Effect of Macroeconomic Factors on Stock Prices in Ghana: A Vector Error Correction Model Approach

Transcription

1 Vol. 3, No.2, Aprl 2013, pp ISSN: HRMARS Effect of Macroeconomc Factors on Stock Prces n Ghana: A Vector Error Correcton Model Approach Kwame MIREKU 1 Kwaku SARKODIE 2 Kwas POKU 3 1,3 Department of Accountng and Fnance, School of Busness, Kwame Nkrumah Unversty of Scence and Technology, Ghana 1 E-mal: my27cecl@yahoo.com, Telephone Number: Assembles of God Senor Hgh School 2 E-mal:brothersark@yahoo.com Telephone Number: Abstract Ths study examnes the effect of macroeconomc varables on stock prces n Ghana. Analyss was done usng monthly data from to Ths study employed contegraton test and vector error correcton models (VECM) to examne both long-run and short-run dynamc relatonshps between the stock market ndex and the macroeconomc varables. Generalzed mpulse functon (IRF) and forecast error varance decomposton (FEVD) were used to detect the effect of shocks n the macroeconomc factors on complete tme path of stock prces and vce versa. The tme seres propertes of the data were, frst, analyzed usng the Augmented Dckey-Fuller (ADF) and Phllps- Perron tests. The emprcal results derved ndcate that all the varables were statonary after ther frst dfferencng. The paper establshed that there s contegraton between macroeconomc varables and Stock prces n Ghana ndcatng long run relatonshp. The above long term relaton ndcates that Interest Rate (TB) and Exchange Rate (XR) have a negatve effect on Stock Prces whles Inflaton (CPI) showed a postve effect on Stock Prces (DSI). Results of Impulse Response Functon (IRF) and Forecast Error Varance Decomposton (FEVD) ndcate that the macroeconomc varables dentfed a low sgnfcant nfluence on share prce movements n Ghana. Key words Vector Autoregressve Model (VAR), Contegraton, Macroeconomc Varables, Stock Prces 1. Introducton Over the past few decades, the nteracton of share returns and the macroeconomc varables has been a subject of nterest among academcans and practtoners. It s often argued that stock prces are determned by some fundamental macroeconomc varables such as the nterest rate, the exchange rate and nflaton. Evdence from Maghayereh (2002) ndcates that nvestors generally beleve that macroeconomc events have a large nfluence on stock prce. Ths motvates many researchers to nvestgate the relatonshps between share returns and macroeconomc varables. For example, usng the Arbtrage Prcng Theory (APT), developed by Ross (1976), Chen et al. (1986) used some macroeconomc varables to explan stock returns n the US stock markets. The authors fndngs revealed that ndustral producton, changes n rsk premums, and changes n the term structure to be postvely related to the expected stock returns, whle both the antcpated and unantcpated nflaton rates were negatvely related to the expected stock returns. The development of contegraton analyss provded another approach to examne the relatonshps between the macroeconomc varables and stock returns. For example, Mukherjee and Naka (1995) employed the Johansen contegraton test n the Vector Error Correcton Model (VECM) and found that the Japanese stock market s contegrated wth sx macroeconomc varables namely, exchange rate, money supply, nflaton rate, ndustral producton, long term government bond rate and the short term call money rate. The results of the long-term coeffcents of the macroeconomc varables are consstent wth the hypotheszed

2 equlbrum relatonshps. Contegraton s a statstcal property of tme seres varables. Two or more tme seres are contegrated f they share a common type of stochastc drft: that s, to a lmted degree they share a certan type of behavour n terms of ther long-term fluctuatons. In Afrca, Jeffers, Okeahalam and Matome (2001) reported that, the real stock market ndex of the Johannesburg Stock Exchange (JSE) has a postve long-term relatonshp wth real GDP and real exchange rate, and a negatve relatonshp wth real long-term nterest rate over the perod 1985 to Establshng such relatonshp would not only be very useful to polcy makers and nvestors alke but wll also test the effcency of the stock market because establshng a lead-lag relatonshp between stock prces and macroeconomc varables exposes the exstence of arbtrage proft hence the neffcency of the market. Wth ths mportant mplcaton on stock prces, there have been lmted studes on the dynamc lnkage between the macro economy and stock prces n Ghana. Recognzng the mportance of these studes, the authors have wdened the study to capture most of the key macroeconomc ndcators for robust analyses. The man focus of ths research was to examne the mpact of macroeconomc varables, specfcally, exchange rate, nterest rate and nflaton rate on stock prces n Ghana. The emprcal methods nclude contegraton analyss and Vector Error Correcton Model (henceforth VECM). 2. Lterature revew 2.1. Theoretcal revew Macroeconomcs s the study of the behavor of an economy at the aggregate level, as opposed to the level of specfc subgroups or ndvduals whch s called mcroeconomcs (Chen, Roll and Ross, 1986). Macroeconomc factors nclude nflaton, Interest Rate, Exchange Rate, unemployment, and ndustral producton. Stock prces are set by a combnaton of factors that no analyst can consstently understand or predct (Poon and Taylor, 1991). In general, economsts say, they reflect the long-term earnngs potental of companes. Investors are attracted to stocks of companes they expect wll earn substantal profts n the future; because many people wsh to buy stocks of such companes, prces of these stocks tend to rse. On the other hand, nvestors are reluctant to purchase stocks of companes that face bleak earnngs prospects; because fewer people wsh to buy and more wsh to sell these stocks, prces fall Emprcal evdence The relatonshp between the stock market returns and the macroeconomc varables are mostly documented for developed countres. One semnal paper analyzng the determnants of the stock market returns s presented by Chen, Roll and Ross (1986). Ther paper examnes the relatonshp between the market returns and macroeconomc factors wth a dfferent methodology that s based on prcng the systematc macroeconomc rsks. They found a strong relatonshp between the market returns and the macro varables lke ndustral producton, changes n the rsk premum and the expected and unexpected nflaton n Unted States. There are also some papers that found no emprcal evdence that the macroeconomc factors affect the stock returns. Poon and Taylor (1991) n Unted Kngdom and Gjerde and Saettem (1999) n Norway found that macroeconomc varables do not appear to affect share returns. Bnswanger (2000), who used monthly data durng the perod 1953 to1995, also states that the prce movements snce early 1980's cannot be explaned by fundamental factors mplyng that the lnk between stock prces and real economc actvty has broken down n U.S. and G-7 countres. Interest rates are expected to be negatvely related to market returns ether through the nflatonary or dscount factor effect. Cho and Jen (1991) report that the expected returns on common stocks are systematcally related to the market rsk and the nterest-rate rsk. The fndngs of the study ndcate that the nterest-rate rsk for small frms s a sgnfcant source of nvestors' portfolo rsk and the nterest-rate rsk for large frms s "negatve". The effect of nterest rate on stock returns has been studed over emergng markets as well. Al-Sharkas (2004) for Jordan and Adam and Tweneboah (2008) for Ghana ndcate that the relatonshp between stock prces and nterest rates s negatve and statstcally sgnfcant. Gultekn (1983) testng the generalzed Fsher hypothess for 26 countres for the perod of 1947 to 1979, could not fnd a sgnfcant postve relatonshp between nomnal stock returns and nflaton rates. Moreover, the fndngs of the study revealed that regresson coeffcents are predomnantly negatve. A negatve relatonshp between nflaton and stock prces s contended n lterature because an ncrease n the 33

3 rate of nflaton s accompaned by both lower expected earnngs growth and hgher requred real returns. In the US, there s substantal emprcal evdence that hgh nflaton s assocated wth a hgh equty rsk premum and declnng stock prces (Hoguet, 2008). Rsng nflaton s apt to restrctve economc polces, whch n turn ncreases the nomnal rsk-free rate and hence rases the requred rate of return n valuaton models. There s no theoretcal consensus nether on the exstence of relatonshp between stock prces and exchange rates or on the drecton of the relatonshp. However, n the lterature, two approaches have been asserted to establsh a relatonshp between exchange rate and stock prces: the goods market modeland the portfolo balance model. The frst approach accordng to Flannery and Protopapadaks (2001) focused on the assocaton between the current account and the exchange rate. They developed a model of exchange rate determnaton that ntegrates the roles of relatve prces, expectatons, and the assets markets, and emphaszed the relatonshp between the behavour of the exchange rate and the current account. They also argue that there s an assocaton between the current account and the behavour of the exchange rate. It s assumed that the exchange rate s determned largely by a country s current account or trade balance performance. These models post that changes n exchange rates affect nternatonal compettveness and trade balance, thereby nfluencng real economc varables such as real ncome and output. Thus, the goods market model suggests that changes n exchange rates affect the compettveness of a frm, whch n turn nfluences the frm's earnngs or ts cost of funds and hence ts stock prce. On a macro level then, the mpact of exchange rate fluctuatons on stock market would depend on both the degree of openness of the domestc economy and the degree of the trade mbalance. Thus, goods market models represent a postve relatonshp between stock prces and exchanges rates wth drecton of causaton runnng from exchange rates to stock prces. The concluson of a postve relatonshp stems from the assumpton of usng drect exchange rate quotaton (Forest, 2006). 3. Methodology of research Databank Stock Index (DSI) was used to represent the Ghana stock market. The Treasury bll rate (as a measure of nterest rates), the consumer prce ndex (as a measure of nflaton), and the exchange rate were used as the macroeconomc varables. The data set covers nneteen year perod from to on monthly bass. The more frequent the data the more accurate results can be obtaned from model and tests. Ths was the motvaton for takng monthly data. All data are secondary data that were extracted from IMF- World Bank World Development Indctors, September 2010 onlne edton. The analyss was done usng E- vews, 2006 verson. Table 1 presents a bref descrpton of the varables and ther source. Table 1. Descrpton and source of data Varable Concept Descrpton Source LDSI Log of Stock Index Databank Stock Index Databank LCPI Log of nflaton Consumer Prce Index IFS statstcs LXR Log of exchange rate Prncpal Rate (Natonal Currency per USD) IFS statstcs LTB Log of Interest rate 91-day Treasury bll Rate IFS statstcs 3.1. Model specfcaton The strategy for modelng the study s fashoned out n a way such that all varables enterng are treated as endogenous. In order to fully address our research questons, proposed generc equatons were estmated as follows: ΔY t = f(γ(γ(l) t,u, ΔZ t, β ) (1) Where Y t s the vector of observed dependant varables, Y t represent lagged values of Y,Γ(L) s a matrx of parameters Y t, U represent vector of unobserved varable, Zt s a vector of statonary endogenous ndependent varables, β s a matrx of parameters of the endogenous varables. The dependant varable s 34

4 the Databank Stock Index (DSI) n our regresson analyss. The ndependent varables are the macroeconomc factors Method of analyss and nterpretaton In ths study, VAR technque was employed based on contegraton and error correcton model to study the long run relatonshp between macroeconomc factors and stock prces. The study specfcally uses Johansen (1991, 1995) maxmum lkelhood procedure of estmatng contegraton vectors. Generalzed mpulse functon (IRF) and forecast error varance decomposton (FEVD) are used to detect the effect of shocks n the macroeconomc factors on complete tme path of stock prces and vce versa. The frst requrement of estmatng contegratng vector s a test for statonarty; the order of ntegraton of the varables s requred. For ths purpose, Augmented Dckey-Fuller (ADF) and Phllps-Perron tests for unt roots are employed wth Schwartz Bayesan Crteron (SBC) and Akake Informaton Crteron (AIC) as a leadng ndcator for VAR lag selecton. For non-contegratng varable, we examne how the shock n the macroeconomc factors perpetuates tself nto the stock prces through IRF and FEVD from frst dfference VAR Vector Autoregressve (VAR) model The VAR s the expanded form of Autoregressve (AR) model and was popularzed n econometrcs by Granger and Newbold (1974). In general, a kth order VAR for a 2x1 vector of jontly determned (endogenous) varables X t s wrtten as: k X t = ϕ + Π X t + ε = 1 t (2) Where the resdual vector ε t s nterpreted nnovaton (shock) n X that s predcted from past values of varables n the system. Havng n mnd that most of fnancal tme seres are non- statonary at level we wrte (7) n frst dfference form; k ΔX t = ϕ + Π ΔX t + ε = 1 t (3) The VAR system gves very lttle nformaton about the dynamc nteracton among the seres, we therefore represent t n movng average representaton (MAR) form whch present equvalent nformaton contaned n the orgnal estmaton and allow s tme path of the varous shocks to be traced (Granger and Newbold, 1974). By polynomal lag dvson, MAR s derved and represent n the form: ΔX t Π 1 (ΒΒ) t = Ψ(Β)ε t = ε t + Ψ ε t = (4) Where Ψ = jψ and Ψ Ι = j 0 = n j 1 The advantage of usng ths model s that t estmates the dependence among the varables dentfed, and also allows shockng a specfc varable and analyzng how shocks carry on themselves through mpulse responses Johansen Multvarate Contegraton Test The multvarate contegraton approach s based on error correcton representaton of the p order Vector Autoregressve model wth Gaussan error: p 1 ΔX t = ϕ + Γ ΔX t 1 ΠX t p ε = t (5) 35

5 Where Δ s the frst dfference operator, Γ = (Ι A 1... A ) s coeffcent matrx representng shortrun dynamcs, and defned by Π = (Ι A 1... A ) s n n matrx, where I s an dentty matrx, whose rank determnes the number of contegratng vectors. If rank ( ) =r, then X t has r contegratng relaton or n-r common stochastc trends. The extent of long-run relatonshp s determned by the number of contegratng vectors. Here, there are three possble cases that may arse: If n-r=0 (r=n) (full rank), Xts statonary [I (0)] and there are no stochastc trends. Contegraton s sad not to be defned n such cases. Rank ( ) =0 ndcatng no statonary long-run relatonshps among the elements ofx t. Reduced rank (.e. Rank ( ) <n) mplyng presence of at least one common stochastc trend, and error correcton representaton ofx t such that =αβ', where α and β are n r matrces. The rows of β matrx gve the contegratng vectors and the columns of matrx α are adjustment factors. Two dfferent lkelhood rato tests were developed by Johansen for testng the number of contegraton vectors (r): the trace test gven by: g λtrace (r) = T ln(1 λ ) = r + 1 (6) and Maxmum egenvalue test statstcs gven by: λ max (r,r + 1) = Tln(1 λ + 1. (7) The null hypothess of the trace statstcs tests s no contegraton H 0 : r = 0aganst the alternatve of more than 0 contegraton vectorh 1 : r > 0 whereas the maxmum Egenvalue statstcs test the null hypothess of r contegratng vectors aganst the alternatve of r + 1 contegratng vectors Impulse Response and Varance Decomposton Analyss The two varables ytand z t standard VAR model wth two types of reduced form shocks εyand εzcan be represented n Vector Movng Average, VMA( ) n terms of εytand ε zt n matrx notaton of the form: () y t y φ = + 11 z z = 0 () t φ 21 () φ ε 12 y,t 1 () φ ε 22 z,t 1 (8) The coeffcents ofytand zt through tme. The coeffcent φ jk () can be used to trace the mpact of one unt change n εytand εzton the path φ jk () s nterpreted as change n j th varable due to a unt shock to k th varable at th perod. The mpulse response apples n ths study s generalzed mpulse response functon (GIRF), gven by: GIRF(n, ε t, ω t 1 ) = E[y t + n /ε j,t, ω t 1 ] E[y t + n /ω t 1 ] (9) Where n s the forecast horzon, ε t + s a random shock, y t s a random vector, and ω t 1 s a specfc realzaton of the nformaton set Ω t 1. Followng Pesaran and Smth (1995) by constructng orthogonal sets of nnovatons that do not depend on the VAR orderng, GIRF was derved from an nnovaton to the jth varable by applyng a varable specfc Cholesky factor computed wth the varable at the top of the Cholesky orderng. FEVD measures the 36

6 percentage of varaton n each varable that can be explaned by ts own shock and the shocks to all the varables n the system. Followng Enders (1995), we defne forecast error varance decomposton (FEVD) by: σ [ a 2 z 12 (0) + a 12 2 (1) a 2 σ ( m) y 12 ( m 1) 2 ] (10) It can be sad that as σ 2 y (m) ncreases, m perod ncreases. Addtonally, the varance can be splt nto two seres: y t and z t seres. Therefore, e yt and e zt composes the error varance for y. Seres y t s ndependent of seres z t f e yt approaches unty (ndcates e zt approaches zero) and therefore we can say that y t s exogenous wth respect to z t. In contrast, we say that y t s endogenous relatve to z t f e yt approaches zero (Enders, 1995). 4. Data Presentaton and Analyss 4.1. Testng for Unt Root and Statonarty A necessary but not suffcent condton for contegraton s a test for unt root for examnng the statonarty of the varable used n ths study. Statonarty s mportant for standard econometrc theory. For the purpose of ths study the Augmented Dckey-Fuller (ADF) and Phllps-Perron tests for unt roots are employed. The results of these tests can be seen n Tables 2 and 3. Table 2. Augmented Dckey -Fuller Unt Root Test ( ) Varable Levels Frst Dfference Order of Integraton Intercept Intercept CPI [0.4637] [0.0000] DSI [ ] [0.0000] TB [0.5436] [0.0000] XR [0.0971] [ ] Table 3. Phlp-Perron Unt Root Test ( ) Varable Levels Frst Dfference Order of Integraton Intercept Intercept CPI [0.3046] [0.0000] DSI [0.7454] [0.0000] TB [0.5253] [0.0000] XR [0.1980] [0.0000] Augmented Dckey-Fuller (ADF) test and Phlp-Perron (PP) test were appled to all the varables n levels and n frst dfference and presented n Tables 2 and 3 respectvely. The results ndcate that all the data were non-statonary at levels but become statonary after frst dfferencng, whch provded a necessary, but not suffcent ratonale for estmatng contegraton and error correcton model. Knowng ths leads to the testng of Long Run relatonshp between the macroeconomc Varables and Stock Prces. 37

7 4.2. Long Run Relatonshp - Lag Length Selecton Ths step nvolves estmatng the model and determnng the rank, r to fnd the number of contegratng relatons n our model. The model lag length selecton was determned by both Schwarz (SIC) and Akake (AIC) Informaton Crteron usng 8 lags n the general VAR model. The am s to choose the number of parameters whch mnmzes the value of the nformaton crtera. That lag length selecton determnes whch year selecton would have sgnfcance on the current results. The SIC has the tendency to underestmate the lag order, whle addng more lags ncreases the penalty for the loss of degrees of freedom. To make sure that there s no remanng autocorrelaton n the VAR model, AIC s selected as the leadng ndcator. The model lag length reported n Table 4 ndcates approprate lag length as 2. Table 4. VAR Lag Order selecton Crtera Lag LogL LR FPE AIC SC HQ NA e e-13* * * * e e * 9.56e e e e * ndcates lag order selected by the crteron LR: sequental modfed LR test statstc (each test at 5% level) FPE: Fnal predcton error AIC: Akake nformaton crteron SC: Schwarz nformaton crteron HQ: Hannan-Qunn nformaton crteron 4.3. Contegraton Test (Long Run Test) A test for the presence of contegratng long-run relatonshp among the varables usng Johansen's Maxmum Lkelhood approach among the Databank stock Index (DSI), Treasury Bll (TB), Exchange Rate(XR) and Consumer Prce Index(CPI) usng lag length of 2 was done. An ntercept and no trend are specfed for the contegraton test. Both trace test and maxmum egenevalue test are performed on the varance usng general VAR lag length of 2, whch s selected from AIC and SBIC lag selecton test. The results of both tests are reported n Tables 5 and 6 respectvely. Table 5. Unrestrcted Contegraton Rank Test (Trace) Hypotheszed Trace 0.05 No. of CE(s) Egenvalue Statstc Crtcal Value Prob.** None * At most At most At most Trace test ndcates 1 contegratng eqn(s) at the 0.05 level * denotes rejecton of the hypothess at the 0.05 level **MacKnnon-Haug-Mchels(1999) p-values 38

8 Table 6. Unrestrcted Contegraton Rank Test (Maxmum Egenvalue) Hypotheszed Max-Egen 0.05 No. of CE(s) Egenvalue Statstc Crtcal Value Prob.** None * At most At most At most Max-egenvalue test ndcates 1 contegratng eqn(s) at the 0.05 level * denotes rejecton of the hypothess at the 0.05 level **MacKnnon-Haug-Mchels(1999) p-values The trace statstcs and the maxmum egenvalue statstc tests suggest one contegratng vector at 5% sgnfcance level. Gven evdence n favour of at least one contegraton relaton, a test of zero coeffcents on each factor n the contegratng vector s conducted to determne whether the coeffcents for all factors n the contegratng VAR model are sgnfcantly dfferent from zero. The tests ndcate that the entre varables are sgnfcantly dfferent from zero. We therefore reject the null hypothess and conclude that there exst a long run relatonshp between stock prces and macroeconomc factors n Ghana. The long-run contegratng relatonshp between the macroeconomc factors and stock prces normalzed on LDSI s gven by: DSI= -1.25TB CPI XR C [-2.437] [ ] [ ] The coeffcents of Interest Rate (TB) and Exchange Rate (XR) are negatvely sgned. Inflaton (CPI) on the other hand has a postve sgn The Short Run Relatonshp (Error Correcton Model) Gven the evdence n favour of at least one contegratng vector, the Vector Error correcton model (VECM) was estmated to examne the short run relatonshp between the market ndex and macroeconomc varables. The result of the VECM estmaton s reported n Table 7. Table 7. Error Correcton Model D(LNDSI) D(LNTB) D(LNCPI) D(LNXR) ContEq ( ) ( ) (9.7E-05) ( ) [ ] [ ] [ ] [ ] % Corrected per month Tme taken to revert to long run relatonshp n the absence of changes n ndependent varables Months Table 7 shows vector error correcton model for DSI wth sgnfcant error correcton term n the Databank stock ndex equaton, showng explct nformaton on the short-run dynamc nteractons among those varables. The sgn and magntude of these error correcton parameters gve nformaton about the short-run mechancs of the process. They ndcate the drecton and speed of adjustment towards the longrun equlbrum course. A quck look at Table 7 shows that, the adjustment coeffcents are negatve and statstcally sgnfcant n the entre models rrespectve of whch varable s dependent varable n the model. The negatve sgn ndcates that wth absence of any change n the ndependent varables, devaton from the long run relaton s offset by ncrease n the dependent varable. For nstance wth the absence of changes n TB, CPI and XR, devaton of the model from long run relatonshp s offset by about 0.11 percent ncrease n DSI. Ths means devaton from long run equlbrum relatonshp takes 909 months from Table 6 for the DSI to revert t n the absence of changes n other varables. 39

9 4.5. Effect of Shock on Macroeconomc Factors to Stock Prces (Impulse Response) In ths secton mpulse response functons were nvestgated for DSI n order to see the effect of macroeconomc factors on the DSI n terms of reacton over tme. Fgure 1 shows the generalzed mpulse response functons whch trace the effects of a shock to one endogenous varable on to the other varables n the system. It should be noted that a shock to one varable does affect tself and other endogenous varables n the system due to the dynamc lag structure. Response to Generalzed One S.D. Innovatons Response of LNDSI to LNDSI Response of LNDSI to LNTB Response of LNDSI to LNCPI Response of LNDSI to LNXR Fgure 1. Impulse Response The fgure above shows that DSI ndex respond postvely to nnovatons n ts own lags. However, t changes n steady response after some perods. Economcal explanaton to ths evdence can be that when the Ced (natonal currency n Ghana) deflates the exports are motvated and the Ghana ndustry whch s manly mnng and agrculture export s also motvated, consequently the expectatons of future cash flows of such companes ncrease whch n turn push up the share prces. The story wth CPI s dfferent. There was a negatve response at frst perods to CPI. Ths s probably reasoned by the hgh cash flows related wth prce ncreases. Inflaton s the predcton of lower real cash flows that pushes down the share prces. Shocks to LNXR nstantaneously affect the DSI postvely from the fgure. Ths s attrbuted to the fact that snce most companes mport goods wth the major foregn currences, an ncrease n exchange rate nstantly results n an ncrease n share prces n order to mantan share prces. The mpact of Interest Rate on DSI s almost neglgble Varance Decomposton Varance decomposton or forecast error varance decomposton ndcates the amount of nformaton each varable contrbutes to the other varables n a vector autoregresson (VAR) models. Varance decomposton determnes how much of the forecast error varance of each of the varable can be explaned by endogenous shocks to the other varables. That s n determnng the prce of Stocks whch of the macroeconomc factors s consdered greatly before prcng. 40

10 Varance Decomposton Percent LNDSI varance due to LNDSI Percent LNDSI varance due to LNTB Percent LNDSI varance due to LNCPI Percent LNDSI varance due to LNXR Fgure 2. Varance Decomposton The plot from Fgure 2 above suggests that DSI support largely on ts own prevous prce. Ths s an ndcaton that n determnng the current stock prces, the prevous prce s the major determnng factor to be consdered. Neglgble amount of varances n forecasted errors of DSI can be explaned by CPI from the graph n Fgure 2. The relatvely nfluental varables are consdered to be XR and TB, whch are not themselves sgnfcant to explan stock prces wthn gven case. 5. Summary of Fndngs and Recommendatons The man fndngs are summarzed as follows: share prces cannot be predcted from past macroeconomc varables; there s a long run relatonshp between macroeconomc factors and Stock Prcng; the mpulse results from the analyss suggest that Shocks to XR nstantaneously affect the DSI postvely. For CPI and TB, mpact of the shock s neglgble; and the Varance results from the analyss also suggest that DSI support largely on ts own prevous prce Based on the emprcal fndngs of the study, the followng recommendatons are offered to polcy makers, stakeholders and management of companes n Ghana: For a better stock market performance, polcy makers should put n place measures that wll ensure stable macroeconomc envronment, snce any dsturbances n the macroeconomc envronment may affect the stock market s actvtes. So to attract nvestors means that we should have a stable macroeconomc envronment; It s also recommended that nvestors should take nto consderaton the nature of volatlty n the macroeconomc varables n the economy to make an nformed decson as to where to drect ther nvestments. For example, whenever the local currency deprecates, t s a sgnal that the stock market returns s lkely to deprecate, especally for an mport domnated economy. But ths argument s based on an mprovement n the nternatonal compettveness of the local frms; and Good regulatory framework should be desgned to make t attractve for more companes to be lsted on the stock market. 6. Conclusons Ths study emprcally examnes the relatonshp between macroeconomc factors and stock prces n Ghana. Databank Stock Index (DSI) was used to represent the Ghana stock market. The Treasury bll rate (as a measure of nterest rates), the consumer prce ndex (as a measure of nflaton), and the exchange rate were used as the macroeconomc varables. The data set covers nneteen year perod from to on 41

11 monthly bass. Johansen's multvarate contegraton test was used to examne the long run relatonshp between share prces and the macroeconomc varables. Short run dynamcs were traced usng mpulse response functon and forecast error varance decomposton analyss. Fndngs of ths study may help stakeholders to recognze the lnk between macroeconomc factors and Stock Prces and choosng approprate measures to evaluate and analyze the companes performance. The fndngs from ths study suggest there are both long and short run relatonshp between the macroeconomc varables and stock prces. Also the mpulse results from the analyss also suggest that Shocks to Exchange Rate nstantaneously affect the DSI postvely. For Inflaton and TB, mpact of the shock s neglgble. Fnally, the Varance results from the analyss also suggest that DSI support largely on ts own prevous prce. Neglgble amount of varances n forecasted errors of DSI can be explaned by Inflaton. The relatvely nfluental varables are consdered to be Exchange Rate and Interest Rate. References 1. Adam, M.A. and Tweneboah, G. (2008), Macroeconomc Factors and Stock Market Movement: Evdence from Ghana, MPRA Paper No Adjas, C., Harvey, S.K. and Agyapong, D. (2008), Effect Of Exchange Rate Volatlty on the Ghana Stock Exchange, Afrcan Journal of Accountng, Economcs, Fnance and Bankng Research Vol. 3. No Al-Sharkas, A. (2004), The Dynamc Relatonshp between Macroeconomc Factors and the Jordanan stock market, Internatonal Journal of Appled Econometrcs and Quanttatve Studes, pp Aydemr, O. and Demrhan, E. (2009), The Relatonshp between Stock Prces and Exchange Rates Evdence from Turkey, Internatonal Research Journal of Fnance and Economcs, 23, Bnswanger, M. (2000), Stock returns and real actvty: Is there stll a connecton? Appled Fnancal Economcs, 10(4), Brahmasrene, T. and Jranyakul, K. (2007), Contegraton and Causalty between stock ndex and macroeconomc varables n an emergng market, Academy of Accountng and Fnancal Studes Journal, 11(3), Chen, N.F., Roll, R. and Ross, S.A. (1986), Economc Forces and the Stock Market, Journal of Busness 59, Cho, D. and Jen, F.C. (1991), The Relaton between Stock Returns and Short-Term Interest Rates. Revew of Quanttatve Fnance and Accountng, Vol. 1, pp Flannery, M. J. and Protopapadaks, A. A. (2001), Macroeconomc Factors Do Influence Aggregate Stock Returns. Journal of Fnance 53, Forest, P. (2006), Testng for Granger causalty between stock prces and economc growth, MPRA Paper No Gjerde, O. and Saettem, F. (1999), Casual Relatons among Stock Returns and Macroeconomc Varables n a Small, Open Economy. Journal of Internatonal Fnance Markets Insttutons and Money, 9, Granger, C.W.J., and Newbold, P. (1974), Spurous regressons n econometrcs. Journal of Econometrcs, 2, Gultekn, N. B. (1983), Stock market returns and nflaton: evdence from other countres. Journal of Fnance 38, Hoguet, G. R. (2008), Inflaton and Stock Prce.State Street Global Advsors 15. K. Jeffers, C. Okeahalam & T. Matome, Internatonal Stock Market Lnkages n Southern Afrca, AERC Research Paper 105, Narob: Afrcan Economc Research Consortum. 16. Johansen, S., (1991), Estmaton and Hypothess Testng of Contegratng Vectors n Gaussan Vector Autoregressve Models, Econometrca 59, Johansen, S., (1995), Lkelhood based Inference n Contegrated Vector Auto-Regressve Models, Oxford Unversty Press. 18. Kandr, S. Y. (2008), Macroeconomc Varables, Frm Characterstcs and Stock Returns: Evdence from Turkey. Internatonal Research Journal of Fnance and Economcs. 42

12 19. Maghayereh, A. (2002), Causal Relatons among Stock Prces and Macroeconomc Varables n the Small, Open Economy of Jordan. Journal of Fnance Maysam, R. C. and Koh, T. S., (2000), A Vector Error Correcton Model of the Sngapore Stock Market. Internatonal Revew of Economcs and Fnance 9, Naka, A., Mukherjee, T., (1995), Macroeconomc Varables and the Performance of the Indan Stock Market: Unversty of New Orleans, Department of Economcs and Fnance Workng Paper No Poon, S., and Taylor, S..J. (1991), Macroeconomc Factors and the UK Stock Market. Journal of Busness and Accountng, 18, Ross, S.A., (1976), The Arbtrage Theory of Captal Asset Prcng, Journal of Economc Theory, 13,