Crowding out Effect of Foreign Aid in Selected Developing Countries: Panel Data Evidence

Transcription

1 Iran. Econ. Rev. Vol.17, No., 13. Crowdng out Effect of Foregn Ad n Selected Developng Countres: Panel Data Evdence Sadegh Bakhtar Hojjat Izadkhast Seyed Komal Tayeb Receved: 1/8/11 Accepted: 13/4/16 Abstract ublc sector decson-makers are faced wth the task of allocatng Presources among dfferent alternatve subject due budgetary constrants. In ths paper Offcal Development Assstance (ODA) data have been consdered as foregn ad. ODA s channeled through the publc sector of recpent countres and, hence, the ultmate effect of ODA on savngs or economc growth depends on how governments respond to t. Ths paper tres to explore the mpact of offcal development assstance on publc sector behavor n selected developng countres and contrbute to the fscal response lterature on two man grounds. Frst, t specfes a fscal response model. Second, usng panel data model for a sample of 5 ad recpent countres n Asa and Latn Amerca over the perod Emprcal results ndcate that offcal development assstance has a postve and sgnfcant effect on government nvestment expendture, but t has not sgnfcantly mpact on government current expendtures. Results also show that ODA crowds out both government revenue and publc borrowng. Keywords: Publc Fscal Behavor, Offcal Development Assstance, Panel Data. 1- Introducton Based on the department of Economc and Socal Affars of UN for many low ncome countres, Offcal Development Assstance remans an mportant vehcle for fnancng development gven low levels of domestc Department of Economcs, Islamc Azad Unversty, Khorasgan Branch, Isfahan, Iran Canddate of Economcs, Unversty of Isfahan, Isfahan, Iran (Correspondng Author). Department of Economcs, Unversty of Isfahan, Isfahan, Iran.

2 5/ Crowdng out Effect of Foregn Ad n Selected Developng Countres savngs and lmted access to prvate captal flows. The amount of ODA based on UN data reachng $133 bllon n 11 (UN, 1 P.V). The government plays a consderable role n the plannng and mplementng of projects n developng countres, evdently, a large proporton of offcal development assstance s channeled through the publc sector. The fnal effects of offcal development assstance on economc growth of the country wll depend on how t nfluences fscal polcy (Machado, 9). Net offcal development assstance to developng countres has averaged between3.7 and 6.7 percent of GDP durng 198 9, amountng to around 4 percent of average tax revenues. The relatvely hgh share of ad n government budgets n some countres has rased concerns about the detrmental effects of ad dependency on domestc revenue effort, spendng programs and budget plannng as well as nsttuton-buldng (Gupta, Clements &Tareq, 8). Poltcal economy consderatons provde addtonal support to the argument that offcal development assstance may dscourage taxaton by recpent governments (Benedek, Crvell, Gupta, & Muthoora, 1). A key argument of the ad dependency lterature, for example, s that offcal development assstance lowers tax revenue, because t undermnes the development of domestc nsttutons that support tax admnstraton and good governance (Knack, ;Heller and Gupta, ; Brautgam and Knack, 4; and Mosley, Hudson &Horrell8). The present paper develops a fscal response model based on the recent work n the fscal response lterature and provdes soluton that can be estmated usng more advance econometrc technques. A model for a sample of 5 countres n Asan and Latn Amercan countres over the perod have been estmated. The paper s organzed as follows. The next Secton revew the relevant lterature, Secton III outlnes the specfcaton of a fscal response model and derve ts soluton to analyze the fscal effects of offcal development assstance. Secton IV deals wth data ssues. The man results wll be presented n Secton V and concludng remarks wll be gven n Secton VI. 1- These countres wll be mentoned n Secton IV.

3 Iran. Econ. Rev. Vol.17, No., 13. /53 - Revew of lterature The fscal response lterature poneered by Heller (1975).Fscal behavor means decdng between varous sources or revenue and level of expendture. Specfcally, t looks at how government behaves vs-à-vs of offcal development assstance (ODA). Crtcs argue that ODA have resulted n ncreased publc or prvate consumpton rather than n ncreased nvestment, and have contrbuted less to growth than was antcpated. A data plot suggests a negatve assocaton between total net ODA and total tax revenues (Fgure1). Between 198 and 1995, when offcal development assstance as a share of GDP was ncreasng, average tax revenue n relaton to GDP decreased slghtly. Post-1995, a declne n the share of total net ODA to GDP was accompaned by hgher tax revenues as percentage of GDP. Fgure1: Average Net ODA and Tax Revenue n Low- and Mddle- Income Countres, Source: Benedek, Crvell, Gupta, & Muthoora, (1). Evdence of a negatve relatonshp between net offcal development assstance and tax revenue has been borne out n the data n several studes. Ghura (1998), usng data for a sample of 39 sub-sahran Afrcan countres for the perod , fnds that ad has a statstcally sgnfcant negatve effect on the rato of taxes to GDP. Gupta et al (4) examned the revenue response to net offcal development assstance and found that total net ODA has a negatve

4 54/ Crowdng out Effect of Foregn Ad n Selected Developng Countres assocaton wth government revenues. They also concluded that the effect depends on the composton of ad: whle grants have a negatve effect on revenues, loans have no sgnfcant or postve effect. They argued that ther results may reflect the fact that loans have to be repad eventually, whle grants represent real transfers from abroad. Moreover, they found that although the effect of grants s modest on average, t s hgher n countres wth weak nsttutons. Remmer (4) uses a broader sample of 1 developng countres over the perod She fnds a negatve relatonshp between three dfferent measures of net offcal development assstance and changes n tax revenue to GDP. As noted above, Gupta et al. (4) fnd smlar results. Moreover, they show that the negatve effect of ODA grants s stronger n countres wth weak nsttutons. Knack (9) also fnds a robust negatve relatonshp between soveregn rents(from ad and natural resources) and taxaton. A number of recent contrbutons have however argued that the results are senstve to the composton and scope of the estmaton sample. Clst and Morrssey (11), extendng the tme perod of the Gupta et al. (4) database, and re-estmatng the relatonshp between net offcal development assstance and taxaton from the md-198s to 5 fnd a postve assocaton between net offcal development assstance and taxaton. They argue that ths reflects a break n pattern of allocaton of net offcal development assstance snce the md-198s, when net offcal development assstance became condtonal on mplementaton of good polces or concomtant structural reforms. Ths argument s central to Carter s (11) crtque of exstng studes on net offcal development assstance and taxaton. He argues that the econometrc evdence to date s nconclusve because of falures to adequately address the endogenous nature of net offcal development assstance. Studes such as Mosley et al. (1987), Gang and Khan (1991), Khan and Hoshno (199), Otm (1996), Gupta (1997), Franco- Rodrguez etal. (1998), Gang and Khan (1999, McGllvray (), Franco-Rodrguez (), Mavrotas (),MC-Gllvroy & Ouattara (5), Ouattara (6),Mc Gllvray(9),Machado (1),Clst&Morrssey (11) and Benedek et al (1), usng fscal response modelstry to answer questons such as does net offcal development assstance complement or substtute

5 Iran. Econ. Rev. Vol.17, No., 13. /55 for government revenue? Does t manly fnance government consumpton or publc nvestment? Does t crowds-out publc borrowng? 3- A Fscal Response Model Fscal response studes assume that publc sector decson-makers are faced wth the task of allocatng resources among dfferent alternatve subject to budgetary constrants. These decson-makers n the publc sector behave as a sngle ndvdual wth a well-behaved, homothetc preference map wth followng utlty functon followng Heller (1975): U F INEX, CUEX, GREV, ODA, BORR ) (1) ( Where INEX s government nvestment expendture, CUEX s government current expendture, GREV s government revenue, ODA s net offcal development assstance and BORR s net publc borrowng from other sources. The standard approach n the fscal response lterature s to defne the publc sector utlty functon as a quadratc loss functon and mnmze t as follows (Ouattara, 6): a1 a a3 U a ( INEX INEX ) ( CUEX CUEX ) () a4 ( GREV GREV ) ( BORR BORR ) The starred varables represent exogenous target and represent the weght attached to each component of the utlty functon. All other varables n equaton (), represent actual varables, and treated as endogenous. The ratonale for specfyng the utlty functon n ths manner s that publc decson-makers set annual targets (represented by the starred varables) for each of the decson varables and conscously strve to acheve

6 56/ Crowdng out Effect of Foregn Ad n Selected Developng Countres these targets subject to budgetary constrant 1. The utlty functon s specfed n a way that f the decson-makers try to derve from the targets t wll result n a loss of utlty. Followng ouattara (6), the decson-maker s assumed to maxmze utlty functon () subject to the followng budget constrant: INEX CUEX GREV ODA BORR (3) The above budget constrant assumes that expendture must equal total government's recepts. The lagrangan functon s appled to maxmzaton problem, as follows: L 1 ( INEX INEX 4 ( BORR BORR ) ) ( CUEX CUEX ) [ INEX COEX GREV ODA BORR] 3 ( GREV GREV ) (4) Where s the Lagrange multpler? Takng the dervatve of L wth respect to the choce varables and, and solvng the frst-order condton through leads to the followng sem-reduced form equatons: INEX 1INEX [ ODA BORR GREV CUEX ] (5) CUEX 3CUEX 4[ ODA BORR GREV INEX ] (6) GREV 5GREV 6[ ODA BORR IENX COEX ] (7) BORR 7BORR 8[ ODA GREV INEX CUEX ] (8) 1- Recent papers such as Franco-Rodrguez et al. (1998), McGllvray () have endogensed ad on the bass that t s a government choce varable. However, gven that the target for ad s generally set as ad commtments ths mples that the mpact of ad obtaned n the reduced form equatons wll be that of the commtment values and not the dsbursement values. Ths wll tend to overestmate the mpact of ad, as the amounts commtted are generally hgher than those dsbursed. -Bhn and McGllvray (1993) provde detal dscusson about the specfcaton of the utlty functon.

7 Iran. Econ. Rev. Vol.17, No., 13. /57 Where the j ( j= 1,,., 8), are combnatons of ( =1, 4) 1 One problem faced wth most fscal response studes has been to obtan the target varables. It s assumed that the target for publc captal expendture can be peroxde by the followng: INEX 1GDPC DESE 3 ODA (9) Where GDP c s GDP per capta, DESE s debt servce and ODA s net offcal development assstance. GDP measures the country's level of economc development. Publc captal expendture wll also depend on the debt burden of the country measured here by debt servcng. Target for government consumpton expendture s also defned by a smlar economc relatonshp: CUEX 1GDP DESE c 3 ODA (1) The target for government revenue s approxmated by: GREV 1GDPc EXPR 3 IMPR (11) GDP C s defned as above, EXPR and IMPR denotes exports and mports respectvely. The ncluson of EXPR and IMPR s based on the fact that many developng countres get revenue from exports and mports. Fnally, the target for borrowng s specfed by the followng relatonshp: BORR 1GDP c ODA (1) Borrowng wll depend on GDP c and ODA. Substtutng Esq. (9)-(1) nto Esq. (5)-(8) gves the followng full reduced form equatons: 1- These coeffcents can be obtaned from the authors upon request.

8 58/ Crowdng out Effect of Foregn Ad n Selected Developng Countres INEX t 1 GDP ct DESE t 3 ODA t 4 EXPR t 5 IMPR t 1 e 1t (13) CUEX t 1 GDP c t DESE t 3 ODA t 4 EXPR t 5 IMPR t e t (14) GREV t 1 GDP ct DESE t 3 ODA t 4 EXPR t 5 IMPR t 3 e 3t (15) BORR t 1 GDP ct DESE t 3 ODA t 4 EXPR t 5 IMPR t 4 e 4 t (16) s, s, s and s can be traced back to the s representng the wegh attached to each element of utlty functon ().In ths regard panel data technques provde us wth more powerful tools to test these relatonshps among economc varables. In ths paper, the above equatons wll be estmated usng panel data econometrcs technque. Gven that equatons derved from the soluton of the model. For ths purpose each of Eqs. (13)- (16) can be represented by the followng general reduced-form regresson: Y X e (17) t t t Where the Y t s dependent varable and X t denotes a set of explanatory varable for country n tme t. These two types of estmators are desgned to handle the systematc tendency of e t to be hgher for some ndvduals than for others (ndvdual effects) and possbly hgher for some tme perods than for others (tme effects). The fxed effect estmator does ths by (n effect) usng a separate ntercept for each ndvdual or tme perod. The random effect estmator s based on the followng decomposton of e t where ε s the ndvdual effect, µ the tme effect, and η the purely random effect. β s estmated by the structure mposed upon e t by ths assumpton. Both fxed and random effects adjust for heteroskedastcty. However, the two effects have ther own advantages and dsadvantages. The random effects model treats the

9 Iran. Econ. Rev. Vol.17, No., 13. /59 ndvdual effect as part of the error term, t suffers from the possblty of bas due to a correlaton between t and the repressors. Hausman (1978) provdes a test for dscrmnatng between the fxed effects and the random effects estmators. The test s based on comparng the dfference between the two estmators of the coeffcent vectors, where the random effects estmator s effcent and consstent under the null hypothess and nconsstent under the alternatve hypothess. The fxed effects estmator s consstent under both the null and the alternatve hypothess. If the null s true then the dfference between the estmators should be close to zero. The calculaton of test statstcs requres the computaton of the covarance matrx of β 1 - β. In the lmt the covarance matrx smplfes to Var(β 1 ) Var(β ), where β 1 s the fxed effects estmator and β s the random effects 4- Sources of Data The data used here come from two man sources and cover the perod Countres n Southland West Asa are: Iran, Turkey, Pakstan, Jordan, Oman, Syra, Inda, and Yemen. In South and East Asa are: Indonesa, Malaysa, Thaland, Phlppne, Bangladesh, Nepal, Vetnam. In Latn Amerca are: Argentna, Brazl, Bolva, Venezuela, Colomba, Peru, Chle, Paraguay, Uruguay 1. Data on government nvestment expendture s obtaned from the World Bank Global Development Network Growth Database. Data on government current expendture, government revenue, foregn ad, debt servce, exports and mports of goods and servces (all expressed as a percentage of GDP), GDP C s GDP (market prce) per capta were obtaned from the World Bank World Development Indcator, 1. Data on borrowng s the net publc borrowng, obtaned as a resdual from Eq. (3) n the model. 5- Emprcal Results Tables 1-4, summarze the results related to the estmaton of Eqs.(13)- (16). Each equaton appled for the same perod (1991-1). The szes of the samples are 5 countres n Asan and Latn Amercan countres. Accordng to F-Lemer test, H hypothess (Poolng approach) rejected and all equatons estmated wth panel data approach. Accordng to Husman test 1- Because other countres n these regons are not reported varables such as debt servce, ODA, are not consdered.

10 6/ Crowdng out Effect of Foregn Ad n Selected Developng Countres n government nvestment expendture equaton, H hypo the ss accepted and ths equaton estmated based on random effects estmator. In other equatons, H hypothess rejected and these equatons estmated based on fxed effects estmator. The mpact of net offcal development assstance on the government nvestment expendture s assessed n Table 1. Table1: Estmated Results for Government Investment Expendture Equaton, Based on Random Effects Varable Coeffcent GDP C DESE.7(1.81) -.4 (-.11) ODA.17 (1.99) EXPR.154 (.61) IMPR.734 (.51) Constant 5.3 (4.17) F- Lemer 5.6 (.) + Husman-test 5.3(.39) + Observatons 5 Number of countres 5 Robust z-statstcs n parentheses. sgnfcant at 5%; sgnfcant at 1%& sgnfcant at 1%. + Indcate P-value for F and Husman test, respectvely. Source: Authors The results n Table 1 show that net offcal development assstance bears a postve and statstcally sgnfcant sgn n government nvestment expendture. Whch s consstent wth fndngs of Heller (1975), Khan and Hoshno ( 199), Franco- Rodrguez et al. (1998), Ouattara (6), Machado (1). The mpact of foregn ad on the government current expendture s assessed n Table. Table : Estmated Results for Government Current Expendture Equaton, Based on Fxed Effects Varable Coeffcent GDP C DESE -.1(-.8).11 (.3) ODA.47 (.8) EXPR -.41 (-.43) IMPR.69 (3.41) Constant 1.99 (.45) F- Lemer 57.5 (.) + Husman-test 18.5(.3) + Observatons 5 Number of countres 5 Robust t-statstcs n parentheses. sgnfcant at 5%; sgnfcant at 1%. + Indcate P- value for F and Husman test, respectvely. Source: Authors

11 Iran. Econ. Rev. Vol.17, No., 13. /61 The results n Table show that net offcal development assstance has nsgnfcant effect on government current expendtures, then contrary to the fndng of Boone (1996) and others, we fnd no evdence that ad flows are used to fnance government consumpton expendture. The mpact of foregn ad on the government revenue s examned n Table 3. Table 3: Estmated Results for Government Revenue Equaton, Based on Fxed Effects Varable Coeffcent GDPC -. (-1.75) DESE -.63 (-1.) ODA (-.5) EXPR.19 (.8) IMPR. (.71) Constant 19.3 (5.56) F- Lemer 57.5 (.)+ Husman-test 17.8(.3)+ Observatons 5 Number of countres 5 Robust t-statstcs n parentheses. sgnfcant at 5%; sgnfcant at 1%& sgnfcant at 1%.+Indcate P-value for F and Husman test, respectvely. Source: Authors The coeffcent of net offcal development assstance on revenue s negatve and statstcally sgnfcant at the conventonal level, thus suggestng that developng countres publc sector authorty reduce ther revenue collecton efforts when net offcal development assstance s made avalable to them. Ths fndng corroborates fndng by Heller (1975), Franco-Rodrguez et al. (1998), Mc- Gllvray, Ouattara (5).Fnally and Benedek et al (1). Table 4 whch summarze results related to the net publc borrowng n equaton (16). Table 4: Estmated Results for Publc Borrowng Equaton, Based on Fxed Effects Varable Coeffcent GDPC.4 (.88) DESE.65 (1.18) ODA -.56 (-6.61) EXPR -.43 (-1.74) IMPR.13 (4.51) Constant (-4.5) F- Lemer 3.94 (.)+ Husman-test.11(.1)+ Observatons 5 Number of countres 5 Robust t-statstcs n parentheses. sgnfcant at 5%; sgnfcant at 1%. +Indcate P- value for F and Husman test, respectvely. Source: Authors

12 6/ Crowdng out Effect of Foregn Ad n Selected Developng Countres The results shows that net offcal development assstance have a negatve and statstcally sgnfcant effect on publc sector borrowng. Overall ths results support the vew that government n developng countres are lkely to reduce other forms of borrowng Concludng Remarks Ths paper has nvestgated the relaton between net offcal development assstance and publc sector fscal behavor usng a panel data n Asan and Latn Amercan countres over the perod The results ndcate that net offcal development assstance has a postve and sgnfcant effect on nvestment expendture, whch n turn mght crowds n prvate nvestment. The evdence also shows that ODAloans are not used to ncrease current expendture. Results also show that net offcal development assstance crowds out government revenue, then publc sector authortes n developng countres reduce ther efforts to collect revenue when ODA s avalable to them. Fnally, net offcal development assstance does nduce a reducton n publc sector borrowng. Ths paper has also some research mplcatons. Frst, as argued by McGllvray and Morrssey (1) ODA effectveness studes need to frst consder the queston of government behavor vs-à-vs of ODA loans before studyng ther broad macroeconomc mpacts. The relatonshp between ODA and growth mght not be a straghtforward one, as many ad growth studes seem to assume. Therefore, understandng how these ODA loans affect fscal aggregates mght shed new lght on the ad debate. Second, t mght also be worth analyzng whether the fscal behavor of the recpent government depends on the nature or the type of ODA loans. Ths mplcaton of our fndngs s that developng country should attract more foregn captal by lberalzng ther fnanced markets. References 1- Bakhtar, S., Tayeb, S.K., &Izadkhast, H. Impact of Offcal Development Assstance on Government Fscal Behavor n The Selected 1 - The argument s based on the fact that ad s cheaper compared to borrowng.

13 Iran. Econ. Rev. Vol.17, No., 13. /63 Asan Developng Countres, Quarterly Journal of The Economc Research, 1 (), 1. - Benedek, D., Crvell, E., Gupta, S., and Muthoora, P. Foregn Ad and Revenue: Stll a Crowdng Out Effect?,IMF Workng Paper, WP/1/186, Bnh, T.N., McGllvray, M. Foregn ad tax and publc nvestment: a comment. Journal of Development Economcs, 41, 1993, Brautgam, D., and Knack, S. Foregn Ad, Insttutons, and Governance n Sub-Saharan Afrca, Economc Development and Cultural Change, 5(), 4, pp Clst.P., & Morrssey, O. Ad and Tax Revenue, Journal of Internatonal Development, 3 (), 11, Franco-Rodrguez, S. Recent advances n fscal response models wth an applcaton to Costa Rca. Journal of Internatonal Development, 1,, Franco-Rodrguez, S., Mcgllvray, &M., Morrssey, O. Ad and the publc sector n Pakstan: evdence wth endogenous ad. World Development, 6, 1998, Gang, I., Khan, H. Foregn ad, taxes and publc nvestment. Journal of Development Economcs, 34, 1991, Gang, I., Khan, H. Foregn ad and fscal behavour n a bounded ratonalty model: dfferent polcy regmes. Emprcal Economcs, 4, 1999, Gupta, S., Clements, B., Pvovarsky, A., and Tongson, E. R. Foregn Ad and Revenue Response: Does the Composton of Ad Matter?, Chapter 14 n S. Gupta, B. Clements,and G. Inchauste (ed.), Helpng Countres Develop: The Role of Fscal Polcy(Washngton: Internatonal Monetary Fund); Gupta, S., Clements, B.and Tareq, S. Moblzng Revenue, Fnance and Development, Vol. 45, Issue 3 (Washngton: Internatonal Monetary Fund); Gupta, K.L. Publc fscal behavour and foregn ad: some model solutons. Economc Modellng, 14, 1997, Heller, P. A model of publc fscal behavor n developngcountres: ad, nvestment and taxaton. Amercan Economc Revew, 65, 1975, Heller, P., and Gupta, S. More Ad Makng t Work for the Poor, WorldEconomcs, 3, No. 4,. 15- Khan, H.A., & Hoshno, E. Impact of foregn ad on fscal behavourof LDC governments. World Development,, 199,

14 64/ Crowdng out Effect of Foregn Ad n Selected Developng Countres 16- Knack, S. Ad Dependence and the Qualty of Governance: A CrosscountryEmprcal Analyss, Polcy Research Workng Paper Seres 396 (Washngton: TheWorld Bank),. 17-,. Soveregn Rents and the Qualty of Tax Polcy and Admnstraton, Journal of Comparatve Economcs, Elsever, 37(3), 9, pp Machado.R. Ad and Fscal Polcy n Ncaragua; A Fscal Response.Journal of Internatonal Development,, 1, Mavrotas, G. Foregn ad and fscal response: does ad dsaggregaton matter? Weltwrtschaftlches Archve, 138,, McGllvray, M. Ad and publc sector behavour n developng countres.revew of Development Economcs, 4,, Mc Gllvray, M. Ad, Economc Reform and Publc Sector Fscal Behavor n Developng Countres, Revew of Development Economc, 13(3), 9, McGllvray, M., Morrssey O. A Revew of Evdence on the Fscal Effects of Ad.Credt Research Papers. Unversty of Nottngham; McGllvray, M., & Ouattara, B. Ad debt burden and government fscal behavor n Cote d Ivoˆ re. Journal of Afrcan Economes, 14, 5, Mosley, P., Hudson, J. & Horrell, S. Ad, the publc sector and the market n less developed countres.economc Journal, 97, 1987, OECD,. DAC Onlne Statstcs (OECD, Pars); Moss, T., Pettersson, G. and Walle, N. An Ad-nsttutons Paradox? A RevewEssay on Ad Dependency and State Buldng n Sub-Saharan Afrca, In W. Easterly(Ed), Renventng Foregn Ad, Cambrdge, MA: The MIT Press; Otm, S. Foregn ad and government fscal behavour n low-ncome South Asan countres. Appled Economcs, No. 8, 1996, Ouattara, B. Froegn Ad and Government Fscal Behavor n Developng Countres: Panel Data Evdence, Economc Modelng, No. 3, 6, Remmer, K. Does Foregn Ad Promote the Expanson of Government?, Amercan Journal of Poltcal Scence, 48(1), 4, UN. Department of Economc and Socal Affars, World Economc and Socal Survey 1: Unted Naton Publcaton; World Bank, Global Development Indcators; World Bank, World Development Indcators; 1.