tunsa where he skewness s negave, suggesng he effecveness of symmercal dsrbuon. Then, he Jarque Bera es reveals a low value for he NEER, (P/P*) and REER, leadng o accep he normaly hypohess for hese varables. Then, o confrm he esence of asymmercal effecs on he condonal varance of echange raes and dfferenal prce, we follow Zvo (008) by deermnng he correlaon beween reurns and lagged squared reurns (see Table ). We noe ha he correlaons beween reurns and lagged squared reurns have a negave values for he NEER, (P/P*) and REER for Moroccan case and posve values for Tunsan case. Ths ndcaes he presence of leverage effec n Morocco and s absence n Tunsa. These resuls seem unepeced, because Morocco and Tunsa are very close n erms of echange and rade polces (e.g. Baccouche e al. (008) and Emmono and Rey (008)). Remans o check wheher hese prelmnary resuls are sold and do no change subsanvely when we move o hsorcal evaluaon Hsorcal evaluaon 6

8 The applcaon of GARCH models allows us o deermne he echange rae volaly n nomnal and real erms. To choose he bes model among many eensons GARCH, we used varous nformaon crera (.e. Akake, Schwarcz and Hannan-Qunn). These crera evaluae models based on hsorcal behavor of each varable. The model wh he lowes values s mos preferred 8. The dscrmnaon funcon dffers from one creron o oher. I should be noed ha hese crera are suffcen o udge he qualy of our esmaes (see Bouoyour e al. 0). For nomnal echange rae volaly, he bes model chosen s he N-GARCH (, ) for Morocco and he GARCH (, ) for Tunsa. The specfcaons chosen o deermne he relave commody prces uncerany are respecvely, P-GARCH (, ) for Morocco and T- GARCH (, ) for Tunsa. The bes models chosen n real erms n Morocco and Tunsa are respecvely he E-GARCH (, ) and GARCH-M (, ) 9. In Append C, we mach he Kernel densy of seleced models. The -as shows a posve value of nformaon crera, larger values mean beer opons assocaed wh chosen model. Ths mehod favors he Gaussan dsrbuon for Morocco and Tunsa, wh leverage effec for he frs case and s absence for he second case. Imporanly, Table 3 shows ha for boh economes n queson, he duraon of perssence, he nensy of shock, he leverage effec and ARCH and GARCH effecs of nomnal echange volaly and prce dfferenal are less mporan han real echange volaly. More precsely, for Morocco, he leverage effec s posve for boh nomnal and real erms, ndcang ha bad news have greaer mpac han good news, bu hs effec s much more mporan for he real echange rae. The effec of a posve shock for he NEER s less mporan han he REER, and smlarly for he negave shock. I s observed ha he bad and good news have he same mpac on he condonal varance. The perssence of he condonal volaly s mporan for boh cases wh less mporance for he NEER. The sum of ARCH and GARCH s more mporan for he REER han he NEER. The REER volaly s more sensve o he (P/P*) uncerany han ha of NEER n erms of perssence, nensy of shock and ARCH and GARCH effecs. For Tunsa, he duraon of perssence and he ARCH and GARCH effecs of NEER, (P/P*) and REER are more mporan han Morocco. Ths resul can no be confrmed wh he 8 The creron of Schwarcz s more parsmonous han ha of Akake snce nroduces more parameers n he model. 9 For dealed eplanaons of each specfcaon, see Anderson e al. (009). 7

9 absence of asymmercal effec n he model chosen o measure he Tunsan echange uncerany. Bu esng he same models for boh counres, we show he same resul,.e. Tunsan volaly s more perssen. Thus, he condonal varance of nomnal echange rae behaves beer han ha of he real echange rae. Fgure 8 confrms he more volale behavor of he real echange rae han he nomnal echange rae and n Tunsa han n Morocco. Therefore, he hsorcal evaluaon of echange rae volaly n nomnal and real erms confrms our prelmnary analyss. Remans hereafer o be seen wheher he lnkage beween echange volaly and epors n Morocco and Tunsa vary dependng o he duraon of perssence, nensy of he shock, leverage effec and ARCH and GARCH effecs assocaed o echange uncerany. To do so, we eplore a vas array of regressors. 4. Esmaes of he lnkage beween echange volaly, relave commody prce uncerany and epors Mos of he sudes on he effec of echange volaly and rade performance assume ha he correlaon beween hese wo varables s conssenly ambguous bu wh magnude less han 0% (e.g. McKenze (999), Vergl (00), Achy and Sekka (003), Sadkov e al. (004), Rey (006) and Eger and Zumaquero (007) ). Assumng away hs queson, our purpose here s o regress he oal and secoral epors on echange volaly n nomnal and real erms and oher conrol varables. We use hen he followng model. XPR = α 0 +α LnREER +α Ln VOLR - + α 3 Ln (X ) R +ξ (3) XPN = α 0 +α LnNEER +α Ln VOLN - + α 3 Ln (P/P*) + α 4 Ln VOL (P/P*) - +α 5 Ln(X ) N +ξ (4) Where ξ and ξ are he error erms. -The logarhm of nomnal and real effecve echange raes, respecvely, noed NEER and REER. There s no a pror reason ha a varably of REER maches eacly ha of he NEER, especally because he REER depends on boh nomnal echange rae and relave 8

10 commody prces 0. Followng our epor funcon specfcaon, he domesc and foregn commody prce ndces are producer prces. -The logarhm of he volaly n nomnal and real erms, appromaed respecvely by he volaly of nomnal echange rae (VOLN) and ha of real effecve echange rae (VOLR). Echange volaly s deermned usng opmal GARCH model chosen by nformaon crera. Accordng o Nabl e al. (004) and Eger and Zumaquero (007), echange volaly ncrease he degree of uncerany n erms of epor compeveness and can have an ambguous effec dependng o he secor of epors. -The volaly of dfferenal prce (Ln (P/P*)) s deermned usng he opmal GARCH specfcaon seleced by nformaon crera and Kernel densy. The dfferenal prce uncerany dampens rade performance and can almos dsrup he raecory of epors n prmary producng counres (e.g. Blaman e al. (007) and Arezk e al. (0)). -The logarhm of (X) presens conrol varables. We chose as eplanaory varables he naonal GDP and he GDP of rade parners whch have a pullng role n epors (e.g. Nabl and Varoudaks, 00). For he GDP of mporng counres, we used he weghed average of he man parners of Morocco and Tunsa, where he European zone corresponds o he share of epors o he euro area, he wegh for he Amercan zone represens he share of epors o he Amercan counres... Quarerly daa from he Inernaonal Moneary Fund and Econsas, coverng he perod from 996 o he las quarer of 009 were used o assess he een o whch echange volaly may affec he oal and secoral epors n Morocco and Tunsa, and o check f dfferenal prce uncerany really plays an mporan role n generang he resuls. All hese varables are aken a me unless he volaly s aken a me -. 0 I presens he dfferenal beween naonal commody prces and foregn commody prces. Some sudes use consumer prces, bu accordng o Eger and Zumaquero (007), hs nde conans admnsered prces, ndrec aes and mpored producs, whch can no reflec eacly he effec of relave commody prce on epors. I s obvous ha eporers need a delay o adus her prces. 9

11 4.. Toal of epors Our resuls summarzed n Table 4 reveal ha he mpac of echange rae volaly on he oal of epors s ambguous. For Morocco, he relaonshp beween hese wo varables appears negave and sascally sgnfcan. Ths resul s valuable n boh nomnal and real erms. Ths may be arbuable o he lack of developed hedgng facles n Morocco ha can proec eporers agans echange rae uncerany (e.g. Achy and Sekka, 003). However, he same relaonshp appears posve and sgnfcan for Tunsan case. If we lnk hs resul o he rsk averson (e.g. Ozurk and Acaravc, 006), hs can be far from rue for he smple reason ha we has no a clear ndcaor for he degree of rsk averson. Alhough, he real echange rae s heorecally deermned by many facors (e.g. Rogoff, 996), sudes on s fundamenals n developng counres emphasze he mporan role of erms of rade whch are closely lnked o ol prces (e.g. Eger and Zumaquero, 007). Hence, he real echange can be nfluenced by he movemens of crude of ol. We ry hen o check hs vewpon by subracng he share of energy from oal epors. Our resuls show ha overall epors s more affeced by echange volaly n Tunsa han n Morocco,.e. an ncrease by 0% n echange rae volaly reduces Moroccan epors by.47% and ncrease of Tunsan epors by 5.70%. By subracng he energy epors from he oal of epors, we show a negave neracon beween he consdered seres, mplyng ha he relaonshp beween echange volaly and Tunsan epors s sensve o he changes n ol prces, whch s no he case for Morocco. No surprsngly, he lack of sensvy owards he subracon of he energy secor from he Moroccan oal epors s nensely due o he low proporon of hese producs n he global epors of hs counry (see Fgure 4). In addon, we observe clearly ha an ncrease by 0% of nomnal echange volaly promps a drop n nomnal epors by.78% n Morocco and 4.43% n Tunsa. We noce also ha he effec of dfferenal prce on epors s always sronger han ha of nomnal echange rae,.e. an ncrease n 0% of dfferenal prce uncerany yelds o a decrease n nomnal epors by 3.0% n Morocco and by 5.85% n Tunsan case. 0

12 Besdes, he effec of foregn echange and dfferenal prce unceranes on he oal of epors has ncreased wh he nroducon of srucural breakpons 3. We fnd also ha he Moroccan and Tunsan GDP eplan he compeveness of epors more han he GDP of her mporng counres, whch means ha he domesc producon of hese counres s less mporan han he foregn addressed demand. Ths resul s heavly epeced because he problem n hese counres s her nably o mee eernal demand. More precsely, n Morocco and Tunsa, he domesc demand plays an mporan role as deermnans of epors. Conversely, he foregn demand has a mnor role n deermnng epors performance. We arbue hs nensely o he weakness of producon capacy (e.g. Baccouche e al. 008). 4.. Manufacurng secor Despe he compeveness of manufacurng secor on he oal of epors n boh Morocco and Tunsa (e.g. Sekka, 0), we fnd a negave and sgnfcan effec of volaly on manufacured epors for Moroccan case, whch mples ha he companes specalzed n manufacured producs would no be able o modfy s npus opmally o adap ecessve echange volaly. However, hs relaonshp s posve and sgnfcan for Tunsa. As oal epors, we hough o subrac energy s share. Ths causes a change n sgns (See Table5). Tha s o say, an ncrease by 0% n real echange volaly leads o a decrease of he level of manufacurng epors by 3.4% n Morocco and an ncrease by 4.7% n Tunsa. For hs las counry, a subracon of he energy share mples a decrease n manufacurng epors by 3.3%. Ineresngly, he effec of dfferenal prce volaly on nomnal epors sll more mporan han ha of nomnal echange rae and for Tunsa han Morocco. Inroducng he dummy varable of economc crss, he lnkage beween echange volaly and epors becomes sronger. 3 The srucural break corresponds o a dummy varable, noed DV, represenng he economc crss whch akes he value 0 before he economc crss,.e. before he second quarer of 008 and for he perod wh hgher volaly assumed afer he economc crss.

13 4.3. Agrculural secor Many emprcal sudes have been conduced on he effec of commody prce and ehange rae volales on secoral rade, n parcular, agrculure (e.g. Maskus (986), Larapes and Koray (990), Kumar and Dhawan (99), ec ). Accordngly, Kwanashe e al. (994) argue ha a deprecaon of he effecve echange rae leads o an ncrease of agrculural prces of epors and boos of domesc producon. Our resuls don suppor hs fndng revealng a posve and nsgnfcan neracon beween he wo varables (see Table 6). Ths resul s epeced because seems normal ha agrculure s eporers have a neural aude o echange uncerany snce here are pershable producs Mnng secor Our resuls ndcae ha mnng secor s negavely affeced by he volaly of echange rae n boh nomnal and n real erms,.e. an apprecaon by 0% n echange rae volaly affecs he flows of mnng rade leadng o a drop of epors by 7.30% n Morocco and by 4.73% n Tunsa (see Table 7). Ths can be eplaned by he fac ha he relaonshp beween echange volaly and mnng secor s hghly correlaed wh he nernaonal prces of mnng and phosphaes (e.g. Varangs e al. 004). We noe also ha he mpac of dfferenal prce volaly on epors sll more nense han ha of nomnal echange rae. Addonally, all he coeffcens assocaed o echange rae uncerany eher n nomnal or n real erms appear small and que comparable o hose assocaed o oher secors. Ths s manly due, based on he works of Bouoyour and Rey (005) and Bouoyour and Selm (03) o he hgh degree of compeveness assocaed o hs secor n boh counres Energy secor Hghly flucuaons of he crude of ol and s mpac on echange rae have rekndled varous quesons regardng he dynamc behavor of echange rae afer changes n ol prce (e.g. Bénassy e al. (005) and Couder e al. (008)). In our paper, we evaluae only he relaonshp beween echange rae volaly and energy secor n Tunsan case, as he share assocaed of Moroccan energy epors s very low (see Fgure 3). We fnd ha hs lnkage s posve and sascally sgnfcan. More precsely, an ncrease by 0% n real echange rae volaly leads o an ncrease by 4.73% n energy epors (see Table 8). The effec of nomnal

14 echange uncerany s negave and mnor comparable o ha of dfferenal prce. Ths fndng may be nensely arbuable he lack of a rough gudance abou how echange polcy should cope wh ol prce uncerany. 5. Robusness check In hs sudy, we performed 880 esmaes based on dfferen GARCH eensons (lnear versus nonlnear, symmercal versus asymmercal). These specfcaons follow varous dsrbuons (Gauss, Suden, GED), can be zero mean, consan mean or n mean, wh movng average erm or whou and based on Maquard or Bernd-Hall-Hall-Hausman algorhmc opmzaons. Our resuls were gven for only he bes models. To avod overloadng of our presenaon, we hough ha s no crucal o presen he resuls of mnng, agrculural and energy secors and o concenrae only on hose of oal epors and manufacurng secor. The resuls summarzed n Table 9 reveal ha: () () For Morocco, n 77.77% of cases, he volaly of nomnal echange rae has a negave and sgnfcan effec on oal of epors 4, n 73.88% of cases for he dfferenal commody prce uncerany 5 and n 77.% for he real echange volaly 6. Almos smlarly for manufacurng secor. For Tunsa (afer subracng energy s share), n 83.33% of cases, he nomnal echange rsk has a negave and sgnfcan mpac on overall epors 7. In 88.44% of cases for he relave prce uncerany and n 89.99% for he real echange volaly. As Morocco, he manufacurng secor n Tunsa s characerzed by average effecs of VOLN, VOL (P/P*) and VOLR almos equal o hose of oal epors. No surprsngly, hs fndng may be due o he large share of hs secor n he oal of epors. 4 Posve and sgnfcan n only.66% of cases, posve and nsgnfcan n 6.0% of cases and negave and nsgnfcan n 4.99% of esmaes. 5 Posve and sgnfcan n 4.44% of cases, posve and nsgnfcan n 6.66% of esmaes and negave and nsgnfcan n 4.99% of esmaes. 6 Posve and sgnfcan n only.77% of cases, posve and nsgnfcan n 6.66% of cases and negave nsgnfcan n 9.6% of cases. 7 Posve and sgnfcan n only.77% of esmaes, posve and nsgnfcan n 9.44% of cases and negave nsgnfcan n 3.88% of esmaes. 3

15 6. Concluson The obecve of hs sudy s wofold: Frs, o choose he bes model able o deermne he volaly of nomnal echange rae, dfferenal prce and real echange rae of Morocco and Tunsa. Second, o regress oal and secoral epors on echange volaly n boh nomnal and real erms and oher varables epeced o nfluence rade performance. I appears from our resuls neresng nsghs: () () () The nomnal echange rae s less volale han ha of dfferenal commody prce and real echange rae wh regard o he duraon of perssence, he nensy of shock, he leverage effec (.e. sgn of nnovaons) and ARCH and GARCH effecs and her effecs hen on oal and secoral epors are mnor. The share of energy n oal and manufacurng epors can occur an nverse effec of echange volaly. The effec of echange volaly on epors depends o srucural breaks. Ths resul does no change dependng o secor-o-secor varaon. These fndngs are robus accordng o he dfferen specfcaons used. Ineresngly, f we esmae echange volaly-epors relaonshp usng he naïve models (.e. average absolue devaon and movng average absolue devaon), he assocaon (no repored here) sll robus and sable. In a nushell, our resuls make an nuve conrbuon. We show ha he sudy-o-sudy varaon on he focal relaonshp s manly arbuable o secors specalzaon, asymmeres and srucural breaks n volaly process. Thus and o he een ha echange rae and commody prce unceranes are cosly, a harder mplcaon here s wheher Moroccan and Tunsan polcymakers should aemp o mgae her effecs on rade performance. 4

16 References Achy, L. and Sekka, K. (003). The European sngle currency and MENA s epors o Europe. Revew of Developmen Economcs 7, p Agahe, C. (994). Echange rae volaly and rade: A survey. Documen de raval n 94-5 de la Banque de Canada. Anderson, T-G., Davs, R-A., Kreb, J-P. and Mkosh, T. (009). Handbook of fnancal me seres. Sprnger publcaons. Arezk, R., Lederman, D. and Zhao, H. (0). The relave volaly of commody prces: a reapprasal. IMF Workng Papers /68, Inernaonal Moneary Fund. Arezk, R., Dumrescu, E., Freyag, A. and Qunyn, M. 0. Commody Prces and Echange Rae Volaly: Lessons from Souh Afrca Capal Accoun Lberalzaon. IMF Workng Papers /68, Inernaonal Moneary Fund. Baccouche, R., Bouoyour, J. M Henn, H. and Mouley, S. (008). Dynamque des nvesssemens, muaons secorelles e converblé du compe de capal: mpacs des mesures de lbéralsaon e epérences comparées Tunse Maroc. MPRA paper Baley, M-J., Tavlas, G-S. and Ulan, M. (986). Echange Rae Varably and Trade Performance: Evdence for he Bg Seven Indusral Counres. Revew of World Economes, p Bollerslev, T. (986). Generalzed auoregressve condonal heeroskedascy. Journal of Economercs 3, p Bollerslev, T., Engle, R.F and Nelson, D.B., (993). ARCH models n Handbook of Economercs IV, Elsever Scence. Bouoyour J. and Rey S., (005). Echange Rae Regme, Real Echange Rae, Trade Flows and Foregn Drec Invesmens: Morocco. Afrcan Developmen Revew 7(), p Bouoyour J., Jellal M. and Selm R., (0). Es-ce que les flu fnancers rédusen la volalé de la crossance marocane? CATT workng paper, Unversy of Pau. Bouoyour, J. and Selm, R. (03). Echange rae uncerany and epor performance: wha mea-analyss reveals? MAER-NET 03 Colloquum, Unversy of Greenwch. Clark, P-K. (973). A Subordnaed Sochasc Process Model wh Fne Varance for Speculave Prces. Economerca, Economerc Socey, vol. 4(), p , January. 5

17 Chowdhury, A. (993). Does Echange Rae Volaly Depress Trade Flows? Evdence from Error Correcon Models. The Revew of Economcs and Sascs 75, p Ćorć, B. and Pugh, G. (00). The effecs of echange rae varably on nernaonal rade: a mea-regresson analyss. Appled Economcs, 4, p Dell Arcca G., (999). Echange Rae Flucuaons and Trade Flows: Evdence from he European Unon. IMF Saff Papers 46(3), p Éger, B. and Morales-Zumaquero, A. (007). Echange Rae Regmes, Foregn Echange Volaly and Epor Performance n Cenral and Easern Europe: Jus Anoher Blur Proec? Wllam Davdson Insue Workng Papers Seres wp78, Unversy of Mchgan. Honroyanns G., Swamy P-A., Tavlas G-S. and Ulan M., (005). Some furher evdence on echange rae volaly and epors.workng paper of he Bank of Greece n 8. Hossen M-R. e Moghadas R., (00). Echange rae volaly and Iranan epors. publcaons IDOSI. Koray, F. and Lasrapes W., (989). Real Echange Rae Volaly and U.S. Blaeral Trade: a VAR Approach. The Revew of Economcs and Sascs 7-4, p McKenze, M-D., (999). The Impac of Echange Rae Volaly on Inernaonal Trade flows. In: Journal of Economc Surveys 3, p Koksal, B. (009). A comparson of condonal volalyesmaors for he ISE naonal nde reurns. Journal of Economc and Socal Research (), p. -8. McKenze, M-D., (998). The Impac of Echange Rae Volaly on Ausralan Trade Flows. Journal of Inernaonal Fnancal Markes. Insuons and Money 8, p Nabl M-K., Keller J. and Véganzonès-Varoudaks M-A. (004). Echange Rae Managemen whn he Mddle Eas and Norh Afrca Regon: he Cos o Manufacurng Compeveness. Workng paper n 8, Amercan Unversy of Beru, p.-3. Nelson D-B., (99). Condonal heeroskedascy n asse reurns: A new approach. Economerca 59, p Ozurk, I. (006). Echange Rae Volaly and Trade: A Leraure Survey. Inernaonal Journal of Appled Economercs and Quanave Sudes, Euro-Amercan Assocaon of Economc Developmen, vol. 3(), p

18 Sadkov A., Zeng L., Clark P., We S-P. e Tamrsa N., (004). A New Look a Echange Rae Volaly and Trade Flows. IMF Occasonal Papers 35, Inernaonal Moneary Fund. Tong H., (990). Nonlnear me seres analyss snce 990: Some personal reflecons. Unversy of Hong Kong and London school economcs. Vergl, H., (00). Echange rae volaly n Turkey and s effecs on rade flows. Journal of Economc and Socal Research 4 (), p Zakoan, J.M. (994). Threshold Heeroskedasc Models. Journal of Economc Dynamcs and Conrol 8, p Zvo, E., Cho, K. and Yu W., (008). Long Memory versus Srucural Breaks n Modelng and Forecasng Realzed Volaly. Workng Papers UWEC FC, Unversy of Washngon. 7

19 Fgure. Echange rae n nomnal and real erms Morocco NEER REER Tunsa NEER REER Source : Econsas TM. The daa of Morocco are dsponble on : hp:// The daa of Tunsa are dsponble on : hp:// 8

20 Fgure. Evoluon of dfferenal prce nde (P/P*)_Morocco (P/P*)_Tunsa Source : Econsas TM. Fgure 3. Share of he prncpal parners desnaon n he oal of parners Europe USA Ohers Morocco Tunsa Source: CIA, World Facbook (009). hps:// 9

21 Fgure 4. Share of each secor n he oal of epors Morocco Q-998Q 998Q3-00Q 00Q3-004Q 004Q3-007Q 007Q3-009Q Manufacurng secor Agrculural secor Mnng secor Energy secor Tunsa Q-998Q 998Q3-00Q 00Q3-004Q 004Q3-007Q 007Q3-009Q Manufacurng secor Agrculural secor Mnng secor Energy secor Source : UN Comrade hp://comrade.un.org/db/ 0

22 Fgure 5. Evoluon of oal epors as a % of GDP Morocco Tunsa Source: Usherbrooke daa and auhors' calculaons. hp://perspecve.usherbrooke.ca/blan/sasques/7 Fgure 6. Average share of quarerly secoral epors n he oal of rade from 996 o Manufacurng secor Agrculural secor Mnng secor Energy secor Morocco Tunsa Source : UN Comrade. hp://comrade.un.org/db/

23 Fgure 7. Nomnal echange rae, dfferenal prce and real echange rae reurns MOROCCO <r REER < <r NEER < <r (P/P*) < r NEER r (P/P*) r REER TUNISIA <r REER < <r NEER < <r (P/P*) < r NEER r (P/P*) r REER Noe: r NEER : Nomnal effecve echange rae reurn; r (P/P*) : relave commody prces reurns; r REER : Real effecve echange rae reurn; Source : Econsas TM and auhors calculaons.

24 Table. Descrpve sascs Morocco Tunsa r REER r (P/P*) r NEER r REER r (P/P*) r NEER Mean Medan Mamum Mnmum Sd. Dev Skewness Kuross Jarque-Bera Noe : r NEER : Nomnal effecve echange rae reurn; r (P/P*) : relave commody prces reurns; r REER : Real effecve echange rae reurn; Source : Econsas TM. Table. Tes of leverage effec n he condonal volaly Morocco Tunsa REER (P/P*) NEER REER (P/P*) NEER ( r, r ) Noe : NEER : Nomnal effecve echange rae; REER : Real effecve echange rae; (P/P*) : dfferenal of prces; Source : Econsas TM 3

25 Table 3. Hsorcal evaluaon: Parameers of varance equaon Morocco Tunsa VOLR VOL (P/P*) VOLN VOLR VOL (P/P*) VOLN w 3.05E-06* (.44) 3.8E-05* (.49) -.893** (-.665).9E-05 (0.748).64E-05 (0.0) (0.06) -0.0** (-.37) -0.48* (-.84) 0.09 (0.058) 0.489* (.65) 0.33 (0.36) ** (.3570) 0.0 (0.6) * (-.446) (0.057) *** (0.3) 0.839*** (5.990) 0.785* (.45) *** (4.006) 0.7 (0.375).388*** (6.59) (-0.560) *** (-0.759) 0.7** (.94) -0.00*** (-3.637) 0.083*** (6.7807) Duraon of perssence Symmerc : q p q p Asymmerc : Inensy of shock q Symmerc : Negaveshock : q q Posveshock : Leverage effec ARCH and GARCH effecs q p Noe : VOLN : Volaly of nomnal effecve echange rae ; VOL (P/P*) : Volaly of dfferenal prces of commodes ; VOLR : Volaly of real effecve echange rae ; w : The reacon of condonal varance; α : ARCH effec; β : ARCH effec; Y : Leverage effec. 4

26 Fgure 8. Hsorcal evaluaon : Echange and dfferenal prce volales Morocco Tunsa VOLN.004 GARCH VOL(P/P*) VOLR.0005 GARCH

27 Table 4. Esmaon of he relaonshp beween echange volaly, relave commody prces uncerany and oal epors Morocco Tunsa () () (3) (4) (5) (6) In real erms C *** ** * (0.085) (-3.38) (0.96) (-.0) (0.64) (-.04) REER -.784* -.54* * (-.03) (-.37) (-.764) (-0.704) (-0.409) (-0.97) VOLR -0.47* * 0.570** -0.38* -0.93*** -0.93** (-.068) (-.568) (.009) (-.99) (-5.09) (-.) GDP 0.608* 0.86***.09*.68***.534*.68*** (.90) (6.45) (.3) (4.709) (.567) (4.033) GDP* 0.09** 0.3* 0.03* 0.04* 0.04* 0.043* (.084) (.748) (.33) (.507) (.533) (.50) DV * (-.55) *** (-3.059) ** (-.474) R In nomnal erms C.34**.76** **.3-3.0* (.37) (.8) (.000) (-.) (0.77) (-.044) NEER -.93* -.004* -0.86** (-.000) (-.3) (-.7) (-0.866) (-0.504) (-.45) (P/P*) -.3* -.90* * -.330* -.35* (-.334) (-.003) (-.66) (-.6) (-.08) (-.099) VOLN -0.78** -0.89* * -0.57** * * (-.46) (-.5) (-.664) (-.000) (-.09) (-.4) VOL (P/P*) -0.30* * 0.585** -0.36* 0.95*** -0.9** (-.4) (-.995) (.063) (-.07) (5.8) (-.67) GDP 0.7* 0.904* 0.74*.75***.00*.75** 6

28 (.57) (.80)) (0.804) (3.074) (.03) (.6) GDP* 0.053* 0.094* 0.030* 0.054** 0.054** 0.057** (.468) (.634) (.784) (.00) (.055) (.04) DV * (-.350) * (-.44) -0.43* (-.00) R Noe : DV : dummy varable equal o unl 008 :Q o 00 :Q4 (economc crss) and equal o 0 oherwse; ***, **, * : sgnfcan respecvely a %, 5% e 0% ; () : Moroccan epors whou dummy varable ; (): Moroccan epors wh dummy varable; (3): Tunsan epors wh energy share and whou dummy varable ; (4) : Tunsan epors whou energy share and whou dummy varable ; (5) : Tunsan epors whou energy share and wh dummy varable ; (6) : Tunsan epors wh energy share and wh dummy varable. 7

29 Table 5. Esmaon of he relaonshp beween echange volaly, relave commody prces uncerany and manufacurng secor Morocco (5%, 48%) Tunsa (53%, 50%) () () (3) (4) (5) (6) In real erms C * * 5.45*.054 (-0.60) (-.543) (-0.00) (-.09) (.033) (0.677) REER -0.90*** -.680** * -.04** -.49** * (-.) (-.390) (-.37) (-.39) (-.595) (-.3) VOLR -0.34* -0.83* 0.47* -0.34*** -0.33* -0.80** (-.80) (-.756) (.835) (-5.40) (-.05) (-.78) GDP 0.890*.87***.370*.95**.39***.44*** (.65) (3.90) (.48) (.009) (4.003) (3.006) GDP* 0.358* 0.36* 0.00* 0.050* 0.0* 0.076* (.467) (.30) (.88) (.664) (.07) (.9) DV * (-.5) *** (-3.40) ** (-.903) R In nomnal erms C *** ** * (0.085) (-3.38) (0.96) (-.0) (0.64) (-.04) NEER -.784* -.54* * (-.03) (-.37) (-.764) (-0.704) (-0.409) (-0.97) (P/P*) -.0** * * ** -0.59*** (-.34) (-.309) (-0.765) (-.84) (-.55) (-4.63) VOLN * -0.09** -0.03* -0.68* -0.87* -0.68* (-.) (-.043) (-.007) (-.5) (-.43) (-.6) VOL (P/P*) -0.47* * 0.570** -0.38* -0.93*** -0.93** (-.068) (-.568) (.009) (-.99) (-5.09) (-.) GDP 0.608* 0.86***.09*.68***.534*.68*** 8

30 (.90) (6.45) (.3) (4.709) (.567) (4.033) GDP* 0.09** 0.3* 0.03* 0.04* 0.04* 0.043* (.084) (.748) (.33) (.507) (.533) (.50) DV * (-.55) *** (-3.059) ** (-.474) R Noe : DV : dummy varable equal o unl 008 :Q o 00 :Q4 (economc crss) and equal o 0 oherwse; ***, **, * : sgnfcan respecvely a %, 5% e 0% ; () : Moroccan manufacured epors whou dummy varable ; (): Moroccan manufacured epors wh dummy varable ; (3): Tunsan manufacured epors wh energy and whou dummy varable ; (4) : Tunsan manufacured epors wh energy and dummy varable ; (5) : Tunsan manufacured epors whou energy and dummy varable; (6) : Tunsan manufacured epors whou energy and wh dummy varable. The frs percenage s he average of he marke share of each secor n oal epors for oal of perod; he second s he percenage marke share of each secor for he pas year. 9

31 Table 6. Esmaon of he relaonshp beween echange volaly, relave commody prces uncerany and agrculural secor Morocco (0%, %) Tunsa (7%, 9%) () () (3) (4) In real erms C 0.064** -.6** (.533) (-.300) (0.340) (0.093) REER -0.8*** -0.89* -.56** -.3 (-3.55) (-.79) (-.339) (-.05) VOLR (0.940) (0.696) (.089) (0.578) GDP 0.875* 0.99***.959***.080* (.436) (6.544) (6.58) (.843) GDP* 0.00* 0.57* * (.804) (.95) (0.34) (.95) DV ** (-.04) ** (-.08) R In nomnal erms C 0.97**.5* * (.8) (.78) (0.79) (-.603) NEER -0.66*** -0.83*** * -.53 (-3.49) (-3.30) (-.465) (-0.58) (P/P*) -.04* -.306* -.00* -0.9* (-.055) (-.) (-.53) (-.) VOLN (0.76) (0.89) (0.763) (0.56) VOL (P/P*) (0.893) (0.86) (0.669) (0.83) GDP 0.875* 0.97** 0.9* 0.70*** 30

32 (.436) (.63) (.03) (3.93) GDP* 0.00* 0.079* 0.09* 0.037* (.804) (.8) (.) (.5) DV *** (-3.09) (-0.45) R Noe : DV : dummy varable equal o unl 008 :Q o 00 :Q4 (economc crss) and equal o 0 oherwse; ***, **, * : sgnfcan respecvely a %, 5% e 0% ; () Moroccan agrculural secor whou dummy varable; () : Moroccan agrculural secor wh dummy varable ; (3): Tunsan agrculural secor whou dummy varable; (4) : Tunsan agrculural secor wh dummy varable. The frs percenage s he average of he marke share of each secor n oal epors for oal of perod; he second s he percenage marke share of each secor for he pas year. 3

33 Table 7. Esmaon of he relaonshp beween echange volaly, relave commody prces uncerany and mnng secor Morocco (3%, 3%) Tunsa (8%, 7%) () () (3) (4) In real erms C -.7** -.537* * (.3) (-.33) (0.79) (-.603) REER -0.66*** -.35** * -.53 (-3.49) (-.40) (-.465) (-0.58) VOLR ** * * * (-.80) (-.49) (-.669) (-.083) GDP 0.908* 0.873** 0.9* 0.8*** (.903) (.08) (.03) (3.93) GDP* 0.00* 0.079* 0.09* 0.034* (.804) (.8) (.06) (.5) DV *** (-3.09) (-.54) R In nomnal erms C 0.980* ** * (.960) (-.03) (0.66) (-.447) NEER * ** -0.79* -0.4 (-.764) (-.704) (-.409) (-0.97) (P/P*) -.093** -.6* -.039* -0.9* (-.099) (-.003) (-.53) (-.07) VOLN -.09* ** *** * (-.33) (-.64) (-5.09) (-.) VOL (P/P*) ** -0.88* * (-.009) (-.99) (-.576) (-.60) GDP.09*.6*** 0.43*.044* 3

34 (.63) (4.709) (.49) (.70) GDP* 0.33* 0.44* 0.93* 0.05*** (.33) (.607) (.533) (4.033) DV ** (-.000) ** (-.474) R Noe : DV : dummy varable equal o unl 008 :Q o 00 :Q4 (economc crss) and equal o 0 oherwse; ***, **, * : sgnfcan respecvely a %, 5% e 0% ; () Moroccan mnng secor whou dummy varable; () : Moroccan mnng secor wh dummy varable ; (3): Tunsan mnng secor whou dummy varable; (4) : Tunsan mnng secor wh dummy varable. The frs percenage s he average of he marke share of each secor n oal epors for oal of perod; he second s he percenage marke share of each secor for he pas year. 33

35 Table 8. Esmaon of he relaonshp beween echange volaly, relave commody prces uncerany and energy secor Tunsa (%, %) () () In real erms C -.5* (-.076) REER -0.48* (-.398) VOLR 0.473* (.669) GDP 0.9* (.03) GDP* 0.09* (.06) -.94*** (-7.733) -0.59*** (-0.05) 0.37 (5.886) 0.686*** (7.563) 0.007* (.35) DV ** (-.7) R In nomnal erms C (0.33) NEER -0.48* (-.398) (P/P*) -.07** (-.34) VOLN -0.* (-.4) VOL (P/P*) 0.7* 4.30) (-0.83) * (-.774) -0.8* (-.4) -0.63* (-.78) 0.00* (.754) GDP 0.774* 0.80*** 34

36 (.30) (.660) GDP* 0.034* (.73) 0.035* (.54) DV ** (-.038) R Noe : DV : dummy varable equal o unl 008 :Q o 00 :Q4 (economc crss) and equal o 0 oherwse; ***, **, * : sgnfcan respecvely a %, 5% e 0% ; (): Tunsan energysecor whou dummy varable; () : Tunsan energy secor wh dummy varable. The frs percenage s he average of he marke share of each secor n oal epors for oal of perod; he second s he percenage marke share of each secor for he pas year. 35

37 Table 9. Robusness check: Sgnfcance of parameers assocaed o echange rae and dfferenal prce unceranes Toal of epors Morocco (77.77% ; 73.88% ; 77.%) Tunsa (83.33% ; 88.44% ; 89.99%) Posve+ Negave+ Posve+ Negave+ Posve+ Negave+ Posve+ Negave+ sgnfcan Sgnfcan nsgnfcan nsgnfcan sgnfcan sgnfcan nsgnfcan nsgnfcan Gauss : VOLN.66% 80.00% 6.66%.66% 3.33% 86.66% 6.66%.66% VOL (P/P*) 3.33% 78.33% 5.00% 3.33% 0.00% 9.66% 8.33% 0.00% VOLR 0.00% 86.66%.66%.66%.66% 9.66% 5.00% 3.33% Suden: VOLN 0.00% 80.00% 5% 5.00% 0.00% 90.00% 6.66% 3.33% VOL (P/P*) 5.00% 76.66%.66% 6.66% 0.00% 93.33% 6.66% 0.00% VOLR 3.33% 75.00% 8.33% 3.33% 0.00% 93.33% 5.00%.66% GED: VOLN.66% 73.33% 6.66% 8.33% 5.00% 73.33% 5.00% 6.66% VOL (P/P*) 5.00% 66.66% 3.33% 5.00% 6.66% 80.00% 0.00% 3.33% VOLR 5.00% 70.00% 8.33% 6.66% 5.00% 85.00% 0.00% 0.00% Manufacurng secor Morocco (69.44% ; 74.99% ; 78.88%) Tunsa (75.55% ; 79.44% ; 8.%) Gauss : VOLN.66% 68.33% 8.33%.66% 3.33% 83.33%.66%.66% VOL (P/P*) 3.33% 78.33% 6.66%.66% 5.00% 86.66% 0.00% 6.66% VOLR.66% 80.00% 5.00% 3.33% 3.33% 90.00%.66% 5.00% Suden: VOLN 0.00% 76.66% 3.33% 0.00%.66% 76.66% 8.33% 3.33% VOL (P/P*) 3.33% 8.66% 5.00% 0.00% 3.33% 80.00% 3.33% 3.33% VOLR.66% 85.00% 5.00% 8.33% 3.33% 8.66% 0.00% 5.00% GED: VOLN.66% 63.33%.66% 5.00%.66% 66.66% 5.00% 6.66% 36

38 VOL (P/P*).66% 65.00% 3.33% 8.33% 5.00% 7.66% 3.33% 0.00% VOLR 3.33% 7.66% 5.00% 0.00%.66% 75.00% 5.00% 8.33% Noe: Gauss s Gaussan dsrbuon; suden s a sandardzed Suden dsrbuon; GED s a generelzed error dsrbuon; he frs percenage s he average of sgnfcance assocaed o he volaly of nomnal echange rae; he second percenage s he average of sgnfcance assocaed o he volaly of dfferenal commody prce; he hrd percenage s he average of sgnfcance assocaed o he volaly of real echange rae. 37

39 Append A. Echange and rade polces by counry Echange reforms Perod Morocco Tunsa Modfcaon of he weghs of drham s baske. Peg of he dnar o a baske epanded he dollar Deprecaon of he drham by % (e.g. Emmono and Rey, 008). Recovery perod of real overvaluaon and srucural adusmen program Adopon of a deprecaon polcy. Lberalzaon of he curren accoun Converbly of he curren accoun. Creaon of spo marke Changes of weghs of drham s baske makng a favor o euro. Devaluaon of he dnar by 5% o promoe epors (e.g. Bouoyour and Selm, 03) Undervaluaon of he drham by 9% (e.g. Baccouche e al. 008). A perod of sably of he dnar condoned by he delberae polcy of he cenral bank. Trade reforms Srucural adusmen hghlghng mechansms o promoe epors. Fnancal suppor for nvesors by seng up an epor promoon fund Accesson of Morocco o he General Agreemen on Tarffs and Trade. A new sraegy focusng on he developmen of economy Novel rade law compable wh he measures General Agreemen wh World Trade Organzaon. aken by he General Agreemen wh World Trade Organzaon Grans by 30% for nvesors n agrculure o overcome he vagares of he weaher. Parnershp and co-developmen wh Europe and eles and clohng agreemen Assocaon agreemen Europe-Morocco leadng o Progressve dsmanlng of arffs and Agadr a sgnfcan arff reducons on he varous Agreemen made o lberalze progressvely rade. producs and Agadr agreemen wh man obecve o reduce arffs and make he counry more aracve o foregn nvesors Dsmanlng of he mul fber agreemen and he accesson of Chna o he world rade organzaon. The dsmanlng of he mul fber agreemen and he esablshmen of a fund access for epors o help companes o dversfy her markes. 38

40 39 Append B. Measures of volaly Naïve models. Average absolue devaon / ) ( / n e e n VOL. Movng average absolue devaon / ) ( / m e e m VOL GARCH models Lnear Nonlnear Symmerc Asymmerc 3. GARCH (Bollerslev, 986) p q 4. GARCH-M (GARCH n mean, Bollerslev e al. 993) r 5. C-GARCH (Componen GARCH, Dng e al. 993) ) ( ) ( ) ( 6. QGARCH (Quadrac GARCH, Senana, 995) p q b ) ( 7. IGARCH (Engle and Bollerslev, 986 and Nelson, 99) ) ( ) ( p q 8. AGARCH (Asymmerc GARCH, Engle e al. 987) p q ) ( 9. TGARCH (Threshold GARCH, Zakoan, 994) p q _ ) (

41 40 0. GJR-GARCH (Glosen e al., 993) p q I ( ) ( 0. GJR-PARCH (GJR power GARCH, Glosen e al. 993) p q I ( ) ( 0. EGARCH (Eponenal GARCH, Nelson, 99) p q z z ) log( )) / ( ( ) log( 3. PGARCH (Power GARCH, Hggns and Bera, 99) p q 4. A-PGARCH (Asymmerc power GARCH, Dng e al., 993) p q ) ( 5. NGARCH (Nonlnear GARCH, Bollerslev e al.993) p q ) ( 6. NGARCHK (Nonlnear GARCH wh one shf, specal case of NGARCH) p q ) ( 7. NPGARCHK (Nonlnear Power GARCH wh shf, Bollerslev e al., 993) p q ) ( Noe: e : The logarhm of echange rae and ha of relave commody prce, m : movng average order (m=8, smlar o he value used n he maory of sudes on hs subec, lke Larape and Koray (985). : condonal varance, : condonal sandard devaon, : reacon of shock, 0 : reacon of shock, : ARCH erm, : GARCH erm, : error erm; I : denoes he nformaon se avalable a me ; I - : denoes he nformaon se avalable a me -;z : he sandardzed value of error erm where / z ; : nnovaon, : leverage effec; ) /( : corresponds o he uncondonal varance ; b : quadrac order, : power parameer, : level shf. For more deals, see Anderson e al. (009) and Bouoyour e al. (0).

42 Append C. Hsorcal evaluaon: Choce of he bes model Morocco VOLN : Models chosen usng Akake creron N o Model q p Dsrbuon MA erm Mean Leverage Opmzaon Dsrbuon NGARCH Gauss No Zero Yes Maquardl A-PARCH Gauss Yes In Yes BHHH NGARCH Yes Consan Yes Maquardl NGARCH GED No Consan Yes Maquardl E-GARCH Yes Consan Yes BHHH 4.35 VOL (P/P*) : Models chosen usng Akake creron N o Model q p Dsrbuon MA erm Mean Leverage Opmzaon Dsrbuon P-GARCH Gauss Yes In No Maquardl 5.45 GARCH-M Yes In No Maquardl C-GARCH No Consan No Maquardl I-GARCH Gauss No Consan No Maquardl GARCH-M Gauss Yes Consan No Maquardl 5.4 VOLR : Models chosen usng Akake creron N o Model q p Dsrbuon MA erm Mean Leverage Opmzaon Dsrbuon E-GARCH Gauss Yes Zero Yes Maquardl I-GARCH No Zero No BHHH GARCH-M No In No BHHH I-GARCH Gauss No Consan No Maquardl C-GARCH Gauss Yes Zero No BHHH