Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu Peer-revewed and Open access journal ISSN: 1804-5006 www.academcpublshngplatforms.com The prmary verson of the journal s the on-lne verson BEH - Busness and Economc Horzons Volume 9 Issue 4 2013 pp.86-96 DOI: http://dx.do.org/10.15208/beh.2013.24 Fnancal crss and exchange rates n emergng economes: An emprcal analyss usng PPP- UIP-Framework Abdul Rashd 1, Mashael Bn Saedan 2 1 Internatonal Insttute of Islamc Economcs (IIIE), Islamabad, Pakstan 2 Dar al Uloom Unversty, Ryadh, Saud Araba correspondng e-mal: abdulrashd@u.edu.pk correspondng address: Room D-13, Sector H-10, Islamabad, Pakstan Ths paper emprcally nvestgates the effects of 2008 fnancal crss on exchange rate determnaton n PPP-UIP framework for four emergng countres, usng monthly date over the perod 1981-2012. The results suggest that the recent fnancal crss led to change the role of exchange rate determnants n exchange rate determnaton. The fndngs also reveal that the effects of fnancal crss on the exchange rate are dfferent n all the four emergng economes. The fndngs of the study are of sgnfcant for polcy makers n desgnng effectve polces n order to reduce the effects of fnancal crss on exchange rates. JEL Classfcatons: C39; F29; F31 Keywords: Exchange rates; fnancal crss; PPP; UIP; emergng economes Introducton Globalzaton makes the world more correlated through many channels such as exports, mports, and foregn captal flows. More economc and fnancal ntegraton spread the mpact of 2008 fnancal crss faster globally. Ths fnancal crss was one of the worst dsasters n the hstory snce the crss of 1929, whch causes a huge fall n the household wealth and the whole fnancal market (Brunnermeer, 2008; Crotty, 2009; Galbrath, 2009). Moreover, t caused several serous effects n the global exchange rate regme (Fratzscher, 2009). Therefore, t would be sgnfcant to nvestgate the mpact of 2008 fnancal crss on exchange rate determnaton. The foregn exchange market s consdered as one of the hghest tradng value n all fnancal markets. Thus, the fluctuaton n the exchange rate mght lead to a sgnfcant mpact on the underlyng economy, n partcular when rsk and uncertanty are growng n the fnancal markets. The recent fnancal crss causes several doubts and uncertanty regardng the sustanablty of exstng fnancal system across the world. Along wth numerous other socoeconomc effects, the fnancal crss also affects the determnaton of exchange rates n both short and long run (Keblowsk and Welfe, 2011). The man am of ths paper s to examne how 2008 fnancal crss affects exchange rate determnaton n purchasng power party (PPP) and uncovered nterest rate party (UIP) framework. We study the mpact of fnancal crss on exchange rates n PPP-UIP framework. We also examne whether the mpact of fnancal crss on the exchange rate dffers across countres wth dfferent economc and socal backgrounds. Thus, the emprcal analyss s curred for four emergng economes, namely Egypt, Inda, Turkey, and Thaland, usng monthly data coverng the perod from 1981 to 2012. The rest of the paper s organzed as follows. In subsequent secton we provde revew of the lterature related to PPP, UIP, and fnancal crss. Secton 3 explans the PPP, UIP, - 86-2013 Prague Development Center
Busness and Economc Horzons Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu and the combned form of PPP and UIP. Secton 4 descrbes data and presents the emprcal fndngs. Secton 5 concludes the paper. Lterature revew Several researchers, such as Fratzscher (2009), Blanchard et al. (2010), Wong and Wa L (2010), Keblowsk and Welfe (2011), and Tsangardes (2012),examne the mpact of the subprme fnancal crss upon the global exchange rate system usng dfferent methods. Most of these studes have used ether the PPP or UIP whle explorng fnacal crss effects on exchnage rates.however, other studes, such as Rashd (2009),Keblowsk and Welfe (2011), andjaramllo and Servanb (2012),testthe jont form of the PPP and the UIP. These studes provde strong evdence that both PPP and UIP condtons together play a sgnfcant role n the determnaton of exchnage rate. Thus, consderng only one of these parates whle examnng the behavor of exchnage rate may yeld bas results. Jaramllo and Servan (2012) usng trade-weghted exchange rates test whether the PPP and UIP hold for the Peruvan economy.ther study covers the perod1997-2011. They document that the mxture of PPP and UIP sgnfcantly explans the dynamcs of the nomnal effectve exchange rate n Peru. They also argue that although the central bank s nterventon s sgnfcant for smoothng the exchange rate short term volatlty, t dose not have a long term nfluence on the exchange rate. Keblowsk and Welfe (2011) propose a new modellng of exchange rate that extends the captal enhanced equlbrum exchange rate (CHEER) model, whch s the combnaton of the PPP and UIP. Specfcally, they nclude a ndependent credt default rsk nto ther specfcaton to take nto account the decsons of fnancal nvestors. They usecontegrated vector autoregressve (VAR) system and monthly data from Poland and Euro area. Ther results suggest that the soveregn credt rsk s an mportant factor that determnes the exchange rate along wth the prce and the nterest rate dfferentals. As a result of the fnancal crss, there was uncertanty n the fnancal market, whch may affect the determnatons of exchange rates (Keblowsk and Welfe, 2011). The recent fnancal crss caused abrupt fluctuatons n the global exchange rate regme (Fratzscher, 2009) that had an unfavorable mpact n the emergng countres manly through external shocks; mostly by two channels: net foregn captal flows and exports (Blanchard et al., 2010). Consequently, the experence of recent fnancal crss left several lesson for emergng economes, partcularly, regardng the choce of the exchange rate regme (Tsangardes, 2012). Most of the emergng countres have constructed consderable postve holdngs of US dollar treasury blls from the tme of the crses of the late 1990s, whereas they face a boom n the FDI captal nflows at the same perod (Devereux and Sultherland, 2009). Nonetheless, Fratzscher (2009) hghlghts that the subprme fnancal crss breakdown the dea that the US dollar plays a vtal role n the nternatonal adjustment process because of the sharp declne n the assets prce and the huge deleveragng procedure amd fnancal organzatons. Thus, the economes went to recesson, whch led to huger hazard to human securty that became global fnancal nsecurty and had several serous effects upon the emergng countres, especally the poorest (Fukuda-Parr, 2008). Fratzscher (2009) analyzes the data from 50 advance and emergng countres to nvestgate the change n the global exchange rate durng the recent fnancal crss perod. He states that a sharp fluctuaton n the global exchange rate confguratons has caused by the recent fnancal crss. He strongly recommends the mportance of the macroeconomc fundamental, n specfc, suffcent foregn exchange reserves and sound current account postons to pawn the captal flow reversal. Blanchard et al. (2010) study the mpact of 2008 fnancal crss n the emergng countres dong a case study of three emergng countres (Latva, Russa, and Chle). They used a 2013 Prague Development Center - 87 -
Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu smple of cross-country specfcaton, connectng unexpected trade and fnancal varables over two quarters. Ther results do not support the hypothess that holdng more foregn reserves helps lmt the drop n output n the dsaster. Fukuda-Parr (2008) also observed that even though some developng countres ncreased the reserves and surpluses, they badly affected from the recent fnancal crss. Revewng prevous emprcal studes, we fnd that there s not enough emprcal evdence on how the recent fnancal crss affects exchange rate determnaton n the PPP-UIP framework. Therefore, n ths paper, we study the mpacts of 2008 fnancal crss on the exchange rate n four emergng economes after takng nto account the factors that related to both PPP and UIP. Economc theory Purchasng Power Party (PPP) Under the PPP, the change n the prce levels between any two countres determnes the exchange rate for these countres when expressed n same currency, whch s the assumpton of low of one prce (Mshkn, 2010; Plbeam, 2006). The relatve form of PPP s as: e t d f pt pt t t =1,,T and =1,,N (1) e t = log nomnal exchange rate for th domestc country s defned as the number of domestc currency unts needed to purchase one foregn currency unt. p = log domestc prce level at tme t d t f p t = log of foregn country prce level at tme t e t = trade shock wth zero mean and fnte varance = constant T = total observatons N = total number of countres ncluded n the analyss. Uncovered nterest rate party (UIP) The UIP theory allows the captal movements and t states that the change of the nterest rate between any two countres determnes the exchange rate for these countres (Plbeam, 2006). The UIP can be expressed as: t d f t t t e 1 t =1,,T and =1,,N (2) d t = log domestc nterest rate at tme t f t = log foregn nterest rate at tme t D= dfference operator - 88-2013 Prague Development Center
Busness and Economc Horzons Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu m t = error term wth zero mean and fnte varance Combnng PPP and UIP MacDonald and Taylor (1992) and Rashd (2009) state that there s not enough emprcal evdence supportng the PPP and UIP separately as many researchers faled to fnd t. They also argue that there are several factors that caused the falure of PPP; for nstance, trade barrers, relatve prce levels, mperfect market, and transport costs. Whle the lmted captal moblty and rsk premum are examples of the factors that cause the falure of UIP. Therefore, the two models mght not be evaluated ndvdually when explorng the determnants of exchange rate. The man advantage n the combned form of PPP and UIP s that both party condtons complete each other. The approach, whch combned PPP and UIP n a sngle equaton, s the captal enhanced equlbrum exchange rate (CHEER) model. A key dea of the CHEER model s that a statonary connecton relable wth the assets and good markets nterdependence adjustment nto the equlbrum s shaped by non-statonary devaton from PPP and UIP ( Stephens, 2004; Rashd, 2009). Hence, PPP s a long-term crcumstance, whch supposed that the PPP forms n the expectatons foundaton n the UIP crcumstance. So, ths lnk s transferred to the UIP equaton by pluggng equaton (1) nto equaton (2), whch yeld the followng model: d f d f pt pt et t t t (3) After rearrangng, we obtan the followng equaton: e t d f t t d f pt pt (4) t Where, Fnally, to examne the mpact of fnancal crss, we augment equaton (4) by addng the nteractons between explanatory varables and a fnancal dummy whch takes value one after the crss perod and zero otherwse. Specfcally, the equaton takes the followng form. e d f d crss f crss d f crss t t pt D pt D t t D d f t pt pt (5) Econometrcs Framework Data The monthly data coverng the perod from 1981-2012 are taken from the Internatonal Fnancal Statstcs (IFS) database for four emergng countres. We select four dfferent emergng countres: frst a bg Asan economy (Inda), second an economy heavly reles on toursm (Thaland), thrd a country (Turkey), whch s closest to Europe geographcally, 2013 Prague Development Center - 89 -
Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu and fnally one Mddle East ol produced country that s Egypt. The selecton of these countres allows us to make an nterestng comparson. The varables ncluded n the analyss are the exchange rate whch s defned as the prce of one unt of foregn currency (US dollar) n domestc currency of the underlyng country, the nterest rate whch s money market rates and s defned as average of rates on deposts wth maturty of one to 90 days n natonal currency, consumer prce ndex (CPI), producer prce ndex (PPI), and share prce ndex (SPI) whch refers to the ndex of securty prces and s computed based on prces of ordnary shares n all ndustres traded on the major stock exchange of respectve country. All varables are n log form. Addtonally, we create a dummy varable (takng value 1 for post crss perod (11/2007 untl 12/2012) and 0 for pre crss perod) n order to dentfy the mpact of the fnancal crss. Table 1 presents the summary statstcs. TABLE 1. SUMMARY STATISTICS Egypt Inda Turkey Thaland Panel Varables Mean S.D Mean S.D Mean S.D Mean S.D Mean S.D Exchange rate 0.906 0.8496 3.299 0.5895-3.204 3.4140 3.409 0.2259 1.102 3.2210 Interest rate 2.495 0.2161 20144 0.2663 3.705 0.4259 1.684 0.9062 2.507 0.9172 CPI 3.892 0.9082 4.024 0.6596 0.567 3.8220 4.290 0.3301 3.193 2.5110 PPI 3.860 0.8517 4.052 0.5841 4.311 0.3208 4.258 0.3505 4.114 0.6010 In January 2003, the Central Bank of Egypt (CBE) announced floatng exchange rate for the Egyptan pound (LE). Snce the 1960s untl fscal year 2003, Egypt s exchange rate has been pegged wth US$ and exhbted a large extent of rgdty. Wth the announcement of the float, Egypt s currency has been deprecated untl October 2004. However, for next four years, the exchange rate has contnuously and reached LE/US$ 5.32. Ths apprecaton s manly attrbuted to foregn exchange earnngs from ol exports, toursm, Suez Canal, and FDI. After gettng ndependence 1947, Inda has shfted from a par value system to a basket-peg exchange rate regme. In December 1971, wth the breakdown of the Bretton Woods System, Indan rupee was pegged wth pound sterlng. In the early 1990s, Inda recognzed that Indan macroeconomc polces and structural factors had sgnfcantly created balance of payment problems. To overcome balance of payment dffcultes, Inda shfted to a flexble exchange rate system n 1992-93. Before 1963, there was a float of Tha Baht. However, the floatng exchange rate regme was ended on October 20, 1963, and the Tha Baht was pegged wth US$. In March 1978, nstead of only US dollar, Thaland pegged Baht to a weghted basked currency of major tradng partners of Thaland. Afterwards, Thaland adopted controlled floatng rate and allowed Baht to float wthn a narrow range. For the perod of 1984-1997, Thaland has used monetary and fnancal measures to sheld the Baht value aganst US currency. In December, 2006, the bank of Thaland mposed several controls on the exchange rate whch resulted n a consderable spread between offshore and onshore exchange rates. However, on March 3, 2008, the controlled were manly lfted and thus, there s now sgnfcant dvergence between offshore and onshore exchange rates. Before the perod of 1980, there was a fxed exchange rate regme n Turkey. The fxed exchange rate system was replaced wth crawlng peg exchange rate system n 1980. However, Turkey adopted a floatng exchange rate regme for the perod 1989-1993. Durng the perod of 1994-1996, the Turksh Lra exchange rate has been determned wthn crawlng band. In 2000, Turkey started the stablzaton program and under ths program, Turkey adopted currency-peg regme. Fnally, snce 2001, there s a floatng exchange rate regme n Turkey. - 90-2013 Prague Development Center
Busness and Economc Horzons Emprcal results Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu Ths secton presents emprcal fndngs that examne the mpacts of fnancal crss on exchange rate determnaton under two well-know partes (PPP and UIP) for four emergng economcs. To test the effect of fnancal crss on exchange rate determnaton, we run several specfcatons followng prevous emprcal studes, such as Rashd (2009), Jaramllo and Servan (2012). In ths paper, we consder the USA as a foregn country. We begn our emprcal analyss by testng the order of ntegraton of each varable ncluded n the study. Specfcally, we apply the augmented Dckey-Fuller (hereafter ADF) unt root test on both levels and frst dfferences of the varables and test whether they are ntegrated of order zero or one. We also apply the panel unt root test, namely Fsher-type tests. Ths test s based on the ADF test. Specfcally, Fsher-type tests estmate unt root tests for each ndvdual ncluded n the panel and combne the p-value from these tests to construct an overall test. The hypothess for these tests s that all panels contan unt roots, whle the alternatve hypothess s that at least one panel s statonary. The results from the ADF tests for levels and frst dfferences of the varables are presented n Tables 2 and 3, respectvely. TABLE 2. THE ADF UNIT TEST RESULTS FOR LEVELS Varables Constant Constant + trend Egypt Exchange rate 1.089 (0.719) -1.006 (0.943) Interest rate -1.475 (0.545) -2.454 (0.351) CPI -3.085 (0.027) -1.595 (0.794) PPI -1.705 (0.428) -1.573 (0.803) Inda Exchange rate -2.091 (0.248) -1.126 (0.924) Interest rate -0.044 (0.954) -1.934 (0.366) CPI -1.024 (0.744) -1.892 (0.658) PPI -1.205 (0.671) -1.174 (0.915) SPI -1.286 (0.635) -2.12 (0.534) Turkey Exchange rate -2.587 (0.095) 0.726 (1.000) Interest rate -0.572 (0.877) -1.576 (0.801) CPI -2.849 (0.051) 2.274 (1.000) PPI -1.103 (0.713) -6.066 (0.000) SPI -2.226 (0.197) -1.535 (0.816) Thaland Exchange rate -1.885 (0.339) -1.781 (0.713) Interest rate -1.492 (0.537) -2.723 (0.226) CPI -0.971 (0.763) -1.211 (0.908) PPI 0.581 (0.987) -3.058 (0.116) SPI -1.521 (0.522) -3.549 (0.034) USA Interest rate 0.242 (0.974) -1.087 (0.931) CPI 0.734 (0.990) -1.903 (0.653) PPI -3.524 (0.007) -2.962 (0.143) SPI -1.438 (0.563) -1.275 (0.894) Note: The fgures gven n parentheses are p-value. The null hypothess for the ADF unt root test s that the seres n non-statonary, whle the alternatve hypothess s that the seres s statonary. For Egypt, both exchange rate and nterest rate are non-statonary at ther levels. These results hold regardless we estmate the ADF equaton wthout or wth a lnear tme trend. 2013 Prague Development Center - 91 -
Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu However, consumer prce ndex s statonary when the ADF equaton s estmated wthout a lnear tme trend. It appears non-statonary when the trend s ncluded n the equaton. Lookng at results for Inda we fnd that only the nomnal exchange rate s statonary at ts level when only constant s ncluded n the ADF equaton. However, when the trend s added nto the equaton, t becomes non-statonary. All others varable are non-statonary regardless of whether a lnear tme trend s ncluded or not except the nterest rate, whch s non-statonary wthout the trend, and become statonary wth the trend. Turnng to Turkey we observe that shares prce ndex s statonary when the ADF equaton runs wthout the trend but all the other varables are non-statonary. However, the consumer prce ndex, prce producton ndex, and the exchange rate are statonary when the trend s controlled but the SPI turns non-statonary. The nterest rate n both cases appears non-statonary. For Thaland, all the underlyng varables appear non-statonary at ther levels, as we do not reject the null of unt root at any acceptable level of sgnfcant. These fndngs hold for all varables except share prce ndex even when we nclude a lnear tme trend n the ADF equaton. Share prce ndex appears statonary when we consder a lnear tme trend. Fnally, n case of the USA, the unt root test results provde evdence that all the underlyng varables are also non-statonary at ther levels. The results from the ADF unt root tests for frst dfferences of the varables are gven n Table 3.It s clear from the table, for all varables, we reject the null of non-statonary n favor of the alternatve hypothess of statonary for all countres. Ths mples that all the varables are ntegrated of order one. TABLE 3. THE ADF UNIT TEST RESULTS FOR FIRST DIFFERENCES Varables Constant Constant + trend Egypt Exchange rate -9.350 (0.000) -9.376 (0.000) Interest rate -6.950 (0.000) -6.946 (0.000) CPI -9.698 (0.000) -10.266 (0.000) PPI -10.023 (0.000) -10.214 (0.000) Inda Exchange rate -7.778 (0.000) -7.944 (0.000) Interest rate -11.417 (0.000) -11.535 (0.000) CPI -9.933 (0.000) -9.978 (0.000) PPI -9.262 (0.000) -9.337 (0.000) SPI -9.005 (0.000) -9.030 (0.000) Turkey Exchange rate -7.373 (0.000) -7.959 (0.000) Interest rate -8.939 (0.000) -9.128 (0.000) CPI -5.568 (0.000) -6.412 (0.000) PPI -12.558 (0.000) -12.535 (0.000) SPI -7.828 (0.000) -8.112 (0.000) Thaland Exchange rate -9.830 (0.000) -9.880 (0.000) Interest rate -6.807 (0.000) -6.798 (0.000) CPI -9.403 (0.000) -9.438 (0.000) PPI -9.365 (0.000) -9.460 (0.000) SPI -6.823 (0.000) -6.893 (0.000) USA Interest rate -8.198 (0.000) -8.288 (0.000) CPI -8.756 (0.000) -8.831 (0.000) PPI -9.146 (0.000) -9.584 (0.000) SPI -8.609 (0.000) -8.691 (0.000) Note: The fgures gven n parentheses are p-value. The null hypothess for the ADF unt root test s that the seres n non-statonary, whle the alternatve hypothess s that the seres s statonary. - 92-2013 Prague Development Center
Busness and Economc Horzons Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu Table 4 presents the results for panel unt root tests for the underlyng varables at both levels and frst dfferences. Most of the varables appear non-statonary when a lnear trend term s ncluded n the equaton. However, the results from estmatng the Fshertype tests for frst dfferences of the varables show that all the varables are statonary. These results hold when we even nclude a lnear tme trend n the equaton. Overall, the results from panel unt root tests suggest that all varables are ntegrated of order one. TABLE 4. PANEL UNIT TEST RESULTS FOR LEVELS Varables Constant Constant + trend Exchange rate -3.814 (0.000) 2.064 (0.980) Interest rate -1.780 (0.036) 0.033 (0.513) CPI -3.942 (0.000) 1.479 (0.930) PPI -1.710 (0.043) -1.811 (0.035) SPI -2.894 (0.001) -0.479 (0.316) Panel Unt Root Test results for frst dfferences Varables Constant Constant + trend Exchange rate -15.419 (0.000) -13.972 (0.000) Interest rate -14.789 (0.000) -13.131 (0.000) CPI -14.914 (0.000) -14.1448 (0.000) PPI -16.255 (0.000) -15.496 (0.000) SPI -12.699 (0.000) -11.580 (0.000) Note: The fgures gven n parentheses are p-value. After confrmng the order of the ntegraton of the varables, we apply the Johansen (1995) contegraton test to dentfy whether the varables ncluded n the exchange rate model are contegrated n the long run. Specfcally, we apply the trace statstc to examne the number of contegrated vectors. The results are presented n Table 5. The astersk ndcates the maxmum sgnfcant number of contegrated vectors. TABLE 5. RESULTS FOR COINTEGRATION TESTS Rank Egypt Inda Turkey Thaland 0 76.771 80.457 105.589 71.777 1 36.653* 46.997* 52.74 38.284* 2 17.352 21.624 24.838* 15.689 3 7.265 9.969 7.897 3.697 4 2.527 1.488 0.914 0.015 Note: * ndcates the number of sgnfcant contegrated vectors. As t can be seen from the table, there s only one contegrated vector for all countres except Turkey. For Turkey, there are two contegrated vectors. However, we select the frst one, as t s assocated wth the hghest egenvalue when we estmate the vector error correcton model to examne the mpact of fnancal crss on the determnaton of the exchange rate. The exstence of the contegraton between the exchange rate and domestc nterest rate, foregn nterest rate, domestc prce levels, and foregn prces suggest that these varables have a co-movement n the long run. In other words, there s a unque long-run equlbrum. After confrmng the exstence of the long-run relatonshp between the exchange rate and ts determnants, we estmate the vector error correcton model for each country to examne the mpact of fnancal crss on the exchange rate. The results are presented n Table 6. 2013 Prague Development Center - 93 -
Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu TABLE 6. RESULTS FROM VECTOR ERROR CORRECTION MODEL Dependent varable: log(exchange Rate) Varables Egypt Turkey Thaland Inda Coef. P-V Coef. P-V Coef. P-V Coef. P-V Error term -0.002 (0.442) 0.001 (0.084) -0.006 (0.320) -0.005 (0.000) log(exchange rate)t-1 0.060 (0.250) 0.185 (0.001) 0.158 (0.002) 0.051 (0.345) log(nterest rate)t-1 0.309 (0.018) -0.027 (0.566) 0.013 (0.573) -0.069 (0.166) log(cpi)t-1-0.017 (0.894) 0.120 (0.301) 0.025 (0.933) 0.355 (0.021) log(foregn nterest rate)t-1-0.069 (0.117) -0.006 (0.891) -0.022 (0.439) 0.025 (0.318) log(foregn CPI)t-1-0.624 (0.057) 0.078 (0.819) 0.343 (0.141) 0.11 (0.541) Dummy crss log(nterest -0.287 (0.208) 0.057 (0.683) -0.057 (0.188) -0.046 (0.969) rate)t-1 Dummy crss log(cpi)t-1-0.422 (0.392) 0.266 (0.590) 0.351 (0.503) 0.541 (0.122) Dummy crss log(foregn 0.087 (0.096) -0.008 (0.892) 0.018 (0.602) 0.017 (0.575) nterest rate)t-1 Dummy crss log(foregn CPI)t-1 0.532 (0.300) -0.317 (0.526) -0.347 (0.506) -0.533 (0.129) log(exchange rate)t-2 0.022 (0.693) log(nterest rate)t-2 0.003 (0.944) log(cpi)t-2-0.148 (0.278) log(foregn nterest rate)t-2-0.006 (0.834) log(foregn CPI)t-2-0.078 (0.664) Dummy crss log(nterest -0.794 (0.521) rate)t-2 Dummy crss log(cpi)t-2 0.348 (0.381) Dummy crss log(foregn nterest rate)t-2 0.003 (0.915) Dummy crss log(foregn CPI)t-2-0.047 (0.890) log(exchange rate)t-3 0.052 (0.346) log(nterest rate)t-3 0.028 (0.572) log(cpi)t-3-0.077 (0.617) log(foregn nterest rate)t-3-0.025 (0.318) log(foregn CPI)t-3-0.038 (0.832) Dummy crss log(nterest 1.541 (0.260) rate)t-3 Dummy crss log(cpi)t-3-0.066 (0.851) Dummy crss log(foregn nterest rate)t-3 0.039 (0.196) Dummy crss log(foregn -0.527 (0.230) CPI)t-3 Trend -0.001 (0.002) 0.312 (0.000) -0.001 (0.000) Constant 0.006 (0.017) 0.105 (0.000) 0.018 (0.315) 0.068 (0.000) Note: The values gven n parentheses are p-value. Lookng at the coeffcent of error term, we observe that the sgn of the estmated coeffcent s negatve for three countres, namely Egypt, Inda, and Thaland. The negatve sgn s consstent wth the theory. Ths mples that there s a sgnfcant convergence to the long-run equlbrum. The p-values ndcate that ths convergence s statstcally meanngful only for the case of Inda. Interestngly, the estmated coeffcent of the error term for Turkey s postve and statstcally sgnfcant at 10% level of sgnfcance, ndcatng that there s dvergence from the long-run equlbrum. The results also ndcate that the one-perod lagged value of exchange rate has a postve and statstcally sgnfcant mpact on the current level of exchange rate for Turkey and Thaland. Nonetheless, for remanng two countres, whle the estmated mpact s postve, t s not sgnfcant statstcally. Based on the vector error correcton model, we derve the long-run estmates that are presented n Table 7. - 94-2013 Prague Development Center
Busness and Economc Horzons Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu TABLE 7. RESULTS FROM VECTOR ERROR CORRECTION MODEL Dependent varable: log(exchange Rate) Varables Egypt Turkey Thaland Inda Coef. P-V Coef. P-V Coef. P-V Coef. P-V Log(exchange rate) 1.000 1.000 1.000 1.000 Log(nterest rate) -1.653 (0.002) 0.537 (0.459) 9.166 (0.000) 0.063 (0.902) Log(CPI) 0.467 (0.138) -5.633 (0.008) -1.285 (0.217) 1.898 (0.749) Log(foregn nterest rate) 0.142 (0.418) 0.113 (0.546) 0.444 (0.629) -0.199 (0.721) Log(foregn CPI) -7.743 (0.000) -5.276 (0.153) 52.144 (0.004) 4.164 (0.562) Dummy crss log(nterest rate) 0.057 (0.982) 15.390 (0.000) -36.940 (0.000) 17.375 (0.000) Dummy crss log(cpi) -23.024 (0.001) 73.496 (0.000) 77.770 (0.090) 72.970 (0.000) Dummy crss log(foregn nterest rate) -3.534 (0.000) 3.914 (0.000) 9.633 (0.000) -4.039 (0.000) Dummy crss log(foregn CPI) 23.570 (0.004) -114.980 (0.000) -56.750 (0.000) -47.540 (0.000) Trend 0.019 (0.000) -0.037 (0.000) -0.017 (0.000) Constant 34.893 (0.000) 29.243 (0.000) -29.300 (0.000) -7.496 (0.000) Note: The values gven n parentheses are p-value. Consstent wth the theory, domestc nterest rate s negatvely and statstcally sgnfcantly related wth the exchange rate for only Egypt. For Turkey, t s sgnfcantly postvely related to the exchange rate. Nevertheless, for remanng two countres, there s no statstcally sgnfcant relatonshp between domestc nterest rate and the exchange rate. Lookng at the nteracton between domestc nterest rate and fnancal crss dummy, we observe that the coeffcent s postve and statstcally sgnfcant for Indan and Thaland. On the other hand, t s negatve and statstcally sgnfcant for Turkey, whle for Egypt the estmate s postve but t appears statstcally nsgnfcant. It should be noted that n Egypt, the nterest rate mpact on exchange rates becomes nsgnfcant after fnancal crss. In contrast to the case of Egypt, the role of domestc nterest rate has become sgnfcant n the determnaton of the exchange rate after fnancal crss n Inda and Thaland. Surprsngly, for the case of Turkey, the mpact of nterest rate on the exchange rate was postve before fnancal crss, whle t turns negatve after fnancal crss. The mpact of domestc prces on exchange rate s statstcally nsgnfcant for all countres except Inda. For Inda, t s negatve and statstcally sgnfcant at acceptable level of sgnfcance. However, after the fnancal crss, the mpact of domestc prces s sgnfcant for all the four countres. Specfcally, t s negatve for Egypt, whereas t s postve for remanng three countres. Turnng to the mpact of foregn nterest rate and prce levels, we fnd that the mpact of foregn nterest s statstcally sgnfcant after fnancal crss, whle t was statstcally nsgnfcant before fnancal crss for all the four countres. Specally, after fnancal crss, the exchange rate s postvely affected by the foregn nterest rate n the case of Egypt and Thaland. Nonetheless, the exchange rate s negatvely affected by the foregn nterest rate for the remanng two cases. The foregn prce level s negatvely related to exchange rate for Egypt and Inda, whle t s postvely related to the exchange rate for Turkey and Thaland. However, the relatonshp s statstcally sgnfcant only for the Egypt and Turkey. The estmated coeffcent of the nteracton between foregn prces and fnancal crss dummy s postve and sgnfcant for Egypt, whle t s negatve and sgnfcant for remanng three countres. Overall, the results presented n Table 7 suggest that the role of exchange rate determnants has been sgnfcantly changed n terms of both ther sgn (mpact) and statstcal sgnfcance after 2008 fnancal crss. These fndngs are consstent wth the prevous emprcal evdence that ndcate the sgnfcant mpact of fnancal crss on exchange rate determnatons. These results also confrm the dea that the mpact of fnancal crss on the exchange rate sgnfcantly dffers across countres wth dfferent economc and socal backgrounds. 2013 Prague Development Center - 95 -
Concluson Fnancal crss and exchange rates n emergng economes BEH: www.beh.pradec.eu Ths study examnes the mpact of 2008 fnancal crss on the exchange rate n PPP-UIP framework for four emergng countres, namely Egypt, Inda, Turkey, and Thaland. The study uses monthly data coverng the perod from 1981-2012. The results reveal that the mpact of recent fnancal crss led to change the role of determnants of exchange rates n exchange determnaton. Moreover, we observe that the effects of fnancal crss on the exchange rate are dfferent n dfferent emergng economes. The fndngs of the study are of sgnfcant for polcy makers n desgnng effectve polces n order to reduce the effects of fnancal crss on exchange rates. The fndngs are also sgnfcant n makng decsons for the exchange rate regme; especally n the rsky tme n order to mtgate the adverse effects of fnancal crss. References Blanchard, O. J. et al., 2010. The ntal mpact of the crss on emergng market countres, Brookngs Papers on Economc Actvty, pp.263-323, http://dx.do.org/10.1353/eca.2010.0005 Brook, A and Hargreaves, D., 2001. PPP based Analyss of New Zealand's equlbrum exchange rate, Reserve Bank of New Zealand Dscusson Paper No. DP2001/01 Brunnermeer, M.K., 2008. Decpherng the lqudty and credt crunch 2007-08, Workng Paper, Natonal Bureau of Economc Research Crotty, J., 2009. Structural Causes of the Global Fnancal Crss: A crtcal assessment of the new fnancal archtecture, Cambrdge Journal of Economcs, 33, pp.563-580 Devereux, M. and Sultherland, A., 2009. A portfolo model of captal flows to emergng market, Journal of Development Economcs, 89 (2), pp.181-193, http://dx.do.org/10.1016/ j.jdeveco.2008.07.002 Dsyatat, P and Galat,G., 2007. The effectveness of foregn exchange nterventon n emergng market countres: Evdence from the Czech Koruna, Journal of Internatonal Money and Fnance,26 (3), pp.383-402, http://dx.do.org/10.1016/j.jmonfn.2007.01.001 Fratzscher, M., 2009. What explans global exchange rate movements durng the fnancal crss?, Journal of Internatonal Money and Fnance, 28 (8), 1390-1407, http://dx.do.org/10.1016/j.jmonfn.2009.08.008 Fukuda-Parr, S., 2008. The human mpact of the fnancal crss on poor and dsempowered people and countres, UN General Assembly: Interactve Panel on the Global Fnancal Crss 30 Galbrath, J., 2009. The Great crash of 1929, New York: Houghton Mffln Harcourt Jaramllo, M. and Servan, S., 2012. Modelng exchange rate dynamcs n Peru: A contegraton approach usng the UIP and PPP, SBS Documents de Trabajo Johansen, S., 1995. Lkelhood based nference n contegrated vector autoregressve models, USA: Oxford Unversty Press, http://dx.do.org/10.1093/0198774508.001.0001 Keblowsk, P and Welfe, A., 2011. A rsk-drven approach to exchange rate modelng, Workng Paper Warsaw School of Economcs MacDonald, R and Taylor, M., 1992. Exchange rate economcs: A survey, IMF Staff Papers, 39 (1), pp.1-57, http://dx.do.org/10.2307/3867200 Mshkn, F. S., 2010. The economcs of money, bankng and fnancal markets, 9th ed. Boston: Pearson Educaton Plbeam, K., 2006. Internatonal fnance, 3rd ed. London: Copyrght Lcensng Agency Rashd, A., 2009. Testng the Modfed-Combned PPP and UIP hypothess n South Asan economes, Appled Econometrcs and Internatonal Development, 9 (1), pp.199-218 Tsangardes, C., 2012. Crss and recovery: Role of the exchange rate regme n emergng market economes, Journal of Macroeconomcs, 34 (1), pp.1-252 Wollf, M., 2009. Fxng global fnance, Hampshre: Yale Unversty Press Wong, D.K and Wa L, K., 2010. Comparng the performance of relatve stock return dfferental and real exchange rate n two fnancal crses, Appled Fnancal Economcs, 20 (1-2), pp.137-150, http://dx.do.org/10.1080/09603100903266468-96 - 2013 Prague Development Center