of 26

The Natural Resource Curse, Fiscal Decentralization, and Agglomeration Economies

Munch Personal RePEc Archve The Natural Resource Curse, Fscal Decentralzaton, and Agglomeraton Economes Raveh, Ohad Hebrew Unversty of Jerusalem, Economcs Depratment January 0 Onlne at
0 views26 pages
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Munch Personal RePEc Archve The Natural Resource Curse, Fscal Decentralzaton, and Agglomeraton Economes Raveh, Ohad Hebrew Unversty of Jerusalem, Economcs Depratment January 0 Onlne at MPRA Paper No. 9664, posted 9 Mar 0 9:06 UTC The Natural Resource Curse, Fscal Decentralzaton, and Agglomeraton Economes a Ohad Raveh b Hebrew Unversty of Jerusalem Abstract Natural resource abundance s a blessng for some countres, yet s a curse for others. The degree of fscal decentralzaton may account for ths dvergent outcome. Resources tend to locate n remote, non-agglomerated, and sparsely populated areas; a hgh degree of fscal decentralzaton gves a resource abundant regon an advantage n the nter-regonal tax competton over captal so that t attracts some captal from agglomerated and densely populated regons. Gven a suffcently hgh agglomeraton level, any such movement of captal would brng a loss of output n the agglomerated regon that outweghs the sum of gans from resource ncome and ncreased output n the remote regon so that aggregate product n the economy drops. Ths theory s emprcally tested -and confrmed- buldng on Sachs and Warner s nfluental works on the resource curse, employng the World Bank s Fscal Decentralzaton Indcators, and takng the Unted States as a case study. eywords: Natural Resources, Economc Growth, Resource Curse, Fscal Decentralzaton, Agglomeraton Economes, Tax Competton JE classfcaton: C, O3, O8, O57, Q33 a I thank Nathan Sussman and Omer Moav for ther gudance and support. Also, I thank the Socal Scences and Humantes Research Councl of Canada for provdng fnancal support. b Correspondence: Ohad Raveh, Economcs Department, Hebrew Unversty of Jerusalem, Mt. Scopus, Jerusalem, 9905, Israel . INTRODUCTION Snce the nfluental works of Sachs and Warner (995, 997, 999, 00) the so-called resource curse, descrbng an nverse relatonshp between resource abundance and economc growth, has attracted consderable attenton; further studes provded addtonal emprcal evdence of ths phenomenon as well as varous potental explanatons for ts occurrence. Nonetheless, the queston of why resource endowments s a blessng for some countres yet s a curse for others, remans a puzzle. ane and Tornell (996) suggest that the exstence of powerful groups n conjuncton wth weak nsttutons provde an explanaton, Mehlum et al. (006) argue t s rather strctly the qualty of nsttutons that matter, Hodler (004) provdes a smlar argument for the level of fractonalzaton, and fnally Andersen and Aslaksen (008) pont at consttutonal arrangements as a vable determnant. Ths paper contrbutes to ths strand of the lterature, as t presents an addtonal potental explanaton the level of fscal decentralzaton. 3 As a frst step, I follow Mehlum et al. (006) and plot n Fgure the average annual real per capta GDP growth from 970 to 990 versus resource abundance. 4 Panel (A) s based on all 48 countres of the data set, and provdes a strong ndcaton for an The meanng of resource-abundance should be properly defned, as t may carry some confuson wth t. For an extensve dscusson over the precse termnologes of natural resources see arou and Van der Zwaan (00). In ths paper the defnton used follows that whch s usually employed by economsts studyng the Dutch Dsease resource abundance refers to the amount of already exploted natural resources and reserves proven to be economcally explotable. For further dscusson see Auty 00, Gylfason 000, 00, Gylfason et al. 999, ane and Tornell 996, ete and Wedmann 999, Papyraks and Gerlagh 004, Rodrquez and Sachs 999, Sachs and Warner 997, 999, Fscal decentralzaton comprses the fnancal aspects of devoluton to regonal and local governments, and t covers two man nterrelated ssues; the frst s the dvson of spendng responsbltes and revenue sources between levels of government, and the second s the amount of dscreton gven to regonal and local governments to determne ther expendture and revenues. The defnton adopted n ths paper concerns both ssues, yet emphaszes the latter. 4 Note that G represents average annual real per capta GDP growth, whle R represents resource abundance (measured by prmary exports as share of total exports n 970; an elaborated dscusson over the resource measures used appears n the emprcal part of the paper). 3 occurrence of a resource curse. In panels (B) and (C), the sample s splt to two equal sub-samples accordng to the degree of fscal decentralzaton (a measure to be dscussed n-detal n the emprcal part of the paper); results ndcate that a resource curse appears n countres wth a relatvely hgher degree of fscal decentralzaton (panel (B)), yet t (A) All countres Average annual real per capta GDP growth Resource dependency G = R R-squared = 0.7, p 0.0 (B) Decentralzed countres (C) Centralzed countres Average annual real per capta GDP growth Resource dependency Average annual real per capta GDP Growth Resource dependency G = R R-squared = 0.89, p 0.03 G = R R-squared = 0.0, p 0.39 FIGURE. Resource-abundance and economc growth n: (A) all countres (B) decentralzed countres (C) centralzed countres completely dsappears n countres wth a relatvely lower degree of fscal decentralzaton (panel (C)). 5 Ths prelmnary result s mantaned when I control for 5 The countres n panel (B) are: Austra, Costa Rca, Ecuador, Fj, Fnland, Gamba, West Germany, Greece, Honduras, Iran, enya, Madagascar, Malaysa, Mexco, Norway, Panama, Paraguay, Senegal, Sr anka, Sweden, Swtzerland, Uruguay, Venezuela, and Zamba. The countres n panel (C) are: Australa, Canada, Chle, Colomba, Denmark, Domncan Republc, France, Inda, Indonesa, Ireland, Israel, Italy, 4 varous factors, as wll be evdent n the emprcal part of the paper. On ths bass, I assert that the degree of fscal decentralzaton provdes an addtonal explanaton for the dvergent effects of resource endowments. Based on the premse that resources tend to locate n remote, non-agglomerated, and sparsely populated regons, I construct a smple two-regon model of captal tax competton (extendng Zodrow and Meszkowsk s (986) basc captal tax competton model), where one of the regons exhbts agglomeraton economes. 6 The model shows that once an economy s fscally decentralzed a resource endowment n the remote regon gves t an advantage n the nter-regonal tax competton over the naton s captal, so that some amount of captal s attracted to t from the agglomerated regon; gven a suffcently hgh agglomeraton or populaton densty levels n the resource scarce regon, any such captal movement wll cause output loss n the resource scarce regon n magntude that outweghs the sum of gans of the resource ncome and the output ncrease n the resource rch regon, so that aggregate product n the economy drops. In addton, the model shows ths mechansm only ntensfes as nsttutonal qualty weakens (trggered by rent-seekng behavor rather than tax competton, as corrupton ncreases), allevatng concerns of ts exclusve applcablty to economes wth hgh nsttutonal qualty. To emprcally test the man hypothess, I use the World Bank s Fscal Decentralzaton Indcators n conjuncton wth Sachs and Warner s (997) data set. Addng the fscal decentralzaton measure and an nteracton term of t wth the resource abundance proxy to the regressons made by Sachs and Warner (997) provde sgnfcant results (n the expected drectons) that are robust to usng dfferent fscal decentralzaton measures as well as dfferent resource share proxes; these results hold when controllng for nvestment, openness, nsttutonal qualty, ethncty, terms of trade, secondary school enrollment, and (followng Hodler s (004) and Mehlum et al. s (006) fndngs) orea Republc, Malaw, Maurtus, Netherlands, Portugal, South Afrca, Span, Thaland, Trndad and Tobago, Tunsa, Unted ngdom, and Unted States. 6 Introduced by Marshall (90), the concept of agglomeraton economes refers to the postve externaltes of economc ntegraton at the local level, especally wth respect to ncreased labor market poolng, shared nputs, and knowledge spllovers. 5 nteracton terms of ethncty and nsttutonal qualty wth the resource share proxy. In addton, I construct a cross-country agglomeraton ndex and use t to show that decentralzed economes wth greater shares of non-agglomerated areas are potentally more vulnerable to the resource curse, as the model suggests. Fnally, emprcal evdence that support the man predctons of the model s provded, takng the Unted States as a case study. I fnd that resource abundance s assocated wth more compettve tax envronments, greater amounts of physcal captal per capta, and lower agglomeraton levels. Thus, movement of captal from agglomerated to non-agglomerated regons (due to resource abundance and a tax competton that follows) s observed, whch confrms the mechansm suggested. The paper s structured as follows Secton presents the setup of the model, and goes through the theoretcal analyss. Secton 3 provdes emprcal evdence for both the man hypothess as well as for the mechansm of the model, and undertakes robust checks to the emprcal fndngs. Secton 4 concludes.. THE MODE et us consder the benchmark settng of the model, under the framework of the basc captal tax competton model of Zodrow and Meszkowsk (986), n ts smplest form. There exsts an economy wth two symmetrc regons, each havng a manufacturng sector. Producton n each regon s undertaken by captal () and labor (), employed through a constant returns to scale neoclasscal producton functon that follows the Inada Condtons (F(,)); t takes place n the manufacturng sector, to produce a fnal good (Y) that s ether consumed (X) or converted to a pure publc good (G). The startng populaton sze of each regon s (where N and ). 7 abor market s nelastc so that each resdent s employed and provdes one unt of labor. Thus, we have: Y F(, ) X G () 7 Note that throughout the paper represents the regon, where (, ). Also n terms of notaton, subscrpts represent the regon, whle superscrpts represent the sector; n addton, captal letters represent level varables, whle small ones represent per capta terms. 6 There s a fxed supply of captal n the economy (where * ), that s equally owned by ts resdents (so that each owns: * / N k * ). To be able to focus on the effects of captal moblty (followng the emprcal observatons) I assume captal s perfectly (and costlessly) moble, whle labor s completely mmoble. 8 Snce the focus s on decentralzed economes, I take the smplfyng assumpton that regons have complete autonomy on determnaton of tax rates well as on tax retenton and usage; thus, each regon has a government that leves a per-unt, source-based, captal tax to fnance a pure publc good, so that: G T () The after-tax rate of return on captal s ; although determned endogenously (by the free captal moblty condton, whch wll be presented later), s taken as gven by each regon. Followng that, the pre-tax rate of return on captal would be T. There are many frms (each beng a prce taker) operatng n each of the regons, and there s free entry to the market. Captal markets are compettve so that proft maxmzaton by each frm yelds: 9 f k ( k ) T (3) Also, the free entry condton yelds: 0 w f ( k ) f k (4) k Resdents of ths economy have dentcal preferences, represented by a strctly quasconcave utlty functon, U(X,G), wth the followng propertes: 8 In a prevous paper I show that on a regonal bass captal flows to regons rch n pont-source resources, whle labor does not flow to regons rch n dffuse-source resources (results avalable from the author). Therefore, n ths paper I focus on captal moblty, leavng labor mmoble for smplcty. In addton, note that n the context of ths model, captal refers specfcally to physcal captal, rather than to other types of captal (such as fnancal captal, or human captal). 9 Proft of a representatve frm n ether of the regons s: f ( k ) ( T ) k w ) Therefore, proft would be maxmzed at: d / dk 0 0 The free entry condton mposes 0, for all frms n the naton. ( 7 U X, U 0, U, U 0, U 0; n addton, they own equal shares of the frms (n G XX GG XG ther respectve regons). Therefore, gven that resdents spend all ther ncome on prvate consumpton, a representatve resdent s budget constrant would be: x * f ( k ) ( T ) k k (5) Each regon competes for the economy s captal stock, by means of tax competton (so that a captal tax competton arses, modeled along Cournot-Nash lnes). Ths s a statc, one-perod model, where the order of events s as follows each regon sets ts captal tax level, based on whch captal s reallocated across the economy; ths, n turn, determnes the regonal wage and publc goods levels (whch sets the per-capta utlty levels). That sad, by equaton (3) each regon derves k T ) ( so that t can vary k by ts choce oft. Totally dfferentatng equaton (3) wth respect to k andt, we get: dk dt 0 f k k (6) By equaton (), we get: dg dt dk k T (7) dt Also, by dfferentatng equaton (5) wth respect to T and substtutng equaton (6), we get: dx dt k (8) Each regon ams to set the tax level that would maxmze the welfare of ts resdents. eepng ths objectve n mnd, each regon would, thus, maxmze the utlty of a representatve resdent, subject to the budget constrants of the regon and the resdent. In effect, makng X and G normal goods wth dmnshng returns. In addton, t s assumed that margnal utltes of X and G go to nfnty as each approaches zero, or otherwse go to zero as each approaches nfnty (smlar to the Inada Condtons of the producton sde). 8 Therefore, n ts smplest form the problem of each of the regons would be expressed as follows: MaxU ( x, G ) { T } et us denote U U by m x, G ); thus, we get: 3 dx dt G X dg m( x, G ) 0 dt ( Substtutng equatons (7) and (8) to equaton (9) and rearrangng, we get: 4 m( x, G ) (0) T dk k dt In equlbrum, the followng captal moblty condton must hold: 5 f k T f k T () Therefore, n equlbrum equatons (0) (for each of the regons) and () must hold, whch means that captal tax rates wll be postve n each of the regons, n equlbrum. (9). Agglomeraton Economes Departng from the benchmark settng let us now assume there exsts agglomeraton economes n Regon. 6 Followng Cccone and Hall (996), ths agglomeraton effect s Note that gven the assumptons made on the utlty functon, as well as based on the settng of the problem, there would be an nteror soluton to the gven problem, n each of the regons, such thatt, k, G, x 0. Therefore, corner solutons are not consdered n ths case. 3 Ths was derved by totally dfferentatng U ( x, G ) wth respect to x and. 4 The followng result replcates that whch was derved by Zodrow and Meszkowsk (986). It can be gven a Modfed Samuelson Condton (Batna 990) nterpretaton, showng how the publc good s undersuppled n each of the regons due to the non-cooperatve behavor. To emphasze ths pont further a MCPF (Margnal Cost of Publc Funds) nterpretaton can be adopted here as well (Brownng, 976), showng how n equlbrum each of the regons wll face excess costs when rasng an addtonal unt of revenue, caused by the usage of dstortonary taxes and the tax competton. 5 Captal wll place where ts margnal product s hgher, untl t s equated across regons. G effects. 9 Followng the analyss of the benchmark settng, nothng changes n the behavor 9 modeled by addng ( 0, and s exogenously determned) 7 to the producton ( undertaken n Regon (such that Y F, ) ). In terms of descrbng agglomeraton effects, ths addton has three nterpretatons 8 frstly t ndcates labor s more complementary to captal n the agglomerated regon, secondly t mples that the agglomerated regon presents hgher wage rates, and thrdly t says that producton n the agglomerated regon exhbts ncreasng returns to scale n labor and captal (thus, makng workers n Regon more productve than those of Regon ). In effect, represents the degree of agglomeraton, where a hgher means greater agglomeraton of Regon, whereas n Regon equatons (3) through (6) now become: f ( k) T k w k ( f ( k ) f k ) () (3) x * f ( k) ( T ) k k (4) dk dt 0 (5) f k k Thus, gong through the maxmzaton problem of Regon, and pluggng equatons ()- (5) as was done prevously, we get the followng reacton functon: m( x, G ) T k f k k (6) 6 Ths extenson relates to Bucovetsky s (99) work; however, whle he sourced the asymmetry between the regons n ther relatve populaton szes, n ths work no assumpton s made over the relatve regonal populaton szes and the asymmetry s manly sourced n the agglomeraton effects. 7 The assumpton that s exogenously determned captures the dea that once resources are dscovered the economy s at a certan agglomeraton level that s exogenous to the resource dscovery. 8 Each such nterpretaton follows the defnton of agglomeraton as descrbed by rugman (99), Cccone and Hall (996), and others. 9 Note that n terms of Cccone and Hall (996) represents both the effects of congeston and agglomeraton, because n ths model acreage does not vary between regons. 0 Also, equaton () now becomes: f k T f k T As before, equatons (6) (for each of the regons) and (7) must hold n equlbrum, so that regonal tax rates would be postve, n equlbrum. (7). The Natural Resource Curse n Decentralzed Countres et us assume Regon receves an exogenous and mmoble resource endowment (Q), whch t taxes on a lump-sum bass (z). 0 Therefore, n ths case the regonal budget constrant would be: G T z (8) Total output of Regon would now be: Y X G F(, ) Q (9) * The resource s equally owned by resdents of Regon (so that: q Q / ) and t provdes an exogenously-determned rate of return of. Therefore, the budget constrant of a representatve resdent n Regon would be: x k * * f ( k ) f k z / k q (0) Once agan, the regons engage n a captal tax competton. Note that Regon behaves accordng to the analyss presented prevously; therefore, let us see how the stuaton changes n Regon, as ts problem s analyzed as follows: MaxU ( x, ) G { T, z} Substtutng equatons (0) and (8) to the gven problem, we get the followng frst order condtons: U () X U G dx dg U X U 0 G () dt dt 0 Note that results do not change f otherwse a per-unt, source based (and thus dstortonary), tax s mposed nstead of the lump-sum one. Usage of lump-sum tax smplfes the analyss. Ths, n fact, follows the reasonng of Zodrow and Meszkowsk (986) who used a lump sum tax as well (wth the ntroducton of local publc servces) n ther analyss. Note that dx dg, dt dt are dentcal n computaton to equatons (7) and (8), only wth the correspondng notaton. Thus, f we substtute these to the frst order condtons and solve, we get the followng: T 0 (9) Ths means that f the lump sum tax on the resource rents s unrestrcted or that otherwse the dscovered resource s substantal enough (n the sense that suffcent taxes can be leved on the resource rents so that the effcent level of publc good s suppled) then Regon can, n fact, effcently lower ts captal tax rates to zero, whle as was seen n the prevous analyss, the tax rate of Regon remans postve. Ths emphaszes the fscal advantage the resource gves to the regon n whch t was found. f Therefore, the captal moblty condton now becomes: T k k f Comparng equaton (0) to (7) we see that once the resource s ntroduced to Regon (so that t can engage n the nter-regonal tax competton more aggressvely) at least some postve amount of captal wll be drawn to t (sncet T 0 ); let us mark ths amount as m. In effect, ths means that m amount of captal moves from the agglomerated regon to the non-agglomerated one; therefore, n case the output loss from the msusage of those m from both the usage of the captal n the agglomerated regon outweghs the output gan m captal n the non-agglomerated regon as well as the resource revenues themselves, then ths explans why resource abundance can be a curse n decentralzed countres. et us show ths n terms of the model, and solve for the agglomeraton level (n Regon ) above whch a resource endowment (n Regon ) would lower the overall output level of the economy. as ~ et us denote the equlbrum level of captal n each regon (under scenaro (0)). Ths means that once the resource s ntroduced and result (0) s derved (so that (0) The cases of a restrcted z or a relatvely small resource dscovery a
View more
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!