Analyzing the Relationship between Income Inequality and Economic Growth: Does the Kuznets Curve Exist in Thailand?

Transcription

1 Analyzing he Relaionship beween Income Inequaliy and Economic Growh: Does he Kuznes Curve Exis in Thailand? Paravee Meneejuk and Woraphon Yamaka Absrac. This sudy aemps o explore he relaionship beween income inequaliy and economic growh, and paricularly he exisence of he Kuznes curve in Thailand s economy. The esimae resuls show ha he Kuznes hypohesis does no hold in he case of Thailand bu we can discover an alernaive way o reduce income inequaliy ha is he inflaion and he share of privae secor. Addiionally, his sudy applies he same algorihm for invesigaing he Kuznes curve o he regions o answer he quesion Are here regional Kuznes Curves in Thailand? Our model shows some ineresing resuls abou income disribuion in he differen regions ha may challenge he convenional heory of Kuznes.. Inroducion The relaionship beween income dispariy and economic growh has been a criical issue in macroeconomics over several decades. Thailand is one of he counries ha need economic growh o improve sandard of living, reduce povery, and also o survive in he modern economy. Bu in he meanime a rise in growh can also cause he inequaliy which may lead o a long-sanding problem which in urn can become desrucive o he growh. Alan Greenspan, he chairman of he Federal Reserve of he Unied Saes from 987 o 006, gave noion on he negaive consequences of oo much income equaliy on he economic sabiliy and he democraic capialis sysem ha Income inequaliy is where he capialis sysem is mos vulnerable. You can have he capialis sysem if an increasing number of people hink i is unjus. (Greenspan, 007). While we seem o consider he inequaliy as a consequence of economic growh, some economiss end o argue ha inequaliy may affec he growh a he same ime and perhaps i is necessary o propel he growh. Two sides of he same coin is wha Osry and Berg, IMF saff, used o define he effec of inequaliy on economic growh hrough heir paper in 0. They reckoned ha he impac of unequal disribuion of income is spli ino wo sides. Inequaliy can influence he growh negaively and become a problem for sociey. For example, i can obsruc he human capial accumulaion especially in educaion. Tha is because individuals having higher income can inves more in educaion han poor people can and have los of choices in occupaion. Recenly, IMF has released one more discussion paper abou inequaliy and growh eniled Causes and Consequences of Income Inequaliy: A Global Perspecive (Dabla-Norris e al., 05). One par of his paper refers o a negaive effec of inequaliy on growh and reckons ha inequaliy can also lead o poliical and economic insabiliy, resuling in corrupion, resource misallocaion, and nepoism. In Faculy of Economics, Chiang Mai Universiy, Chiang Mai, Thailand Address: mparavee@gmail.com, woraphon.econ@gmail.com

2 conras, inequaliy can also be good for economy since i provides incenives for people o inves and save for heir fuure livelihood and also for enrepreneurs o run heir businesses successfully (Lazear and Rosen, 98). Inequaliy does no insanly propel economy bu wihou inequaliy economy is like socialism where he growh of economy hardly exiss. Trade-off beween increasing growh and reducing inequaliy While growh is needed, inequaliy sill maers. Some economiss sugges ha an improvemen in income disribuion and economic growh can be achieved a he same ime (Boonyamanond, 007). Simon Kuznes, an American economis who received he Nobel Memorial Prize in Economic Sciences in 97, hypohesizes ha as a counry develops he economic inequaliy would firs increase and hen laer decrease (Kuznes, 955). This hypohesis is graphed as an invered-u shape called Kuznes Curve where he horizonal x- axis is income per capia and he verical y-axis is economic inequaliy. This invered-u shape curve implies ha income inequaliy following he curve is gradually increased resuling from economic evoluion. Tha is, a he iniial level of low economic growh, income inequaliy is also low. Then, as counries seek o increase he rae of economic growh, indusrial secor ofen plays a key role o achieve he goal. This even leads o a big gap beween rich and poor, and hence income inequaliy increases. According o he Kuznes Hypohesis, he level of inequaliy keeps rising as long as he economic growh increases. However, as he same hypohesis, inequaliy will decrease afer he growh reaches some cerain level, a hreshold, hen he economy is developed efficienly resuling in an increase in average income, an improvemen in social welfare of sociey, higher educaed populaions, and so on. Moivaed by his reasoning, his paper aims o invesigae wheher he Kuznes hypohesis holds in Thailand s economic sysem where Thai governmen has been rying o promoe he economic growh and especially wih a concern abou inequaliy a he same ime. This is crucially imporan because, if we could prove ha he hypohesis holds and he resul could illusrae he hreshold accuraely, i means we can find he proper rae of economic growh ha will no hur income disribuion in he long erm. Our findings should be useful for he Thai governmen and policy makers and can conribue o longer-run benefis for sociey.. Lieraure Reviews As in he lieraure, many sudies aimed o measure he effec of economic growh on he income inequaliy and paid aenion on an invesigaing of he Kuznes hypohesis. We found he differenial resuls depending on he counry s or region s characerisic, he daa used (ype and period of ime), and he mehod for esimaion. However, he resuls can be separaed ino wo groups. The firs group is he empirical evidences ha quie clearly accep a Kuznes hypohesis in favor of he invered U-shaped relaionship beween income inequaliy and economic growh. We found he exisence of he Kuznes hypohesis in many counries such as he sudy of Heshmai (004), Khalifa and El Hag (00) and Ogus (0). However, here are sill some sudies ha rejec his hypohesis or could no find he invered U-shaped relaionship beween hese wo facors. For example, Hsing and Smyh (994) sudied he invered U-shaped curve of income inequaliy and economic growh and found an insignifican effec of hese wo variables hrough he U.S. daa. Moreover, he sudies of Deininger and Squire (996), Easerly (999) and Dollar and Kraay (00) also showed some evidences abou he nonexisence of he Kuznes curve. As, we menion above, he heerogeneous resuls were obained in he previous lieraures. We canno insanly conclude ha he Kuznes hypohesis is held for Thailand. In

3 he case of Thailand, he effec of economic growh on income inequaliy has been invesigaed by various sudies, for example Ikemoo and Uehara (000), Noppakhun (007), and more recen Preechamea, (05). Some of hem could find a lile suppor for he Kuznes hypohesis in Thailand; however, mos of he sudies were limied by he lack of he efficien saisical ools. These sudies ypically rely on linear regression in which some of hem decide o spli he observaion ino differen group and esimae differen coefficiens. These previous algorihms may face o a bias resul easily and, someimes, canno capure he complexiy of he process. We suspec ha he lack of he efficien ools and he esimaion mehod are he imporan reason ha leads o he various resuls. Firs and foremos i has o be said ha here is no such hing as he righ measuremen of he effec of economic growh on inequaliy. The quesion which mehod should be chosen is dependable on how suiable he mehod is explaining he given problem. We found ha many empirical sudies are now based on he linear rigid quadraic regression model such as he auoregressive disribued lag model (ARDL) in he sudy of Cheng and Wu (04) and error correcion model (ECM) in he sudy of Shahbaz (00). However, he regression resuls cauioned abou esimaion of only parameric forms which produce sricly concave funcions o es Kuznes hypohesis. Therefore, non-parameric and semi-parameric approaches are proposed o examine how income inequaliy boh wihin and across counries has evolved a differen levels of developmen (see, Frazer, 006). For a brief review of he mehodology used o deec he Kuznes hypohesis, refer o Ganaie and Kamaiah (05). However, he srong assumpion of simple linear relaionship beween inequaliy and subsequen growh may lead he low accuracy of he resul and i migh no be appropriae o invesigae he relaionship beween inequaliy and growh under he Kuznes hypohesis. Banerjee and Duflo (003) examine he daa wihou imposing a linear srucure and i becomes clear ha he daa does no suppor he linear srucure and exis he non-lineariies in his relaionship. Recenly, few sudies have invesigaed relaionship beween inequaliy and subsequen growh. One of he excepions is he sudy carried ou by Savvides and Segnos (000). They employed a hreshold regression model and perform a es of invered-u hypohesis. The advanage of his model is ha we can achieve our es wihou spliing he sample o esimae differen coefficiens for poor and rich counries or regions and for differen ime periods. In his paper, he non-lineariy relaionship is aken ino accoun, we employ a Kink regression approach wih unknown hreshold of Hansen (05) insead of using hreshold regression model since he funcional form of he model is disconinuiy. Moreover, he Kink regression is exended o smooh Kink regression in order o provide a beer fi o invered U- shaped relaionship beween inequaliy and growh. This new model will become one of he conribuions of his sudy. 3. Mehodology: Economeric Modeling Framework Invesigaing he invered U-shaped curve or he so-called Kuznes curve is a profoundly complicaed problem and difficul o do in he real world because economic growh and income inequaliy hemselves acually depend on many facors. (We will discuss abou he facors affecing growh and income inequaliy laer in he fifh par: model specificaion and variables.) Therefore, an appropriae ool for invesigaing his phenomenon wih special concern abou he real mechanism of growh and income inequaliy becomes difficul, bu imporan. As in he lieraure, numerous sudies failed o find he Kuznes curve especially he empirical sudies on ime series, perhaps, due o a lack of an effecive ool (See e.g. Gallup (0)).Therefore, his paper is going o presen a novel ool for invesigaing he Kuznes hypohesis ha is he simulaneous kink equaion (SKE) model. 3

4 As we are dealing wih a simulaneous relaionship beween income inequaliy and economic growh where each equaion conains several specific facors, and we speculae ha some covariae variables can produce effecs boh negaive and posiive since many sudies promoe he idea ha he economy behaves differenly if values of cerain variables lie in one sae raher han in anoher. Therefore, we modify a sudden swiches kink regression model wih unknown hreshold of Hansen (05) o smooh swiches kink regression model and apply o he simulaneous equaion model in order o capure our consideraions idenifying he unequal effec of growh on he disribuion of income or he so-called Kuznes hypohesis. Wha is he kink regression? Briefly, i is simply a regression disconinuiy model. The regression funcion iself is coninuous bu he slope is disconinuous a a hreshold poin, hereby a kink and a smooh ransiion beween regimes is ofen more convenien and realisic han jus sudden swiches ransiion. Wih ha in mind, we have he simulaneous smooh ransiion kink equaion (SKE) model as an economeric model for his work. 3. Inroducion o he Simulaneous Smooh Transiion Kink Equaion (SKE) The SKE model wih m -simulaneous equaion can ake he form as he following equaion where ( y,..., y m ) and (,..., m ) are T vecor of dependen variable and error, respecively. The erm X is m k marix of exogenous regressors, and j Y i is a T G marix of h G endogenous regressors on he righ-hand side of he i equaion. (,..., m) is he sigma parameer of margin,, m and (,..., m ) are he margin error ha follows some disribuions. G G k k i ( F(r i,s i )) i i F(r i,s i ) i j( F(r j,s j)) j j F(r j,s j) j i i j j y Y Y X X G G k k m m mi mi mi i mi mi mi i mj mj mj j mj mj i i j j y ( F(r,s )) Y F(r,s ) Y ( F(r,s )) X F(r, s ) X mj j m m () Following Teräsvira (998), we separae each regressor variable ino wo regimes using he funcional form of he ransiion funcion which akes he form as F(r, s), () r ( s Xi ) e where r is a kink or hreshold parameer, s is smooh ransiion parameer and hus F(r,s) h has a monoonously increasing funcion. For i equaion, he coefficiens wih respec o he righ hand side variables equal o ( i,.., mi) and ( j,..., mj ) for value of Y and i X imes i ( F(r,s)). Conversely, hey are equal o ( i,.., mi) and ( j,..., mj ) for value of Y and i X i imes ( F(r, s)). Ye he regression funcion is coninuous in all variables. Imporanly, his model is assumed o have some correlaion beween he errors of he m -equaions, so ha he Copulas are employed o model he nonlinear dependence srucure for our model. Here we have he Copula based SKE model wih differen marginal disribuions as suggesed by Wichiaksorn, Choy, and Gerlach (006) and Paspipakul, e al. (06). Therefore, in he nex secion, he general idea of copula approach will be presened in brief. 4

5 3. Copulas Copulas are paramerically specified join disribuions which are generaed from any given marginal disribuions (Trivedi and Zimmer, 007). The main advanage of copulas is ha i can separae he marginal behavior and he dependence srucure of he variables from heir join disribuion funcion (Mahfoud and Michael, 0). This paper conduced boh bivariae Ellipical Copulas and Archimedean Copulas o model he dependence srucure of he SKE model( m ). Thus, le u and u be he marginal disribuion of and, wih u F( ) and u G( ), hen according o Sklar s Theorem here is a copula funcion C such ha: C( u, u ) C( F( ), G( )) (3) where u and u are bound on [0,] and are uniform disribuion Here, he Ellipical class consiss of he symmeric Gaussian and -Copula. The Archimedean Copula consiss of he asymmeric Gumbel, Fank, Clayon, and Joe. (See Cherubini, Luciano and Vecchiao, 004 and Wichiaksorn, Choy, and Gerlach, 006) 3.3 Esimaion of he Copula based SKE model In his paper, wo equaions SKE model is considered. Thus, he bivariae Copula is conduced for joining hese wo equaions ogeher. Prior o he model esimaion, we need o check he saionariy of our daa using he Augmened Dickey-Fuller uni roos es. The esimaion echnique ha we employ here is a maximum likelihood mehod where he log likelihood funcion of our model can be defined as: T log L( y, Y, y, Y, X, X, u, u,, ) (log l( y, Y, X ) log( y, Y, X ) log c( u, u ) i (4) where, i, i, r i, r j,s i,s j, j, j,, i, i, r i, r j,s i,s j, j, j and are he esimaed parameers of equaion, equaion, and copula funcion, respecively. In his sudy, we proposed four differen marginal disribuions consising Normal, Suden-, skewed Normal, and skewed Suden- disribuions. Copula families, i.e. Gaussian, Suden-T, Clayon, Gumbel, Frank, and Joe are proposed o join hose marginals in order o consruc he join disribuion funcion of a bivariae random variable wih he univariae marginal disribuion. 4. Simulaion Sudy We conduc a simulaion sudy o explore he performance and accuracy of our proposed model Copula based SKE. We apply he Suden- copula o model he dependence srucure of he bivariae SKE model and se he rue value for he correlaion coefficien a 0.5 while he addiional degrees of freedom parameer c is equal o 4. Then, he obained uniform daa are ransformed o be and using he quanile funcion of marginal disribuion choices where and are equal o, he addiional degrees of freedom parameer of quanile funcion of he Suden- disribuion,, is equal o 6. Thus, we can wrie our simulaed Suden- copula based SKE wih normal and Suden- margins as in he following equaions: 5

6 y ( F(r,s )) X F(r,s ) X. y ( F(r,s )) y F(r,s ) y To generae he dependen y of he firs equaion, he independen variable X are randomly simulaed from N (5,0) where he kink poin or hreshold value for Xis r =5, he smooh parameer s =0.5, and he rue coefficiens for, 0.5 and. To simulae y of he second equaion, independen y are simulaed from he firs equaion, he y is r =0, he smooh parameer s =0.5, and he rue kink poin or hreshold value for coefficiens for, 0. and. Table. Simulaion Resuls of Copula-based SKE wih normal and suden- margins Copula True Esimae S.E. Copula True Esimae S.E Gaussian Suden v v r r r r s s s s Joe Clayon v v r r r r s s s s Gumbel Frank v v r r r r s s s s Source: Calculaion. 6

7 Table shows he resuls of he Mone Carlo simulaion invesigaing he maximum likelihood esimaion of he Copula based SKE model. We found ha our model can perform very well hrough simulaion sudy. I is found ha he mean parameers are very close o heir rue values wih low and accepable sandard errors. The resuls demonsrae he accuracy of esimaion. For example, Table represening Normal and Suden s margins shows ha he mean value of he coefficien is.84 wih sandard error equals 0.5 while he rue value is. Overall, he Mone Carlo simulaion suggess ha our proposed model Copula based SKE is reasonably accurae. 5. Model Specificaion and Variables Apar from he difficuly of economeric framework, his par is abou an economic model and some discussion abou he variables of he model. To es a relaionship beween income inequaliy and economic growh, we use he following wo equaions which are modeled simulaneously. The firs equaion represens he economic growh in which we consider no only he rae of increase in GDP bu also many oher facors influencing he growh, such as moneary base, human capial accumulaion, insiuional change, and governmen deb. These variables will be discussed laer. The second equaion represens inequaliy in income disribuion. We measure he inequaliy using he bes known index Gini coefficien and consider all possible facors ha end o have a major impac on disribuion of income, such as subsidies, privae secor share of GDP, inflaion, governmen expendiure on educaion, and rae of unemployed labor. A key poin of his model is ha he growh equaion conaining Gini coefficien and he rae of growh is also enered ino he Gini equaion in order o find a direc link beween hese wo facors. GDP GDP ln( M ) ln( HC ) ln( Ins ) 3 4 Gini Deb u 5 6 Gini GDP ln(sub ) ln( PS ) ln( Inf ) 3 4 Edu Unemp u 5 6 (5) (6) A his sage, we consruc his model and decide which economic variables should be included by combining a lo of informaion from exper advices and lieraures. The firs par of he growh equaion includes change in nominal GDP lagged by one period ( GDP ) and moneary base ( M ). The moneary base refers o a par of he money supply which is highly liquid including noes, coins, and commercial bank deposis wih he cenral bank. Economis John Taylor of Sanford Universiy has quaniaively shown he posiive impac of he moneary base growh on GDP growh hrough he U.S. daa. The work of Kimura, Kobayashi, Muranaga, and Ugai (003), he saff of he Bank of Japan, also suppors his idea. They found ha an increase in he moneary base can simulae Japan s economy a zero ineres rae. We also consider his variable and expec o see he significan role of he moneary base in simulaing he oupu in Thailand s economy. The nex variable is human capial accumulaion ( HC ) which comes from an argumen from Gregory, Romer, and Weil (99). They argued ha he radiional framework of Solow growh model should include human capial accumulaion, in erms of educaion and raining, healh saus, and experience, since hey experimenally found ha human capial has a grealy posiive impac on economic growh and may induce susainable growh. Also, insiuion ( Ins ) should be considered as a The Gini coefficien is a number on a scale measured from 0 o, where 0 represens perfec equaliy and represens oal inequaliy (Ray, 993). 7

8 cause of he long-run growh (Acemoglu, Johnson, and Robinson, 005). The insiuion in a sociey is commonly relaed o such social mechanisms as democracy, religion, propery righs, and rule of law. The auhors grealy explained hrough heir work ha insiuion deermines he incenives and creaes consrains in he economy. For example, as he impac of very differen economic and poliical insiuions, Souh Korea has much higher economic growh rae and beer sandard of living han Norh Korea. The las included variable in he growh equaion is change in governmen deb ( Deb ). Ghazanchyan, Sosky, and Zhang poined hrough heir recen work in 05 ha high governmen spending may induce growh, bu he excessive amoun of spending may cause deb, and hen lead o lower growh. Similarly, Bivens and Irons (00) also comprehensively explained in he same way ha deb maers for economic growh, as found from heir experimen ha 90% of deb o GDP will lead o a negaive growh effec and slower economic growh. Nex, we will discuss abou poenial sources influencing income inequaliy. I is no easy o choose which facors should be included in he inequaliy equaion because inequaliy is somewha complicaed and ofen unclear o define. Therefore, following he sudy of Anneli Kaasa (003), we are able o separae he facors ino four groups. The firs group is economic growh and general developmen of a counry in which we selec he growh in GDP ( GDP ) o represen his group and o describe a paricular counry s developmen level. In addiion, bu imporanly, hey are chosen jus o ac in keeping wih he Kuznes Hypohesis. The nex variable is subsidies ( Sub ) or he ransfer. OECD (0) repored ha he ransfer sysems can help reduce overall income inequaliy in all counries. For example, on average across he OECD, hree quarers of he reducion in inequaliy is due o he ransfers. And hence, we ake ino accoun his variable and aemp o see he effec of his variable in he case of Thailand. The second group comprises he poliical facors which are defined as privaizaion and he public secor [Anneli Kaasa (003)]. Privaizaion is believed o cause wealh concenraion which finally leads o income inequaliy. Similarly in he public secor facor, Gusafsson and Johansson (999) poined ou ha he wage gap in he public secor is ypically lower han in he privae secor. Therefore, he bigger share of privae secor ( PS ) may imply he higher income inequaliy; hence we should have a special concern and include his facor ino our inequaliy equaion. The hird group is macroeconomic facors. We especially consider inflaion rae ( Inf ) and he number of unemployed people ( Unemp ) o represen macroeconomic influence on inequaliy. The effec of inflaion is found o increase inequaliy since high inflaion leads o a redisribuion of resources wih fixed nominal income. To suppor his idea, we refer o an empirical sudy on income disribuion by Bulir (998) which indicaed ha inflaion affecs levels of income inequaliy as measured by Gini coefficien. The influence of unemploymen is commonly undersood in is own righ as i obviously has inequaliy increasing effec. Finally, he fourh group is demographic facors; here we focus on some par of demographic developmen in which educaion is our special concern. We decide o use a variable expendiure on educaion ( Edu ) o illusrae his group because educaion is grealy imporan for reducing uneven income. I is found ha a counry wih high invesmen in educaion can significanly reduce income inequaliy hrough a developmen in level of human capial (Debla-Norris e al., 05). 6. Daa We use a quarerly daa se of all considered variables, spanning from 993:Q o 05:Q4. The economic growh daa include a variey of differen facors as described previously: GDP, human capial index, insiuion index, moneary base, and he governmen 8

9 deb. In addiion, he income inequaliy daa include he Gini coefficien, he governmen subsidy and ransfer paymen, privae secor share of GDP, inflaion rae, he governmen expendiure on educaion, and he rae of unemployed labor. We colleced all he daa from Thomson Reuers DaaSream, from Financial Invesmen Cener (FIC), Faculy of Economics, Chiang Mai Universiy, excep he Gini coefficiens which were colleced from he Naional Saisical Office of Thailand and World Bank daa. The regional Gini coefficiens were colleced from Office of he Naional Economic and Social Developmen Board (NESDB), covering he years 997 o 03. Noe ha, human capial index, Gini coefficien and insiuion index were disaggregaed from he annual o he quarerly series by he emporal disaggregaion mehod of Chow and Lin (97). Table. Daa descripion Variable GDP EDU Inf Unemp Sub PS Mean Median Maximum Minimum Sd. Dev Skewness Kurosis Variable Ins Gini HC M Deb Mean Median Maximum Minimum Sd. Dev Skewness Kurosis Source: Calculaion. The descripive saisics of all daa series are presened in Table. In addiion, as we use ime series daa, esing for saionariy of daa is hen required. Here we apply he Augmened Dickey-Fuller (ADF) uni roos es o he subses of daa o examine wheher or no he daa series conains a uni roo. We do no show he resul here in his paper bu we found ha all variables passed he es a level wih probabiliy equal o zero, meaning all of hem are saionary 7. Empirical Resuls In his secion, we will presen he resuls of a se of experimens sequenially. The experimen begins wih ploing he simple link beween income inequaliy and economic growh and finding he curren posiion of Thailand among all counries around he world. Then, he esimaed resuls regarding he quesion of wheher or no here is he Kuznes curve in Thailand s economy will be indicaed in he following par. We are going o sar wih he es for a kink effec on each variable, and hen, as a sequenial esimaion mehod, we will selec he bes-fi Copula for our model before obaining he esimaed resuls in he las par. Addiionally, in he las wo pars of his empirical resul, we will exend his algorihm for invesigaing he Kuznes curve o he regions o answer he quesion Are here regional Kuznes Curves in Thailand? In addiion, we will evaluae he efficiency of some Thailand s economic and social developmen plans by looking a heir accomplishmens of boosing he economy and reducing income inequaliy. 9

10 7. The simple link beween inequaliy and growh: Wha is he posiion of Thailand? In his secion, he Gini coefficien and he GDP growh in he year 05 of 44 counies around he world including Thailand are used o indicae he simple link beween inequaliy and economic growh and also invesigae he posiion of our counry. To achieve hese goals, he fuzzy c-means algorihm inroduced by Gusafson and Kessel (978) is conduced o pariion he Gini coefficien and GDP growh in a flexible manner. We need o deermine he pariion of he variables regarding he number of clusers. These clusers have fuzzy boundaries, ha is, each daa value belongs o he cluser o some degree. Their modelling echnique involved clusering he daa ino fuzzy ses, esimaing he relaionship of Gini and GDP growh over each se, and hen combining each sub-model ino a single overall model by using he membership funcions associaed wih he clusered daa. According o he resul shown in Figure and Table 3, he analysis reveals ha he counries can be separaed ino wo clusers. The differen cluser can reflec he economic sabiliy and he membership saus wihin he world. We define Cluser (Red do) as he counries wih low income inequaliy and low economic growh and Cluser (Blue do) as he counries wih high income inequaliy and high economic growh. Fig.. Plo of Gini coefficiens agains GDP growh, 05 Table 3. Cluser descripion Fuzzy Observaion Mean of Growh Mean of Gini Cluser (Red do) Cluser (Blue do) Source: Calculaion. The resul shows ha Thailand locaes in he cluser (based on his daa se), meaning ha Thailand is in he group of counries wih lower income inequaliy and low economic growh. To compare he abiliy of Thailand wih oher counries, we generae he horizon dashed line meaning he same level of GDP growh and he verical dashed line meaning he same level of Gini coefficiens as Thailand. We found ha here are some counries, e.g. Cambodia, Lao, Indonesia, and Turkey, have higher level of economic growh when hey have he same level of income inequaliy as Thailand. On he oher hand, some counries e.g. Japan, Liberia, Neherland, and Belgium, have lower income inequaliy when hey have he same level of economic growh as Thailand. This resuls indicae ha Thailand seems no o be a a good posiion of boh inequaliy and economic growh when compare wih he oher counries. 0

11 7. Exploring he Relaionship beween Growh and Inequaliy in Thailand In his subsecion, we will presen he resuls sequenially. The firs par is he es for a kink effec on each variable. The second par is abou selecing he bes-fi Copula or he join disribuion o be a linkage beween economic growh and income inequaliy models. Then, he esimaes of he growh and income inequaliy models will be shown in he las par of his subsecion. 7.. Tesing for a Kink (Threshold) Effec The imporan quesion is wheher or no he kink (or hreshold) exiss wih respec o our model. Even hough we are given a clue by he exper and lieraures ha some covariaes have boh negaive and posiive effecs on economic growh and even on income inequaliy (See Secion 5), we sill need o do an experimenal es o verify his hypohesis. Recall, he kink effec means (or refers o) he unequal effec of he covariae on he dependen variable. I can be viewed as a regression disconinuiy model where he regression funcion iself is coninuous bu he slope is disconinuous a a hreshold poin, hereby a kink. Thus, we expec o obain he kink effec of GDP growh on he disribuion of income or he so-called Kuznes hypohesis. For his purpose, we conduc he Likelihood raio es (LR-es) o es he kink effec following a recommended algorihm of Hansen (000), he originaor of he kink regression. Tha is, he kink effec on a relaionship beween dependen variable, y, and is covariae, x i, is examined as a single equaion. This algorihm is kep using for each pair of he covariae and he dependen variable. We assume he null hypohesis is linear regression and he alernaive is kink regression. The resuls are shown in Table 4. Table 4. Resuls of Likelihood Raio Tes Covariae Growh Equaion GDP In( M ) In( HC ) In( Ins ) Gini Deb GDP Inequaliy Equaion GDP In( Sub ) In( PS ) In( Inf ) Edu Unemp Gini * 5.8*** Source: Calculaion. Noe: * and *** denoe rejecions of he null hypohesis a he 0% and % significance levels, respecively. Assuming he null hypohesis is rue, Table 4 indicaes ha only he covariaes PS and In( Inf ) rejec he null hypohesis of lineariy a he 0% and % levels, In( ) respecively, and hence heir oucomes are said o be in favor of he kink regression for he inequaliy equaion. On he oher hand, we found ha oher covariaes canno rejec he null hypohesis of lineariy, herefore, hey are said o be in favor of he linear regression for he growh equaion. As he resul, we failed o find he significan kink effec of GDP growh on he income inequaliy, meaning ha he Kuznes hypohesis does no hold for Thailand s economy hrough his daa se. However, here sill is a kink effec on income inequaliy due o he changes in he share of privae secor ( In( PS ) ) and inflaion ( In( Inf ) ). Therefore, our simulaneous kink equaion (SKE) model wih specified kink poins can ake he form as:

12 GDP GDP ln( M ) ln( HC ) ln( Ins ) Gini Deb ( F(r,s )) F(r,s ) Gini GDP ln( sub ) ln( PS ) ln( PS ) j 4( F(r 4,s 4)) ln( Inf ) 4 F(r 4,s 4) ln( Inf ) 5Edu 6Unemp (7). (8) The model is se similarly o he equaions (7) and (8) in which growh in GDP and he Gini coefficien are he dependen variables influenced by a variey of independen variables. The parameers and can be explained by he same manner as he convenional i j regression, bu he erms 3, 3, 4 and represen he coefficiens of negaive and posiive 4 r and r 4 as he kink poins which pars in he kink regression. Here, we rea he parameers 3 have o be esimaed. This model is said o be he SKE model because hese wo equaions are simulaneously esimaed where he dependen variable of one equaion also plays a role as he independen variable in he oher equaion, bu he slope wih respec o he share of privae secor ( In( ) In( PS ) r3 PS ) and inflaion ( In( Inf ) ) are disconinuous since hey have he kink a and In( Inf ) r4, respecively. 7.. Selecing Copulas for he SKE Model As a sequenial esimaion mehod, he Copula has o be seleced before esimaion. Therefore, his par is abou selecing he bes-fi Copula or he join disribuion o be a linkage beween economic growh and income inequaliy equaions. Given a se of Copula families i.e. Gaussian, Suden s, Clayon, Gumbel, Frank, and Joe, as described in Secion 3., we hen selec he Copula using a classical way of selecing model, he Akaike Informaion Crierion (AIC). In addiion, we also employ anoher measure, he Bayesian Informaion Crierion (BIC) o srenghen he resul. We assume four differen disribuions for he marginals, namely Normal, Suden s, skewed Normal, and skewed Suden s disribuions. Therefore, we have go sixeen cases for a pair of marginal disribuions of he growh and inequaliy equaions, respecively. The resuls are illusraed in Table 5. According o he resuls shown in Table 5, he minimum values of AIC and BIC are and , respecively, from which Frank Copula is chosen o be a linkage beween skewed Suden- of he growh equaion and skewed normal margin of he inequaliy equaion.

13 Table 5. AIC and BIC Crieria for Model Choice Marginal Disribuion Family of Copula Gaussian Suden- Clayon Gumbel Frank Joe [] Normal/Normal [] Normal/Suden [3] Normal/Skewed Normal [4] Normal/Skewed Suden [5] Skewed Normal/Normal [6] Skewed Normal/ Suden [7] Skewed Normal/ Skewed Normal [8] Skewed Normal/ Skewed Suden [9] Suden-/Normal [0] Suden-/Suden [] Suden-/Skewed Normal [] Suden-/Skewed Suden [3] Skewed Suden-/ Normal [4] Skewed Suden-/ Suden [5] Skewed Suden-/ Skewed Normal [6] Skewed Suden / Skewed Suden- Source: Calculaion Esimaes of he Growh and Income Inequaliy models To explore he relaionship beween growh and income inequaliy in Thailand, equaions () and () are esimaed by MLE and he resuls are presened in Table 6. In general, he model can perform well hrough his daa se and i can capure a nonlinear effec of some variables. Mos of he parameers are righly signed and saisically significan a he convenional levels. The esimaes of he growh equaion The esimaes of he growh equaion, shown a he beginning of Table 6, reveal ha only some variables suggesed by he empirical growh lieraures are significan in he case of Thailand. I is found ha Thailand s economic growh depends significanly on he insiuion, Gini coefficien, and he governmen deb, bu he impac of oher variables on he economic growh are no significanly found. 3

14 Table 6 illusraes a negaive relaionship beween GDP growh and he insiuion index in which he coefficien is significanly equal o Typically, he insiuion is considered as a cause of he long-run growh (Acemoglu, Johnson, and Robinson, 005). I is commonly relaed o such social mechanisms as democracy, religion, propery righs, and rule of law and i also deermines he incenives and creaes consrains in he economy. The differen ype of economic (and poliical) insiuion can bring abou he differen level of oucome of he economy. For example, as he impac of very differen economic insiuions, Souh Korea has much higher economic growh rae and beer sandard of living han Norh Korea. The negaive correlaion beween GDP growh and he insiuion index happened in Thailand may imply ha he curren economic (and poliical) insiuion can be desrucive o he growh. Addiionally, Table 6 also shows he posiive effec of he governmen deb on he GDP growh in which he coefficien is equal o , meaning ha an addiional % of he governmen deb leads o 0.064% increase in economic growh. This resul provides numerical evidence consisen wih wha some economiss have found in he lieraures, for example he works of Ghazanchyan, Sosky, and Zhang (05) and Bivens and Irons (00). We know ha he governmen spending can help induce he economy as well as he governmen deb, bu he deb can become desrucive o he growh if i is excessive. Bivens and Irons (00) found from heir experimen ha 90% of deb o GDP will lead o a negaive growh effec and slower economic growh, bu before ha cerain level he deb can be good for he economy. However, Thailand s daa provides jus one way of his consequence; i is found ha Thailand governmen deb creaes only he posiive effec on he economic growh. Imporanly, we also found a posiive relaionship beween growh and he Gini coefficien. The coefficien is equal o 0.05, meaning ha an increase in he income inequaliy by % leads o an increasing in GDP growh in Thailand over 0.0%. This resul implies ha during he sudied period (he years ), he income inequaliy measured by Gini coefficien- can creae a posiive impac on Thailand s economy. The possible reason for his siuaion is ha inequaliy may provide he incenives for people o inves, o make money, and o save for heir fuure livelihood and also for progress ha drive he capialis sysem. Inequaliy can creae he incenives for innovaion and produciviy ha provide he broader economic and social benefis of capialism. In fac, inequaliy may no insanly propel he economy bu wihou inequaliy economy is like socialism where he growh of economy hardly exiss. The esimaes of he income inequaliy As also repored in Table 6, income inequaliy in Thailand depends significanly on GDP growh, share of privae secor, inflaion, and he expendiure on educaion. We found ha boh share of privae secor and inflaion have nonlinear impacs on income inequaliy. These happen due o he kink effec and we will discuss compleely abou hese effecs laer. As we failed o find he kink effec of he GDP growh on income inequaliy across his daa se (See Secion 7.. Tesing for a Kink Effec), we may conclude ha he Kuznes hypohesis does no hold for Thailand s economy. However, we sill found a linear relaionship among hese wo variables, he GDP growh is found o be posiively correlaed wih income inequaliy. The coefficien of he GDP growh is 0.039, meaning ha an addiional % of he growh rae of GDP leads o 0.04% increase in Gini coefficien. This esimae resul confirms he previous works and also he empirical lieraures (as repored in Secion 5), ha is a rise in GDP growh can bring abou he income inequaliy. 4

15 Table 6. Esimaion resuls of Copula based SKE model Variables Esimaed Value S.E. ( 0 ) Growh Inercep erm *.6 GDP growh lagged by one period Moneary base Human Capial Accumulaion Insiuion *** 0. Gini Coefficien 0.05** 0.95 Governmen deb * 3.97 Sigma ( ) **.97 Degree of freedom Suden-.4739*** Degree of freedom Skewed Suden *** 9.7 Income Inequaliy Inercep erm GDP growh 0.039* 0.63 Governmen subsidy and ransfer paymen Share of Privae Secor (posiive par) 0.005*** 0.0 Share of Privae Secor (negaive par) *** 0. Kink parameer of Share of Privae Secor 4.56*** Smooh parameer of Share of Privae ***.46 Secor Inflaion (firs posiive par) 0.05* Inflaion (second posiive par) 0.045* 0.7 Kink parameer of Inflaion.809*** 4.39 Smooh parameer of Inflaion 0.758** 5.4 Expendiure on Educaion *.4 Unemploymen Sigma ( ) 0.08*.06 Degree of freedom Skewed Suden-.4833** 8.64 Join Esimae Frank Copula Dependency ( ).9833*** 7.36 Source: Calculaion. Noe: *, **, and *** denoe rejecions of he null hypohesis a he 0%, 5%, and % significance levels, respecively. Addiionally, Table 6 shows ha he expendiure on educaion is also imporan for reducing uneven income. The coefficien is equal o , meaning ha an addiional % of he expendiure on educaion leads o 0.04% decrease in Gini coefficien of Thailand. Our resul provides numerical evidence consisen wih wha IMF has poined ou hrough he discussion paper abou inequaliy and growh eniled Causes and Consequences of Income Inequaliy: A Global Perspecive (Dabla-Norris e al., 05). The expendiure on educaion can play a pivoal role in reducing income inequaliy because educaion deermines occupaional choice, access o jobs, and he level of pay. However, we failed o find significance for oher imporan variables across his daa se, i.e. he governmen subsidy and ransfer paymen and he growh rae of unemployed worker. The Kink effecs of he share of privae secor and inflaion on income inequaliy A specific capabiliy of our mehod allows us o preserve he nonlinear impacs of he share of privae secor and inflaion on income inequaliy. The impacs are spli ino wo 5

16 groups based on kink poins. In Figure (a) we display he relaionship beween Gini coefficien and he share of privae secor (% of GDP) hrough a regression line (red line), on which he horizonal axis is he share of privae secor and he verical axis is income inequaliy measured by he Gini coefficien. We can see ha he regression shows a small posiive slope for low share of privae secor wih a significan kink around 4.56 (4.5%), displayed as he blue do, swiching o a negaive slope for he share of privae secor beyond ha kink value. This means share of privae secor displays a posiive impac on he Gini coefficien when he share is below 4.5%. The coefficien of his posiive par is 0.005, meaning ha an addiional % of share of privae secor leads o 0.005% increase in Gini coefficien. To explain his posiive correlaion, we refer o he lieraure on he privae secor and he disribuion of income. The lieraure shows ha privaizaion can cause wealh concenraion which finally leads o income inequaliy (Anneli Kaasa, 003). To make i clearer, if we compare he wage gap of he privae secor wih public secor, we will find ha earnings inequaliy in he public secor is ypically smaller han in he privae secor. The public secor adminisraively ses wages, hereby he low income inequaliy wihin he secor, bu in he privae secor whose income usually depends on, for example, invesmen, producion, and oupu. Therefore, he disribuion of earnings wihin he privae secor is usually unequal. This process is likely o increase income inequaliy due o he wage differenial beween public and privae secors and he unequal disribuion of income wihin he privae secor. So ha is why he share of privae secor should deserve a special aenion from he policy maker. Apar from he previous knowledge of he privae secor, our sudy argues ha he growing size of he privae secor beyond he kink value can negaively affec income inequaliy. As shown in he same Figure (a), he regression line is swiching from he posiive slope o he negaive slope for he share of privae secor beyond he kink value around 4.5%, displayed as he blue do. The coefficien of he share of privae secor in he negaive rend is , meaning ha an increase in he share of privae secor by % can cause a reducion in Gini coefficien by 0.003%. According o his evidence, he share of privae secor may no longer be desrucive he disribuion of income. Insead, he proper level of he share of privae secor can help reduce uneven income in Thailand. Share of privae secor and Gini smooh kink plo Inflaion and Gini smooh kink plo fied line fied line Gini kink poin (a) Share of privae secor (b) Inflaion Fig.. Plo of Gini coefficien and is covariaes and he rend lines esimaed by Copula based Kink regression 6

17 In Figure (b) we illusrae he relaionship beween Gini coefficien and inflaion hrough anoher regression line. We can see ha he regression shows a seeply posiive slope for low inflaion wih a significan kink around.809 (.8%), displayed as he blue do, swiching o a smaller posiive slope for he inflaion beyond ha kink value. The resul observes he posiive relaionship beween inflaion and income inequaliy, in which he higher inflaion he more income inequaliy. However, our work can do more han ha; we find a salien resul ha he impacs of inflaion on inequaliy are spli ino wo levels based on he kink poin. In he firs sage, a any level of inflaion below.8%, an addiional % of inflaion leads o 0.% (0.05) increase in Gini coefficien. Bu in he second sage, he effec is much less han he firs sage. We found ha an addiional % of inflaion jus leads o 0.05% (0.045) increase in Gini coefficien. To explain his difference, we refer o he discovery of Chris Crowe, CEP researcher and IMF s officer who firs menions abou posiive relaionship beween inflaion and inequaliy. Inflaion here is a consequence of raising governmen revenue by prining money. Tha is a cenral bank issues new currency in order o le he governmen collec more ax raher han collecing ax paid ou of he exising money sock, bu his even can cause higher inflaion rae. No all people suffer from he same inflaion ax rae because he ax his only a group of people holding more nominal asses such as cash as a fracion of oal income, who are ypically poor and middle class. As a resul, inflaion below ha kink value creaes a big gap beween he wealhy and poor people, and hen leads o he greaer income inequaliy. We can see ha income inequaliy keeps rising as long as inflaion increases, bu beyond ha kink value he effec of inflaion seems o be less han before, bu sill exiss. Chris Crowe suggess ha income inequaliy in his second sage is influenced by democraic insiuions, and happens among he wealhy people. The greaer inequaliy in his sage means greaer inequaliy in poliical paricipaion in which he poliical voices depend on income. The wealhies people have high power in adopion of policy, such as inflaion, more favorable o heir own business, and his leads o higher income inequaliy among hose wealhy people. Bu, from our opinions, we hink in he oher way round. I is rue ha he greaer inflaion may cause he higher income inequaliy in boh sages; bu wha we sugges here is o use he proper level of inflaion o reduce income inequaliy in he case of Thailand. Our experimen found ha in he firs sage, under he kink value.8%, he elasiciy of he Gini coefficien wih respec o he inflaion is low; herefore a small decrease in he inflaion rae resuls in a large decrease in he Gini coefficien. Unlike he second sage, he elasiciy is much higher, meaning ha even he cenral bank provides a big decrease in he inflaion rae; we can observe jus a minor decrease in he Gini coefficien. Hence, i is reasonable o preserve he inflaion rae below ha kink value. 7.3 Are here he Kuznes Curves in he regions of Thailand? In his secion, we exend he same algorihm of invesigaing he Kuznes curve o he regions o es wheher he Kuznes Hypohesis holds in some regions of Thailand. We consider five regions (i=5) consising of Middle, Norh, Norheas, and Souh regions, and Bangkok since each region s characerisic is dissimilar, and hence he relaionship beween GDP growh and income inequaliy may be differen across regions. We conduc he Likelihood raio es (LR-es) o es he kink effec on a relaionship beween he regional Gini coefficien (dependen variable) and he regional GDP growh (covariae). This algorihm is kep using for each pair separaely. Again, we assume he null hypohesis is linear regression and he alernaive is kink regression. The resuls are shown in Table 7. 7

18 Table 7. Resuls of Likelihood Raio (LR) es of each region Covariae GDP i Bangkok Gini B Middle Gini M 8 Dependen variable Norh Gini N Norheas Gini NE Souh Gini.46*** 0.79*** 3.3*** Source: Calculaion. Noe: *** denoes rejecion of he null hypohesis a he % significance levels. Table 7 indicaes ha he Middle region, he Norh regions, and Bangkok can rejec he null hypohesis of lineariy a he % levels, meaning ha he relaionship beween GDP growh and Gini coefficien of hese hree regions have he kink effec. The esimaed resuls are shown in Table 8. I is found ha he regional Gini coefficiens depend posiively on he regional GDP growh in which he Middle region, he Norh regions, and Bangkok have he nonlinear impacs of GDP growh on income inequaliy due o he kink effec. As we found he exisence of he kink effec, we expec o obain he Kuznes curve in hese hree regions. However, he ruh ha we canno find he Kuznes curve in Thailand s economy is also rue for he regions. As we can see, he impacs are spli ino wo levels based on he kink poin, whereas all of hem are posiive. The Norheas region has a posiive linear relaionship beween GDP growh and Gini coefficien. Is coefficien is equal o.3, meaning ha an addiional % of GDP growh leads o.3% increase in Gini coefficien of he Norheas region. Noe ha he parameer of he Souh regions is no saisically significan. Table 8. Esimae resuls for each region Covariae GDP i Bangkok Gini B Middle Gini M Dependen variable Norh Gini N Norheas Gini NE S Souh Gini Inercep erm 0.337*** 0.334*** 0.356*** 0.35*** 0.364***.3*** 0.47 GDP i GDP i (firs sage) 0.00* 0.49*** 0.09* GDP i (second sage) 0.03***.55** 0.0** Kink parameer of GDP (%) 0.73*** 0.805** 0.945** Smooh parameer of GDP.05* 0.755**.554** Source: Calculaion. Noe: *, **, and *** denoe rejecions of he null hypohesis a he 0%, 5%, and % significance levels, respecively. As shown in Table 8 and Figure 3, he impacs of he growh on he inequaliy are spli ino wo levels based on he kink poin. For insance, he resul of Bangkok shows ha a any level of GDP growh below 0.73% per quarer, an addiional % of he growh leads o 0.00% increase in Bangkok s Gini coefficien. Bu in he second sage, he effec is much larger han he firs sage. We found ha an addiional % of he growh can lead o 0.03% increase in he Gini coefficien. The resuls are he same for he Middle and Norh regions. These resuls argue he convenional heory of Kuznes he invered U-shaped curve of income concenraion- ha he decreases do no necessarily happen, and in fac, inequaliies S

19 have grown sharply in hese hree regions afer reaching hose kink poins. Our finding is raher linked wih he work of Thomas Pikey, a well-known French economis. Pikey's work refers o he capial-oupu raio and explains ha he rae of capial reurn (r) in developed counries is persisenly greaer han he rae of economic growh (g), and his will cause income (wealh) inequaliy o increase in he fuure. Moivaed by his reasoning, we hink ha he posiive nonlinear relaionship beween he growh and inequaliy of he hree regions can be explained hrough he same idea of Pikey. As hese hree regions play a key role in Thailand s economy especially Bangkok. Bangkok and he surrounding areas is he economic cenre of our counry, i is he hear of he counry's invesmen and developmen as well as Chiang Mai which is in he Norh. I is viewed as he second-larges province of Thailand ha conribues a lo o Thailand s economy. This causes he rae of capial reurns of hese hree regions o be high and greaer han he growh rae of GDP, and hence i leads o an increasing in he income inequaliy. Bangkok Cener Bangkok Gini coefficien fied line Kink Poin Cener Gini coefficien fied line Kink Poin GDP growh (%) Norh GDP growh (%) Norh Gini coefficien fied line Kink Poin - 0 GDP growh (%) Fig. 3. Plo of Gini coefficien agains GDP growh and he rend lines esimaed by Copula based Kink regression 9

20 7.4 Policy effec on he economic growh-gini coefficien In his secion, we aemp o examine wheher Thailand s economic and social developmen plans, covering he plans 7-, can accomplish boosing he economy and reducing income inequaliy a he same ime. We focus on he overall counry and he five regions of Thailand, consising of Middle, Norh, Souh, Norheas, and Bangkok. Firs of all, le consider he overview impac of economic growh on he Gini coefficiens of Thailand as shown in Table 9. The esimaed regression coefficiens indicae ha GDP growh creaes he posiive effec on he Gini coefficien, whereas many parameers are no saisically significan. The finding also illusraes ha he income inequaliy ends o increase along he Sevenh Plan o he Elevenh Plan. This resuls show he evidence of a relevan impac of GDP growh rae o income inequaliy during he Eighh Plan and he Elevenh Plan. Table 9. Economic plan summary Economic Plan Inercep Slope Plan 7 (99-996) *** (0.003) (0.09) Plan 8 (997-00) 0.48*** 0.006* (0.006) (0.070) Plan 9 (00-006) 0.47*** 0.03 (0.003) (0.89) Plan 0 (007-0) 0.396*** (0.00) (0.0869) Plan (0-05) * (0.0054) (0.54) Source: Calculaion. Noe: *, and ***, denoe rejecions of he null hypohesis a he 0 and % significance levels, respecively. The numbers in blankes are he sandard errors. Furhermore, i would also be ineresing o invesigae in deail he relaionship beween inequaliy and GDP growh in five differen regions or areas of Thailand. A imevarying regression model is adoped o esimae ime-varying effecs of economic growh on income inequaliy. By using his model, i allows he effecs o evolve as he level of income inequaliy changes along our sample sudy. Noe ha he sudy uses Gini coefficien of each region as a proxy for income inequaliy, and he GDP growh of each region as a proxy for economic growh. The resul of ime varying coefficien of he GDP growh effec on Gini for each region is illusraed in Figure 4. In mos of he regions, he effecs of GDP growh are eiher posiive for he whole sample period or urn from negaive o posiive, excep for Souh region. In his figure, we can observe ha Norh, Norheas, and Middle regions presen a similar movemen of he ime varying coefficiens and mosly ake place above he zero line. This indicaes ha here is a posiive effec of economic growh on Gini coefficiens during he implemenaion of he Eighh Plan o he Elevenh Plan. In he case of BKK and Souh regions, we observe ha here is a negaive effec of economic growh on inequaliy during he Eighh Plan before moving o posiive effec when approaching he Ninh Plan for BKK and he Tenh Plan for Souh. We find an ineresing main issue ha is an economic growh can reduce he inequaliy during he Eighh Plan in some regions. We found ha he Eighh Plan has adjused he developmen concep, shifing from growh orienaion o peoplecenered developmen. In he pas plans, he naional and social developmen concep has focused on he acceleraion of economic growh uilizing comparaive advanage in erms of naural resources and low-cos labor o produce goods and services for expor. However, he success in economic growh migh no mean ha he Thai people are enjoying a greaer 0

21 wealh and qualiy of life. Thus, he Eighh plan aimed o improve he well-being raher han he economic growh. And his migh be he reason ha economic growh could bring a lower income inequaliy during his plan. In conrary o he Eighh Plan, he effec of economic growh on inequaliy is posiive along he Ninh o he Elevenh Plan. How could i have happened? To answer his quesion, we look back o a he main poins of hose plans and found ha hese plans adoped he Sufficiency Economy philosophy o guide he developmen and adminisraion of he counry, a he same ime as coninuing he holisic approach of people-cenered developmen from he Eighh Plan. Moreover, hese plans also buil an economy wih srong inernal foundaions and flexible o exernal changes and aiming for balanced developmen wih respec o people, sociey, economy, and environmen in order o achieve susainable developmen and he wellbeing of he Thai people. The main idea of hese hree plans seem o be similar o he Eigh plan bu he ime varying coefficiens resuls provide a posiive effec of growh on inequaliy. We expec ha Thailand had encounered more complicaed domesic and global changes, and flucuaions ha would presen a hrea o naional developmen. Thailand was in ransiion from a rural o an urban sociey. As a resul of he comparmenalized developmen of urban and rural areas, here is imbalance in he developmen of rural communiies. These, in urn, had weakened he economy, he sociey and democracy. The NESDB (0) repored ha he disribuions of income by region and he reurns o he facors of producion have been skewed. Reurns o labor have declined coninuously, from 30. percen during he Eighh Plan o 8.9 percen during he firs hree years of he Tenh Plan, while reurns o capial, enrepreneurship and land have increased. These evidences indicae he lack of economic and social opporuniy, access o resources, bargaining power, and fairness wih respec o righs among groups of people ha affecs boh income and accumulaion of wealh. In addiion, major obsacles have remained, including poliical unres, environmenal and ecological degradaion, low educaional qualiy, and severe drug problems. Q3/ Q4/03 Fig. 4. Time-varying effecs (coefficiens) of GDP growh on income inequaliy of each region 3 3 Due o he esimaion mehod of ime varying regression model need he iniial daa o compue a ime varying coefficien herefore, he ime varying coefficiens in Sevenh plan canno be obained.