The Relationship between External Financial Flows and Economic Growth in the Southern African Development Community (SADC): The Role of Institutions

Transcription

1 The Relatonshp between External Fnancal Flows and Economc Growth n the Southern Afrcan Development Communy (SADC): The Role of Instutons Kapngura Forget Mngr, Unversy of Fort Hare, East London Campus, South Afrca. Prof S.I Ikhde, Unversy of Stellenbosch Busness School, Cape Town, South Afrca. Abstract The study examned the relatonshp between external fnancal flows, domestc savngs and economc growth n the SADC regon for the perod from 1980 to 2009 specfcally lookng at the role played by nstutons. The majory of countres n the SADC regon are experencng low levels of savngs, whch has led to them relyng more on external fnancal flows to brdge the gap between domestc demand for fnance and domestc supply. However the relatonshp between external fnance and economc growth s stll a contentous ssue. Gven ths, the study has thus examned the lnk between growth and external fnance n the regon, specfcally focusng on the mpact of the dfferent forms of external fnancal flows on economc growth n the regon ncorporatng the role played by nstutons. The emprcal results revealed that all the three types of external fnancal flows have a sgnfcant mpact on economc growth n the SADC regon except ODA; however when all the dfferent types of external fnancal flows were nteracted wh the measure of nstutons, all types of external fnancal flows became sgnfcant n explanng economc growth n the regon. Ths supports the hypothess that good nstutons are necessary n promotng economc growth n developng countres. Keywords: External Fnancal Flows, Economc Growth, SADC, Panel Model, Instutons JEL Codes: F43, F21, F32 1

2 1. Introducton and Background to the Study Generally s accepted that foregn capal nfluences the course of the real economy, and ths accounts for the populary of the subject of capal flows n academc and offcal polcy dscourse. Several studes (e.g. Papaoannou 2005; Prasad, Rajan and Subramanan 2007; Cal, Massa and Wllem te Velde 2008; Obstfeld 2008; Adams 2009; Macas and Massa 2009) documents the pvotal role of prvate capal flows n promotng economc performance. However, the conclusons have been mxed. The avalable studes on the relatonshp between external fnancal flows and economc growth documents that foregn capal flows may mpact posvely on both the country of orgn and recpent countres. A study by Schoenmaker and Wagner (2011) ndcates that the presence of external fnance n the form of foregn banks n the domestc economy brng about dversfcaton effects. Foregn banks allow domestc frms to have a multple lendng relatonshps wh domestc and foregn banks. In cases n whch domestc banks are lendngconstraned, frms wll have the opton of the foregn banks. In addon to the benefs whch come wh cross border bankng, there are studes whch argue that FDI provdes benefs that domestc nvestment does not. FDI nflows are seen as vehcles for the nternatonal transfer of technology and knowhow. Borenszten et al. (1998) and Choong et al. (2010) suggest that foregn capal nflows can provde addonal capal and augment domestc savngs, promotng capal accumulaton and ncrease the growth rate. Also, external fnancal flows n the form of portfolo nvestment are thought to help n the modellng of fnancal markets n developng countres through knowledge spllover and market effcent effects (Choong et al., 2010). Whn the SADC regon, the Regonal Interactve Strategc plan states that savngs and nvestments are central determnants of the rate and pattern of economc growth n the regon. Through ncreasng domestc savngs and usng the resources n productve domestc nvestment, SADC countres wll strengthen the regon's prospects for accelerated economc growth, poverty eradcaton and sustanable development. The report shows that between 2008 and 2012, regonal Gross Natonal Savngs (GNS) fell short of regonal Gross Domestc Capal Formaton (GDCF). In 2012, the average SADC GDCF was 20.9 percent of GDP aganst a regonal GNS of 16.6 percent of GDP, leavng a resource gap of 4.3 percent and among ndvdual countres, there were wde dspares between savng and nvestment rates, wh most countres recordng negatve resource balances. Gven the low savngs rates n the regon, there s need for external resources to support the domestc nvestments. However a number of studes such as Acemoglu et al (2001); Djurovc (2012); Drffled and Jones (2013) have stressed that foregn capal flows are effectve gven that there are good nstutons n the recpent country. 2

3 Nevertheless, despe the posve effects assocated wh foregn capal flows, the ECA (2006) report shows that external fnancal flows may have negatve effects on Afrcan economes through the Dutch Dsease effects n whch the ncrease n the flow of capal ncreases the demand for the domestc currency whch wll n turn reduce the competveness of a country s export ndustry and make mports cheaper deteroratng the country s external poson. In addon, external fnancal flows may ncrease the vulnerably of a country whch s caused by the hgh volatly and unpredctably of capal flows. In addon, a number of studes such as Macas and Massa (2009) argue that the 2008 global fnancal crss mpacted negatvely on the volume of external fnancal flows to developng countres such as those n the SADC regon as the source countres focused on ther domestc economy. Also the SADC (2009) report ndcates that Oversees Development Assstance to the regon was on the declne as donor countres budgets were under pressure. Ths dlemma was further enhanced by a declne n employment levels, real ncomes and remtances from developed countres to the regon (SADC Report, 2009). Gven the uncertantes assocated wh external fnancal flows hghlghted n the prevous paragraph, and the crcal role of nvestment and nfrastructure development as a factor n regonal ntegraton whch gven the low levels of savngs has to be supplemented by external fnancal flows, the study seeks to establsh the relatonshp between external fnancal flows and economc growth n the regon. 2. Overvew of Macroeconomc Varables and Trends n Fnancal Flows n the SADC regon Assessng the state of macroeconomc condons n the regon, as ndcated n table 2, on average the SADC regon s growth have averaged around 5%. In addon the SADC countres on average have managed to reduce nflaton to sngle dg fgures. However government expendure stll surpasses government revenue. In addon publc debt s stll huge n the regon. Another nterestng fndng s the defc between nvestment and savngs. From 2008 to 2012 there s a shortfall n that nvestment surpasses savngs. Ths suggests the mportance of external fnance to brdge the gap between the two. Ths s further llustrated n fgure 2. Table 1: SADC Economc Indcators Perod Real GDP (% Growth) Inflaton (%) Government Revenues (% of GDP) Government Expendures (% of GDP) Overall Fscal Balance (% of GDP) General Publc Debt (% of GDP) Total Investment (% of GDP) Natonal Savngs (% of GDP) Per Capa GDP based on PPP (% Growth)

4 Proceedngs of the Thrd Mddle East Conference on Global Busness, Economcs, Fnance and Bankng Volume of Exports (% of Growth) Volume of Imports (% Growth) Current Account Defc (% of Growth) Source: SADC Facts and Fgures (2013) Fgure 1: Trends of Savngs and Capal Formaton n the SADC regon GCF Savngs Source: World Bank Afrcan Development Indcators Fgure 1 shows the composon of the dfferent capal flows n the SADC regon for each country. The four types of external fnance analysed are remtances, Foregn Drect Investment, Cross Boarder Bank Flows and Oversees Developemnt Assstance. Fgure 2 shows all the four types of foregn capal normalsed aganst GDP. It s clearly evdent that ODA, FDI and remtances are the major sources of foregn capal n the SADC regon. For countres such as Lesotho, remtances surpasses all other types of capal. However for Mozambque ODA s the major source of foregn capal. The same apples to Malaw and Zamba and Zmbabwe. However countres such as South Afrca and Maurus seem to depent less on these types of foregn capal, on a closer analyss Foregn Portfolo nvestment whch s not coporated due to lack of data on other coutres seem to be the major component Fgure 2: 13 Composon of Capal Flow REM FDI CBF Net ODA Source: World Development Indcators (2013) 4

5 3. Revew of Emprcal Lerature It s nterestng to note that the major focus of the avalable studes on the relatonshp between external fnancal flows and economc growth have concentrated on establshng the mpact of foregn capal on economc growth at macroeconomc level eher applyng crosssectonal or panel data to a number of countres (Borenszten, ; De Gregoro, and Lee, (1998); Carkovc, and Levne, (2002);). In addon to those focusng at cross-sectonal or panel data, there are studes based at mcro level data or case studes. These studes nclude Atque, Ahmad, and Azhar (2004); Ayanwale (2007); Khalq and Noy (2007). However, the emprcal lerature on the mpact of foregn capal on economc growth s largely nconclusve. Of noteworthy s that the majory of the cross-sectonal and panel data studes, rrespectve of the forms of foregn capal, tend to establsh a sgnfcant posve relatonshp between the measure of external fnance and economc growth (see Borenszten, De Gregoro, and Lee, 1998; Week, 2012). On the other hand the majory of tme seres country-specfc studes have establshed eher a negatve relatonshp or nsgnfcant relatonshp (see Khalq and Noy 2007; Duasa and Kassm, 2009). In addon to the posve mpact of foregn capal, there has been a prolferaton of studes whch have emphassed the mportance of nstutons n harnessng the posve mpact of foregn capal. Based on the work of North (1981), La Porta et al (1997) and Acemoglu et al (2001), Drffled and Jones (2013) argue that even though the nstutonal varables are not strongly lnked to economc growth they are however strongly correlated wh all three types of capal flows analysed by the author (FDI, Remtances and Ad). In addon the authors argue that nvestor protecton, good governance and the mantenance of law and order ncrease the extent to whch nward capal flows stmulate economc growth. The authors concluded therefore that FDI and remtances may be the trgger that generates a szeable growth payoff. In the same ven, Djurovc (2012) found that nward FDIs were attracted to the developng natons wh hgher avalably of educated labour, hgher government spendng and more effcent qualy of governance. The study pont to an exstence of a connecton between FDI and economc growth whch s strengthened when other factors are ncluded n the analyss. In lne wh the prevous studes of North (1981), La Porta et al (1997) and Acemoglu et al (2001), Drffled and Jones (2013) revewd ths further shows the mportance of other nstutonal factors when comes to realsng the posve effects of FDI. The argument of nstutons s further supported by Moura and Forte (2010). Revewng the avalable studes on the mpact of foregn capal on economc growth, the authors concluded that a common feature n most of the studes analysed suggest that the effects of foregn capal, specfcally FDI depend on the most vared condons exstng n each 5

6 country, when FDI occurs or s provded subsequently, whether they can be economc, polcal, socal, cultural or other. The authors suggest that the reasons most frequently mentoned derved from the way the country can benef from the presence of multnatonals and the advantages they carry and that can be used to mprove the host country s economy performance. Among these, the most mentoned s how the host country can gan by usng more advanced technologes and knowledge. 4. Theoretcal framework and Methodology The study s underpnned by the standard endogenous growth model n whch economc growth s nfluenced by the four forms of capal, Foregn Drect Investment (FDI), Oversees Development Assstance (ODA), Portfolo Investment (PI) and Remtances (R). Based on the work of Catrnescu et al (2009) and Drffeld and Jones (2013) s assumed that each type of fnancal flow fnances s own form of nvestment that determnes economc growth. Investment s regarded as an aggregate of publc and prvate nvestment, wh publc nvestment generally fnanced by ad n part, whlst prvate nvestment s nfluenced by FDI, R and FPI. The producton functon n ths case can be stated accordngly as: Y t Tt ( Gt At DIt FDI t Rt ) 4.1 where Y t represents output, Gt represents government nvestment spendng, At s ODA, s share of ODA used for publc nvestment purposes, DIt s domestc nvestment, FDI t s foregn drect nvestment whch also ncorporates Portfolo nvestment (PI), Rt s remtances and s the share of remtances devoted to prvate nvestment. Equaton 4.1 was expanded to take nto account Cross boarder bank flows. Thus the model becomes: Y t Tt ( Gt At DIt FDI t Rt CBF ) 4.2 In ths case CBF represents Cross Boarder Bank flows. Drffeld and Jones (2013) argue that Ad can nfluence economc growth drectly or va publc nvestment, whlst FDI and remtances nfluences economc growth through external prvate sources. The same also apples to CBF consderng that s a prvate source. In addon to the dfferent forms of capal, nstutonal factors have emerged also as mportant factors determnng the contrbuton of the forms of capal towards economc growth. Though Portfolo nvestment has featured as a very mportant source of capal as ndcated n the lerature revew secton, s nterestng to note that the majory of countres n the SADC regon fnancal markets are stll at ther nfancy. The stock markets are promnent but stll underdeveloped. The bond market n some of the countres s non-exstent. Gven ths, ths varable was dropped n the analyss therefore. The study wll therefore be based on the 6

7 above framework examnng four types of external fnance whch are FDI, CBF, ODA and remtances. 4.1 Model Specfcaton Based on the theoretcal framework dscussed n 4.1 and 4.2, the study wll employ a model n whch the four conventonal forms of external fnance are ndependent varables. The followng equaton encapsulates the estmated model: Y X 4.3 Where: Y = measure of economc growth for country. X = the dfferent forms of external fnance for country. In addon to the conventonal model specfed above, the study wll also consder other varables that determne economc growth and augments them wh the dfferent forms of external fnancal varables. Ths s mportant takng nto account the heterogeney of SADC countres. The revew of lerature has ndcated that factors such as nstutonal qualy, corrupton, law and order are other varables whch have been levelled out as other determnants of capal flows and hence economc growth. Thus the model shall ncorporate these varables to come up wh the followng model: Y Where: X Z 4.4 Y = measure of economc growth for country. X conventonal explanatory varables representng the dfferent forms of capal flows (FDI, R, ODA and FPI) for country. Z vector of explanatory varables affectng economc effcency such as Openness as a percentage of GDP, human capal, domestc nvestment, exchange rate, nflaton rate, government expendure, measure of fnancal development, poly. The model can thus be wrten as: Gdp FDI REM _ POL BCP / GDP POL REM OPE FDI _ POL ODA 3 GDS 10 CBF ODA _ POL 15 4 GFCF M 3/ GDP INF EXT _ DBT 11 5 CBF _ POL The model wll nclude two nteractve dummy varables; and followng the work of De Wet and Van Eyden (2005). The two authors argue that s mperatve to take nto account the state of development of the South Afrcan economy when carryng out a cross- 7

8 country study n whch s ncluded. In ths case represents the South Afrcan nteractve dummy varable whch wll be constructed by multplyng the savngs rate for each country by the South Afrcan dummy varable to take nto account that South Afrca has a dfferent level of capal mobly as well as the level of savngs as compared to other countres n the SADC regon. De Wet and Van Eyden (2005) show that South Afrca can be regarded as a developed country n the context of SADC or lookng at Sub-Saharan Afrca. Therefore falure to take ths nto account would bas the rest of the regon upwards. In the event that s sgnfcant, the actual coeffcent of the savngs rate would be the sum of 10 whch represents Gross domestc savngs and 17. The tme nteractve dummy ( ) on the other hand was constructed by multplyng a tme trend wh the savngs rate for each country. Ths s supported by Isaksson (2001) who argues that captures the change n capal flows for the tme perod n order to evaluate the effectveness of polcy and nstutonal changes whch are amed at capal mobly. De Wet and Van Eyden (2005) argue that a negatve value for 18 wll be an ndcaton of a decrease n the savngs rate, mplyng that polcy changes are effectve and capal s moble n the regon. 4.2 Estmaton Technques The study focused on modellng the mpact of the dfferent macroeconomc and external fnancal varables establshed n the model through the smple panel modellng technque. However pror to estmatng the panel technque, the propertes of the data wll be analysed through panel un root tests Panel Un Root tests To obtan robust results, the study utlsed several methods to carry out the un root tests n the panels used n the study. These methods are dscussed n ths secton The Levn and Ln (LL) test Ths test was orgnally developed by Levn and Ln (1992). The test s an extenson of the DF test. The model s of the followng form: Y n, t a Y, t 1 k Y, t k t u 4.6 k 1 Asterou and Hall state that the model allow for two-way fxed effects, one comng from and the other from t. Therefore, both un-specfc fxed effects and un specfc tme effects are ncluded. The un-specfc fxed effects allow for heterogeney snce the lagged restrcted to beng homogenous across all uns of the panel. The null and the alternatve hypotheses are stated as: a Y s 8

9 H 0 H a : 0 : 0 The LL test also assumes that the ndvdual processes are cross-sectonally ndependent mplyng that the pooled OLS estmator of wll follow a standard normal dstrbuton under the null hypothess. However the LL test was crcsed on restrctng to be homogenous across all The Im, Pesaran and Shn (IPS) test Im, Pesaran and Shn (1997) extended the LL test takng nto account heterogeney on the coeffcent of the Y, t 1 varable and proposng a procedure whch reles on the average of the ndvdual un-root test statstcs. The IPS test provdes separate estmatons for each secton, allowng dfferent specfcatons of the parametrc values, the resdual varance and the lag lengths. The model s gven as: Y n, t a Y, t 1 k Y, t k u 4.7 k 1 The null and the alternatve n ths case wll be formulated as: H a H : 0 for all 0 : 0 for at least one The null n ths case assumes that all seres are non-statonary process under the alternatve that a fracton of the seres n the panel are assumed to be statonary. Ths contradcts the LL test whch assumes that all seres are statonary under the alternatve hypothess. It s also mportant to note that Im et al formulated ther model under the assumpton that T s the same for all cross-sectons. Ths therefore requres a balanced panel n the computaton of the t -test statstc whch s the average of the ndvdual ADF t -statstc computed as: t 1 N N 1 t 4.8 The authors also constructed the IPS statstc for testng for un root n panels gven by: t IPS N ( t 1 N Var[ t N 1 T E[ t T 0] 0]) Panel Data Estmaton Methods The mpact of the dfferent varables to be employed n the study were analysed through panel data model. The Panel data analyss has a number of advantages over the other 9

10 approaches. These nclude s ably to ncrease the sample sze and hence obtanng better estmates. Also Asterou and Hall (2011) argue that under certan crcumstances the problem of omted varables whch at tmes cause based estmates n a sngle ndvdual regresson s not lkely to occur. Also, Wooldrdge (2002) argues that panel data controls for unobserved cross-secton heterogeney. There are three basc panel data estmaton methods namely Pooled OLS, Random Effects Model (REM) and Fxed effects Model (FEM) Pooled Regresson Model The pooled regresson model emphasses the jont estmaton of coeffcents usng the ordnary least squares (OLS) thereby ncreasng degrees of freedom and decreasng standard errors of the coeffcents (Baltag, 2008). Gven an equaton wh two explanatory varables x2 and x3 a pooled model can be wrten as: y e x2 x 3 3 The coeffcents ( 1, 2, 3 ) are assumed to be constant for all ndvdual countres n all tme perods, and do not allow for possble ndvdual heterogeney. Ths characterstc s what led to equaton 4.10 beng called a pooled model. However s major weakness s that pools all cross-sectons and tme seres data and estmates a common regresson dsregardng possble country-specfc dfferences Random Effects Model In the random effects model (REM) or the error component model (ECM), the ndvdual effects are treated as random draws from a larger populaton. The random effects model can be wrten as: y x u, u ~ IID(0, u ); ~ IID(0, ) 4.11 Where u s treated as an error term consstng of two components; an ndvdual specfc component, whch does not vary over tme, and a remander component, whch s assumed to be uncorrelated over tme. In other words all correlaton of the error terms over tme s attrbuted to the ndvdual effects. It s assumed that and ndependent and ndependent of 0 and n equaton 4.10 s unbased and consstent. u are mutually x js (for all j and s ). Ths mples that the OLS estmator for However, Gujarat (2004, 648) states that f the cross-secton or ndvdual-specfc error component s zero there wll be no dfference between the pooled and the random effects model n whch case one would smply run a pooled regresson. Baltag (2008, 17) mantans that the random effects model s approprate f draws are made randomly from a large populaton where N s generally large. 10

11 4.3.3 Fxed Effects Model (FEM) A Fxed Effects Model s a lnear regresson model n whch the ntercept terms vary over the ndvdual countres. Thus: y 2 x u, u ~ IID(0, ), 4.12 u where s usually assumed that all x are ndependent of all u. Gujarat and Porter (2009), states that these ntercepts capture the dfferent specal features of each country. In addon, Grffhs et al (2008) also states that these fxed effects can be analysed to study the extent of country heterogeney and to examne any partcular countres of nterest. However, n the event that these ndvdual effects are dentcal, one can then use a pooled least squares regresson model. The fxed effects model can be estmated by employng the Whn Q estmaton technque or the least square dummy varable (LSDV) technque. Wrng equaton 4.11 n a usual regresson framework ncludng a dummy varable for each un n the model: y N jd j 1 j x u 4.13 where d j = 1 f = j and 0 otherwse. There are a set of N dummy varables n the model. The parameters 1..., N and n equaton 4.12 can be estmated by ordnary least squares. The mpled estmator for s referred to as the least squares dummy varable (LSDV) estmator. The estmator of can be obtaned by performng the regresson as devatons from ndvdual means. Ths mples elmnatng all the ndvdual effects frst by transformng the data to obtan: y y x x ) ( u u ) 4.14 ( Equaton 4.13 does not nclude the ndvdual effects. The transformaton that produces observatons n devatons from ndvduals means n equaton 4.13 s called the whn transformaton. The OLS estmator for obtaned from the transformed model s called the whn estmator or fxed effects estmator and s dentcal to the LSDV. It s gven by: 1 N T N T ˆ FE ( x x )( x x ) ( x x)( y y ) t 1 Assumng that all x are ndependent of all u, the fxed effects estmator wll be unbased for. Both fxed effects modelng technques namely LSDV and WITHIN /Q estmaton methods bascally produces the same results. For makng nferences, the LSDV s reported as 11

12 estmates as opposed to calculatng the ntercept coeffcents and as such produces standard errors, t-statstc and p-values. 4.4 Choosng the approprate model: To decde between a pooled regresson model (restrcted) and a fxed effects model (unrestrcted), the F-test was performed F-test The F-test s constructed as follows: H 0 : N 1 H A : Not all equal to 0 0 F statstc RSS URSS / n 1 URSS / nt n k ~ F n, nt n k Where: RSS - Resdual sum of squares from the pooled model (restrcted model) URSS - Resdual sum of squares from the FEM (unrestrcted model) n - Number of cross sectons (countres n the SADC regon) t - Number of tme perods k - Number of explanatory varables The F-test essentally nvolves a comparson of the resdual sum of squares (RSS) of both models. Ths method s justfed on the bass that OLS entals mnmsng the resdual sum of squares. The ratonale of the F-test s that f after mposng restrctons a much greater RSS s obtaned then the restrcted model s not supported by the data. Smlarly, f the RSS decreases sgnfcantly followng the restrctons then s concluded that the restrcted model s supported by the data. The null hypothess of cross-secton homogeney s rejected f F- statstc > F-crcal value at 0.05 level of sgnfcance The Hausman Test The Hausman test was also carred out as a robust check to choose the approprate technque. The general dea of the Hausman test s that two estmators are compared n whch one s consstent under both the null and alternatve hypothess and one that s consstent under the null hypothess only. A sgnfcant dfference between the two estmators ndcates that the null hypothess s unlkely to hold. Assumng that E { u x s} 0 for all s, t so that the fxed effects estmator ˆ FE s consstent for rrespectve of the queston as to whether effects estmator x and are uncorrelated, whle the random ˆ RE s consstent and effcent only f x and are not correlated. 12

13 In examnng the dfference between the two whch s: ˆ FE and RE ˆ requres estmatng the covarance between V{ ˆ ˆ } { ˆ } { ˆ FE RE v FE v RE}, 4.17 The Hausman test statstc wll thus be computed as: ( ˆ H FE ˆ RE ) [ Vˆ{ ˆ } Vˆ{ ˆ }] FE RE 1 ( ˆ FE ˆ RE ), 4.18 where the Vˆ s denotes estmates of the true covarance matrces. Under the null hypothess, whch mplcly states that plm( ˆ ˆ ) 0, the statstc H has an asymptotc Ch-squared dstrbuton wh K degrees of freedom, where K s the number of elements n. The Hausman test thus wll test whether the fxed effects and random effects estmators are sgnfcantly dfferent. 5. Presentaton of Emprcal Results and Dscusson As part of the analyss of the data pror to estmaton, un root tests were performed on the data to establsh the level of ntegraton of the data. The study utlsed two tests, Levn and Ln (1992) and Im, Pesaran and Shn (2003) s IPS panel un root test so as to obtan robust results thought the IPS test was consdered more robust as compared to other tests as has the ably to preserve sample propertes of the data n the case of small samples. The results are reported n table 2. However all tests ndcate that the varables are statonary at level seres. Havng establshed the level of statonary of the varables the study proceeds wh the panel estmatons n the next secton. FE RE Table: 2: Panel Un Root tests Tests Statstc P-value Null Hypothess Each ndvdual seres contans a un root Im, Pesaran & Shn (IPS) *** ADF - Fsher Ch-square *** PP - Fsher Ch-square ** Null Hypothess Assumes common un root process Levn, Ln & Chu (LLC) *** * / [**]/ (***) denotes sgnfcance at 10%, / [5%] / (1%) level of sgnfcance respectvely. Source: Author s computaton usng Evews 8 Econometrc Software 5.1 Panel Model Estmatons To obtan robust results, all the three models under panel model were estmated to see f the results are consstent. However the best model was chosen utlsng the Hausman and F-test to be dscussed later Pooled OLS Model The pooled OLS model was estmated frst, and the results are shown n table 3. 13

14 Table 3: Pooled OLS Model Model 1 Model 2 Model 3 Model 4 Model 5 Const (2.0379) (0.8881) (1.1162) (1.1678) (0.8881) FDI (0.0302) (0.1286) (0.0288) REM (0.0663) (0.0649) (0.0286) ODA (0.0347) (0.0266) (0.0266) (0.0271) CBF (0.0462) (0.1709) (0.0309) GFCF (0.0363) (0.0367) (0.3429) (0.0345) (0.2059) M2/GDP BCP/GDP (0.0191) (0.0264) (0.1193) (0.0125) (0.0156) (0.0125) (0.1715) (0.0120) (0.1494) (0.0214) POL (0.1070) (0.0962) (0.0649) (0.0677) (0.0630) OPE (0.0136) (0.1365) (0.0086) (0.0091) (0.0817) GDS (0.0616) (0.0623) (0.0398) (0.1783) (0.3267) INF EXT_DBT (0.0059) (0.0059) (0.0050) (0.0051) (0.0039) REM_POL (0.0043) (0.0036) (0.0268) (0.0274) (0.2257) FDI_POL (0.0072) (0.0072) (0.0286) (0.0068) (0.0594) ODA_POL CBF_POL SA dummy (0.0047) (0.0048) (0.2152) (0.0047) ( ) (0.1243) (0.0409) ( ) (0.1254) (0.0041) ( ) (0.4512) (0.4054) (0.0015) (0.5321) Tme_Dummy (0.9331) (0.9693) (0.9830) (0.9759) (0.9872) Adjusted R Notes: Standard errors reported n parentheses. */(**)/[***] ndcates sgnfcance of the coeffcents or rejecton of the null hypothess on a 10%/(5%)/[1%] level of sgnfcance. Source: Author s computaton usng Evews 8 Econometrc Software The pooled OLS estmaton s the most restrctve of all the estmatons as does not take nto account the cross-sectonal heterogeney whn the SADC regon, assumng that there s a common ntercept n the whole panel. The results ndcate that all the dfferent types of external fnance nfluences economc growth posvely n the SADC regon. The mpact becomes pronounced gven that nstutons are taken nto account. The nsgnfcance of the tme dummy varable suggests that there s no dfference n the aggregate of the dfferent types of external fnance n the regon between South Afrca and other countres n the regon. 14

15 Ths may be explan n part that Foregn Portfolo Investment (FPI) forms the greater component of FPI n South Afrca as compared to the types examned n the study Fxed Effects Model The fxed effects model takes nto account the heterogeney n the cross-secton and also assumes a dfferent ntercept for each country whch s ncluded n the sample. In ths case fxed effects may represent dfferences n polcal, nstutons, economc systems amongst others whch are not explcly ncluded n the specfcaton, whch are however accounted for when estmatons are carred out. The model was estmated and the results are reported n table 4. Model 1 Const *** (0.9509) Table 4: Fxed Effects Model Results Dependent varable: GDP Model 2 Model 3 Model 4 Model * (0.9290) (1.9290) (1.9436) FDI ** ** ** (0.0286) (0.0298) (0.0301) REM ** ** (0.0232) (0.0671) (0.0166) ODA ** ** (0.0301) (0.0350) (0.0346) CBF ** ** ** ** (0.0360) (0.0170) (0.0170) (0.0172) (0.8516) GFCF ** ** ** *** (0.0214) (0.0354) (0.0368) (0.0364) (0.0348) M2/GDP ** ** ** ** *** BCP/GDP (0.0157) ** (0.0021) (0.0193) ** (0.0124) (0.0192) ** (0.0131) (0.0194) ** (0.0312) (0.0182) *** (0.0012) POL *** ** ** *** OPE (0.0070) GDS (0.0359) INF EXT_DBT ** (0.0049) REM_POL ** (0.0902) FDI_POL ** (0.0464) (0.0723) (0.0962) (0.0962) (0.0961) (0.0961) ** (0.0136) ** (0.0621) * *** (0.0059) *** (0.0036) ** (0.0164) *** (0.1365) ** (0.0622) * *** (0.0059) ** (0.0127) ** (0.0729) * (0.0137) ** (0.0627) * *** (0.0052) ** (0.0366) ** (0.1730) * (0.0136) ** (0.0623) ** *** (0.0052) ** (0.2486) *** (0.0165) 15

16 ODA_POL ** (0.0640) CBF_POL (0.0239) SA dummy ** (0.0120) * (0.0047) ** (0.0048) ** (0.0147) ** (0.0048) ** (0.0125) ** (0.0481) ** (0.0048) ** (0.0214) (0.0048) ** (0.1240) ** (0.0312) ** (0.0241) Tme_Dummy (0.1306) (0.9461) (0.9450) (0.9430) (0.9461) F ** ** ** * 4.112** Notes: Standard errors reported n parentheses. */(**)/[***] ndcates sgnfcance of the coeffcents or rejecton of the null hypothess on a 10%/(5%)/[1%] level of sgnfcance. Source: Author s computaton usng Evews 8 Econometrc Software The results ndcate that agan all the coeffcents of the dfferent forms of external fnance are posvely related to economc growth n the regon. The tme dummy varable s posve and statstcally sgnfcant n ths case suggestng that changes whch have occurred n the dfferent countres have nfluenced the flow of capal n the dfferent countres. Ths suggests also that there s a dfference n the amount external fnance whch s attracted between South Afrca and the other countres n the regon. The model has an adjusted R 2 whch s hgher than the other models Random Effects Model The Random effects model does acknowledge cross sectonal heterogeney; however dffers from the fxed effects model n that these effects are generated by a specfc dstrbuton. The model assumes cross-secton dfferences; however does not explcly model them. The model was estmated and results are reported n table 5. Constant (0.9509) FDI ** (0.0286) REM ** (0.0232) ODA ** (0.0301) CBF ** (0.0360) GFCF ** (0.0214) Table 5: The Random Effects Model Results Model 1 Model 2 Model 3 Model (0.9506) * (0.0221) ** (0.0246) * (0.0143) ** (0.0323) (0.8542) ** (0.0285) * (0.0244) ** (0.0144) ** (0.0304) (0.8363) ** (0.0282) * (0.0217) * (0.0247) * (0.0331) Model (0.7280) ** (0.2872) 16

17 M2/GDP ** (0.0157) BCP/GDP ** (0.0124) POL ** (0.0723) OPE ** (0.0070) GDS ** (0.0359) INF * EXT_DBT ** (0.0049) REM_POL ** (0.0902) FDI_POL ** (0.0464) ODA_POL ** (0.0640) CBF_POL *** (0.0239) SA dummy (0.2354) Tme_Dummy (0.1306) ** (0.0155) ** (0.0012) ** (0.0562) ** (0.0071) ** (0.0331) * ** (0.0050) ** (0.0025) ** (0.0059) ** (0.0039) ** (0.0037) (0.2300) (0.9551) ** (0.0157) ** (0.0012) ** (0.0571) ** (0.0072) ** (0.0332) * ** (0.0049) ** (0.0023) ** (0.0065) ** (0.0142) ** (0.0034) (0.4531) (0.9693) ** (0.0143) ** (0.0012) ** (0.0566) ** (0.0071) ** (0.0337) * ** (0.0049) ** (0.0025) ** (0.0065) ** (0.0039) ** (0.0038) (0.1021) (0.9693) ** (0.1382) *** (0.0012) *** (0.0565) *** (0.0070) *** (0.0134) * ** (0.0042) *** (0.2250) *** (0.0058) ** (0.0398) *** (0.0015) (0.3214) (0.9693) Notes: Standard errors reported n parentheses. */(**)/[***] ndcates sgnfcance of the coeffcents or rejecton of the null hypothess on a 10%/(5%)/[1%] level of sgnfcance. Source: Author s computaton usng Evews 8 Econometrc Software Tests were conducted to choose the best model. The results are reported n the appendx. The results ndcate that the calculated F-statstc exceeds the F-crcal value, we reject the null hypothess that the sample of the SADC countres s homogenous. Ths mples that the FEM model should be utlsed snce allows for country specfc effects n the specfcaton. The Husman test was also carred out to test the better model between FEM and REM, 2 (Ch-Square) tests results confrmed that the FEM was the approprate model that f our data. The FEM Model s therefore nterpreted n detal. The results agan ndcate that the dfferent types of external fnance nfluences economc growth posvely n the SADC regon. The tme dummy varable n ths case ndcates that there s no sgnfcant dfference between South Afrca and the rest of the countres n the SADC regon. However despe all the dfferent models beng estmated, formal tests to decde whch model s approprate and are dscussed below. 17

18 The emprcal results reveal that there s a posve and sgnfcant relatonshp between FDI and economc growth. The coeffcent of FDI s 0.22, suggestng that a un ncrease n FDI wll result n economc growth ncreasng wh about Ths s consstent wh the apror expectaton as well as a number of avalable studes. The result n n consonance wh Rvera-Batz and Romer (1991), Borenszten et al (1998), Balasubramanyam et al (1996), Worke (2005) and Petrakos et al (2007). Trpkova and Tashevska (2011) argues that ths s lkely to arse as FDI nflows foster nternatonalzaton of producton through the transmsson of deas that can be traded ndependently of goods and trade n ntermedate and capal goods that embody technology, ncreases trade openness as well as ncreasng favourable spllover effects and fnance external current account defcs whch s a common feature n many of the SADC countres. The emprcal results also suggest that there s a posve relatonshp between remtances and GDP n the SADC countres. The coeffcent for remtances s 0.36, whch s relatvely hgh. Ths result s consstent wh Taylor (1992), Woodruff and Zenteno (2001), Gulano and Ruz-Arranz (2009), Adams and Cuecuecha (2010). These authors argue that remtances act as a developmental tool through ther effect on nvestment. Ths s further supported by Ratha (2013) who argues that an ncrease n remtances apart from benefng ndvduals at the household level benefs the natonal economy as well through ncreased nvestment. The author shows that a number of studes n Mexco have ndcated that recept of remtances s assocated wh greater accumulaton of assets n farm equpment, hgher levels of selfemployment and ncreased small-busness nvestments n mgrant-sendng areas. In addon, remtances also boost domestc savngs and mprove fnancal ntermedaton whch s a very mportant factor n adng access to the fnancal system. A good case n pont s Zmbabwe. There are a number of studes (Maphosa 2004; Brackng & Sachkonye 2007; Tevera & Chkanda 2009) whch ndcates that the nflow of remtances supported development programmes n the country as the country was not able to access nternatonal fundng. The ODA varable was found to be posve though economcally nsgnfcant n the base model. Ths s not n lne wh the apror expectaton, however s supported by a number of emprcal studes such as Burnsde and Dollar (2000) who argues that ODA on s won cannot nfluence growth n Afrca. However the author argues that becomes effectve when nteracted wh another polcy varable such as nstuton or a proper monetary or fscal polcy. However other studes such as Alesna and Weder (2002), Easterly et al (2004) and Brautgam and Knack (2004) argue that there s evdence suggestng that hgh levels of AID s assocated wh deteroraton n governance n Afrca. Cross Boarder bankng was found to be sgnfcant and posve. The results suggest that a un ncrease n CBF wll amount to about 0.37 ncrease n economc growth n the SADC 18

19 regon. Ths result s consstent wh Macas and Massa (2009) on a study of the Sub-Saharan Afrca. The authors argue that cross boarder bank flows allows the realsaton of potental benefs of fnancal ntegraton whch ncludes rsk sharng, dversfcaton and better allocaton of capal among nvestment opportunes. In addon, cross boarder bankng allows banks to take advantage of economes of scale and scope, dversfy actves as well as spread rsk and revenues. Ths wll result n banks beng able to mprove resource allocaton, rsk management and mprove on ther profably. Also, an ncrease n cross boarder bank flows augments domestc fnancal capacy and ncreases the pool of borrowng. However Macas and Massa (2009) argue that nternatonal banks should not control a large secton of the bankng sector as ths would expose the bankng sector to crses, for example the global fnancal crses of 2008 whch crppled other countres bankng sector. However ths result s antagonstc to Durham (2003). The four types of fnancal flows were nteracted wh nstutons to examne f nstutons play a role n determnng ther magnude. The emprcal results show that all the coeffcents of the dfferent types of external fnance became more sgnfcant as compared when they are on ther own. The emprcal results reveal that the coeffcent for FDI became 0.54, from Ths suggests that nstutonal framework plays a very mportant role n the determnaton of the effectveness of FDI flows. Ths s n lne wh Ades and D Tella (1999), Dreher et al (2007) and Boerner and Hanz (2009) who suggest that nstutonal reforms reduce nvestment rsk, generates greater returns to the prvate sector nvestors as well as mprove on nnovaton. At the same tme nstutonal reform s lkely to elmnate sources of corrupton. Corcell et al (2012) also argues that the qualy of nstutons determne capal market effcency n terms of property rghts and n freedom from corrupton. Where property rghts are enforced and there are low levels of corrupton, ths wll encourage nvestment, reduce transacton costs and therefore encourage FDI flows. Ths s further supported by Drffeld and Jones (2013) who pont out that where nstutons fal to protect ntellectual property rghts of nvestors; ths wll nfluence the extent to whch FDI wll mpact on growth. Ths occurs manly through a reducton n the propensy of frms to nvest n a country as well as mpactng negatvely on technology transfer as well as the extent to whch nvestors wll engage wh the country s frms. The emprcal results also suggest that the magnude of remtances becomes more pronounced when s nteracted wh nstutons. Ths agan stresses the mportance of nstutons n determnng the extent to whch remtances mpact on economc growth. Ths s supported by a number of studes such as Catrnescu et al (2009). The authors argue that good nstutons are mportant to encourage people to rem to ther countres. In a suaton where nstutons are not n order, for example the case of an nformal economy, there s 19

20 lkely to be hgh transacton costs. Also n such nstances there maybe theft of remtances. In such cases ther effect on economc growth wll be reduced. The results thus suggest that SADC countres should mprove on ther nstutonal frameworks so as to realse the benefs of remtances. The effect of AID on the other hand when nteracted wh nstutons becomes sgnfcant suggestng that good nstutons are a prerequse for the effectveness of ad. Ths s consstent wh Burnsde and Dollar (2000) who establshed that ad on s own has no effect on growth; however when s nteracted wh nstutons wll have a condonal effect. Ths s also consstent wh Burnsde and Dollar (2004), Chauvet and Gullaumont (2004). However cross boarder bank flows become more sgnfcant when nteracted wh nstutons. Ths agan supports the mportance of nstutons n determnng the effectveness of external fnance. The measure of nstutonal qualy on s own was found to have a posve relatonshp wh growth. Ths s supported by Acemoglu et al (2001) and Rodrk (2004). The results suggest that nstutons are of great mportance to the success of the economes n the SADC regon. Drffed and Jones (2013) argue that nstutons act as the bedrock of a successful functonng mxed economy as markets are embedded n the nstutons. The role of nstutons as ndcated earler on mpacts also on other mportant determnants of economc growth. Thus becomes mportant to always mantan qualy nstutons. Another mportant fndng n the study s the negatve value of the tme nteractve dummy whch was negatve n all the estmatons suggestng that nstutonal changes and polces amed at attractng external fnance has been successful. In addon to the dfferent types of fnancal flows examned n the study, a number of macroeconomc varables were also examned whch nclude nflaton, openness, fnancal development, Gross fxed capal formaton, nstutonal qualy, and domestc savngs. All these macroeconomc varables have proved to have a sgnfcant mpact on economc growth n the SADC regon. All the varables had the expected sgn. 6. Conclusons and Polcy Implcatons of the Study The study focused on analysng the mpact of the dfferent forms of external fnancal flows on economc growth n the SADC regon emphassng the mportance of nstutons utlsng a panel model. The analyss n the study hghlghts that the mportance of external fnancal flows to countres n the SADC regon cannot be underestmated gven that countres have low levels of savngs to fnance nvestment. Thus the gap between savngs and nvestment can only be flled n wh external fnancal flows. The emprcal results revealed that three types of external fnance, FDI, CBF and remtances have a posve mpact on economc growth n the regon. On the other hand ODA was found to be nsgnfcant. 20

21 However when all the types of external fnancal flows were nteracted wh the measure of nstutons, they all became sgnfcant ncludng ODA. For the other three types of fnancal flows the mpact was more sgnfcant stressng the mportant role played by nstutons n enhancng growth n the regon. The growth regresson results have strongly ndcated that the qualy of nstutons matter most for growth n the SADC regon. Apart from mproved nstutonal qualy attractng foregn fnancal flows, also has a posve mpact on growth. Ths shows the mportance of the nstutonal varables n promotng growth n the regon. Countres where there s no nvestor protecton and mantenance of law and order are lkely to lose out on attractng fnancal flows. Ths wll retard economc growth. A good case s Zmbabwe whch based on s law of ndgensaton, there has been capal flght coupled wh a shrunk n GDP. References Acemoglu, D., Johnson, S. and Robnson, A. (2001) The colonal orgns of comparatve development: An emprcal nvestgaton. Amercan Economc Revew 91(5): Adams Jr, R.H. and Cuecuecha, A. (2010) Remtances, household expendure and nvestment n Guatemala. World Development 38(11): Adams, S (2009). Can foregn drect nvestment (FDI) help to promote growth n Afrca? Afrcan Journal of Busness Management Vol.3 (5), pp , May Ades, A. and D Tella, R. (1999) Rents, competon and corrupton. Amercan Economc Revew 89(4): Alesna, A. and Weder, B. (2002) Do corrupt governments receve less foregn ad? Amercan Economc Revew 92(4): Asterou, S. G and Hall, D (2011) Appled Econometrcs, A modern approach usng Evews and Mcrof (revsed edon). Palgrave Macmllan. Atque, Z., Ahmad, M. H., & Azhar, U. (2004). The mpact of fd on economc growth under foregn trade regmes: A case study of Pakstan. The Pakstan Development Revew, 43 (4), Ayanwale A. B. (2007) FDI and Economc Growth:Evdence from Ngera, Afrcan Economc Research Consortum Research Paper No. 165,2007. Balasubramanyam, V.N., Salsu, M. and Sapsford, D. (1996) Foregn drect nvestment and growth n EP and IS countres. The Economc Journal 106(434): Baltag, B. H. (2008) Econometrcs, 4 th Edon, Sprnger-Verlag Berln Hedelberg Boerner, K. and Hanz, C. (2009) The polcal economy of corrupton and the role of economc opportunes. Economcs of Transon 17(2): Borenszten, E., De Gregoro, J. and Lee, J.W. (1998) How does foregn drect nvestment affect economc growth? Journal of Internatonal Economcs 45(1): Borenszten, E., De Gregoro, J. and Lee, J.W. (1998) How does foregn drect nvestment affect economc growth? Journal of Internatonal Economcs 45(1): Brackng, S and Sachkonye, L (2006), Remtances, poverty reducton and the nformalsaton of household wellbeng n Zmbabwe, Economc & Socal Research Councl, GPRG-WPS-045. Brautgam, D. and Knack, S. (2004) Foregn ad, nstutons, and governance n Sub-Saharan Afrca, Economc Development and Cultural Change 52(2): Cal, M; Massa, I & Wllem te Velde, D (2008). The Global Fnancal Crss: fnancal flows to developng countres set to fall by one quarter, Oversees Development Instute. 21

22 Carkovc M, Levne R (2002). Does Foregn Drect Investment Accelerate Economc Growth? n H. T. Moran, E. Graham and M. Blomstrom (eds.), Does FDI Promote Development? Washngton, DC: Instute for Internatonal Economcs. Catrnescu, N., Leon-Ledesma, M., Pracha, M. and Qulln, B. (2009) Remtances, nstutons, and economc growth. World Development 37(1): Choong, C, Ahmad Z B, Zulkornan Y and Habbullah, M, S (2010), Prvate capal flows, stock market and economc growth n developed and developng countres: A comparatve analyss, Japan and the World Economy 22: Corcell, F., Drffeld, N.L., Roland, I and Pal, S. (2012) When does leverage hurt productvy growth? A frm level analyss. Journal of nternatonal Money and Fnance 31(6): De Wet, A. H and Van Eyden, R (2005) Capal Mobly n Sub-Saharan Afrca: A Panel data Approach, South Afrcan Journal of Economcs, Vol.73.1: Djurovc, A (2012) The Impact of Foregn Drect Investment on the Economc Growth of Developng Countres ( ) (October 10, 2011). Economc Development, 2012, 2-3, Dreher, A., Kotsoganns, C. and McCorrston, S. (2007) Corrupton around the world: Evdence from a structural model. Journal of Comparatve Economcs 35(3): Drffeld, N and Jones, C (2013) Impact of FDI, ODA and Mgrant Remtances on Economc Growth n Developng Countres: A System Approach, European Journal of Development Research 25,2: Duasa, J. & Kassm, S.H. (2009). Foregn portfolo nvestment and economc growth n Malaysa. The Pakstan Development Revew, 48(2), Durham, J.B. (2003), Foregn Portfolo Investment, Foregn Bank Lendng, and Economc Growth. Internatonal Fnance Dscusson Papers 757. Washngton, DC: Board of Governors of the Federal Reserve System. Easterly, W., Levne, R. and Roodman, D. (2004) Ad, polces, and growth: Comment. Amercan Economc Revew 94(3): Economc Commsson for Afrca (2009) Economc Report on Afrca 2009; Developng Afrcan Agrculture Through Regonal Value Chans; Uned Natons Economc Commsson for Afrca, Adds Ababa, Ethopa. Gulano, P & Ruz-Arranz, M, (2009). "Remtances, fnancal development, and growth," Journal of Development Economcs, Elsever, vol. 90(1), pages , September. Gujarat, 2004 D.N. Basc Econometrcs (Fourth ed), McGraw-Hll/Irwn, New York (2004). Im, K., M.H. Pesaran, and Y. Shn (2002), Testng for Un Roots n Heterogeneous Panels, Journal of Econometrcs, 115, Isaksson, A. (2001). Fnancal Lberalzaton, Foregn Ad, and Capal Mobly: Evdence from 90 Developng Countres. Journal of Internatonal Fnancal Markets, Instuton, and Money 11: Khalq, A & Noy, I (2007). "Foregn Drect Investment and Economc Growth: Emprcal Evdence from Sectoral Data n Indonesa," Workng Papers , Unversy of Hawa at Manoa, Department of Economcs. La Porta, R., Lopez-de-Slanes, F., Shlefer, A. and Vshny, R.W. (1997) Legal determnants of external fnance. Journal of Fnance 52(3): Levn, A., Ln, C.-F., May (1992). Un root tests n panel data:asymptotc and "ne-sample propertes. UC San Dego, Workng Paper Macas, J B and Massa, I (2009) The Global Fnnacal crss and Sub-Saharan Afrca; The effects of slowng prvate capal nflows on growth, Overseas Development Instute, Workng Paper 304. Maphosa, F. (2005) The Impact of Remtances from Zmbabweans Workng n South Afrca on Rural Lvelhoods n the Southern dstrcts of Zmbabwe, Forced Mgraton Workng Paper Seres #14 Forced Mgraton Studes Programme, Unversy of the Wwatersrand [onlne] (accessed 15 September 2012). 22